Skip to content

Instantly share code, notes, and snippets.

@vanbasten23
Created December 4, 2025 18:29
Show Gist options
  • Select an option

  • Save vanbasten23/abc3b2db22748e1f16aa0e92c6b91404 to your computer and use it in GitHub Desktop.

Select an option

Save vanbasten23/abc3b2db22748e1f16aa0e92c6b91404 to your computer and use it in GitHub Desktop.
exec ${PAGER:-/usr/bin/less -R} "$0" || exit 1
Test settings: forge with network access
Host details: itmm4.prod.google.com Linux 6.6.65-smp-1300.170.0.0 x86_64 astoria-genoa-base
executor.INFO: analog/view?storage=borgremote&bns=/bns/it/borg/it/bns/build-forge-executor-tpu/prod-cbf-ghostlite.forge-executor/0&min_time=1764872604000000&ts=1764872614000000
Test command:
cd /build/work/aef67bf50706fee86777a93cc065340a246c/google3/runfiles/google3 && \
env - \
BORG_CELL=it \
CUSTOM_METRICS_DIR=/build/work/aef67bf50706fee86777a93cc065340a246c/google3/../custom_metrics \
GCOV_PREFIX=/build/work/aef67bf50706fee86777a93cc065340a246c/google3/coverage \
GCOV_PREFIX_STRIP=3 \
GUNIT_OUTPUT=xml:/build/work/aef67bf50706fee86777a93cc065340a246c/google3/blaze-out/k8-fastbuild/testlogs/experimental/users/jevinjiang/ullm/tests/ragged_paged_attention_v3_test_gl/shard_1_of_50/test.xml \
HOME=/build/work/aef67bf50706fee86777a93cc065340a246c/google3/tmp \
JAVAHASHSEED=random \
JAVA_RUNFILES=/build/work/aef67bf50706fee86777a93cc065340a246c/google3/runfiles \
JVM_FLAGS=-Xverify:none -XX:+UnlockExperimentalVMOptions -XX:SymbolTableSize=2221217 -XX:-UnlockExperimentalVMOptions -Xms256m -Xmx2048m -XX:+ErrorFileToStderr \
LD_LIBRARY_PATH=/build/work/aef67bf50706fee86777a93cc065340a246c/google3/_solib:/build/work/aef67bf50706fee86777a93cc065340a246c/google3/runfiles/google3/_solib__third_Uparty_Scrosstool_Sv18_Sstable_Ccc-compiler-k8-llvm.k8:/build/work/aef67bf50706fee86777a93cc065340a246c/google3/runfiles/google3/_solib_k8 \
LOCKSERVICE_LOCALITY=test \
LOCKSERVICE_USE_PROXY=preferred \
LOGNAME=forge-00 \
NPROC=4 \
OLDPWD=/build/work/aef67bf50706fee86777a93cc065340a246c/google3 \
PATH=/usr/buildtools/buildhelpers/v5/bin:/bin:/usr/bin:/usr/sbin:/sbin:/build/work/aef67bf50706fee86777a93cc065340a246c/google3/tools/test \
[email protected] \
PORTSERVER_PORT=10010 \
PWD=/build/work/aef67bf50706fee86777a93cc065340a246c/google3/runfiles/google3 \
PYTHONPATH=/usr/local/buildtools/current/sitecustomize \
PYTHON_RUNFILES=/build/work/aef67bf50706fee86777a93cc065340a246c/google3/runfiles \
SHLVL=1 \
SVELTE_SERVERS=test \
TESTBRIDGE_TEST_ONLY=basic \
TEST_BINARY=experimental/users/jevinjiang/ullm/tests/ragged_paged_attention_v3_test_gl \
TEST_COMPONENTS_DIR=/build/work/aef67bf50706fee86777a93cc065340a246c/google3/blaze-out/k8-fastbuild/testlogs/experimental/users/jevinjiang/ullm/tests/ragged_paged_attention_v3_test_gl/shard_1_of_50/test.test_components \
TEST_DIAGNOSTICS_OUTPUT_DIR=/build/work/aef67bf50706fee86777a93cc065340a246c/google3/blaze-out/k8-fastbuild/testlogs/experimental/users/jevinjiang/ullm/tests/ragged_paged_attention_v3_test_gl/shard_1_of_50/test.test_diagnostics \
TEST_FAILURE_UNDECLARED_OUTPUTS_DIR=/build/work/aef67bf50706fee86777a93cc065340a246c/google3/blaze-out/k8-fastbuild/testlogs/experimental/users/jevinjiang/ullm/tests/ragged_paged_attention_v3_test_gl/shard_1_of_50/test.outputs/failure_only \
TEST_INFRASTRUCTURE_FAILURE_FILE=/build/work/aef67bf50706fee86777a93cc065340a246c/google3/blaze-out/k8-fastbuild/testlogs/experimental/users/jevinjiang/ullm/tests/ragged_paged_attention_v3_test_gl/shard_1_of_50/test.infrastructure_failure \
TEST_LOGSPLITTER_OUTPUT_FILE=/build/work/aef67bf50706fee86777a93cc065340a246c/google3/blaze-out/k8-fastbuild/testlogs/experimental/users/jevinjiang/ullm/tests/ragged_paged_attention_v3_test_gl/shard_1_of_50/test.raw_splitlogs/test.splitlogs \
TEST_PREMATURE_EXIT_FILE=/build/work/aef67bf50706fee86777a93cc065340a246c/google3/blaze-out/k8-fastbuild/testlogs/experimental/users/jevinjiang/ullm/tests/ragged_paged_attention_v3_test_gl/shard_1_of_50/test.exited_prematurely \
TEST_SHARD_INDEX=0 \
TEST_SHARD_STATUS_FILE=/build/work/aef67bf50706fee86777a93cc065340a246c/google3/blaze-out/k8-fastbuild/testlogs/experimental/users/jevinjiang/ullm/tests/ragged_paged_attention_v3_test_gl/shard_1_of_50/test.shard \
TEST_SIZE=medium \
TEST_SRCDIR=/build/work/aef67bf50706fee86777a93cc065340a246c/google3/runfiles \
TEST_TARGET=//experimental/users/jevinjiang/ullm:tests/ragged_paged_attention_v3_test_gl \
TEST_TIMEOUT=900 \
TEST_TMPDIR=/build/work/aef67bf50706fee86777a93cc065340a246c/google3/tmp \
TEST_TOTAL_SHARDS=50 \
TEST_UNDECLARED_OUTPUTS_ANNOTATIONS_DIR=/build/work/aef67bf50706fee86777a93cc065340a246c/google3/blaze-out/k8-fastbuild/testlogs/experimental/users/jevinjiang/ullm/tests/ragged_paged_attention_v3_test_gl/shard_1_of_50/test.outputs_manifest \
TEST_UNDECLARED_OUTPUTS_DIR=/build/work/aef67bf50706fee86777a93cc065340a246c/google3/blaze-out/k8-fastbuild/testlogs/experimental/users/jevinjiang/ullm/tests/ragged_paged_attention_v3_test_gl/shard_1_of_50/test.outputs \
TEST_UNUSED_RUNFILES_LOG_FILE=/build/work/aef67bf50706fee86777a93cc065340a246c/google3/blaze-out/k8-fastbuild/testlogs/experimental/users/jevinjiang/ullm/tests/ragged_paged_attention_v3_test_gl/shard_1_of_50/test.unused_runfiles_log \
TEST_WARNINGS_OUTPUT_FILE=/build/work/aef67bf50706fee86777a93cc065340a246c/google3/blaze-out/k8-fastbuild/testlogs/experimental/users/jevinjiang/ullm/tests/ragged_paged_attention_v3_test_gl/shard_1_of_50/test.warnings \
TEST_WORKSPACE=google3 \
TZ=US/Pacific \
UNITTEST_ON_BORG=1 \
UNITTEST_ON_FORGE=1 \
USER=forge-00 \
XML_OUTPUT_FILE=/build/work/aef67bf50706fee86777a93cc065340a246c/google3/blaze-out/k8-fastbuild/testlogs/experimental/users/jevinjiang/ullm/tests/ragged_paged_attention_v3_test_gl/shard_1_of_50/test.xml \
_=/usr/buildtools/buildhelpers/v5/bin/env \
experimental/users/jevinjiang/ullm/tests/ragged_paged_attention_v3_test_gl --xla_tpu_enable_log_recorder
-- 2025-12-04 10:23:34 PST Forge runner: Test started
-----------------------------------------------------------------------------
Running tests under Python 3.12.11: embedded.
I1204 10:23:36.782620 2807 init.cc:78] Remote crash gathering hook installed.
I1204 10:23:36.784829 2807 profilez_handler.cc:59] /censusprofilez handler registered
I1204 10:23:36.786599 2807 python_stack_size.cc:37] EventManager thread stack size of 131072 might be too small for Python callbacks. Crash? Read go/py-cpp-stack.
I1204 10:23:36.787881 2807 python_stack_size.cc:37] EventManager thread stack size of 131072 might be too small for Python callbacks. Crash? Read go/py-cpp-stack.
I1204 10:23:36.789636 2807 profilez.cc:124] Census profiling enabled
I1204 10:23:36.794162 2807 census-init.cc:276] Census enabled
I1204 10:23:36.796262 2807 init_experiments.cc:49] Stubby experiments status: monitoring_experiment: enabled, test_stubby_experiment_debug_wave_and_step: enabled, test_stubby_experiment_default_false: enabled
I1204 10:23:36.799571 2807 registration.cc:61] go/serverthrottler-in-stubby is disabled by configuration.
I1204 10:23:36.799948 2807 add_ptoken_logwriter.cc:82] Registering AddPTokenLogWriter factory
I1204 10:23:36.836049 2807 migration_state_tracking.cc:23] Migration not enabled - not starting notification watcher.
I1204 10:23:36.836084 2807 tpu_hal.cc:436] Registered plugin from module: breakpoint_debugger_server
I1204 10:23:36.837022 2807 util.cc:262] Bweclient: Not running in Borg: using defaults
I1204 10:23:36.839251 2807 tpu_runtime_type_flags.cc:50] TFRT TPU Runtime enabled for JAX.
I1204 10:23:36.839342 2807 init-domain.cc:126] Fiber init: default domain = switchto, concurrency = 422, prefix = switchto-default
I1204 10:23:36.840129 2807 privacy_context_defaults.cc:3717] PrivacyContext default attribute configuration initialization.
I1204 10:23:36.840153 2807 privacy_context_defaults.cc:2191] Post initialization validating PrivacyContextDefaultConfig: go/debugonly
I1204 10:23:36.840392 2807 privacy_context_defaults.cc:3862] PrivacyContextDefaultConfig has no origin_product_id.
I1204 10:23:36.840400 2807 privacy_context_defaults.cc:3879] PrivacyContext default attribute configuration module initialization complete.
I1204 10:23:36.840407 2807 fireaxe_privacy_context_optout_listener_init.cc:12] FireaxePrivacyContextOptoutListener initialization.
I1204 10:23:36.840465 2807 fireaxe_privacy_context_optout_listener_init.cc:16] FireaxePrivacyContextOptoutListener initialization complete.
I1204 10:23:36.840472 2807 fireaxe_privacy_context_breakglass_watcher_init.cc:12] FireaxePrivacyContextBreakglassWatcher initialization.
I1204 10:23:36.840510 2807 fireaxe_privacy_context_breakglass_watcher_init.cc:16] FireaxePrivacyContextBreakglassWatcher initialization complete.
I1204 10:23:36.850556 2807 python_stack_size.cc:37] EventManager thread stack size of 131072 might be too small for Python callbacks. Crash? Read go/py-cpp-stack.
I1204 10:23:36.850798 2807 python_stack_size.cc:37] EventManager thread stack size of 131072 might be too small for Python callbacks. Crash? Read go/py-cpp-stack.
I1204 10:23:36.851173 2807 python_stack_size.cc:37] EventManager thread stack size of 131072 might be too small for Python callbacks. Crash? Read go/py-cpp-stack.
I1204 10:23:36.851907 2807 chrometracing.cc:168] Writing Chrome trace_events (for chrome::tracing) to /build/work/aef67bf50706fee86777a93cc065340a246c/google3/blaze-out/k8-fastbuild/testlogs/experimental/users/jevinjiang/ullm/tests/ragged_paged_attention_v3_test_gl/shard_1_of_50/test.outputs/chrometracing/ctrace.ragged_paged_at.2807.trace
I1204 10:23:36.852472 2807 lamon.cc:155] Assuming running on corp, analytics enabled.
I1204 10:23:36.853667 2807 loas2_securewrapper_multihop_aes_128_gcm_openssl.cc:60] Failed to get registered cross universe signing client. This is only a problem if your binary needs to send RPCs across TPC universes; otherwise, no action is needed. If your binary needs to send RPCs across TPC universes, rebuild your binary with the required LOAS artifacts dependency for TPC: [ //security/loas/l2/public/multihop:loas2_multihop_artifacts ]. This requirement is part of the go/tpc-rpc-integrity-thinmint-direction rollout.
I1204 10:23:36.854031 2807 chubby_util.cc:217] LOCKSERVICE_LOCALITY env: test
lockservice_locality flag:
machine locality: it
BORG_PHYSICAL_CELL env:
I1204 10:23:36.854058 2807 host_command_handler_factory.cc:31] Skipping registration of host command handler for opcode Log because it is not in the allowlist.
I1204 10:23:36.854573 2807 lockservice_accurate_time.cc:36] Not using TrueTime, fall back to regular impl.
I1204 10:23:36.854820 2807 tpu_hal.cc:436] Registered plugin from module: megascale_sync_flag_logger
I1204 10:23:36.855009 2807 net_uid.cc:46] GOOGLE_BW_UID available
I1204 10:23:36.855024 2807 lockservice_accurate_time.cc:59] AccurateTime was already initialized.
I1204 10:23:36.855443 2807 pmuprofiler.cc:382] Kernel supports perf_events
I1204 10:23:36.857913 2807 config_parser.cc:116] Parsing Network Predicates (version 20251117.0)
I1204 10:23:36.858560 2807 api.cc:181] Network Predicates loading 2734 good zones.
I1204 10:23:36.878069 2807 string-registry.cc:32] Registering message 'scaffolding.TempusId ("util/scaffolding/internal/tempus-id.proto")'
I1204 10:23:36.878248 2807 pjrt_api.cc:96] PJRT_Api is set for device type tpu
I1204 10:23:36.880002 2807 tf_tpu_flags.cc:63] DeepseaPlatform is NOT registered.
I1204 10:23:36.880047 2807 thin_uls_logwriter.cc:192] Registering thin-client logwriter factory
I1204 10:23:36.880216 2807 tix.cc:223] Registering TixRepository for type RecordIO
I1204 10:23:36.880262 2807 tpu_hal.cc:436] Registered plugin from module: sdc_checker_callback
I1204 10:23:36.880398 2807 logwriter_uls_init.cc:57] No Envelope. Disabled initialization of LogWriterULS. LogWriter prefixes /sawmill{,-transitional,-dual-logging,-localhost,-dryrun,-devnull,-forwarding-proxy,-lite-test} will be interpreted as Google3 File paths instead of LogWriter implementations, so /sawmill will not use ULS.
I1204 10:23:37.058224 2807 tpu_hal.cc:436] Registered plugin from module: xsc_explicit_checksum_tracing_callback
I1204 10:23:37.058637 2837 lamon.cc:312] Starting the 'FlushCountersToMonarch' thread.
I1204 10:23:37.058696 2817 default_exporter.cc:63] [streamz] constructing default BulkExporter instance (for go/bulk-streamz).
I1204 10:23:37.058725 2817 default_exporter.cc:41] [streamz] hermetic behavior of the default BulkExporter instance: NOT_UNDER_BORGLET
I1204 10:23:37.059106 2807 app.py:425] C++ logging enabled; logging.debug will only be visible when --verbosity/-v are 1 or higher
I1204 10:23:37.060863 2807 app.py:335] Process id 2807
I1204 10:23:37.060941 2807 app.py:337] Current working directory /build/work/aef67bf50706fee86777a93cc065340a246c/google3/runfiles/google3
I1204 10:23:37.061166 2807 app.py:340] Built on Wed Dec 31 15:59:59 1969 (-1)
Built by
Built as //experimental/users/jevinjiang/ullm:tests/ragged_paged_attention_v3_test_gl
Build platform:
Currently running under Python 3.12.11: embedded.
W1204 10:23:37.061202 2807 app.py:342] DEBUG BINARY -- Performance may suffer
I1204 10:23:37.061223 2807 app.py:343] Command line arguments:
I1204 10:23:37.061243 2807 app.py:345] argv[0]: '/build/work/aef67bf50706fee86777a93cc065340a246c/google3/runfiles/google3/experimental/users/jevinjiang/ullm/tests/ragged_paged_attention_v3_test_gl'
I1204 10:23:37.061260 2807 app.py:345] argv[1]: '--xla_tpu_enable_log_recorder'
[ RUN ] RaggedPagedAttentionKernelTest.test_ragged_paged_attention_basic0 (dtype=<class 'jax.numpy.float32'>)
I1204 10:23:37.070154 2807 pjrt_client.cc:550] PjRt-IFRT device count: total=1, addressable=1
I1204 10:23:37.070173 2807 pjrt_client.cc:554] Addressable PjRt-IFRT device: CpuDevice(id=0)
I1204 10:23:37.070324 2807 pjrt_api.cc:167] The PJRT plugin has PJRT API version 0.82. The framework PJRT API version is 0.82.
I1204 10:23:37.072815 2807 singleton_tpu_states_manager.cc:96] TpuStatesManager::GetOrCreate(): no tpu system exists. Creating a new tpu system.
I1204 10:23:37.082681 2807 device_util.cc:160] Found 1 ghostlite chips.
I1204 10:23:37.082706 2807 tpu_version_flag.cc:62] Using auto-detected TPU version ghostlite
I1204 10:23:37.091097 2807 device_util.cc:160] Found 1 ghostlite chips.
I1204 10:23:37.099368 2807 device_util.cc:160] Found 1 ghostlite chips.
I1204 10:23:37.107507 2807 device_util.cc:160] Found 1 ghostlite chips.
I1204 10:23:37.107525 2807 flags_util.cc:197] Using default chip configuration.
I1204 10:23:37.107564 2807 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:37.108683 2807 pending_event_logger.cc:928] Enabling PjRt/TPU event dependency logging
I1204 10:23:37.117256 2901 device_util.cc:160] Found 1 ghostlite chips.
I1204 10:23:37.117448 2901 flags_util.cc:315] Picked unused port 10011 as SliceBuilder worker service port.
I1204 10:23:37.125604 2901 device_util.cc:160] Found 1 ghostlite chips.
I1204 10:23:37.125628 2901 tpu_network_factory.cc:104] Using TpunetdClient
I1204 10:23:37.125683 2901 tpunetd_client.cc:132] Creating TpunetdClient with topology go/debugonly
x_dimension: 1
y_dimension: 1
x_wrapping: false
y_wrapping: false
I1204 10:23:37.674173 2902 async_driver.cc:476] [/dev/vfio/131 itmm4:pe2:0] vf_id:0 Driver opened.
W1204 10:23:37.681697 2902 async_driver.cc:1791] All cores not supported.
I1204 10:23:37.814792 2901 flatbuffer_acl_reader_feature_util_google3.cc:85] Successfully loaded FlatbufferAclReader from file at /etc/security-realms/membership_set_fb.bin
I1204 10:23:37.831835 2901 config.cc:256] gRPC experiments enabled: cache_loas_context_in_connection_context, callv3_census_client, callv3_census_server, channelz_use_v2_for_v1_api, channelz_use_v2_for_v1_service, channelz_zviz, chttp2_bound_write_size, deprecate_keep_grpc_initialized, directpath_rpc_header_sanitization_filter, error_flatten, event_engine_callback_cq, event_engine_channelz_socket_info, event_engine_client, event_engine_dns, event_engine_dns_non_client_channel, event_engine_listener, event_engine_max_read_chunk_size, event_engine_secure_endpoint, fuse_filters, google_no_envelope_resolver, graceful_external_connection_failure, lbns_support_in_address_resolver, loas2_protect_memory_optimization, loas_server_auth_decode_uris_from_grpc, max_age_filter_float_to_top, max_inflight_pings_strict_limit, monitoring_experiment, namecheck_core_lib, privacy_context_remove_double_encoding, privacy_context_single_encoding, prod2cloud_w3c_trace, rr_wrr_connect_from_random_index, sleep_use_non_owning_waker, subchannel_wrapper_cleanup_on_orphan, track_writes_in_resource_quota, tsi_frame_protector_without_locks
I1204 10:23:37.832180 2901 print_trace_every_n.h:24] DNS resolver being used from google3. No-envelope resolver linked in: 1
I1204 10:23:37.832189 2901 init.cc:350] Google3 binary running without envelope: the default URI scheme for gRPC client channels will be 'dns' instead of 'google-no-envelope'. If you need google3 naming features, please link the appropriate envelope loader (e.g. //tech/env:envelope_loader) and //net/grpc/no_envelope:grpc++_google_resolver.
I1204 10:23:37.842763 2824 generic-stream-socket.h:2296] cannot enable TCP_INQ on fd=172 err=2
W1204 10:23:37.842956 2823 traffic_class_context.cc:41] Could not retrieve TOS of the connection: INVALID_ARGUMENT: fd 172 is a non-IP socket: Protocol wrong type for socket
=== Source Location Trace: ===
util/task/errno_mapping.cc:14
I1204 10:23:37.863456 2901 psp_machine.cc:613] Local IPs detected: {127.0.0.1, ::1, 2002:a05:660d:4:b0:9da:1b2c:c356, 2002:a05:660d:c:b0:9da:1b2c:c356, 2002:a05:660d:e:b0:9da:1b2c:c356, 2002:a05:660d:10:b0:9da:1b2c:c356, fd14:988a:50ee:10c:b0:9da:1b2c:c356}; netns: true
I1204 10:23:37.863560 2901 psp_machine.cc:877] Found PSP NIC instance /sys/devices/pci0000:e0/0000:e0:01.2/0000:e1:00.0/psp for local address 2002:a05:660d:10:b0:9da:1b2c:c356
I1204 10:23:37.863603 2901 psp_machine.cc:877] Found PSP NIC instance /sys/devices/pci0000:00/0000:00:01.1/0000:01:00.0/psp for local address 2002:a05:660d:4:b0:9da:1b2c:c356
I1204 10:23:37.863627 2901 psp_machine.cc:877] Found PSP NIC instance /sys/devices/pci0000:20/0000:20:01.1/0000:21:00.0/psp for local address 2002:a05:660d:c:b0:9da:1b2c:c356
I1204 10:23:37.863645 2901 psp_machine.cc:877] Found PSP NIC instance /sys/devices/pci0000:c0/0000:c0:01.1/0000:c1:00.0/psp for local address 2002:a05:660d:e:b0:9da:1b2c:c356
I1204 10:23:37.864404 2901 psp_support.cc:405] HTTPServer2 initially intends to listen on {[::]:10011,PSP-incapable}, After the PSP IP expansion, it attempts to listen on {[2002:a05:660d:4:b0:9da:1b2c:c356]:10011,PSP-capable; [2002:a05:660d:c:b0:9da:1b2c:c356]:10011,PSP-capable; [2002:a05:660d:e:b0:9da:1b2c:c356]:10011,PSP-capable; [2002:a05:660d:10:b0:9da:1b2c:c356]:10011,PSP-capable; [fd14:988a:50ee:10c:b0:9da:1b2c:c356]:10011,PSP-incapable; 127.0.0.1:10011,PSP-incapable; [::1]:10011,PSP-incapable} with an accept queue length of { 2147483647 }
I1204 10:23:37.865081 2969 source_selector.cc:41] Reading ACLs from Chubby: flag ganpati_acl_client_force_chubby_acls is set.
I1204 10:23:37.865131 2901 expvar_global_linux_google.cc:135] [varz] process-gcu-seconds is not defined: Value of BORG_MILLIGCUS_PER_CPU is undefined
I1204 10:23:37.865257 2901 ip_validation.cc:183] Skipping initialization of special IPs, since dirinfo_root does not exist.
I1204 10:23:37.865738 2901 rpc-internals-version.cc:158] RPC_VERSION: Saw use of RPC version 2 server
I1204 10:23:37.865870 2901 rpc-internals-rate-acl.cc:331] Default rate ACL config: # Default rate ACL is OPEN
0 0 unknown unknown unknown unknown
I1204 10:23:37.865952 2901 rate_acl_impl.cc:972] new rate_acl checksum: 0xc31012a739f03e1b version: -1
I1204 10:23:37.866343 2901 rate_acl_impl.cc:2009] File rpcacl_forge-00_experimental%users%jevinjiang%ullm%tests%ragged_paged_attention_v3_test_gl is disabled by kRateAclExcludedPatterns
I1204 10:23:37.866491 2901 action_watcher.cc:141] Fireaxe: Not watching Chubby: INVALID_ARGUMENT: Unknown task identity: Running under Forge
=== Source Location Trace: ===
production/fireaxe/cpp/watcher/identity.cc:144
I1204 10:23:37.866736 2970 privacy_context_defaults.cc:940] Processing default configuration update for source: 2001
I1204 10:23:37.867073 2970 privacy_context_defaults.cc:940] Processing default configuration update for source: 2001
I1204 10:23:37.867120 2970 privacy_context_defaults.cc:940] Processing default configuration update for source: 2001
I1204 10:23:37.867250 2961 label_acl_impl.cc:708] File rpcacl_forge-00_experimental%users%jevinjiang%ullm%tests%ragged_paged_attention_v3_test_gl is disabled by kAclExcludedPatterns.
I1204 10:23:37.867476 2901 rpc_service_registered_collector.cc:274] UNIMPLEMENTED: RPC Inventory Envelope service is disabled.
=== Source Location Trace: ===
security/dataprotection/rpcinventory/rpc_service_registered_collector.cc:292
I1204 10:23:37.867833 2901 uberproxy_groups_cpp_dynamic.cc:16] We are running in GDU
I1204 10:23:37.868303 2901 internals-datagram-socket.cc:98] Binding RPC2 UDP socket to [2002:a05:660d:4:b0:9da:1b2c:c356]:10011
I1204 10:23:37.868403 2901 internals-datagram-socket.cc:98] Binding RPC2 UDP socket to [2002:a05:660d:c:b0:9da:1b2c:c356]:10011
I1204 10:23:37.868469 2901 internals-datagram-socket.cc:98] Binding RPC2 UDP socket to [2002:a05:660d:e:b0:9da:1b2c:c356]:10011
I1204 10:23:37.868498 2901 internals-datagram-socket.cc:98] Binding RPC2 UDP socket to [2002:a05:660d:10:b0:9da:1b2c:c356]:10011
I1204 10:23:37.868522 2901 internals-datagram-socket.cc:98] Binding RPC2 UDP socket to [fd14:988a:50ee:10c:b0:9da:1b2c:c356]:10011
I1204 10:23:37.868544 2901 internals-datagram-socket.cc:98] Binding RPC2 UDP socket to 127.0.0.1:10011
I1204 10:23:37.868575 2901 internals-datagram-socket.cc:98] Binding RPC2 UDP socket to [::1]:10011
I1204 10:23:37.869104 2971 client_instance.cc:518] [test] new waiter thread: waiters_id: /lbns/chubby/test/svelte;svelte-tasks.test.ls.google.com.:6297;test.svelte.global.ls.google.com.:6297, test, 5719, false, -1
I1204 10:23:37.869153 2901 noop_barrier.h:18] NoopBarrier::Init
I1204 10:23:37.869292 2901 tpunetd_client.cc:225] Disabling ICI session monitoring on Forge.
I1204 10:23:37.869487 2901 tpunetd_client.cc:243] Session manager starting a new session...
I1204 10:23:37.869497 2901 noop_barrier.h:25] NoopBarrier::SyncWithTimeout: PRE_START_SESSION_BARRIER
I1204 10:23:37.871576 2975 lockservice_loas.cc:81] LOASUtil says I am <forge-00>, LOASUtil says my security realm is campus-sln
I1204 10:23:37.871824 2975 server_pool.cc:908] Attempt DNS lookup for svelte-tasks.test.ls.google.com.
I1204 10:23:37.872996 2975 waiter_thread.cc:833] [test] contacting svelte server at [2002:a05:663a:64c8:b0:41d:9192:5b3c]:6297 (svelte-tasks.test.ls.google.com.) discovery_type: DNS, svelte_cell: null, good_server: true, subset: nullopt, all_resolved_servers: []
I1204 10:23:37.874119 2975 rpc-internals-version.cc:158] RPC_VERSION: Saw use of RPC version 2 client
I1204 10:23:37.875069 2901 noop_barrier.h:25] NoopBarrier::SyncWithTimeout: POST_START_SESSION_BARRIER
I1204 10:23:37.875079 2901 tpunetd_client.cc:245] Session manager started the new session.
I1204 10:23:37.875083 2901 tpunetd_client.cc:249] Session master starting a new session...
I1204 10:23:37.875096 2901 session_master.cc:258] Starting a new ICI network session...
I1204 10:23:37.875284 2974 session_worker_service.cc:104] Session master notifies the worker in a new session: a170413005d7ad56
I1204 10:23:37.877078 2901 session_master.cc:302] Successfully started ICI network session with session id: a170413005d7ad56
I1204 10:23:37.877139 2901 tpunetd_client.cc:251] Session master started the new session.
I1204 10:23:37.906974 2901 psp_support.cc:405] HTTPServer2 initially intends to listen on {[::]:10085,PSP-incapable}, After the PSP IP expansion, it attempts to listen on {[2002:a05:660d:10:b0:9da:1b2c:c356]:10085,PSP-capable; [2002:a05:660d:4:b0:9da:1b2c:c356]:10085,PSP-capable; [2002:a05:660d:c:b0:9da:1b2c:c356]:10085,PSP-capable; [2002:a05:660d:e:b0:9da:1b2c:c356]:10085,PSP-capable; [fd14:988a:50ee:10c:b0:9da:1b2c:c356]:10085,PSP-incapable; 127.0.0.1:10085,PSP-incapable; [::1]:10085,PSP-incapable} with an accept queue length of { 2147483647 }
I1204 10:23:37.907832 2901 internals-datagram-socket.cc:98] Binding RPC2 UDP socket to [2002:a05:660d:10:b0:9da:1b2c:c356]:10085
I1204 10:23:37.907883 2901 internals-datagram-socket.cc:98] Binding RPC2 UDP socket to [2002:a05:660d:4:b0:9da:1b2c:c356]:10085
I1204 10:23:37.907920 2901 internals-datagram-socket.cc:98] Binding RPC2 UDP socket to [2002:a05:660d:c:b0:9da:1b2c:c356]:10085
I1204 10:23:37.907949 2901 internals-datagram-socket.cc:98] Binding RPC2 UDP socket to [2002:a05:660d:e:b0:9da:1b2c:c356]:10085
I1204 10:23:37.907973 2901 internals-datagram-socket.cc:98] Binding RPC2 UDP socket to [fd14:988a:50ee:10c:b0:9da:1b2c:c356]:10085
I1204 10:23:37.907996 2901 internals-datagram-socket.cc:98] Binding RPC2 UDP socket to 127.0.0.1:10085
I1204 10:23:37.908033 2901 internals-datagram-socket.cc:98] Binding RPC2 UDP socket to [::1]:10085
I1204 10:23:37.908380 2901 callback_internal.cc:127] Running in hermetic environment. Coresweep NG Oracle stubby channel unreachable
I1204 10:23:37.908484 2901 tpu_inspectz.cc:324] Registering /tpuz?id=0
I1204 10:23:37.908583 2901 system.cc:1143] tpu::System initialized, current host id: 0, logical device ids: 0
I1204 10:23:37.908633 2807 tpu_system_state.cc:219] CreateTpuSystemState: TPU initialization is successful and it took 799.785749ms
I1204 10:23:37.908670 2807 tpu_host_allocator.cc:66] Premapped buffer is using alignment 32
I1204 10:23:37.909071 2807 tpu_linearizer.cc:80] Inflight linearization byte limit configured to : 1099511627776
I1204 10:23:37.909081 2807 tpu_linearizer.cc:93] Inflight delinearization byte limit configured to : 1099511627776
I1204 10:23:37.909478 2807 pjrt_c_api_client.cc:164] PjRtCApiClient created.
I1204 10:23:37.909612 2807 pjrt_client.cc:550] PjRt-IFRT device count: total=1, addressable=1
I1204 10:23:37.909617 2807 pjrt_client.cc:554] Addressable PjRt-IFRT device: TpuDevice(id=0, process_index=0, coords=(0,0,0), core_on_chip=0)
I1204 10:23:37.910293 2977 waiter_thread.cc:1927] [test] client id set to 1764872617910046023 (server [2002:a05:663a:64c8:b0:41d:9192:5b3c]:6297) shuffle sharding result: nullopt
I1204 10:23:37.910552 2972 client_sharding_config.cc:163] No sharding config needed to be applied.
I1204 10:23:37.910604 2971 general_sharding_config.cc:85] Starting svelte client reshard 0x10b038dde460
I1204 10:23:37.910731 2971 general_sharding_config.cc:219] Redirecting perfile /ls/any/acls/SIGNED/master/fo/forge-00 to test
I1204 10:23:37.910817 2971 general_sharding_config.cc:239] Svelte client reshard complete. Resharded 0 watchers and redirected 1 existing files in 198.006us
I1204 10:23:37.976401 2807 autofdo_agent.cc:127] The passed profile_dir (/placer/prod/home/xla-autofdo/fdo_profiles) is overridden to an empty string.
I1204 10:23:37.976435 2807 autofdo_agent.cc:198] xla_tpu_autofdo_profile_dir updated to
W1204 10:23:37.976440 2807 autofdo_agent.cc:201] xla_tpu_autofdo_use_remote_repo is overridden to false because xla_tpu_autofdo_profile_dir is not set.
W1204 10:23:37.998311 2807 xsymbol_util.cc:123] Skipping XSymbol unoptimized HLO module upload to the XSymbol server (since XSymbolManager is null) and skipping export to side outputs (since the side_outputs_dir is empty). Normally this shouldn't happen, as symbols should either be uploaded to the XSymbol server or exported to side outputs to be uploaded to the XSymbol server later.
I1204 10:23:38.004917 2807 autofdo_profile_utils.h:84] Loading AutoFDO profile entries from /memfile/xla_autofdo/module_flag_fdo_profiles.sstable
I1204 10:23:38.056404 2807 autofdo_profile_utils.h:129] Loaded 13024 AutoFDO profile entries for TPU family TPU_VERSION_GHOSTLITE (5) out of 39239 entries from /memfile/xla_autofdo/module_flag_fdo_profiles.sstable load_entire_sstable false TotalUncompressedValueSize 21598035
I1204 10:23:38.068456 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 11.87716ms
I1204 10:23:38.851726 2819 python_stack_size.cc:37] EventManager thread stack size of 131072 might be too small for Python callbacks. Crash? Read go/py-cpp-stack.
I1204 10:23:39.199498 3011 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:39.200285 3011 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:39.206182 3011 autofdo_profile_utils.h:84] Loading AutoFDO profile entries from /memfile/xla_autofdo/fdo_profiles.sstable
I1204 10:23:40.013749 3011 autofdo_profile_utils.h:129] Loaded 259183 AutoFDO profile entries for TPU family TPU_VERSION_GHOSTLITE (5) out of 720025 entries from /memfile/xla_autofdo/fdo_profiles.sstable load_entire_sstable true TotalUncompressedValueSize 26751651
I1204 10:23:40.014305 3011 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:40.298484 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 1.09959926925s
I1204 10:23:40.319330 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 20.57502925ms
I1204 10:23:40.319466 3001 isa_program_util_common.cc:520] (HLO module jit_convert_element_type): Executable fingerprint:9297e550b3093ca2e31ac252072a7f446b4bc23ff77b9d713daf7adc38c2706f
I1204 10:23:40.319474 3001 isa_program_util_common.cc:524] (HLO module jit_convert_element_type): Executable fingerprint (including data segments):7fe9ccd5c88e10fe89738967b2652070f470f077b1f8cbbc725e7a8d09973130
I1204 10:23:40.319477 3001 isa_program_util_common.cc:527] (HLO module jit_convert_element_type): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
W1204 10:23:40.319695 2807 xsymbol_util.cc:81] Skipping XSymbol Compiler Metadata upload to the XSymbol server (since XSymbolManager is null) and skipping export to side outputs (since the side_outputs_dir is empty). Normally this shouldn't happen, as symbols should either be uploaded to the XSymbol server or exported to side outputs to be uploaded to the XSymbol server later.
I1204 10:23:40.319711 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 2.32309384925s
I1204 10:23:40.319715 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:40.329051 2862 client_comparison_tool.cc:54] [Comparison Tool] Initializing singleton ClientComparisonTool.
I1204 10:23:40.359113 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.1985ms
I1204 10:23:40.360534 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:40.361110 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:40.363094 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:40.408515 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 48.01239075ms
I1204 10:23:40.420064 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 11.41320925ms
I1204 10:23:40.420131 3001 isa_program_util_common.cc:520] (HLO module jit_convert_element_type): Executable fingerprint:4a11b7386c58911d47a3c537cacd03692717b217afd2c40fcd0051b80fd445cc
I1204 10:23:40.420135 3001 isa_program_util_common.cc:524] (HLO module jit_convert_element_type): Executable fingerprint (including data segments):e10561c1c85099ed3da50acd9d44321e63d21c67097cb7f26d6c9d1a59eaecd6
I1204 10:23:40.420138 3001 isa_program_util_common.cc:527] (HLO module jit_convert_element_type): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:40.420353 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 72.12592ms
I1204 10:23:40.420366 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:40.456136 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.24509075ms
I1204 10:23:40.457553 3016 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:40.458168 3016 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:40.459349 3016 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:40.814329 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 356.82480925ms
I1204 10:23:40.862455 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 47.95089075ms
I1204 10:23:40.862650 3001 isa_program_util_common.cc:520] (HLO module jit_convert_element_type): Executable fingerprint:049b2b8f31d9989c10d254a03449fbcc024c4bf25ef9e30aae43d7ccf600b2d5
I1204 10:23:40.862655 3001 isa_program_util_common.cc:524] (HLO module jit_convert_element_type): Executable fingerprint (including data segments):ed0801a639cf54b4dafe3efc9438b9df9d142a966a1302b181aa7f046c552193
I1204 10:23:40.862657 3001 isa_program_util_common.cc:527] (HLO module jit_convert_element_type): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:40.862829 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 417.78129075ms
I1204 10:23:40.862842 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:40.905527 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.36235075ms
I1204 10:23:40.907062 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:40.907621 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:40.909309 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:41.737083 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 830.06606925ms
I1204 10:23:41.802201 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 64.9178ms
I1204 10:23:41.802405 3001 isa_program_util_common.cc:520] (HLO module jit__pad): Executable fingerprint:a69925c3e35460f027ee709200f8059a834b6e151bad75ed1aac4193ec985a52
I1204 10:23:41.802409 3001 isa_program_util_common.cc:524] (HLO module jit__pad): Executable fingerprint (including data segments):322bc19d1262e0542014a77af43ed8fa05c5bae247cb17d49e5413635b25785e
I1204 10:23:41.802412 3001 isa_program_util_common.cc:527] (HLO module jit__pad): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:41.802678 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 909.55586925ms
I1204 10:23:41.802687 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:41.840894 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.25271075ms
I1204 10:23:41.842435 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:41.843016 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:41.844315 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:42.234381 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 391.9972ms
I1204 10:23:42.286217 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 51.64147075ms
I1204 10:23:42.286429 3001 isa_program_util_common.cc:520] (HLO module jit_reshape): Executable fingerprint:24fa2943c08d6e85d3563eb8055a854b296430fed1598d114793a7dabc3284ce
I1204 10:23:42.286434 3001 isa_program_util_common.cc:524] (HLO module jit_reshape): Executable fingerprint (including data segments):78dc18e4087d7bbde5b83f9d0cdc3f136a4beb0593c86ac2e88ee01f816d0031
I1204 10:23:42.286436 3001 isa_program_util_common.cc:527] (HLO module jit_reshape): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:42.286724 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 457.98552925ms
I1204 10:23:42.286733 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:42.321467 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.3628ms
I1204 10:23:42.322917 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:42.323599 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:42.324689 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:42.331822 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 8.97940925ms
I1204 10:23:42.339625 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 7.68726ms
I1204 10:23:42.339681 3001 isa_program_util_common.cc:520] (HLO module jit_iota): Executable fingerprint:16a90f09be0006e8e246bb66b4e20ec6475a8a93ddd9dcfa11d559392d9cd69f
I1204 10:23:42.339684 3001 isa_program_util_common.cc:524] (HLO module jit_iota): Executable fingerprint (including data segments):158d5dd708cfc2ef37b19db0eec0006c18b80b42d41e7f868c1dd217f0b51552
I1204 10:23:42.339686 3001 isa_program_util_common.cc:527] (HLO module jit_iota): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:42.339895 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 29.50696ms
I1204 10:23:42.339903 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:42.377927 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 6.00201075ms
I1204 10:23:42.379345 3016 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:42.379918 3016 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:42.382122 3016 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:42.391279 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 11.99014ms
I1204 10:23:42.399775 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 8.39998ms
I1204 10:23:42.399830 3001 isa_program_util_common.cc:520] (HLO module jit_add): Executable fingerprint:efff707c196fe871a739cc22bfe5045fab53c8b031b13174a671e447d131dd84
I1204 10:23:42.399833 3001 isa_program_util_common.cc:524] (HLO module jit_add): Executable fingerprint (including data segments):903c8ddc1fadae24e851450c050003125343a322ddb93d71b83d7cf8b0600558
I1204 10:23:42.399835 3001 isa_program_util_common.cc:527] (HLO module jit_add): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:42.400073 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 35.02836925ms
I1204 10:23:42.400082 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:42.434690 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.28028ms
I1204 10:23:42.436135 3016 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:42.436698 3016 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:42.437914 3016 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:42.445331 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 9.25302ms
I1204 10:23:42.453163 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 7.72525075ms
I1204 10:23:42.453218 3001 isa_program_util_common.cc:520] (HLO module jit__pad): Executable fingerprint:2465fb80b7e81e394d71acdce5964d521e7f17722fe790d47e342a72d0ee4162
I1204 10:23:42.453221 3001 isa_program_util_common.cc:524] (HLO module jit__pad): Executable fingerprint (including data segments):0d74f996d24d804ac425f2da854d60c51a95459ed55395522afa0de4027e8577
I1204 10:23:42.453224 3001 isa_program_util_common.cc:527] (HLO module jit__pad): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:42.453451 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 29.89918ms
I1204 10:23:42.453463 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:42.485485 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.13516925ms
I1204 10:23:42.486863 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:42.487428 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:42.488559 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:42.682572 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 195.74216ms
I1204 10:23:42.712559 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 29.82224ms
I1204 10:23:42.712690 3001 isa_program_util_common.cc:520] (HLO module jit_convert_element_type): Executable fingerprint:e55c4e283774577a24d8d7bdde4554bcb1e7edb36250fc5be3ff2df6277a85b7
I1204 10:23:42.712693 3001 isa_program_util_common.cc:524] (HLO module jit_convert_element_type): Executable fingerprint (including data segments):ddf575a27a1775652c79d5ba03a52413467c00509cbf88818e2345d25343747c
I1204 10:23:42.712695 3001 isa_program_util_common.cc:527] (HLO module jit_convert_element_type): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:42.712922 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 238.27849075ms
I1204 10:23:42.712932 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:42.747675 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.51551075ms
I1204 10:23:42.749112 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:42.749696 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:42.751192 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:43.211567 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 462.55372ms
I1204 10:23:43.250218 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 38.48024ms
I1204 10:23:43.250355 3001 isa_program_util_common.cc:520] (HLO module jit__pad): Executable fingerprint:e4f3ea7323b667fe21767e8da701c955a38f171e83afc42e277cb1bb2dabc65e
I1204 10:23:43.250365 3001 isa_program_util_common.cc:524] (HLO module jit__pad): Executable fingerprint (including data segments):29e2ae84ca59a0cc678c859e3ac86b0382fd9d1e17b0fdcccb1feda32e189156
I1204 10:23:43.250369 3001 isa_program_util_common.cc:527] (HLO module jit__pad): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:43.250616 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 514.24124ms
I1204 10:23:43.250624 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:43.284432 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.55138ms
I1204 10:23:43.285845 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:43.286428 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:43.287717 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:43.504965 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 219.20614ms
I1204 10:23:43.536884 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 31.75234925ms
I1204 10:23:43.537029 3001 isa_program_util_common.cc:520] (HLO module jit_reshape): Executable fingerprint:3ffd29ee673b480b5a459cfb0db1555f899f1276f8e1d2865205d0aaa5824642
I1204 10:23:43.537033 3001 isa_program_util_common.cc:524] (HLO module jit_reshape): Executable fingerprint (including data segments):dfd022fe3c05eb420b0f105b2c4c97222844336874395e8763413c3b61689b02
I1204 10:23:43.537036 3001 isa_program_util_common.cc:527] (HLO module jit_reshape): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:43.537288 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 264.20636ms
I1204 10:23:43.537297 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:43.570381 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.06222ms
I1204 10:23:43.571756 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:43.572337 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:43.573365 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:43.580427 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 8.73838ms
I1204 10:23:43.588520 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 7.99969075ms
I1204 10:23:43.588576 3001 isa_program_util_common.cc:520] (HLO module jit_iota): Executable fingerprint:16a90f09be0006e8e246bb66b4e20ec6475a8a93ddd9dcfa11d559392d9cd69f
I1204 10:23:43.588579 3001 isa_program_util_common.cc:524] (HLO module jit_iota): Executable fingerprint (including data segments):158d5dd708cfc2ef37b19db0eec0006c18b80b42d41e7f868c1dd217f0b51552
I1204 10:23:43.588581 3001 isa_program_util_common.cc:527] (HLO module jit_iota): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:43.588784 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 29.32262ms
I1204 10:23:43.588793 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:43.624407 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.09822925ms
I1204 10:23:43.625892 3016 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:43.626469 3016 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:43.628432 3016 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:43.637256 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 11.44420925ms
I1204 10:23:43.646022 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 8.66730925ms
I1204 10:23:43.646082 3001 isa_program_util_common.cc:520] (HLO module jit_add): Executable fingerprint:efff707c196fe871a739cc22bfe5045fab53c8b031b13174a671e447d131dd84
I1204 10:23:43.646085 3001 isa_program_util_common.cc:524] (HLO module jit_add): Executable fingerprint (including data segments):903c8ddc1fadae24e851450c050003125343a322ddb93d71b83d7cf8b0600558
I1204 10:23:43.646088 3001 isa_program_util_common.cc:527] (HLO module jit_add): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:43.646322 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 34.13532925ms
I1204 10:23:43.646330 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:43.681942 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.85316ms
I1204 10:23:43.683376 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:43.683936 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:43.685771 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:43.694962 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 11.6568ms
I1204 10:23:43.703719 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 8.65348ms
I1204 10:23:43.703775 3001 isa_program_util_common.cc:520] (HLO module jit__pad): Executable fingerprint:463897d9427f34d05ffdc0843e2b6191b9eb550c331fa4068c958b23c851183c
I1204 10:23:43.703779 3001 isa_program_util_common.cc:524] (HLO module jit__pad): Executable fingerprint (including data segments):d980a84d0e96fd38bb4b0ba42869d190aa273c6735d488885a6d3954e1c88009
I1204 10:23:43.703782 3001 isa_program_util_common.cc:527] (HLO module jit__pad): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:43.704012 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 33.77448ms
I1204 10:23:43.704021 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:43.737365 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.15572ms
I1204 10:23:43.738807 3016 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:43.739379 3016 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:43.740578 3016 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:44.018991 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 280.26778ms
I1204 10:23:44.058588 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 39.42922925ms
I1204 10:23:44.058752 3001 isa_program_util_common.cc:520] (HLO module jit_convert_element_type): Executable fingerprint:0fa5877263f988a819b86d1680b010951a8899714a688f2d74634cb327bc6f8d
I1204 10:23:44.058756 3001 isa_program_util_common.cc:524] (HLO module jit_convert_element_type): Executable fingerprint (including data segments):d8ff194555f7c866d0e59cf74771072bdca5ce912d87a220e5f11cd04438e2fc
I1204 10:23:44.058765 3001 isa_program_util_common.cc:527] (HLO module jit_convert_element_type): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:44.058991 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 332.60310925ms
I1204 10:23:44.058999 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:44.095306 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.91676925ms
I1204 10:23:44.096847 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:44.097551 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:44.099406 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:44.739660 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 642.90456ms
I1204 10:23:44.792373 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 52.51366925ms
I1204 10:23:44.792547 3001 isa_program_util_common.cc:520] (HLO module jit__pad): Executable fingerprint:0fec16cd0a0b830a359aa9e66aa2740fdf9e198854033d1a0c6cc6f7ea320e1a
I1204 10:23:44.792551 3001 isa_program_util_common.cc:524] (HLO module jit__pad): Executable fingerprint (including data segments):9ba0aba775659f17ec06d12565bf7bd8c837bc2a5fbbbf75da30839623573075
I1204 10:23:44.792553 3001 isa_program_util_common.cc:527] (HLO module jit__pad): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:44.792860 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 709.3246ms
I1204 10:23:44.792869 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:44.827130 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.52517075ms
I1204 10:23:44.828603 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:44.829355 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:44.831021 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:45.126963 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 298.46872ms
I1204 10:23:45.168822 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 41.6601ms
I1204 10:23:45.168988 3001 isa_program_util_common.cc:520] (HLO module jit_reshape): Executable fingerprint:dadf519e3f3ce8a15e826a434604227832fc40b271dc7cd685920ce2c30356c0
I1204 10:23:45.168992 3001 isa_program_util_common.cc:524] (HLO module jit_reshape): Executable fingerprint (including data segments):d99f5fd8b80d2133e146bc589b06abcc15e698cf20e4b2248513546d60dbcece
I1204 10:23:45.168994 3001 isa_program_util_common.cc:527] (HLO module jit_reshape): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:45.169261 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 353.45970925ms
I1204 10:23:45.169270 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:45.202283 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 3.99669075ms
I1204 10:23:45.203679 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:45.204410 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:45.205727 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:45.212948 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 9.36625075ms
I1204 10:23:45.221735 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 8.67902ms
I1204 10:23:45.221787 3001 isa_program_util_common.cc:520] (HLO module jit_iota): Executable fingerprint:16a90f09be0006e8e246bb66b4e20ec6475a8a93ddd9dcfa11d559392d9cd69f
I1204 10:23:45.221791 3001 isa_program_util_common.cc:524] (HLO module jit_iota): Executable fingerprint (including data segments):158d5dd708cfc2ef37b19db0eec0006c18b80b42d41e7f868c1dd217f0b51552
I1204 10:23:45.221793 3001 isa_program_util_common.cc:527] (HLO module jit_iota): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:45.222068 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 30.58573075ms
I1204 10:23:45.222099 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:45.257630 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.46210925ms
I1204 10:23:45.259173 3016 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:45.259836 3016 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:45.262012 3016 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:45.270980 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 11.89939075ms
I1204 10:23:45.280500 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 9.40294925ms
I1204 10:23:45.280556 3001 isa_program_util_common.cc:520] (HLO module jit_add): Executable fingerprint:efff707c196fe871a739cc22bfe5045fab53c8b031b13174a671e447d131dd84
I1204 10:23:45.280560 3001 isa_program_util_common.cc:524] (HLO module jit_add): Executable fingerprint (including data segments):903c8ddc1fadae24e851450c050003125343a322ddb93d71b83d7cf8b0600558
I1204 10:23:45.280563 3001 isa_program_util_common.cc:527] (HLO module jit_add): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:45.280790 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 35.53122925ms
I1204 10:23:45.280798 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:45.316437 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.7176ms
I1204 10:23:45.317914 3016 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:45.318535 3016 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:45.320522 3016 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:45.329594 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 11.77386ms
I1204 10:23:45.339084 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 9.37678925ms
I1204 10:23:45.339140 3001 isa_program_util_common.cc:520] (HLO module jit__pad): Executable fingerprint:10664fe2aa6f56d4b270bd26bafe4cdf3370ddb146fd055ff903e06e7b7efcd4
I1204 10:23:45.339143 3001 isa_program_util_common.cc:524] (HLO module jit__pad): Executable fingerprint (including data segments):6e6e994e9d07d781f7efd366150621506dc782617cc4f0a0aa215e7afcc95dda
I1204 10:23:45.339152 3001 isa_program_util_common.cc:527] (HLO module jit__pad): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:45.339410 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 34.6088ms
I1204 10:23:45.339419 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:45.375888 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 6.37968925ms
I1204 10:23:45.377395 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:45.378006 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:45.380514 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:45.632487 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 255.18374925ms
I1204 10:23:45.651523 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 18.83823075ms
I1204 10:23:45.651613 3001 isa_program_util_common.cc:520] (HLO module jit_concatenate): Executable fingerprint:d7139ff3ca6a349187241b02d066ad55ca60e8575812a2b123ce4f2c5124fa5c
I1204 10:23:45.651617 3001 isa_program_util_common.cc:524] (HLO module jit_concatenate): Executable fingerprint (including data segments):4c0fa59b72cc1c3675795f812af2eb7ded87bd2bd9cb6bf8c56d68db3dfa6905
I1204 10:23:45.651619 3001 isa_program_util_common.cc:527] (HLO module jit_concatenate): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:45.651934 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 289.30961075ms
I1204 10:23:45.651946 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:45.690721 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.93694925ms
I1204 10:23:45.692204 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:45.692939 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:45.694714 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:45.779641 2843 estub_channel_helper.cc:109] SmartChannel: Generating Stubs Client / EStubs channel:
go/debugonly
target: "/mbns/local/monitoring/monarch/streamz.mbns"
stub_options {
extensible_stub {
server_name: "Monarch (BulkExporter)"
enable_client_side_throttling: true
enable_retries: true
enable_monitoring: false
traffic_type: BATCH
smoothing_type: NONE
}
}
client_channel_options {
backend_subset: true
subset_size: 5
socket_options {
traffic_class: AF1
}
}
wait_sec: -1
I1204 10:23:45.780010 2843 experiments_flag_data_registry.cc:114] Loaded core and non-core flag data
I1204 10:23:45.780152 2843 experiments_config_mutator_registry.cc:38] Non-core config mutators are registered
I1204 10:23:45.780164 2843 experiments_config_mutator_registry.cc:46] Created 1 core and 1 non-core config mutators
I1204 10:23:45.782602 2843 stubs_client_channel_internal.cc:872] Extensible Stubs doesn't implement WaitUntilNonEmptyWithTimeout -- you very likely want to use WaitUntilReachableWithTimeout. See go/channel-reachability#what-does-waituntilnonempty-do and go/channel-reachability#can_accept for more context.
W1204 10:23:45.784892 2824 internals-datagram-socket.cc:411] udp sendto fd=200, dest=172.20.0.120:9530: Network is unreachable
I1204 10:23:45.795029 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 102.90628ms
I1204 10:23:45.806451 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 11.19388925ms
I1204 10:23:45.806520 3001 isa_program_util_common.cc:520] (HLO module jit__pad): Executable fingerprint:c6b99c4c63f2a3fbd835e0450756f3349388329ec1f87662cb87521dcd1841a5
I1204 10:23:45.806523 3001 isa_program_util_common.cc:524] (HLO module jit__pad): Executable fingerprint (including data segments):89d0d784adcdbbee4c739314ac900b4eaea4e25d24cfb033621cb60151ba2b75
I1204 10:23:45.806526 3001 isa_program_util_common.cc:527] (HLO module jit__pad): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:45.806785 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 127.89071075ms
I1204 10:23:45.806794 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:45.843888 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.92600925ms
I1204 10:23:45.845423 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:45.846124 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:45.847932 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:45.856087 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 10.74262ms
I1204 10:23:45.865309 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 9.07492925ms
I1204 10:23:45.865365 3001 isa_program_util_common.cc:520] (HLO module jit_broadcast_in_dim): Executable fingerprint:f8a8a1bcc14437ec76793f8d974adac664e3c5f5e676f34082c65624e495eac5
I1204 10:23:45.865368 3001 isa_program_util_common.cc:524] (HLO module jit_broadcast_in_dim): Executable fingerprint (including data segments):b7175bd87ce618f7322eb6a7c85b4a92b395ad6e0edb5ac68d5ee005fd07f01b
I1204 10:23:45.865371 3001 isa_program_util_common.cc:527] (HLO module jit_broadcast_in_dim): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:45.865600 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 33.61454925ms
I1204 10:23:45.865608 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:45.892870 2820 monarch_bundle.cc:728] [streamz] warning: Entity does not map to any target: go/debugonly label { key: "stage_name" string_value: "BACKEND_PASSES" } label { key: "binary_name" string_value: "ragged_paged_attention_v3_test_gl" } label { key: "cluster" string_value: "it" } label { key: "container_name" string_value: "action-aef67bf50706fee86777a93cc065340a246c" } label { key: "exclude_from_monarch_borg_task" bool_value: true } label { key: "global_pid" int64_value: 103661114569942127 } label { key: "hostname" string_value: "itmm4" } label { key: "unix_user" string_value: "forge-00" }
I1204 10:23:45.902316 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.77286ms
I1204 10:23:45.903773 3016 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:45.904419 3016 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:45.907136 3016 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:45.915746 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 12.03298ms
I1204 10:23:45.925658 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 9.80482ms
I1204 10:23:45.925727 3001 isa_program_util_common.cc:520] (HLO module jit_concatenate): Executable fingerprint:f15dcb146517f339471f1095ec40fdf8001bb7e41b4d9b049bc5d095ed4e4951
I1204 10:23:45.925736 3001 isa_program_util_common.cc:524] (HLO module jit_concatenate): Executable fingerprint (including data segments):16731fbb97b37de6be3233e5daa9a664e4765c1148e2b56898988f79d9d4d4d6
I1204 10:23:45.925740 3001 isa_program_util_common.cc:527] (HLO module jit_concatenate): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:45.926045 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 36.96879075ms
I1204 10:23:45.926053 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:45.963346 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.8336ms
I1204 10:23:45.964799 3016 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:45.965345 3016 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:45.967172 3016 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:45.975887 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 11.1369ms
I1204 10:23:45.984352 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 8.3566ms
I1204 10:23:45.984411 3001 isa_program_util_common.cc:520] (HLO module jit__pad): Executable fingerprint:f4339e688a6e0427eacb1ea2d4cec46ebd2aac490764a7eebd80b73ad3e0ec37
I1204 10:23:45.984414 3001 isa_program_util_common.cc:524] (HLO module jit__pad): Executable fingerprint (including data segments):4684e37e739f3f53e872e9360c52b3547c3283381ef5d55cf53625f29c3ce733
I1204 10:23:45.984416 3001 isa_program_util_common.cc:527] (HLO module jit__pad): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:45.984672 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 33.09352925ms
I1204 10:23:45.984680 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:46.019159 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.8474ms
I1204 10:23:46.020665 3016 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:46.021354 3016 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:46.022685 3016 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:46.031395 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 10.78662925ms
I1204 10:23:46.039725 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 8.20114ms
I1204 10:23:46.039782 3001 isa_program_util_common.cc:520] (HLO module jit_reshape): Executable fingerprint:ef0479b69cdfad001d139d03513dfa9d34b6e0f68cb790ddfba8100ef7ddb794
I1204 10:23:46.039786 3001 isa_program_util_common.cc:524] (HLO module jit_reshape): Executable fingerprint (including data segments):ac9597e2e8bdf5cc485a0b4a154ee437c7d8bed7b5b205f1865b3305a7c33bdf
I1204 10:23:46.039788 3001 isa_program_util_common.cc:527] (HLO module jit_reshape): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:46.040030 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 32.53315075ms
I1204 10:23:46.040042 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:46.072718 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.19610925ms
I1204 10:23:46.074164 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:46.074879 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:46.076348 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:46.084782 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 10.67802ms
I1204 10:23:46.092951 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 8.06706925ms
I1204 10:23:46.093004 3001 isa_program_util_common.cc:520] (HLO module jit_convert_element_type): Executable fingerprint:2465fb80b7e81e394d71acdce5964d521e7f17722fe790d47e342a72d0ee4162
I1204 10:23:46.093008 3001 isa_program_util_common.cc:524] (HLO module jit_convert_element_type): Executable fingerprint (including data segments):0d74f996d24d804ac425f2da854d60c51a95459ed55395522afa0de4027e8577
I1204 10:23:46.093010 3001 isa_program_util_common.cc:527] (HLO module jit_convert_element_type): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:46.093206 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 31.56127075ms
I1204 10:23:46.093214 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:46.128014 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.51091075ms
I1204 10:23:46.129586 3016 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:46.130147 3016 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:46.131796 3016 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:46.140554 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 10.99978925ms
I1204 10:23:46.148817 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 8.16317075ms
I1204 10:23:46.148872 3001 isa_program_util_common.cc:520] (HLO module jit__pad): Executable fingerprint:ddc65154bef29b7fb015a60e52f2bdbfd0e2ed716f8e2c1eec451d81db55aed7
I1204 10:23:46.148875 3001 isa_program_util_common.cc:524] (HLO module jit__pad): Executable fingerprint (including data segments):881623976ebf4fa50de78b1de1504282e88726150d81b47aca5157456abe676d
I1204 10:23:46.148883 3001 isa_program_util_common.cc:527] (HLO module jit__pad): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:46.149099 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 32.48171075ms
I1204 10:23:46.149107 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:46.181007 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.17188925ms
I1204 10:23:46.182488 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:46.183074 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:46.184194 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:46.191496 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 9.08658ms
I1204 10:23:46.199450 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 7.85923075ms
I1204 10:23:46.199505 3001 isa_program_util_common.cc:520] (HLO module jit_convert_element_type): Executable fingerprint:2465fb80b7e81e394d71acdce5964d521e7f17722fe790d47e342a72d0ee4162
I1204 10:23:46.199514 3001 isa_program_util_common.cc:524] (HLO module jit_convert_element_type): Executable fingerprint (including data segments):0d74f996d24d804ac425f2da854d60c51a95459ed55395522afa0de4027e8577
I1204 10:23:46.199518 3001 isa_program_util_common.cc:527] (HLO module jit_convert_element_type): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:46.199712 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 29.65702ms
I1204 10:23:46.199720 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:46.234873 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.79782925ms
I1204 10:23:46.236396 3016 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:46.237043 3016 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:46.238743 3016 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:46.247578 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 11.22806925ms
I1204 10:23:46.255886 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 8.18948925ms
I1204 10:23:46.255940 3001 isa_program_util_common.cc:520] (HLO module jit__pad): Executable fingerprint:50b24227e242f9e3991477dc6ab80c35eec2a73dd8f3e381c4239bb7bc7ad667
I1204 10:23:46.255944 3001 isa_program_util_common.cc:524] (HLO module jit__pad): Executable fingerprint (including data segments):e469d13e73a2cfc39ea4268e70a7b19c0616c4aa2e683e26211b314683b41fe1
I1204 10:23:46.255946 3001 isa_program_util_common.cc:527] (HLO module jit__pad): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:46.256201 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 33.00663075ms
I1204 10:23:46.256209 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:46.296127 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 6.8528ms
I1204 10:23:46.297611 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:46.298189 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:46.300505 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:46.309450 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 11.89970925ms
I1204 10:23:46.318069 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 8.52441075ms
I1204 10:23:46.318127 3001 isa_program_util_common.cc:520] (HLO module jit__unstack): Executable fingerprint:99bdc5402506ad44c344f97b8e2be26528372eed0ffd6529c115b886921632e4
I1204 10:23:46.318130 3001 isa_program_util_common.cc:524] (HLO module jit__unstack): Executable fingerprint (including data segments):cb0ccbc393421f635f73241862c5bc4618d9e2c7b4394d6ef66b29b3ded310f0
I1204 10:23:46.318133 3001 isa_program_util_common.cc:527] (HLO module jit__unstack): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:46.318398 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 36.3167ms
I1204 10:23:46.318406 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:46.356656 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.01980925ms
I1204 10:23:46.358244 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:46.358879 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:46.360450 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:46.369395 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 11.23592ms
I1204 10:23:46.377659 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 8.14548925ms
I1204 10:23:46.377714 3001 isa_program_util_common.cc:520] (HLO module jit_less_equal): Executable fingerprint:b44327c7d40a86a33fad396cd93bdad30c4bdf690f089c81fe1f9a4c20c1bb70
I1204 10:23:46.377718 3001 isa_program_util_common.cc:524] (HLO module jit_less_equal): Executable fingerprint (including data segments):4d26037fefee5e59828e044a800fd12ee87ffc86c4ba12968b5f7a55c7a1541b
I1204 10:23:46.377720 3001 isa_program_util_common.cc:527] (HLO module jit_less_equal): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:46.377992 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 33.90785075ms
I1204 10:23:46.378010 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:46.413835 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.79681075ms
I1204 10:23:46.415376 3016 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:46.416011 3016 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:46.417600 3016 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:46.426454 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 11.15030925ms
I1204 10:23:46.434681 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 8.12182925ms
I1204 10:23:46.434732 3001 isa_program_util_common.cc:520] (HLO module jit_greater): Executable fingerprint:89c5056ca57246d8da124b5c3ff6f7bc2bdbb9f3cb46d3872b5089ec13e907ab
I1204 10:23:46.434736 3001 isa_program_util_common.cc:524] (HLO module jit_greater): Executable fingerprint (including data segments):608f212e89b21661d0eeb678036783941c118e01832a89281547ad72ed94f6bf
I1204 10:23:46.434738 3001 isa_program_util_common.cc:527] (HLO module jit_greater): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:46.434995 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 32.96137075ms
I1204 10:23:46.435010 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:46.467331 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.42763075ms
I1204 10:23:46.468876 3016 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:46.469458 3016 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:46.470714 3016 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:46.478651 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 9.85243075ms
I1204 10:23:46.486687 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 7.92505075ms
I1204 10:23:46.486739 3001 isa_program_util_common.cc:520] (HLO module jit_convert_element_type): Executable fingerprint:1dc54db94e561e07f7b51f7dbd3593cdb201c92cb525d0e05e5be92a49846064
I1204 10:23:46.486753 3001 isa_program_util_common.cc:524] (HLO module jit_convert_element_type): Executable fingerprint (including data segments):40f165ab5ff2cd3f231aa260924036d5c2146d2976bcce5bd620bfb4902b703e
I1204 10:23:46.486757 3001 isa_program_util_common.cc:527] (HLO module jit_convert_element_type): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:46.486976 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 31.00768ms
I1204 10:23:46.486992 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:46.520721 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.67122925ms
I1204 10:23:46.522278 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:46.522872 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:46.524445 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:46.533266 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 11.06848925ms
I1204 10:23:46.541533 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 8.15692925ms
I1204 10:23:46.541588 3001 isa_program_util_common.cc:520] (HLO module jit_less): Executable fingerprint:65e08e6dea8e33aab77624dc00584b43490928a9c8896be74b1c20cb1e8dce86
I1204 10:23:46.541592 3001 isa_program_util_common.cc:524] (HLO module jit_less): Executable fingerprint (including data segments):0e0da679d6117b834588b357f7d5bfee13d38a6977f4220c819ac7ad72dcb7a0
I1204 10:23:46.541594 3001 isa_program_util_common.cc:527] (HLO module jit_less): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:46.541847 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 32.5683ms
I1204 10:23:46.541862 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:46.575165 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.70775075ms
I1204 10:23:46.576699 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:46.577294 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:46.578835 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:46.587659 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 11.04164925ms
I1204 10:23:46.595889 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 8.12117075ms
I1204 10:23:46.595941 3001 isa_program_util_common.cc:520] (HLO module jit_add): Executable fingerprint:a07fb16f37a51410027642d2558af2b8312641551f14b96b318d1b64d109cd8a
I1204 10:23:46.595945 3001 isa_program_util_common.cc:524] (HLO module jit_add): Executable fingerprint (including data segments):0b69039e2b40f08844d683c9d6ca9507553f24650c1475246c76fa6518b26230
I1204 10:23:46.595947 3001 isa_program_util_common.cc:527] (HLO module jit_add): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:46.596185 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 32.42584ms
I1204 10:23:46.596199 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:46.630762 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.94494ms
I1204 10:23:46.632322 3016 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:46.632941 3016 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:46.634778 3016 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:46.644435 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 12.19646925ms
I1204 10:23:46.652984 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 8.42948ms
I1204 10:23:46.653039 3001 isa_program_util_common.cc:520] (HLO module jit_select_n): Executable fingerprint:2581270d76a08687c8520b0aeee890d4d33298c4cddd094d34a618b6791aff11
I1204 10:23:46.653043 3001 isa_program_util_common.cc:524] (HLO module jit_select_n): Executable fingerprint (including data segments):2be3a1c0e4f83598a904f75522ffa84867fc5fb8548000c16ac6c41c66e6ffef
I1204 10:23:46.653045 3001 isa_program_util_common.cc:527] (HLO module jit_select_n): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:46.653298 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 34.13202925ms
I1204 10:23:46.653306 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:46.688053 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.84709075ms
I1204 10:23:46.689602 3016 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:46.690186 3016 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:46.691979 3016 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:46.701049 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 11.53274ms
I1204 10:23:46.709406 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 8.246ms
I1204 10:23:46.709468 3001 isa_program_util_common.cc:520] (HLO module jit_dynamic_slice): Executable fingerprint:c804a2842fcb06e2eefee97f57c940e41701928cdaa58ae153cdbb66b83e8e5d
I1204 10:23:46.709475 3001 isa_program_util_common.cc:524] (HLO module jit_dynamic_slice): Executable fingerprint (including data segments):2ea1bda4ead727cf06e13c97f0276c273fcaf965d71cce1dfd045ac956d8a7b5
I1204 10:23:46.709477 3001 isa_program_util_common.cc:527] (HLO module jit_dynamic_slice): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:46.709720 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 33.48066ms
I1204 10:23:46.709729 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:46.744443 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.69108ms
I1204 10:23:46.745993 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:46.746574 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:46.747937 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:46.755953 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 10.037ms
I1204 10:23:46.764019 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 7.96398925ms
I1204 10:23:46.764070 3001 isa_program_util_common.cc:520] (HLO module jit_squeeze): Executable fingerprint:1dc54db94e561e07f7b51f7dbd3593cdb201c92cb525d0e05e5be92a49846064
I1204 10:23:46.764073 3001 isa_program_util_common.cc:524] (HLO module jit_squeeze): Executable fingerprint (including data segments):40f165ab5ff2cd3f231aa260924036d5c2146d2976bcce5bd620bfb4902b703e
I1204 10:23:46.764083 3001 isa_program_util_common.cc:527] (HLO module jit_squeeze): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:46.764303 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 31.47456ms
I1204 10:23:46.764317 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:46.798709 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.74106ms
I1204 10:23:46.800219 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:46.800813 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:46.802472 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:46.811570 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 11.41402925ms
I1204 10:23:46.819800 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 8.12446ms
I1204 10:23:46.819851 3001 isa_program_util_common.cc:520] (HLO module jit_dynamic_slice): Executable fingerprint:c804a2842fcb06e2eefee97f57c940e41701928cdaa58ae153cdbb66b83e8e5d
I1204 10:23:46.819854 3001 isa_program_util_common.cc:524] (HLO module jit_dynamic_slice): Executable fingerprint (including data segments):2ea1bda4ead727cf06e13c97f0276c273fcaf965d71cce1dfd045ac956d8a7b5
I1204 10:23:46.819865 3001 isa_program_util_common.cc:527] (HLO module jit_dynamic_slice): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:46.820103 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 32.96545075ms
I1204 10:23:46.820120 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:46.852003 2819 python_stack_size.cc:37] EventManager thread stack size of 131072 might be too small for Python callbacks. Crash? Read go/py-cpp-stack.
I1204 10:23:46.855667 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.07090925ms
I1204 10:23:46.857158 3016 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:46.857756 3016 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:46.859451 3016 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:46.868272 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 11.17988925ms
I1204 10:23:46.877623 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 9.23346ms
I1204 10:23:46.877683 3001 isa_program_util_common.cc:520] (HLO module jit_subtract): Executable fingerprint:76c219bb81478c30d7fb0b0b1ad140e4cc429a59a105bad788c4442db4bf6351
I1204 10:23:46.877687 3001 isa_program_util_common.cc:524] (HLO module jit_subtract): Executable fingerprint (including data segments):a4575fa86e5d7fa75e9c3c288f162118dcc185fe2b314b670e66c09066583c93
I1204 10:23:46.877689 3001 isa_program_util_common.cc:527] (HLO module jit_subtract): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:46.877958 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 34.38245075ms
I1204 10:23:46.877973 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:46.913046 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.58028925ms
I1204 10:23:46.914535 3016 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:46.915206 3016 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:46.917076 3016 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:46.926062 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 11.60249075ms
I1204 10:23:46.935305 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 9.14128925ms
I1204 10:23:46.935360 3001 isa_program_util_common.cc:520] (HLO module jit_dynamic_slice): Executable fingerprint:77dd43fc955b532b16a258b7aaa591a68a5589f8f1b0bd53448945101bcd1af3
I1204 10:23:46.935364 3001 isa_program_util_common.cc:524] (HLO module jit_dynamic_slice): Executable fingerprint (including data segments):0e4c42223355aed208ca2aaf348f8a94b208ef0fd813e2ad038356532420966c
I1204 10:23:46.935366 3001 isa_program_util_common.cc:527] (HLO module jit_dynamic_slice): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:46.935649 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 34.13667075ms
I1204 10:23:46.935658 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:46.970848 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.58756ms
I1204 10:23:46.972358 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:46.972998 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:46.974703 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:46.983510 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 11.25363075ms
I1204 10:23:46.992771 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 9.15092ms
I1204 10:23:46.992826 3001 isa_program_util_common.cc:520] (HLO module jit_add): Executable fingerprint:a07fb16f37a51410027642d2558af2b8312641551f14b96b318d1b64d109cd8a
I1204 10:23:46.992830 3001 isa_program_util_common.cc:524] (HLO module jit_add): Executable fingerprint (including data segments):0b69039e2b40f08844d683c9d6ca9507553f24650c1475246c76fa6518b26230
I1204 10:23:46.992833 3001 isa_program_util_common.cc:527] (HLO module jit_add): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:46.993081 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 33.68086ms
I1204 10:23:46.993088 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:47.026875 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.51006ms
I1204 10:23:47.028389 3016 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:47.029039 3016 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:47.030759 3016 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:47.039534 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 11.23487075ms
I1204 10:23:47.048781 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 9.14902ms
I1204 10:23:47.048840 3001 isa_program_util_common.cc:520] (HLO module jit_subtract): Executable fingerprint:76c219bb81478c30d7fb0b0b1ad140e4cc429a59a105bad788c4442db4bf6351
I1204 10:23:47.048848 3001 isa_program_util_common.cc:524] (HLO module jit_subtract): Executable fingerprint (including data segments):a4575fa86e5d7fa75e9c3c288f162118dcc185fe2b314b670e66c09066583c93
I1204 10:23:47.048850 3001 isa_program_util_common.cc:527] (HLO module jit_subtract): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:47.049096 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 33.57602925ms
I1204 10:23:47.049103 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:47.094927 2807 deepsea_compiler_base.cc:7855] XLA::TPU running hlo passes for 17 instructions, module: jit_floor_divide
I1204 10:23:47.094978 2807 deepsea_compiler_base.cc:7910] Initial HLO module: jit_floor_divide instructions: 17 fingerprint: 1036109573815570274-4035828241379739704-17078542282357815393-1242137874306953380/5
I1204 10:23:47.095389 2807 deepsea_compiler_base.cc:7982] HLO optimizing module: jit_floor_divide instructions: 13
I1204 10:23:47.095396 2807 deepsea_compiler_base.cc:7997] XLA::TPU HLO optimization
I1204 10:23:47.100385 2807 deepsea_compiler_base.cc:6963] XLA::TPU HLO PostOptimizationPipeline
I1204 10:23:47.102480 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 7.50131075ms
I1204 10:23:47.104125 3016 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:47.104824 3016 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:47.107694 3016 window_config_assignment_prod.cc:30] Retrieving backend configs from FDO profiles.
I1204 10:23:47.108034 3016 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:47.119419 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 15.39104ms
I1204 10:23:47.121949 3001 deepsea_compiler_base.cc:3021] final program bundle count: 201 note this count does not reflect cycles spent executing delays.
I1204 10:23:47.128347 3001 deepsea_compiler_base.cc:3021] final program bundle count: 176 note this count does not reflect cycles spent executing delays.
I1204 10:23:47.129527 3001 deepsea_compiler_base.cc:3507] Program divided into 2 overlays without HLO functions (40.0K).
I1204 10:23:47.129667 3001 deepsea_compiler_base.cc:3675] XLA::TPU module name: jit_floor_divide
I1204 10:23:47.129671 3001 deepsea_compiler_base.cc:3677] XLA::TPU program HBM usage: 1.04M / 31.25G
I1204 10:23:47.129677 3001 deepsea_compiler_base.cc:3724] XLA::TPU program VMEM usage: 5.0K / 128.00M
I1204 10:23:47.129716 3001 deepsea_compiler_base.cc:3735] Total hbm usage >= 261.04M:
I1204 10:23:47.129726 3001 deepsea_compiler_base.cc:3735] reserved 260.00M
I1204 10:23:47.129728 3001 deepsea_compiler_base.cc:3735] program 1.04M
I1204 10:23:47.129730 3001 deepsea_compiler_base.cc:3735] arguments 1.0K
I1204 10:23:47.129732 3001 deepsea_compiler_base.cc:3735]
I1204 10:23:47.129734 3001 deepsea_compiler_base.cc:3735] Output size 512B; shares 0B with arguments.
I1204 10:23:47.129735 3001 deepsea_compiler_base.cc:3735]
I1204 10:23:47.129737 3001 deepsea_compiler_base.cc:3735] Total host usage >= 0B:
I1204 10:23:47.129739 3001 deepsea_compiler_base.cc:3735] reserved 0B
I1204 10:23:47.129740 3001 deepsea_compiler_base.cc:3735] program unknown size
I1204 10:23:47.129742 3001 deepsea_compiler_base.cc:3735] arguments 0B
I1204 10:23:47.129747 3001 deepsea_compiler_base.cc:3735]
I1204 10:23:47.129749 3001 deepsea_compiler_base.cc:3735] Output size 0B; shares 0B with arguments.
I1204 10:23:47.129751 3001 deepsea_compiler_base.cc:3735]
I1204 10:23:47.129765 3001 deepsea_compiler_base.cc:3739] Program sflag requirement 212B:
I1204 10:23:47.129771 3001 deepsea_compiler_base.cc:3739] reserved 204B
I1204 10:23:47.129774 3001 deepsea_compiler_base.cc:3739] global 4B
I1204 10:23:47.129776 3001 deepsea_compiler_base.cc:3739] scoped 4B
I1204 10:23:47.129777 3001 deepsea_compiler_base.cc:3739] Program smem requirement 48B:
I1204 10:23:47.129779 3001 deepsea_compiler_base.cc:3739] global 16B
I1204 10:23:47.129781 3001 deepsea_compiler_base.cc:3739] scoped 32B
I1204 10:23:47.129782 3001 deepsea_compiler_base.cc:3739] Program hbm requirement 1.04M:
I1204 10:23:47.129784 3001 deepsea_compiler_base.cc:3739] global 1.00M
I1204 10:23:47.129785 3001 deepsea_compiler_base.cc:3739] overlays 40.0K
I1204 10:23:47.129787 3001 deepsea_compiler_base.cc:3739] Program vmem requirement 5.0K:
I1204 10:23:47.129789 3001 deepsea_compiler_base.cc:3739] scoped 5.0K
I1204 10:23:47.129790 3001 deepsea_compiler_base.cc:3739] Program host requirement 0B:
I1204 10:23:47.129793 3001 deepsea_compiler_base.cc:3747] XLA::TPU program SMEM usage: 1.9K / 1.00M (2 parameters)
I1204 10:23:47.129810 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 10.28034ms
I1204 10:23:47.129858 3001 isa_program_util_common.cc:520] (HLO module jit_floor_divide): Executable fingerprint:28aeabef3111155990e2eeb72b2f561e6d99c90b7d34240223b655237f87b456
I1204 10:23:47.129861 3001 isa_program_util_common.cc:524] (HLO module jit_floor_divide): Executable fingerprint (including data segments):37af056763090cbbf6b41099ffe9b26c0c79db2a28da029bc0113ee1cbcc2923
I1204 10:23:47.129863 3001 isa_program_util_common.cc:527] (HLO module jit_floor_divide): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:47.130216 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 42.36749075ms
I1204 10:23:47.130224 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:47.166257 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.60591075ms
I1204 10:23:47.167741 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:47.168402 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:47.170356 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:47.179584 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 11.92019075ms
I1204 10:23:47.189266 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 9.57096ms
I1204 10:23:47.189322 3001 isa_program_util_common.cc:520] (HLO module jit_dynamic_slice): Executable fingerprint:34b72ddf3ad8cec6bc6263db2be285cdbb8e6ffed72e8335337ac6dd25a5d006
I1204 10:23:47.189328 3001 isa_program_util_common.cc:524] (HLO module jit_dynamic_slice): Executable fingerprint (including data segments):4dfd14712228d2d15e5e32e9dbb549e1434ba06888dab43a97f0a838b8ed66a6
I1204 10:23:47.189331 3001 isa_program_util_common.cc:527] (HLO module jit_dynamic_slice): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:47.189609 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 34.97886ms
I1204 10:23:47.189618 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:47.224354 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.53538ms
I1204 10:23:47.225876 3016 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:47.226531 3016 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:47.228236 3016 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:47.237151 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 11.3534ms
I1204 10:23:47.246304 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 9.04120925ms
I1204 10:23:47.246357 3001 isa_program_util_common.cc:520] (HLO module jit_greater_equal): Executable fingerprint:d067271163a8b50167f309e00f28535fcee3f1356aca2924abce71751e68199b
I1204 10:23:47.246360 3001 isa_program_util_common.cc:524] (HLO module jit_greater_equal): Executable fingerprint (including data segments):07cfc736c007e6222cc467c2c23263c237da9f1c3b78d1d38be25e47a63b59d4
I1204 10:23:47.246363 3001 isa_program_util_common.cc:527] (HLO module jit_greater_equal): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:47.246643 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 33.77224ms
I1204 10:23:47.246653 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:47.360030 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.8005ms
I1204 10:23:47.361586 3016 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:47.362321 3016 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:47.364551 3016 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:47.457250 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 95.77456925ms
I1204 10:23:47.474607 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 17.16197075ms
I1204 10:23:47.474688 3001 isa_program_util_common.cc:520] (HLO module jit_concatenate): Executable fingerprint:79cfdcb080a293efad9fd238d2109150d8e48d8ea03dc3635ddb73ea5f47991e
I1204 10:23:47.474692 3001 isa_program_util_common.cc:524] (HLO module jit_concatenate): Executable fingerprint (including data segments):a4ceb27256d9faa09b27e02ba6a85ea45a8e78a110209b7df78a96b9c05adcca
I1204 10:23:47.474694 3001 isa_program_util_common.cc:527] (HLO module jit_concatenate): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:47.474994 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 127.91425075ms
I1204 10:23:47.475014 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:47.510115 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.59374ms
I1204 10:23:47.511598 3016 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:47.512271 3016 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:47.513762 3016 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:47.551167 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 39.66317075ms
I1204 10:23:47.562733 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 11.43082ms
I1204 10:23:47.562840 3001 isa_program_util_common.cc:520] (HLO module jit_reshape): Executable fingerprint:98e950b3f348ab0adf4057f38666308309238856bcdb4de105f93a35a8b0c8c6
I1204 10:23:47.562844 3001 isa_program_util_common.cc:524] (HLO module jit_reshape): Executable fingerprint (including data segments):d4af74f494afa0e6f5b5836c9af7e8ad2e18cf92ab72bf1267159984ea18b059
I1204 10:23:47.562854 3001 isa_program_util_common.cc:527] (HLO module jit_reshape): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:47.563162 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 64.54542ms
I1204 10:23:47.563171 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:47.597770 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.30672925ms
I1204 10:23:47.599237 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:47.599929 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:47.601498 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:47.685857 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 86.71722ms
I1204 10:23:47.702595 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 16.54446ms
I1204 10:23:47.702671 3001 isa_program_util_common.cc:520] (HLO module jit__pad): Executable fingerprint:ba655bacb8b03b49a7fa7dbe1fb9ab5ea3cad5db50546f10e70288202e74fc00
I1204 10:23:47.702674 3001 isa_program_util_common.cc:524] (HLO module jit__pad): Executable fingerprint (including data segments):9326a68aaf37ae99e1ad0f843fc5b221e5fbf7d73d40a7c6187ef81b5e524799
I1204 10:23:47.702676 3001 isa_program_util_common.cc:527] (HLO module jit__pad): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:47.702936 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 116.24700925ms
I1204 10:23:47.702946 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:47.737056 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.58986ms
I1204 10:23:47.738528 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:47.739254 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:47.740885 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:47.836143 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 97.71078ms
I1204 10:23:47.853957 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 17.61696925ms
I1204 10:23:47.854041 3001 isa_program_util_common.cc:520] (HLO module jit_reshape): Executable fingerprint:f69d5ad88a18e713060d61580688dac1df5791282fc117adc4f4998d497cbe4b
I1204 10:23:47.854049 3001 isa_program_util_common.cc:524] (HLO module jit_reshape): Executable fingerprint (including data segments):c362ae97d07eca6bbb264ee886524be4110a80fe5ba3bcbb77dd2ce18acae07b
I1204 10:23:47.854051 3001 isa_program_util_common.cc:527] (HLO module jit_reshape): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:47.854364 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 128.84420925ms
I1204 10:23:47.854376 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:47.892465 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.02850925ms
I1204 10:23:47.894057 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:47.894778 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:47.896772 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:47.907591 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 13.61396ms
I1204 10:23:47.917029 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 9.2994ms
I1204 10:23:47.917090 3001 isa_program_util_common.cc:520] (HLO module jit_dynamic_slice): Executable fingerprint:e134e0aef1fe6b43a4b9bac4217863429e6c119d61ec4ce85d124aed333612f5
I1204 10:23:47.917094 3001 isa_program_util_common.cc:524] (HLO module jit_dynamic_slice): Executable fingerprint (including data segments):b9efb8ec84916f969cba69529fbd66c1cd2d3a855be41d886b59e5ab08324130
I1204 10:23:47.917096 3001 isa_program_util_common.cc:527] (HLO module jit_dynamic_slice): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:47.917355 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 36.96712ms
I1204 10:23:47.917367 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:47.955020 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.49432ms
I1204 10:23:47.956498 3012 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:47.957254 3012 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:47.959153 3012 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:47.968592 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 12.17106ms
I1204 10:23:47.977853 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 9.15305075ms
I1204 10:23:47.977910 3001 isa_program_util_common.cc:520] (HLO module jit_add): Executable fingerprint:a07fb16f37a51410027642d2558af2b8312641551f14b96b318d1b64d109cd8a
I1204 10:23:47.977913 3001 isa_program_util_common.cc:524] (HLO module jit_add): Executable fingerprint (including data segments):0b69039e2b40f08844d683c9d6ca9507553f24650c1475246c76fa6518b26230
I1204 10:23:47.977915 3001 isa_program_util_common.cc:527] (HLO module jit_add): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:47.978181 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 34.53116925ms
I1204 10:23:47.978191 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:48.012686 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.48847075ms
I1204 10:23:48.014202 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:48.014874 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:48.016538 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:48.024803 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 10.69396ms
I1204 10:23:48.033868 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 8.95256ms
I1204 10:23:48.033922 3001 isa_program_util_common.cc:520] (HLO module jit_broadcast_in_dim): Executable fingerprint:1dc54db94e561e07f7b51f7dbd3593cdb201c92cb525d0e05e5be92a49846064
I1204 10:23:48.033926 3001 isa_program_util_common.cc:524] (HLO module jit_broadcast_in_dim): Executable fingerprint (including data segments):40f165ab5ff2cd3f231aa260924036d5c2146d2976bcce5bd620bfb4902b703e
I1204 10:23:48.033928 3001 isa_program_util_common.cc:527] (HLO module jit_broadcast_in_dim): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:48.034186 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 32.83322ms
I1204 10:23:48.034198 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:48.072824 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 6.7262ms
I1204 10:23:48.074609 3012 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:48.075245 3012 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:48.077292 3012 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:48.086559 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 12.04410925ms
I1204 10:23:48.095997 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 9.30328925ms
I1204 10:23:48.096055 3001 isa_program_util_common.cc:520] (HLO module jit_gather): Executable fingerprint:6dbf22bcc2b6eda4a2575d7f8880971c06b6233acb2402b80705fd5255bf0077
I1204 10:23:48.096059 3001 isa_program_util_common.cc:524] (HLO module jit_gather): Executable fingerprint (including data segments):8695f315604b7024584477e7d24e0582bed544e1a4e4d2043b3a7530b6140327
I1204 10:23:48.096061 3001 isa_program_util_common.cc:527] (HLO module jit_gather): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:48.096393 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 37.61752ms
I1204 10:23:48.096416 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:48.130172 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.44646ms
I1204 10:23:48.131766 3012 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:48.132394 3012 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:48.133734 3012 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:48.141072 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 9.37936ms
I1204 10:23:48.149777 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 8.57460925ms
I1204 10:23:48.149832 3001 isa_program_util_common.cc:520] (HLO module jit_broadcast_in_dim): Executable fingerprint:2465fb80b7e81e394d71acdce5964d521e7f17722fe790d47e342a72d0ee4162
I1204 10:23:48.149835 3001 isa_program_util_common.cc:524] (HLO module jit_broadcast_in_dim): Executable fingerprint (including data segments):0d74f996d24d804ac425f2da854d60c51a95459ed55395522afa0de4027e8577
I1204 10:23:48.149837 3001 isa_program_util_common.cc:527] (HLO module jit_broadcast_in_dim): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:48.150067 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 31.45823075ms
I1204 10:23:48.150083 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:48.189437 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 6.84204925ms
I1204 10:23:48.191158 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:48.191857 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:48.194386 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:48.342333 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 151.26146ms
I1204 10:23:48.359060 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 16.52142ms
I1204 10:23:48.359139 3001 isa_program_util_common.cc:520] (HLO module jit_gather): Executable fingerprint:8993a8a13e3b34ae1de699aab6484152a640f5c806056e26c67d20ce94185165
I1204 10:23:48.359143 3001 isa_program_util_common.cc:524] (HLO module jit_gather): Executable fingerprint (including data segments):b0c809237b36ca31c4ecfd762f3f64b138389845d6373695cb08195178f2d1f9
I1204 10:23:48.359145 3001 isa_program_util_common.cc:527] (HLO module jit_gather): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:48.359448 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 183.89480925ms
I1204 10:23:48.359458 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:48.393637 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.67378925ms
I1204 10:23:48.395245 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:48.395986 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:48.397483 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:48.540050 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 144.89894925ms
I1204 10:23:48.555477 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 15.24837075ms
I1204 10:23:48.555559 3001 isa_program_util_common.cc:520] (HLO module jit_broadcast_in_dim): Executable fingerprint:643b1defc25bd70e9d950a704e34fa47714de96efb21298f2f23581284467a8f
I1204 10:23:48.555563 3001 isa_program_util_common.cc:524] (HLO module jit_broadcast_in_dim): Executable fingerprint (including data segments):92baca98a1a512f9d016cbbaa949cf728b242b5252db7747f4411eb7e17c7f24
I1204 10:23:48.555565 3001 isa_program_util_common.cc:527] (HLO module jit_broadcast_in_dim): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:48.555811 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 173.70204925ms
I1204 10:23:48.555819 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:48.594169 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.43698ms
I1204 10:23:48.595783 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:48.596545 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:48.598878 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:48.607932 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 12.25528925ms
I1204 10:23:48.618291 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 10.23335075ms
I1204 10:23:48.618348 3001 isa_program_util_common.cc:520] (HLO module jit_less): Executable fingerprint:35c4cec8e92e8cc8dca7839e591ad5b5ff3a2bc33a8d416713a1cbe8bfea7e09
I1204 10:23:48.618351 3001 isa_program_util_common.cc:524] (HLO module jit_less): Executable fingerprint (including data segments):b6dfb81f2124f37247c236beaed32223900393979bd055a7cfc141ffa1b764d0
I1204 10:23:48.618353 3001 isa_program_util_common.cc:527] (HLO module jit_less): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:48.618627 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 37.25264ms
I1204 10:23:48.618637 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:48.657894 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.70716925ms
I1204 10:23:48.659423 3012 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:48.660026 3012 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:48.661966 3012 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:48.670944 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 11.57654ms
I1204 10:23:48.681028 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 9.95092ms
I1204 10:23:48.681084 3001 isa_program_util_common.cc:520] (HLO module jit_add): Executable fingerprint:21bda17af4c81842aba49d87f9675c4e48e9986c436edbd6526e3c621eb0c021
I1204 10:23:48.681087 3001 isa_program_util_common.cc:524] (HLO module jit_add): Executable fingerprint (including data segments):fac29855c403755254c9a8b3b4250e9ec8fe78578b05c922bc3264f0ac2fde88
I1204 10:23:48.681090 3001 isa_program_util_common.cc:527] (HLO module jit_add): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:48.681406 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 36.53552ms
I1204 10:23:48.681415 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:48.717022 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.12030925ms
I1204 10:23:48.718583 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:48.719277 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:48.721304 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:48.730428 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 11.9391ms
I1204 10:23:48.740660 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 10.07648ms
I1204 10:23:48.740716 3001 isa_program_util_common.cc:520] (HLO module jit_select_n): Executable fingerprint:1b4af806359ca11a9aeedcab09cec4115d359228b512823f789d069d0a412454
I1204 10:23:48.740720 3001 isa_program_util_common.cc:524] (HLO module jit_select_n): Executable fingerprint (including data segments):101c5260488115e2b7ec811bcd3a241daf5827b1cc7bf76ea72a4ccf9fc17426
I1204 10:23:48.740723 3001 isa_program_util_common.cc:527] (HLO module jit_select_n): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:48.741041 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 36.05921075ms
I1204 10:23:48.741049 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:48.774394 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.3983ms
I1204 10:23:48.775905 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:48.776607 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:48.778001 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:48.786199 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 10.37121075ms
I1204 10:23:48.795096 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 8.79202ms
I1204 10:23:48.795153 3001 isa_program_util_common.cc:520] (HLO module jit__broadcast_arrays): Executable fingerprint:2465fb80b7e81e394d71acdce5964d521e7f17722fe790d47e342a72d0ee4162
I1204 10:23:48.795158 3001 isa_program_util_common.cc:524] (HLO module jit__broadcast_arrays): Executable fingerprint (including data segments):0d74f996d24d804ac425f2da854d60c51a95459ed55395522afa0de4027e8577
I1204 10:23:48.795160 3001 isa_program_util_common.cc:527] (HLO module jit__broadcast_arrays): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:48.795394 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 32.60257075ms
I1204 10:23:48.795403 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:48.830249 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.87455075ms
I1204 10:23:48.831779 3012 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:48.832309 3012 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:48.833736 3012 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:48.841683 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 9.96619075ms
I1204 10:23:48.850723 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 8.91459075ms
I1204 10:23:48.850779 3001 isa_program_util_common.cc:520] (HLO module jit_broadcast_in_dim): Executable fingerprint:f8a8a1bcc14437ec76793f8d974adac664e3c5f5e676f34082c65624e495eac5
I1204 10:23:48.850783 3001 isa_program_util_common.cc:524] (HLO module jit_broadcast_in_dim): Executable fingerprint (including data segments):b7175bd87ce618f7322eb6a7c85b4a92b395ad6e0edb5ac68d5ee005fd07f01b
I1204 10:23:48.850786 3001 isa_program_util_common.cc:527] (HLO module jit_broadcast_in_dim): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:48.851061 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 32.67182ms
I1204 10:23:48.851070 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:48.891833 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 8.41586ms
I1204 10:23:48.893485 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:48.894147 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:48.897155 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:48.960603 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 67.2102ms
I1204 10:23:48.976777 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 15.99620925ms
I1204 10:23:48.976858 3001 isa_program_util_common.cc:520] (HLO module jit_gather): Executable fingerprint:cbde3be4e356e4938bf07c86d90516989f3acf9efc7dbfc1bfbe5b683c04502f
I1204 10:23:48.976862 3001 isa_program_util_common.cc:524] (HLO module jit_gather): Executable fingerprint (including data segments):04eaea747692bfd3b8bfd91f730e85477b775c9d0ee129a5a23463f88e954681
I1204 10:23:48.976864 3001 isa_program_util_common.cc:527] (HLO module jit_gather): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:48.977162 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 101.08448ms
I1204 10:23:48.977171 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:49.012438 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.26598925ms
I1204 10:23:49.013947 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:49.014696 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:49.016214 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:49.413790 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 399.94772925ms
I1204 10:23:49.466015 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 51.99276ms
I1204 10:23:49.466213 3001 isa_program_util_common.cc:520] (HLO module jit_broadcast_in_dim): Executable fingerprint:24fa2943c08d6e85d3563eb8055a854b296430fed1598d114793a7dabc3284ce
I1204 10:23:49.466218 3001 isa_program_util_common.cc:524] (HLO module jit_broadcast_in_dim): Executable fingerprint (including data segments):78dc18e4087d7bbde5b83f9d0cdc3f136a4beb0593c86ac2e88ee01f816d0031
I1204 10:23:49.466221 3001 isa_program_util_common.cc:527] (HLO module jit_broadcast_in_dim): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:49.466480 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 465.24581075ms
I1204 10:23:49.466489 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:49.502565 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.0198ms
I1204 10:23:49.504126 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:49.504846 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:49.506514 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:49.867784 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 363.73726ms
I1204 10:23:49.920864 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 52.87412ms
I1204 10:23:49.921069 3001 isa_program_util_common.cc:520] (HLO module jit_reshape): Executable fingerprint:910cf6757b5536fb12c97d454760da1b79471ffbd77a29d948783b4d33f51fa8
I1204 10:23:49.921073 3001 isa_program_util_common.cc:524] (HLO module jit_reshape): Executable fingerprint (including data segments):255c24a46c262a2185ce2505f33e38ad07b3b674795f573ca2bb14add1ac5bfc
I1204 10:23:49.921075 3001 isa_program_util_common.cc:527] (HLO module jit_reshape): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:49.921342 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 430.98746ms
I1204 10:23:49.921352 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:49.961639 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 6.98327075ms
I1204 10:23:49.963368 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:49.964088 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:49.966655 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:50.210814 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 247.51322925ms
I1204 10:23:50.245180 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 34.16530925ms
I1204 10:23:50.245321 3001 isa_program_util_common.cc:520] (HLO module jit_gather): Executable fingerprint:77322e5d0182b5e1ca049b9fb07cb934196c6b1d33a433c64c0ec6d80debf694
I1204 10:23:50.245327 3001 isa_program_util_common.cc:524] (HLO module jit_gather): Executable fingerprint (including data segments):d3225bf8183ff6c1e511511703ae99b08b88ccafbce1aa34e350cfeb1ce9e462
I1204 10:23:50.245329 3001 isa_program_util_common.cc:527] (HLO module jit_gather): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:50.245649 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 298.05312ms
I1204 10:23:50.245658 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:50.280158 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.21912ms
I1204 10:23:50.281703 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:50.282441 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:50.283962 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
W1204 10:23:50.407964 3126 internals-datagram-socket.cc:411] udp sendto fd=202, dest=172.20.0.120:9530: Network is unreachable
I1204 10:23:50.492347 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 210.73148925ms
I1204 10:23:50.525905 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 33.35810925ms
I1204 10:23:50.526044 3001 isa_program_util_common.cc:520] (HLO module jit_broadcast_in_dim): Executable fingerprint:11969447be7e1e557978b02b38e74a5eeab6b8d9506c2b616319aa7fd2a4053b
I1204 10:23:50.526047 3001 isa_program_util_common.cc:524] (HLO module jit_broadcast_in_dim): Executable fingerprint (including data segments):2856eac5ab542f37950fcd501545a8e2f6706752abe81d42cd2f34d2e6474a1d
I1204 10:23:50.526050 3001 isa_program_util_common.cc:527] (HLO module jit_broadcast_in_dim): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:50.526346 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 257.42726925ms
I1204 10:23:50.526358 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:50.561466 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.24888ms
I1204 10:23:50.562935 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:50.563655 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:50.565199 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:50.774017 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 211.1756ms
I1204 10:23:50.807551 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 33.33932925ms
I1204 10:23:50.807687 3001 isa_program_util_common.cc:520] (HLO module jit__squeeze): Executable fingerprint:11969447be7e1e557978b02b38e74a5eeab6b8d9506c2b616319aa7fd2a4053b
I1204 10:23:50.807691 3001 isa_program_util_common.cc:524] (HLO module jit__squeeze): Executable fingerprint (including data segments):2856eac5ab542f37950fcd501545a8e2f6706752abe81d42cd2f34d2e6474a1d
I1204 10:23:50.807694 3001 isa_program_util_common.cc:527] (HLO module jit__squeeze): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:50.807974 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 257.6446ms
I1204 10:23:50.807985 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:50.852460 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 13.24191075ms
I1204 10:23:50.854118 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:50.854848 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:50.859858 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:51.400909 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 546.86363075ms
I1204 10:23:51.460230 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 59.1212ms
I1204 10:23:51.460553 3001 isa_program_util_common.cc:520] (HLO module jit_scatter): Executable fingerprint:81ef5910e0aa1735563ee4788de44c0536b3fdb18bbf5ac65d4c1de5d57b868a
I1204 10:23:51.460557 3001 isa_program_util_common.cc:524] (HLO module jit_scatter): Executable fingerprint (including data segments):bb43376c584020da43359c3e103bba91854b227bec8a9b344db2430f9279f8a8
I1204 10:23:51.460559 3001 isa_program_util_common.cc:527] (HLO module jit_scatter): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:51.460890 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 628.7013ms
I1204 10:23:51.460900 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:51.502138 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.76688ms
I1204 10:23:51.503697 3012 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:51.504400 3012 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:51.505900 3012 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:51.897016 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 393.41857075ms
I1204 10:23:51.949020 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 51.8072ms
I1204 10:23:51.949233 3001 isa_program_util_common.cc:520] (HLO module jit__squeeze): Executable fingerprint:24fa2943c08d6e85d3563eb8055a854b296430fed1598d114793a7dabc3284ce
I1204 10:23:51.949237 3001 isa_program_util_common.cc:524] (HLO module jit__squeeze): Executable fingerprint (including data segments):78dc18e4087d7bbde5b83f9d0cdc3f136a4beb0593c86ac2e88ee01f816d0031
I1204 10:23:51.949239 3001 isa_program_util_common.cc:527] (HLO module jit__squeeze): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:51.949478 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 459.198ms
I1204 10:23:51.949487 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:51.989774 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 8.63343075ms
I1204 10:23:51.991404 3012 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:51.992159 3012 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:51.995505 3012 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:52.251218 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 259.91002ms
I1204 10:23:52.273349 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 21.92456ms
I1204 10:23:52.273506 3001 isa_program_util_common.cc:520] (HLO module jit_scatter): Executable fingerprint:49769ce01e1ecc83cd707e3cec45a698b957e4dc18030df0d19e0f343c94cb3d
I1204 10:23:52.273512 3001 isa_program_util_common.cc:524] (HLO module jit_scatter): Executable fingerprint (including data segments):a592d1782edcbd7bda62c783d2c8154e763f95a75153cb6e2cb587dad075da02
I1204 10:23:52.273514 3001 isa_program_util_common.cc:527] (HLO module jit_scatter): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:52.273845 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 299.72467075ms
I1204 10:23:52.273854 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:52.312076 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.70138ms
I1204 10:23:52.313559 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:52.314229 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:52.315763 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:52.372218 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 58.73449075ms
I1204 10:23:52.386478 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 14.07044ms
I1204 10:23:52.386560 3001 isa_program_util_common.cc:520] (HLO module jit_reshape): Executable fingerprint:85f12decfdd9dd61290fd15493548c4806429f47b7cfbd5e6a3f9b2188aa3362
I1204 10:23:52.386564 3001 isa_program_util_common.cc:524] (HLO module jit_reshape): Executable fingerprint (including data segments):0aa65461dc5fcff478f1b7b823af7d7c4e23c86fdc95e9dcbb6f4cc16f0a1e9c
I1204 10:23:52.386566 3001 isa_program_util_common.cc:527] (HLO module jit_reshape): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:52.386839 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 86.53136ms
I1204 10:23:52.386850 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:52.422657 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.61926925ms
I1204 10:23:52.424182 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:52.424874 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:52.426385 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:52.514163 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 90.07563075ms
I1204 10:23:52.533163 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 18.80276925ms
I1204 10:23:52.533285 3001 isa_program_util_common.cc:520] (HLO module jit_reshape): Executable fingerprint:08cf8355b57a7e64f0cefb91fb41cb3025e8edd5448022ba6b8970dee295107d
I1204 10:23:52.533291 3001 isa_program_util_common.cc:524] (HLO module jit_reshape): Executable fingerprint (including data segments):7f3b978b2926cfd1a1e94b310f926b20cc54072f2508d884d405fffd36ef9434
I1204 10:23:52.533294 3001 isa_program_util_common.cc:527] (HLO module jit_reshape): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:52.533613 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 122.66715075ms
I1204 10:23:52.533627 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:52.573420 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.73180925ms
I1204 10:23:52.575058 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:52.575771 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:52.578068 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:52.587219 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 12.26106ms
I1204 10:23:52.597154 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 9.80092ms
I1204 10:23:52.597223 3001 isa_program_util_common.cc:520] (HLO module jit_concatenate): Executable fingerprint:26f48f6bba8f53cf38e5ec3e8cea4abb4043ae650e253af1dcc5fe20370edb71
I1204 10:23:52.597229 3001 isa_program_util_common.cc:524] (HLO module jit_concatenate): Executable fingerprint (including data segments):899fe7908c122ad112f42ed34e652ecc650c79ad1d45dc756b9fa9b5d5563985
I1204 10:23:52.597231 3001 isa_program_util_common.cc:527] (HLO module jit_concatenate): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:52.597512 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 37.07199075ms
I1204 10:23:52.597526 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:52.637652 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 7.47284925ms
I1204 10:23:52.639483 3136 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:52.640145 3136 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:52.644856 3136 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:52.744767 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 105.37612ms
I1204 10:23:52.761680 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 16.71682ms
I1204 10:23:52.761784 3001 isa_program_util_common.cc:520] (HLO module jit_gather): Executable fingerprint:9bf0329603661de57f6d5ccccac40f6d6ca8e42d24de6eed4eb54b8f4bad5b7f
I1204 10:23:52.761788 3001 isa_program_util_common.cc:524] (HLO module jit_gather): Executable fingerprint (including data segments):388ad2828ce29a4318c3b95735a7a7f4d272fa983c717747ee6dcf516a3dddb5
I1204 10:23:52.761790 3001 isa_program_util_common.cc:527] (HLO module jit_gather): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:52.762105 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 139.06459075ms
I1204 10:23:52.762118 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:52.797939 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.44887075ms
I1204 10:23:52.799510 3136 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:52.800173 3136 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:52.801631 3136 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:52.832789 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 33.36750925ms
I1204 10:23:52.844427 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 11.49676ms
I1204 10:23:52.844494 3001 isa_program_util_common.cc:520] (HLO module jit_broadcast_in_dim): Executable fingerprint:fad7e4c000697dacf24eeb93ff20a3d2c932536656322c0cb54ec29fc0ff44b2
I1204 10:23:52.844500 3001 isa_program_util_common.cc:524] (HLO module jit_broadcast_in_dim): Executable fingerprint (including data segments):46b973fd87af58407cab1ebce8e811378ebf30ab6d66e93cc652c62029f149da
I1204 10:23:52.844502 3001 isa_program_util_common.cc:527] (HLO module jit_broadcast_in_dim): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:52.844744 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 58.45574ms
I1204 10:23:52.844755 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:52.883445 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.62931075ms
I1204 10:23:52.885025 3136 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:52.885721 3136 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:52.887477 3136 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:52.937576 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 52.63268ms
I1204 10:23:52.950192 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 12.4548ms
I1204 10:23:52.950290 3001 isa_program_util_common.cc:520] (HLO module jit_broadcast_in_dim): Executable fingerprint:59277f23305489ec45aaf709a71d7012390bb9c1069fd892f617f059799afebe
I1204 10:23:52.950294 3001 isa_program_util_common.cc:524] (HLO module jit_broadcast_in_dim): Executable fingerprint (including data segments):24e7e78a202e4761c3fe61503d4c8217f49e540492262b387909dc47915584f6
I1204 10:23:52.950297 3001 isa_program_util_common.cc:527] (HLO module jit_broadcast_in_dim): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:52.950547 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 78.99323075ms
I1204 10:23:52.950556 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:52.984876 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.67611075ms
I1204 10:23:52.986390 3136 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:52.986973 3136 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:52.988218 3136 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:53.318904 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 332.59126ms
I1204 10:23:53.337957 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 18.87338ms
I1204 10:23:53.338072 3001 isa_program_util_common.cc:520] (HLO module jit_reshape): Executable fingerprint:65300bb358e89a3c27d2d8c289499042824f805ac40d3bba1b68860a504167e4
I1204 10:23:53.338076 3001 isa_program_util_common.cc:524] (HLO module jit_reshape): Executable fingerprint (including data segments):e7d1cd33f2a9bd73a73246f2dd766b849f6d7ac8ea9ff2e3fde49ff52196d2b9
I1204 10:23:53.338078 3001 isa_program_util_common.cc:527] (HLO module jit_reshape): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:53.338335 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 364.9667ms
I1204 10:23:53.338347 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:53.378733 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 7.28964ms
I1204 10:23:53.380302 3136 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:53.380954 3136 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:53.383808 3136 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:55.261466 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 1.88123976925s
W1204 10:23:55.363682 2825 internals-datagram-socket.cc:411] udp sendto fd=199, dest=172.20.0.120:9530: Network is unreachable
I1204 10:23:55.416887 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 155.23532ms
I1204 10:23:55.417344 3001 isa_program_util_common.cc:520] (HLO module jit__einsum): Executable fingerprint:769aece15c751438c1c4e0546515020a54076b8a5b38d741d1524b16814f645a
I1204 10:23:55.417350 3001 isa_program_util_common.cc:524] (HLO module jit__einsum): Executable fingerprint (including data segments):708e72a87bbf4d56294954190812b6b6acc0a9f11ff301e00b38984825f17517
I1204 10:23:55.417352 3001 isa_program_util_common.cc:527] (HLO module jit__einsum): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:55.417705 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 2.05328246s
I1204 10:23:55.417719 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:55.460277 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.43815075ms
I1204 10:23:55.461751 3136 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:55.462334 3136 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:55.464261 3136 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:55.621016 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 159.33156ms
I1204 10:23:55.642119 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 20.92194ms
I1204 10:23:55.642210 3001 isa_program_util_common.cc:520] (HLO module jit_multiply): Executable fingerprint:ee55a30d6918eb431ee9ee2bfe6e355055f9febd428c4fcc5f7ab9bfff308570
I1204 10:23:55.642214 3001 isa_program_util_common.cc:524] (HLO module jit_multiply): Executable fingerprint (including data segments):94b368ce984c1d98e7156b4e02ac3d518671728c9d0c6cff165f3bf51f3cca98
I1204 10:23:55.642216 3001 isa_program_util_common.cc:527] (HLO module jit_multiply): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:55.642483 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 194.70094ms
I1204 10:23:55.642493 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:55.677929 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.2353ms
I1204 10:23:55.679465 3136 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:55.680185 3136 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:55.681560 3136 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:55.790922 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 111.52685075ms
I1204 10:23:55.809278 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 18.18192ms
I1204 10:23:55.809410 3001 isa_program_util_common.cc:520] (HLO module jit_iota): Executable fingerprint:1807a72246352dcf96c7ba6ebb28cdc265a4b08450c0a03b30ed61d46fc33524
I1204 10:23:55.809417 3001 isa_program_util_common.cc:524] (HLO module jit_iota): Executable fingerprint (including data segments):ccf951d3cbbd223d5aed4c9c95ee668db8e9ecf233ed0176167d9cb86f67dd2e
I1204 10:23:55.809419 3001 isa_program_util_common.cc:527] (HLO module jit_iota): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:55.809754 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 143.23652ms
I1204 10:23:55.809767 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:55.846981 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.3628ms
I1204 10:23:55.848538 3136 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:55.849274 3136 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:55.851433 3136 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:56.011744 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 163.28094ms
I1204 10:23:56.033392 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 21.46318ms
I1204 10:23:56.033484 3001 isa_program_util_common.cc:520] (HLO module jit_add): Executable fingerprint:ec70923f9c44661a7176f74029da25ba9a66b4cf20cf0377792fd85d063c563a
I1204 10:23:56.033488 3001 isa_program_util_common.cc:524] (HLO module jit_add): Executable fingerprint (including data segments):da4c4595f6f5c1f479d67feec351fe94caa710068aa8201dfea612931a0520ef
I1204 10:23:56.033490 3001 isa_program_util_common.cc:527] (HLO module jit_add): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:56.033754 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 199.27475075ms
I1204 10:23:56.033762 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:56.068602 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.27386ms
I1204 10:23:56.070217 3136 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:56.070953 3136 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:56.072320 3136 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:56.181370 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 111.25098ms
I1204 10:23:56.199493 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 17.94703075ms
I1204 10:23:56.199599 3001 isa_program_util_common.cc:520] (HLO module jit_iota): Executable fingerprint:8a69d3c99e0924b1a36886314e454ab4c96bf2ca25d5aa291c3243fe928b6804
I1204 10:23:56.199605 3001 isa_program_util_common.cc:524] (HLO module jit_iota): Executable fingerprint (including data segments):1bf119b9d7b7f997324f2be660761d8711b3a8f0c5a0cc75406d5bee8bfc6be0
I1204 10:23:56.199608 3001 isa_program_util_common.cc:527] (HLO module jit_iota): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:56.199877 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 142.58133075ms
I1204 10:23:56.199891 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:56.235370 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.69694ms
I1204 10:23:56.236916 3136 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:56.237629 3136 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:56.239324 3136 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:56.305646 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 68.81270925ms
I1204 10:23:56.317337 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 11.5225ms
I1204 10:23:56.317406 3001 isa_program_util_common.cc:520] (HLO module jit_less): Executable fingerprint:1be1faa1886573aaa411fcb756118cc1614557491747ac1444d75e612c7a61b1
I1204 10:23:56.317409 3001 isa_program_util_common.cc:524] (HLO module jit_less): Executable fingerprint (including data segments):950b2ac2a9c9d04b9aa349e28a06cf13fe3410e7c26b0a3a49e9410e49f2503a
I1204 10:23:56.317412 3001 isa_program_util_common.cc:527] (HLO module jit_less): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:56.317675 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 94.06896ms
I1204 10:23:56.317684 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:56.356016 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.83ms
I1204 10:23:56.357582 3136 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:56.358330 3136 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:56.361061 3136 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:56.712784 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 355.30926ms
I1204 10:23:56.750925 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 37.96409075ms
I1204 10:23:56.751063 3001 isa_program_util_common.cc:520] (HLO module jit__where): Executable fingerprint:6c7bf00928a976625dfc03a6cbb6065669e9aa85ec9e0e202cff0eb2e632dee7
I1204 10:23:56.751067 3001 isa_program_util_common.cc:524] (HLO module jit__where): Executable fingerprint (including data segments):ef3e0c77d3ffb013394cd461f5193bc53bee0ee247f56d54f0935ffad80f3e9a
I1204 10:23:56.751069 3001 isa_program_util_common.cc:527] (HLO module jit__where): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:56.751358 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 408.18164925ms
I1204 10:23:56.751368 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:56.789356 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.68084925ms
I1204 10:23:56.790832 3136 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:56.791614 3136 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:56.793361 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:56.840306 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 49.56094925ms
I1204 10:23:56.851767 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 11.2935ms
I1204 10:23:56.851843 3001 isa_program_util_common.cc:520] (HLO module jit_add): Executable fingerprint:19f90b614d31eaac0736648cbe6ad6eddbc63cfb30bc6498da18a22442dd6312
I1204 10:23:56.851849 3001 isa_program_util_common.cc:524] (HLO module jit_add): Executable fingerprint (including data segments):c067307f26bc3aca8cb27a23ce5ea58a56d4f6114cb4f3d67dc23d51dee852c4
I1204 10:23:56.851851 3001 isa_program_util_common.cc:527] (HLO module jit_add): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:56.852150 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 74.6039ms
I1204 10:23:56.852160 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:56.887288 2807 deepsea_compiler_base.cc:7855] XLA::TPU running hlo passes for 10 instructions, module: jit__reduce_max
I1204 10:23:56.887332 2807 deepsea_compiler_base.cc:7910] Initial HLO module: jit__reduce_max instructions: 10 fingerprint: 461925716519770257-4035828241379739704-17078542282357815393-1242137874306953380/5
I1204 10:23:56.887609 2807 deepsea_compiler_base.cc:7982] HLO optimizing module: jit__reduce_max instructions: 10
I1204 10:23:56.887616 2807 deepsea_compiler_base.cc:7997] XLA::TPU HLO optimization
I1204 10:23:56.893498 2807 deepsea_compiler_base.cc:6963] XLA::TPU HLO PostOptimizationPipeline
I1204 10:23:56.895510 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 8.17756925ms
I1204 10:23:56.897532 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:56.898319 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:56.900940 3013 window_config_assignment_prod.cc:30] Retrieving backend configs from FDO profiles.
I1204 10:23:56.901243 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:57.218381 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 320.97034ms
I1204 10:23:57.244308 3001 deepsea_compiler_base.cc:3021] final program bundle count: 2,022 note this count does not reflect cycles spent executing delays.
I1204 10:23:57.250575 3001 deepsea_compiler_base.cc:3021] final program bundle count: 176 note this count does not reflect cycles spent executing delays.
I1204 10:23:57.251723 3001 deepsea_compiler_base.cc:3507] Program divided into 2 overlays without HLO functions (154.0K).
I1204 10:23:57.251859 3001 deepsea_compiler_base.cc:3675] XLA::TPU module name: jit__reduce_max
I1204 10:23:57.251863 3001 deepsea_compiler_base.cc:3677] XLA::TPU program HBM usage: 1.15M / 31.25G
I1204 10:23:57.251867 3001 deepsea_compiler_base.cc:3724] XLA::TPU program VMEM usage: 9.06M / 128.00M
I1204 10:23:57.251878 3001 deepsea_compiler_base.cc:3735] Total hbm usage >= 270.15M:
I1204 10:23:57.251881 3001 deepsea_compiler_base.cc:3735] reserved 260.00M
I1204 10:23:57.251883 3001 deepsea_compiler_base.cc:3735] program 1.15M
I1204 10:23:57.251884 3001 deepsea_compiler_base.cc:3735] arguments 9.00M
I1204 10:23:57.251886 3001 deepsea_compiler_base.cc:3735]
I1204 10:23:57.251888 3001 deepsea_compiler_base.cc:3735] Output size 32.0K; shares 0B with arguments.
I1204 10:23:57.251890 3001 deepsea_compiler_base.cc:3735]
I1204 10:23:57.251892 3001 deepsea_compiler_base.cc:3735] Total host usage >= 0B:
I1204 10:23:57.251894 3001 deepsea_compiler_base.cc:3735] reserved 0B
I1204 10:23:57.251896 3001 deepsea_compiler_base.cc:3735] program unknown size
I1204 10:23:57.251897 3001 deepsea_compiler_base.cc:3735] arguments 0B
I1204 10:23:57.251905 3001 deepsea_compiler_base.cc:3735]
I1204 10:23:57.251908 3001 deepsea_compiler_base.cc:3735] Output size 0B; shares 0B with arguments.
I1204 10:23:57.251910 3001 deepsea_compiler_base.cc:3735]
I1204 10:23:57.251925 3001 deepsea_compiler_base.cc:3739] Program sflag requirement 216B:
I1204 10:23:57.251927 3001 deepsea_compiler_base.cc:3739] reserved 204B
I1204 10:23:57.251929 3001 deepsea_compiler_base.cc:3739] global 4B
I1204 10:23:57.251931 3001 deepsea_compiler_base.cc:3739] scoped 8B
I1204 10:23:57.251933 3001 deepsea_compiler_base.cc:3739] Program vmem requirement 9.06M:
I1204 10:23:57.251934 3001 deepsea_compiler_base.cc:3739] scoped 9.00M
I1204 10:23:57.251936 3001 deepsea_compiler_base.cc:3739] HLO temp 64.0K (0.0% utilization: Unpadded (0B) Padded (0B), 100.0% fragmentation (64.0K))
I1204 10:23:57.251938 3001 deepsea_compiler_base.cc:3739] Program smem requirement 56B:
I1204 10:23:57.251940 3001 deepsea_compiler_base.cc:3739] global 16B
I1204 10:23:57.251942 3001 deepsea_compiler_base.cc:3739] scoped 40B
I1204 10:23:57.251944 3001 deepsea_compiler_base.cc:3739] Program hbm requirement 1.15M:
I1204 10:23:57.251955 3001 deepsea_compiler_base.cc:3739] global 1.00M
I1204 10:23:57.251957 3001 deepsea_compiler_base.cc:3739] overlays 154.0K
I1204 10:23:57.251959 3001 deepsea_compiler_base.cc:3739] Program host requirement 0B:
I1204 10:23:57.251961 3001 deepsea_compiler_base.cc:3747] XLA::TPU program SMEM usage: 1.9K / 1.00M (2 parameters)
I1204 10:23:57.251976 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 33.39018925ms
I1204 10:23:57.252106 3001 isa_program_util_common.cc:520] (HLO module jit__reduce_max): Executable fingerprint:cd039f777659cf5495b2c6f7d590a71aa93c7578bc02796a66bc8e57778990c6
I1204 10:23:57.252109 3001 isa_program_util_common.cc:524] (HLO module jit__reduce_max): Executable fingerprint (including data segments):83f09eebd7f534cac8637956a03a00b6242c406f9143c099a9935f9dee72e572
I1204 10:23:57.252111 3001 isa_program_util_common.cc:527] (HLO module jit__reduce_max): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:57.252461 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 372.43209075ms
I1204 10:23:57.252470 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:57.291366 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 6.19528925ms
I1204 10:23:57.292940 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:57.293541 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:57.295493 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:57.440457 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 147.57453075ms
I1204 10:23:57.461281 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 20.64947075ms
I1204 10:23:57.461390 3001 isa_program_util_common.cc:520] (HLO module jit_subtract): Executable fingerprint:529190172e94da259623aa4cb2dcb955f4a16aab90c951cf05ca110308f4bca7
I1204 10:23:57.461395 3001 isa_program_util_common.cc:524] (HLO module jit_subtract): Executable fingerprint (including data segments):d6d62ac3369591ea16b810bff0c9c487c582bb7902816d9c76427cbc16e0b40c
I1204 10:23:57.461397 3001 isa_program_util_common.cc:527] (HLO module jit_subtract): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:57.461730 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 183.60854ms
I1204 10:23:57.461744 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:57.500180 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.72778ms
I1204 10:23:57.501636 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:57.502212 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:57.503453 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:57.538967 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 37.38452ms
I1204 10:23:57.549202 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 10.09835075ms
I1204 10:23:57.549277 3001 isa_program_util_common.cc:520] (HLO module jit_exp): Executable fingerprint:e37cf2a555dfb704e5c665cc3030cd8ac66ef3f0dc5707fcafdc8939526763a7
I1204 10:23:57.549281 3001 isa_program_util_common.cc:524] (HLO module jit_exp): Executable fingerprint (including data segments):f039d57c4ee01f68cabd811f1ae76555a613de28929289225f5cace6774b7ea9
I1204 10:23:57.549283 3001 isa_program_util_common.cc:527] (HLO module jit_exp): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:57.549532 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 61.37330925ms
I1204 10:23:57.549544 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:57.591058 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 6.18616ms
I1204 10:23:57.592612 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:57.593217 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:57.595005 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:57.927223 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 334.66333075ms
I1204 10:23:57.960835 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 33.45051075ms
I1204 10:23:57.960982 3001 isa_program_util_common.cc:520] (HLO module jit__reduce_sum): Executable fingerprint:a3ee79f60b65bde3ef3f5723bb8539f906f6f66afc02736e0912912665bda158
I1204 10:23:57.960987 3001 isa_program_util_common.cc:524] (HLO module jit__reduce_sum): Executable fingerprint (including data segments):39dee19a25a61facfb3f61aff75e07ba8f7b47ec9cdb4630ec6f70b3ad17d10b
I1204 10:23:57.960990 3001 isa_program_util_common.cc:527] (HLO module jit__reduce_sum): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:57.961309 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 383.80880925ms
I1204 10:23:57.961320 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:58.000580 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 6.77491075ms
I1204 10:23:58.002224 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:58.002916 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:58.005062 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:58.173866 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 171.69138ms
I1204 10:23:58.194903 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 20.85254ms
I1204 10:23:58.195021 3001 isa_program_util_common.cc:520] (HLO module jit_true_divide): Executable fingerprint:2da75e491f697afe9895274d0af0cf82f6ec884f814f6eee0c2a5acac757f6bc
I1204 10:23:58.195026 3001 isa_program_util_common.cc:524] (HLO module jit_true_divide): Executable fingerprint (including data segments):8f76d74d6aec594c0914441adaf3c4a379e4f70f9b8fedbc735caf18f2f79105
I1204 10:23:58.195028 3001 isa_program_util_common.cc:527] (HLO module jit_true_divide): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:58.195454 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 208.62598ms
I1204 10:23:58.195489 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:58.237429 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 7.53512ms
I1204 10:23:58.239027 3136 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:58.239703 3136 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:58.242509 3136 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:59.224111 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 985.17025075ms
I1204 10:23:59.287787 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 63.50346ms
I1204 10:23:59.288103 3001 isa_program_util_common.cc:520] (HLO module jit__einsum): Executable fingerprint:60302d16c556d3356b714083d660442865e40ade05b1147ca9aa2d210c4bd734
I1204 10:23:59.288108 3001 isa_program_util_common.cc:524] (HLO module jit__einsum): Executable fingerprint (including data segments):d8d7bc3d3beeef1fff49c35caa858b0f0b1557bac7cfefd2d5e6861dc6170737
I1204 10:23:59.288111 3001 isa_program_util_common.cc:527] (HLO module jit__einsum): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:59.288446 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 1.06561351075s
I1204 10:23:59.288459 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:59.337326 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 6.30562ms
I1204 10:23:59.339060 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:59.339744 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:59.341622 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:59.351998 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 13.01014925ms
I1204 10:23:59.360820 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 8.70232925ms
I1204 10:23:59.360879 3001 isa_program_util_common.cc:520] (HLO module jit_gather): Executable fingerprint:07a3fb36510ce4fb8b9fc22624f641dd65ac5001819ac91aa55fa34774399800
I1204 10:23:59.360882 3001 isa_program_util_common.cc:524] (HLO module jit_gather): Executable fingerprint (including data segments):317d8982d2a6ff5c98454c687084b459c1cfb791856e0d0175ca65438f3bd04b
I1204 10:23:59.360885 3001 isa_program_util_common.cc:527] (HLO module jit_gather): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:59.361128 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 37.41766925ms
I1204 10:23:59.361137 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:59.395730 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.18222925ms
I1204 10:23:59.397179 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:59.397874 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:59.399306 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:59.408268 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 11.18188ms
I1204 10:23:59.416806 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 8.43690925ms
I1204 10:23:59.416860 3001 isa_program_util_common.cc:520] (HLO module jit_broadcast_in_dim): Executable fingerprint:2465fb80b7e81e394d71acdce5964d521e7f17722fe790d47e342a72d0ee4162
I1204 10:23:59.416863 3001 isa_program_util_common.cc:524] (HLO module jit_broadcast_in_dim): Executable fingerprint (including data segments):0d74f996d24d804ac425f2da854d60c51a95459ed55395522afa0de4027e8577
I1204 10:23:59.416866 3001 isa_program_util_common.cc:527] (HLO module jit_broadcast_in_dim): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:59.417062 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 32.7946ms
I1204 10:23:59.417071 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:59.454596 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 6.4776ms
I1204 10:23:59.456236 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:59.456961 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:59.459485 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:59.547204 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 91.06835075ms
I1204 10:23:59.560333 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 12.96156925ms
I1204 10:23:59.560404 3001 isa_program_util_common.cc:520] (HLO module jit_gather): Executable fingerprint:32b272a0fcc2194de27c40c88be24674e396ace0120217602a85cd1b5b1b60a8
I1204 10:23:59.560407 3001 isa_program_util_common.cc:524] (HLO module jit_gather): Executable fingerprint (including data segments):4cb1edfc8406d6cf99fa715022e112a8f65d2311caa1053eb3d2c3b0e9ab34e6
I1204 10:23:59.560410 3001 isa_program_util_common.cc:527] (HLO module jit_gather): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:59.560665 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 119.47768ms
I1204 10:23:59.560674 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:59.595541 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.31186ms
I1204 10:23:59.597077 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:59.597786 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:59.599332 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:59.679423 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 82.45104ms
I1204 10:23:59.691785 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 12.19306ms
I1204 10:23:59.691855 3001 isa_program_util_common.cc:520] (HLO module jit_broadcast_in_dim): Executable fingerprint:a3e4645ff1fe532606691c13ccad8df74da046b558c21cb768c3c5424dee0c92
I1204 10:23:59.691858 3001 isa_program_util_common.cc:524] (HLO module jit_broadcast_in_dim): Executable fingerprint (including data segments):17bcba144e5a90f058532549d58cd95d54de39790d4cb6b486ce8a94e64f2c62
I1204 10:23:59.691861 3001 isa_program_util_common.cc:527] (HLO module jit_broadcast_in_dim): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:59.692092 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 107.98571075ms
I1204 10:23:59.692101 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:59.729408 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.19374ms
I1204 10:23:59.731017 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:59.732012 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:59.734710 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:59.744804 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 13.94446ms
I1204 10:23:59.754110 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 9.19858ms
I1204 10:23:59.754166 3001 isa_program_util_common.cc:520] (HLO module jit_less): Executable fingerprint:35c4cec8e92e8cc8dca7839e591ad5b5ff3a2bc33a8d416713a1cbe8bfea7e09
I1204 10:23:59.754170 3001 isa_program_util_common.cc:524] (HLO module jit_less): Executable fingerprint (including data segments):b6dfb81f2124f37247c236beaed32223900393979bd055a7cfc141ffa1b764d0
I1204 10:23:59.754172 3001 isa_program_util_common.cc:527] (HLO module jit_less): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:59.754410 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 37.25484ms
I1204 10:23:59.754419 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:59.790735 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.11882ms
I1204 10:23:59.792288 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:59.793005 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:59.795425 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:59.805543 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 13.35847075ms
I1204 10:23:59.815181 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 9.5211ms
I1204 10:23:59.815237 3001 isa_program_util_common.cc:520] (HLO module jit_add): Executable fingerprint:21bda17af4c81842aba49d87f9675c4e48e9986c436edbd6526e3c621eb0c021
I1204 10:23:59.815240 3001 isa_program_util_common.cc:524] (HLO module jit_add): Executable fingerprint (including data segments):fac29855c403755254c9a8b3b4250e9ec8fe78578b05c922bc3264f0ac2fde88
I1204 10:23:59.815243 3001 isa_program_util_common.cc:527] (HLO module jit_add): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:59.815491 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 36.9578ms
I1204 10:23:59.815500 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:59.850218 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.60426ms
I1204 10:23:59.851760 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:59.852520 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:59.854547 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:59.864061 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 12.40926ms
I1204 10:23:59.873571 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 9.40543075ms
I1204 10:23:59.873628 3001 isa_program_util_common.cc:520] (HLO module jit_select_n): Executable fingerprint:1b4af806359ca11a9aeedcab09cec4115d359228b512823f789d069d0a412454
I1204 10:23:59.873632 3001 isa_program_util_common.cc:524] (HLO module jit_select_n): Executable fingerprint (including data segments):101c5260488115e2b7ec811bcd3a241daf5827b1cc7bf76ea72a4ccf9fc17426
I1204 10:23:59.873634 3001 isa_program_util_common.cc:527] (HLO module jit_select_n): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:59.873838 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 35.16476925ms
I1204 10:23:59.873846 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:59.906116 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.23516925ms
I1204 10:23:59.907628 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:59.908377 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:59.909902 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:59.917425 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 9.90926ms
I1204 10:23:59.926178 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 8.64475075ms
I1204 10:23:59.926229 3001 isa_program_util_common.cc:520] (HLO module jit__broadcast_arrays): Executable fingerprint:2465fb80b7e81e394d71acdce5964d521e7f17722fe790d47e342a72d0ee4162
I1204 10:23:59.926233 3001 isa_program_util_common.cc:524] (HLO module jit__broadcast_arrays): Executable fingerprint (including data segments):0d74f996d24d804ac425f2da854d60c51a95459ed55395522afa0de4027e8577
I1204 10:23:59.926235 3001 isa_program_util_common.cc:527] (HLO module jit__broadcast_arrays): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:59.926438 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 31.44416925ms
I1204 10:23:59.926446 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:23:59.959701 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.59035075ms
I1204 10:23:59.961310 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:23:59.962041 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:23:59.963900 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:23:59.972203 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 10.95384ms
I1204 10:23:59.980758 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 8.40164925ms
I1204 10:23:59.980840 3001 isa_program_util_common.cc:520] (HLO module jit_broadcast_in_dim): Executable fingerprint:f8a8a1bcc14437ec76793f8d974adac664e3c5f5e676f34082c65624e495eac5
I1204 10:23:59.980846 3001 isa_program_util_common.cc:524] (HLO module jit_broadcast_in_dim): Executable fingerprint (including data segments):b7175bd87ce618f7322eb6a7c85b4a92b395ad6e0edb5ac68d5ee005fd07f01b
I1204 10:23:59.980848 3001 isa_program_util_common.cc:527] (HLO module jit_broadcast_in_dim): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:23:59.981122 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 32.99846ms
I1204 10:23:59.981131 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:00.019941 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 8.45202925ms
I1204 10:24:00.021605 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:00.022272 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:00.025034 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:00.064236 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 42.67947075ms
I1204 10:24:00.077321 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 12.91845075ms
I1204 10:24:00.077413 3001 isa_program_util_common.cc:520] (HLO module jit_gather): Executable fingerprint:443dd6503799b267bef6d190e199a04990fe21fd9443645e0e1129b53129b472
I1204 10:24:00.077417 3001 isa_program_util_common.cc:524] (HLO module jit_gather): Executable fingerprint (including data segments):3b730a9ed4346fa60f071f8b3c7d4f81417d41888f1e7961bd8e1a22a99dbeea
I1204 10:24:00.077419 3001 isa_program_util_common.cc:527] (HLO module jit_gather): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:00.077726 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 73.05725075ms
I1204 10:24:00.077737 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:00.115766 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.46402925ms
I1204 10:24:00.117286 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:00.117982 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:00.119439 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
W1204 10:24:00.319414 2826 internals-datagram-socket.cc:411] udp sendto fd=198, dest=172.20.0.120:9530: Network is unreachable
I1204 10:24:00.351027 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 233.80234ms
I1204 10:24:00.384042 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 32.83286ms
I1204 10:24:00.384198 3001 isa_program_util_common.cc:520] (HLO module jit_broadcast_in_dim): Executable fingerprint:3ffd29ee673b480b5a459cfb0db1555f899f1276f8e1d2865205d0aaa5824642
I1204 10:24:00.384201 3001 isa_program_util_common.cc:524] (HLO module jit_broadcast_in_dim): Executable fingerprint (including data segments):dfd022fe3c05eb420b0f105b2c4c97222844336874395e8763413c3b61689b02
I1204 10:24:00.384203 3001 isa_program_util_common.cc:527] (HLO module jit_broadcast_in_dim): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:00.384438 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 280.45977075ms
I1204 10:24:00.384454 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:00.420518 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.65799075ms
I1204 10:24:00.421949 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:00.422532 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:00.423825 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:00.644633 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 222.74556925ms
I1204 10:24:00.678020 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 33.21722ms
I1204 10:24:00.678157 3001 isa_program_util_common.cc:520] (HLO module jit_reshape): Executable fingerprint:11969447be7e1e557978b02b38e74a5eeab6b8d9506c2b616319aa7fd2a4053b
I1204 10:24:00.678161 3001 isa_program_util_common.cc:524] (HLO module jit_reshape): Executable fingerprint (including data segments):2856eac5ab542f37950fcd501545a8e2f6706752abe81d42cd2f34d2e6474a1d
I1204 10:24:00.678163 3001 isa_program_util_common.cc:527] (HLO module jit_reshape): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:00.678396 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 269.68982925ms
I1204 10:24:00.678409 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:00.717184 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 6.5619ms
I1204 10:24:00.718801 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:00.719500 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:00.722009 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:00.893176 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 174.48238925ms
I1204 10:24:00.918676 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 25.339ms
I1204 10:24:00.918789 3001 isa_program_util_common.cc:520] (HLO module jit_gather): Executable fingerprint:c85f0fb6e94b1fd91871ecf31e0892a10832990392cd8c1f1c3348e9338a1f99
I1204 10:24:00.918793 3001 isa_program_util_common.cc:524] (HLO module jit_gather): Executable fingerprint (including data segments):ffe8c5e2233a2acc6fd859df87f07e284e4db7ebe3f0162b9dad3137d8c609e7
I1204 10:24:00.918795 3001 isa_program_util_common.cc:527] (HLO module jit_gather): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:00.919057 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 215.49398ms
I1204 10:24:00.919065 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:00.953642 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.4234ms
I1204 10:24:00.955163 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:00.955862 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:00.957389 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:01.104538 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 149.47746ms
I1204 10:24:01.129481 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 24.76705075ms
I1204 10:24:01.129592 3001 isa_program_util_common.cc:520] (HLO module jit_broadcast_in_dim): Executable fingerprint:b398f52911d407d48bffd96eef8fc29780c5b112ac21d24a22edd4f028b92e98
I1204 10:24:01.129596 3001 isa_program_util_common.cc:524] (HLO module jit_broadcast_in_dim): Executable fingerprint (including data segments):7edf1e2ffee6b8d76ca1c14334355477ccbf87232303bf0581e5b1882a1abc38
I1204 10:24:01.129598 3001 isa_program_util_common.cc:527] (HLO module jit_broadcast_in_dim): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:01.129832 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 187.72224ms
I1204 10:24:01.129844 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:01.165221 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.26378ms
I1204 10:24:01.166744 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:01.167498 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:01.169147 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:01.323205 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 156.58408ms
I1204 10:24:01.348710 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 25.31442ms
I1204 10:24:01.348822 3001 isa_program_util_common.cc:520] (HLO module jit__squeeze): Executable fingerprint:b398f52911d407d48bffd96eef8fc29780c5b112ac21d24a22edd4f028b92e98
I1204 10:24:01.348825 3001 isa_program_util_common.cc:524] (HLO module jit__squeeze): Executable fingerprint (including data segments):7edf1e2ffee6b8d76ca1c14334355477ccbf87232303bf0581e5b1882a1abc38
I1204 10:24:01.348827 3001 isa_program_util_common.cc:527] (HLO module jit__squeeze): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:01.349050 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 195.03711075ms
I1204 10:24:01.349059 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:01.392509 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 12.46902925ms
I1204 10:24:01.394160 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:01.394899 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:01.399696 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:01.777699 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 383.62445075ms
I1204 10:24:01.819815 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 41.93840925ms
I1204 10:24:01.820050 3001 isa_program_util_common.cc:520] (HLO module jit_scatter): Executable fingerprint:82c6a1bbfe8c0a1e32661792fdc87dce1f739c99577e237ac32dbbeca0c294c6
I1204 10:24:01.820053 3001 isa_program_util_common.cc:524] (HLO module jit_scatter): Executable fingerprint (including data segments):26349a590ebb7e6963391905fa1f016cb940903521968be72270a9453ac9ff2a
I1204 10:24:01.820056 3001 isa_program_util_common.cc:527] (HLO module jit_scatter): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:01.820341 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 447.36198ms
I1204 10:24:01.820353 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:01.859731 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.395ms
I1204 10:24:01.861164 3012 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:01.861863 3012 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:01.863454 3012 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:02.082533 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 221.43041075ms
I1204 10:24:02.116046 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 33.32163075ms
I1204 10:24:02.116183 3001 isa_program_util_common.cc:520] (HLO module jit__squeeze): Executable fingerprint:3ffd29ee673b480b5a459cfb0db1555f899f1276f8e1d2865205d0aaa5824642
I1204 10:24:02.116187 3001 isa_program_util_common.cc:524] (HLO module jit__squeeze): Executable fingerprint (including data segments):dfd022fe3c05eb420b0f105b2c4c97222844336874395e8763413c3b61689b02
I1204 10:24:02.116189 3001 isa_program_util_common.cc:527] (HLO module jit__squeeze): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:02.116403 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 268.20673075ms
I1204 10:24:02.116412 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:02.155924 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 8.56008ms
I1204 10:24:02.157467 3012 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:02.158199 3012 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:02.161525 3012 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:02.340431 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 183.04152ms
I1204 10:24:02.360078 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 19.47812ms
I1204 10:24:02.360207 3001 isa_program_util_common.cc:520] (HLO module jit_scatter): Executable fingerprint:f75dcc408344b845803b37fda5d2fd3c2139aad29b4cbd2fabb51dffe78e170e
I1204 10:24:02.360211 3001 isa_program_util_common.cc:524] (HLO module jit_scatter): Executable fingerprint (including data segments):7123756afe393d3e562324f9e7468d2c03b03dbeb9ae0e50086766c3b7808a68
I1204 10:24:02.360214 3001 isa_program_util_common.cc:527] (HLO module jit_scatter): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:02.360511 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 220.18418ms
I1204 10:24:02.360520 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:02.397891 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.71988ms
I1204 10:24:02.399368 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:02.399952 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:02.401222 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:02.436802 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 37.51506925ms
I1204 10:24:02.447800 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 10.87588ms
I1204 10:24:02.447872 3001 isa_program_util_common.cc:520] (HLO module jit_reshape): Executable fingerprint:2268ddbb7782d856a27890c8ccbc38699d4f2e6dea81b80eed0ab73b5b919df1
I1204 10:24:02.447877 3001 isa_program_util_common.cc:524] (HLO module jit_reshape): Executable fingerprint (including data segments):9082848ab31db2baf10f26d6c96314d4560a5081177931a6c04691a156495def
I1204 10:24:02.447879 3001 isa_program_util_common.cc:527] (HLO module jit_reshape): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:02.448116 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 61.96564ms
I1204 10:24:02.448124 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:02.483298 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.03888ms
I1204 10:24:02.484831 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:02.485419 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:02.486660 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:02.567964 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 83.21012ms
I1204 10:24:02.585787 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 17.62198ms
I1204 10:24:02.585878 3001 isa_program_util_common.cc:520] (HLO module jit_reshape): Executable fingerprint:92ac86b47b2fcae360a3336c583e3a48c426372fe3437f91e25209bdcd4405e2
I1204 10:24:02.585882 3001 isa_program_util_common.cc:524] (HLO module jit_reshape): Executable fingerprint (including data segments):2cb8c1156efae1ea426743dba1391a30b1660deb70045e775f152545073965c2
I1204 10:24:02.585884 3001 isa_program_util_common.cc:527] (HLO module jit_reshape): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:02.586157 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 114.85102ms
I1204 10:24:02.586165 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:02.625496 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 7.17264925ms
I1204 10:24:02.627139 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:02.627825 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:02.630477 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:02.692881 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 65.8177ms
I1204 10:24:02.706511 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 13.45198ms
I1204 10:24:02.706586 3001 isa_program_util_common.cc:520] (HLO module jit_gather): Executable fingerprint:b1875327c1ac4d4fa3cdbf59767bf5aa56778d1af6d0fb6aa878f1e955d17ed1
I1204 10:24:02.706590 3001 isa_program_util_common.cc:524] (HLO module jit_gather): Executable fingerprint (including data segments):4da0609522454d5684e28889d53c77a8ec51771aa7af7e32082fc6da25e855cb
I1204 10:24:02.706592 3001 isa_program_util_common.cc:527] (HLO module jit_gather): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:02.706873 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 95.73832ms
I1204 10:24:02.706882 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:02.741499 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.27594925ms
I1204 10:24:02.742964 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:02.743666 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:02.745088 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:02.764946 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 22.04733075ms
I1204 10:24:02.774861 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 9.79457075ms
I1204 10:24:02.774923 3001 isa_program_util_common.cc:520] (HLO module jit_broadcast_in_dim): Executable fingerprint:04e159b900d905cf57250dc252c55623c2153e198fc96181d1f5f56962372c42
I1204 10:24:02.774926 3001 isa_program_util_common.cc:524] (HLO module jit_broadcast_in_dim): Executable fingerprint (including data segments):8fd8bc7eb28391bd30879308a0b7da01aa9adc7e915ae56ab855da50d63b905b
I1204 10:24:02.774928 3001 isa_program_util_common.cc:527] (HLO module jit_broadcast_in_dim): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:02.775141 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 44.90704925ms
I1204 10:24:02.775149 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:02.811587 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.42210925ms
I1204 10:24:02.813042 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:02.813726 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:02.815507 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:02.928266 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 115.29985075ms
I1204 10:24:02.948550 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 20.11444925ms
I1204 10:24:02.948645 3001 isa_program_util_common.cc:520] (HLO module jit_broadcast_in_dim): Executable fingerprint:5005132ff0b93eae5d7a6844642ae007cc9976ba254d49e929f5714f6e95a60d
I1204 10:24:02.948648 3001 isa_program_util_common.cc:524] (HLO module jit_broadcast_in_dim): Executable fingerprint (including data segments):2b443f6fc2f862eb674cbcb6302241e64d6a9345b68fb06fee65e31fc82a0bff
I1204 10:24:02.948651 3001 isa_program_util_common.cc:527] (HLO module jit_broadcast_in_dim): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:02.948932 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 148.68614ms
I1204 10:24:02.948943 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:02.985307 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.76306ms
I1204 10:24:02.986807 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:02.987468 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:02.989079 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:03.120296 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 133.57971075ms
I1204 10:24:03.134824 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 14.35599075ms
I1204 10:24:03.134906 3001 isa_program_util_common.cc:520] (HLO module jit_reshape): Executable fingerprint:17b39b4a9831f795478174de5bee31448cef76eca8a21fca7544682afea3e256
I1204 10:24:03.134912 3001 isa_program_util_common.cc:524] (HLO module jit_reshape): Executable fingerprint (including data segments):35f94c5a4def20c334ec3ab0e5efbc40d94f9eec6b497e029e1e8115e6633e5c
I1204 10:24:03.134914 3001 isa_program_util_common.cc:527] (HLO module jit_reshape): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:03.135165 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 161.72586ms
I1204 10:24:03.135174 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:03.173808 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 6.40869075ms
I1204 10:24:03.175347 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:03.176061 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:03.179092 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:03.946177 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 770.93186ms
I1204 10:24:04.039609 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 93.23708ms
I1204 10:24:04.039914 3001 isa_program_util_common.cc:520] (HLO module jit__einsum): Executable fingerprint:bff435d83b6b6987c828443f89e76e9d3b287a8b2046c0f3d4076b46c6634dd6
I1204 10:24:04.039918 3001 isa_program_util_common.cc:524] (HLO module jit__einsum): Executable fingerprint (including data segments):302c70a501fb6db130313e798355269c6fc42168bcfbd05618218ae0e950d915
I1204 10:24:04.039920 3001 isa_program_util_common.cc:527] (HLO module jit__einsum): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:04.040240 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 880.00174ms
I1204 10:24:04.040249 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:04.079365 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.37432ms
I1204 10:24:04.080896 3136 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:04.081603 3136 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:04.083764 3136 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:04.207293 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 126.49548925ms
I1204 10:24:04.224306 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 16.82234ms
I1204 10:24:04.224389 3001 isa_program_util_common.cc:520] (HLO module jit_multiply): Executable fingerprint:eb0c5ce7a113209c59f7b4b600989b804d1bd067b2a3b122dfa25707a729436d
I1204 10:24:04.224393 3001 isa_program_util_common.cc:524] (HLO module jit_multiply): Executable fingerprint (including data segments):97c3be51fc086209ae89577232ec3364be9d637f19224878dba74de3b42be90f
I1204 10:24:04.224396 3001 isa_program_util_common.cc:527] (HLO module jit_multiply): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:04.224659 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 157.72244ms
I1204 10:24:04.224668 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:04.259884 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.23596ms
I1204 10:24:04.261491 3136 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:04.262209 3136 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:04.263517 3136 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:04.303588 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 42.19332ms
I1204 10:24:04.315323 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 11.59682ms
I1204 10:24:04.315393 3001 isa_program_util_common.cc:520] (HLO module jit_iota): Executable fingerprint:ad735c7d46bffb99043359754a13c2b2d97abe8ea1edaa31c5ad240d578d20b5
I1204 10:24:04.315396 3001 isa_program_util_common.cc:524] (HLO module jit_iota): Executable fingerprint (including data segments):e94a94331dc43318c22eadbd3aa15ff8ce3e98a648b941b0a16d7b25aff7edbd
I1204 10:24:04.315399 3001 isa_program_util_common.cc:527] (HLO module jit_iota): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:04.315648 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 67.15302ms
I1204 10:24:04.315660 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:04.351369 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.30736925ms
I1204 10:24:04.352910 3136 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:04.353635 3136 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:04.355701 3136 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:04.482992 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 130.19576ms
I1204 10:24:04.500202 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 17.00955075ms
I1204 10:24:04.500288 3001 isa_program_util_common.cc:520] (HLO module jit_add): Executable fingerprint:46c55c63f19b863cac50a0b334694984eb0dd72bb30dab11b56c95319633f38c
I1204 10:24:04.500293 3001 isa_program_util_common.cc:524] (HLO module jit_add): Executable fingerprint (including data segments):0b232ef122a4cd06216a909bae98016ae9871c1717ec4b2b72bdc71f1d1c219d
I1204 10:24:04.500295 3001 isa_program_util_common.cc:527] (HLO module jit_add): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:04.500596 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 161.54145075ms
I1204 10:24:04.500612 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:04.535869 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.18928ms
I1204 10:24:04.537486 3136 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:04.538206 3136 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:04.539563 3136 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:04.577626 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 40.23583075ms
I1204 10:24:04.590218 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 12.44649075ms
I1204 10:24:04.590291 3001 isa_program_util_common.cc:520] (HLO module jit_iota): Executable fingerprint:4b6035321a0e4f0edd29ba4f6e3665096e235ba52d4625ab73e7af36a3da5968
I1204 10:24:04.590300 3001 isa_program_util_common.cc:524] (HLO module jit_iota): Executable fingerprint (including data segments):f6d5b666555937ee25a4bb349b127a42b859327a7a383273fe126e7a191066de
I1204 10:24:04.590303 3001 isa_program_util_common.cc:527] (HLO module jit_iota): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:04.590566 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 65.892ms
I1204 10:24:04.590574 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:04.626290 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.65444ms
I1204 10:24:04.627865 3136 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:04.628577 3136 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:04.630233 3136 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:04.852198 2819 python_stack_size.cc:37] EventManager thread stack size of 131072 might be too small for Python callbacks. Crash? Read go/py-cpp-stack.
I1204 10:24:04.908164 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 280.41212ms
I1204 10:24:04.934162 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 25.80267075ms
I1204 10:24:04.934275 3001 isa_program_util_common.cc:520] (HLO module jit_less): Executable fingerprint:18b9b16f7d0618215a8af91a691c729931cc62a2db34a0a7a181da6782812002
I1204 10:24:04.934279 3001 isa_program_util_common.cc:524] (HLO module jit_less): Executable fingerprint (including data segments):1200ab69823e2c87a1d1b6a22360250d16177bec2dd2258543d70e468ecade0a
I1204 10:24:04.934281 3001 isa_program_util_common.cc:527] (HLO module jit_less): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:04.934634 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 320.0344ms
I1204 10:24:04.934644 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:04.972723 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.85964ms
I1204 10:24:04.974282 3136 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:04.974995 3136 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:04.977625 3136 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:05.080968 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 106.81174ms
I1204 10:24:05.099955 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 18.80644ms
I1204 10:24:05.100041 3001 isa_program_util_common.cc:520] (HLO module jit__where): Executable fingerprint:8906d2c8d4006465079bc03dc3a748cd3ac7560be7b86c0398f02b1752c58dcb
I1204 10:24:05.100044 3001 isa_program_util_common.cc:524] (HLO module jit__where): Executable fingerprint (including data segments):e1032bad0d95a39b042793097569a4dbe51e59f7e2411fcd78ff1cbb0d4ddac6
I1204 10:24:05.100046 3001 isa_program_util_common.cc:527] (HLO module jit__where): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:05.100330 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 140.54422925ms
I1204 10:24:05.100339 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:05.137632 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.63264ms
I1204 10:24:05.139153 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:05.139873 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:05.141530 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
W1204 10:24:05.275122 3245 internals-datagram-socket.cc:411] udp sendto fd=196, dest=172.20.0.120:9530: Network is unreachable
I1204 10:24:05.384591 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 245.53262925ms
I1204 10:24:05.404113 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 19.3566ms
I1204 10:24:05.404221 3001 isa_program_util_common.cc:520] (HLO module jit_add): Executable fingerprint:2616b834cd48a2b620f6901073a2a086b617effda25687c3d4c9a7ca886b4dca
I1204 10:24:05.404224 3001 isa_program_util_common.cc:524] (HLO module jit_add): Executable fingerprint (including data segments):a99ee4ad1bacfe15197ebbf878883a4ec9a4553126625b59012d97e3ba26bc05
I1204 10:24:05.404227 3001 isa_program_util_common.cc:527] (HLO module jit_add): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:05.404462 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 278.54171075ms
I1204 10:24:05.404474 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:05.439033 2807 deepsea_compiler_base.cc:7855] XLA::TPU running hlo passes for 10 instructions, module: jit__reduce_max
I1204 10:24:05.439075 2807 deepsea_compiler_base.cc:7910] Initial HLO module: jit__reduce_max instructions: 10 fingerprint: 14884049684499439120-4035828241379739704-17078542282357815393-1242137874306953380/5
I1204 10:24:05.439338 2807 deepsea_compiler_base.cc:7982] HLO optimizing module: jit__reduce_max instructions: 10
I1204 10:24:05.439344 2807 deepsea_compiler_base.cc:7997] XLA::TPU HLO optimization
I1204 10:24:05.445019 2807 deepsea_compiler_base.cc:6963] XLA::TPU HLO PostOptimizationPipeline
I1204 10:24:05.446984 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 7.90855075ms
I1204 10:24:05.448900 3136 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:05.449625 3136 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:05.452107 3136 window_config_assignment_prod.cc:30] Retrieving backend configs from FDO profiles.
I1204 10:24:05.452390 3136 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:05.514141 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 65.3618ms
I1204 10:24:05.521425 3001 deepsea_compiler_base.cc:3021] final program bundle count: 598 note this count does not reflect cycles spent executing delays.
I1204 10:24:05.527689 3001 deepsea_compiler_base.cc:3021] final program bundle count: 176 note this count does not reflect cycles spent executing delays.
I1204 10:24:05.528917 3001 deepsea_compiler_base.cc:3507] Program divided into 2 overlays without HLO functions (65.0K).
I1204 10:24:05.529063 3001 deepsea_compiler_base.cc:3675] XLA::TPU module name: jit__reduce_max
I1204 10:24:05.529066 3001 deepsea_compiler_base.cc:3677] XLA::TPU program HBM usage: 1.06M / 31.25G
I1204 10:24:05.529070 3001 deepsea_compiler_base.cc:3724] XLA::TPU program VMEM usage: 2.83M / 128.00M
I1204 10:24:05.529081 3001 deepsea_compiler_base.cc:3735] Total hbm usage >= 263.88M:
I1204 10:24:05.529084 3001 deepsea_compiler_base.cc:3735] reserved 260.00M
I1204 10:24:05.529091 3001 deepsea_compiler_base.cc:3735] program 1.06M
I1204 10:24:05.529093 3001 deepsea_compiler_base.cc:3735] arguments 2.81M
I1204 10:24:05.529094 3001 deepsea_compiler_base.cc:3735]
I1204 10:24:05.529096 3001 deepsea_compiler_base.cc:3735] Output size 16.0K; shares 0B with arguments.
I1204 10:24:05.529098 3001 deepsea_compiler_base.cc:3735]
I1204 10:24:05.529099 3001 deepsea_compiler_base.cc:3735] Total host usage >= 0B:
I1204 10:24:05.529101 3001 deepsea_compiler_base.cc:3735] reserved 0B
I1204 10:24:05.529103 3001 deepsea_compiler_base.cc:3735] program unknown size
I1204 10:24:05.529104 3001 deepsea_compiler_base.cc:3735] arguments 0B
I1204 10:24:05.529106 3001 deepsea_compiler_base.cc:3735]
I1204 10:24:05.529108 3001 deepsea_compiler_base.cc:3735] Output size 0B; shares 0B with arguments.
I1204 10:24:05.529109 3001 deepsea_compiler_base.cc:3735]
I1204 10:24:05.529121 3001 deepsea_compiler_base.cc:3739] Program sflag requirement 216B:
I1204 10:24:05.529123 3001 deepsea_compiler_base.cc:3739] reserved 204B
I1204 10:24:05.529124 3001 deepsea_compiler_base.cc:3739] global 4B
I1204 10:24:05.529126 3001 deepsea_compiler_base.cc:3739] scoped 8B
I1204 10:24:05.529127 3001 deepsea_compiler_base.cc:3739] Program vmem requirement 2.83M:
I1204 10:24:05.529129 3001 deepsea_compiler_base.cc:3739] scoped 2.81M
I1204 10:24:05.529131 3001 deepsea_compiler_base.cc:3739] HLO temp 16.0K (0.0% utilization: Unpadded (0B) Padded (0B), 100.0% fragmentation (16.0K))
I1204 10:24:05.529132 3001 deepsea_compiler_base.cc:3739] Program smem requirement 56B:
I1204 10:24:05.529134 3001 deepsea_compiler_base.cc:3739] global 16B
I1204 10:24:05.529136 3001 deepsea_compiler_base.cc:3739] scoped 40B
I1204 10:24:05.529137 3001 deepsea_compiler_base.cc:3739] Program hbm requirement 1.06M:
I1204 10:24:05.529139 3001 deepsea_compiler_base.cc:3739] global 1.00M
I1204 10:24:05.529140 3001 deepsea_compiler_base.cc:3739] overlays 65.0K
I1204 10:24:05.529142 3001 deepsea_compiler_base.cc:3739] Program host requirement 0B:
I1204 10:24:05.529144 3001 deepsea_compiler_base.cc:3747] XLA::TPU program SMEM usage: 1.9K / 1.00M (2 parameters)
I1204 10:24:05.529157 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 14.8509ms
I1204 10:24:05.529222 3001 isa_program_util_common.cc:520] (HLO module jit__reduce_max): Executable fingerprint:ea083872d9cb42279bd537b98b947e0ba1603a4182fbc86eb98f100cc5e478db
I1204 10:24:05.529229 3001 isa_program_util_common.cc:524] (HLO module jit__reduce_max): Executable fingerprint (including data segments):f9268eb478e44f48effa49c86e52b921ad289960a7658aac2166edd0dc98dd4f
I1204 10:24:05.529231 3001 isa_program_util_common.cc:527] (HLO module jit__reduce_max): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:05.529631 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 97.76624ms
I1204 10:24:05.529642 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:05.568131 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.71492ms
I1204 10:24:05.569773 3136 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:05.570445 3136 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:05.572511 3136 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:05.701637 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 131.9556ms
I1204 10:24:05.718955 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 17.12167075ms
I1204 10:24:05.719044 3001 isa_program_util_common.cc:520] (HLO module jit_subtract): Executable fingerprint:f47be808df76d9c089bbf6b6eb288de9d062d545915d3245280b9a916d7a4cda
I1204 10:24:05.719047 3001 isa_program_util_common.cc:524] (HLO module jit_subtract): Executable fingerprint (including data segments):12de80aec8b01abdd51cefd59490a13871e2f658ef8a16ceb47829b2447e5f79
I1204 10:24:05.719050 3001 isa_program_util_common.cc:527] (HLO module jit_subtract): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:05.719376 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 163.99488ms
I1204 10:24:05.719388 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:05.755368 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.46464ms
I1204 10:24:05.756952 3136 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:05.757646 3136 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:05.759194 3136 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:05.871799 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 114.969ms
I1204 10:24:05.890970 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 18.97419075ms
I1204 10:24:05.891059 3001 isa_program_util_common.cc:520] (HLO module jit_exp): Executable fingerprint:c10540b76b879334e9874ab443ecaeaaa8deaacc66235e2a3094feb342ff30aa
I1204 10:24:05.891063 3001 isa_program_util_common.cc:524] (HLO module jit_exp): Executable fingerprint (including data segments):9bb5ee7a0866cbddf3dca5b5cf5415be6398f0ffc1b182583f5c6d208b1b8ac2
I1204 10:24:05.891065 3001 isa_program_util_common.cc:527] (HLO module jit_exp): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:05.891353 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 147.52382ms
I1204 10:24:05.891366 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:05.930659 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 6.03100925ms
I1204 10:24:05.932244 3136 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:05.933028 3136 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:05.935386 3136 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:06.000713 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 68.59271075ms
I1204 10:24:06.015315 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 14.43201075ms
I1204 10:24:06.015390 3001 isa_program_util_common.cc:520] (HLO module jit__reduce_sum): Executable fingerprint:3b28f0e40fc4216dae7afe65d23fbc0817644d0eaad67421b3d2a08f08df76cb
I1204 10:24:06.015394 3001 isa_program_util_common.cc:524] (HLO module jit__reduce_sum): Executable fingerprint (including data segments):36aa007706910d78d23b32e22eee3791f27b7ad1be27d3897cf7b22915432c16
I1204 10:24:06.015396 3001 isa_program_util_common.cc:527] (HLO module jit__reduce_sum): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:06.015736 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 98.28018ms
I1204 10:24:06.015748 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:06.053560 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.67068ms
I1204 10:24:06.055266 3136 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:06.056011 3136 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:06.058029 3136 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:06.192903 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 137.79488ms
I1204 10:24:06.210273 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 17.1654ms
I1204 10:24:06.210358 3001 isa_program_util_common.cc:520] (HLO module jit_true_divide): Executable fingerprint:37421029aeed54aabdcb1961a585482afe156a957a955ded2208dc6f62fa6799
I1204 10:24:06.210364 3001 isa_program_util_common.cc:524] (HLO module jit_true_divide): Executable fingerprint (including data segments):5f38a5897cfc416bd145940da1991ae606583e6cc3fe46fd59f4d6f867920562
I1204 10:24:06.210373 3001 isa_program_util_common.cc:527] (HLO module jit_true_divide): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:06.210752 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 169.97266ms
I1204 10:24:06.210765 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:06.251390 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 7.04512ms
I1204 10:24:06.253006 3136 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:06.253734 3136 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:06.256902 3136 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:07.201806 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 948.90642ms
I1204 10:24:07.311913 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 109.90156925ms
I1204 10:24:07.312247 3001 isa_program_util_common.cc:520] (HLO module jit__einsum): Executable fingerprint:5362adb56a23543656b3579f0dd647e96f2772ffd01c8279e047003c2dd723cf
I1204 10:24:07.312252 3001 isa_program_util_common.cc:524] (HLO module jit__einsum): Executable fingerprint (including data segments):8395bcf4e1435676576eda488ec53030165f5901ca586519f782d4e416764129
I1204 10:24:07.312254 3001 isa_program_util_common.cc:527] (HLO module jit__einsum): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:07.312632 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 1.07532512925s
I1204 10:24:07.312646 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:07.366382 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 6.47862ms
I1204 10:24:07.368224 3011 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:07.368968 3011 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:07.371126 3011 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:07.381652 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 13.55875075ms
I1204 10:24:07.391473 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 9.67946925ms
I1204 10:24:07.391540 3001 isa_program_util_common.cc:520] (HLO module jit_gather): Executable fingerprint:ffaf6bcec561103e35e770454d9a3f7aab2d6d41564f672146112525e3805f32
I1204 10:24:07.391545 3001 isa_program_util_common.cc:524] (HLO module jit_gather): Executable fingerprint (including data segments):9bd07f1c9612a2e45b53678317a79257e4b045c9b83387b576f8ad9f59ee274b
I1204 10:24:07.391547 3001 isa_program_util_common.cc:527] (HLO module jit_gather): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:07.391807 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 39.90644ms
I1204 10:24:07.391816 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:07.426159 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.19779075ms
I1204 10:24:07.427680 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:07.428387 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:07.429875 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:07.438343 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 10.76202925ms
I1204 10:24:07.446795 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 8.34764ms
I1204 10:24:07.446849 3001 isa_program_util_common.cc:520] (HLO module jit_broadcast_in_dim): Executable fingerprint:2465fb80b7e81e394d71acdce5964d521e7f17722fe790d47e342a72d0ee4162
I1204 10:24:07.446853 3001 isa_program_util_common.cc:524] (HLO module jit_broadcast_in_dim): Executable fingerprint (including data segments):0d74f996d24d804ac425f2da854d60c51a95459ed55395522afa0de4027e8577
I1204 10:24:07.446855 3001 isa_program_util_common.cc:527] (HLO module jit_broadcast_in_dim): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:07.447057 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 32.39644ms
I1204 10:24:07.447065 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:07.488313 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 6.97860925ms
I1204 10:24:07.490126 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:07.490782 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:07.493214 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:07.535716 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 45.69610925ms
I1204 10:24:07.547095 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 11.21089075ms
I1204 10:24:07.547158 3001 isa_program_util_common.cc:520] (HLO module jit_gather): Executable fingerprint:cdc1675fe08b175714484fb634aab970320320bac7f30d5ab8977d97b91a65b3
I1204 10:24:07.547162 3001 isa_program_util_common.cc:524] (HLO module jit_gather): Executable fingerprint (including data segments):b3d07cd7463d575d35e4ca2e67453b897031bea11b52c7ecd027a72e86ae2c7e
I1204 10:24:07.547164 3001 isa_program_util_common.cc:527] (HLO module jit_gather): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:07.547544 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 73.85096ms
I1204 10:24:07.547578 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:07.586639 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.76406925ms
I1204 10:24:07.588332 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:07.588989 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:07.590495 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:07.628331 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 40.10244ms
I1204 10:24:07.638571 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 10.08502925ms
I1204 10:24:07.638632 3001 isa_program_util_common.cc:520] (HLO module jit_broadcast_in_dim): Executable fingerprint:d86e920541c86d97c6b6e2ae0f2817f1c6fdf5dbadf6d0458a2fd90952466016
I1204 10:24:07.638636 3001 isa_program_util_common.cc:524] (HLO module jit_broadcast_in_dim): Executable fingerprint (including data segments):f802f65d0c78051735a80a2ab86699cffb9e1a90a6709dfb5229e17ad8f8c321
I1204 10:24:07.638638 3001 isa_program_util_common.cc:527] (HLO module jit_broadcast_in_dim): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:07.638909 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 65.38983075ms
I1204 10:24:07.638939 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:07.678873 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.72157075ms
I1204 10:24:07.680564 3011 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:07.681293 3011 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:07.683649 3011 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:07.693597 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 13.12699075ms
I1204 10:24:07.702883 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 9.15474ms
I1204 10:24:07.702938 3001 isa_program_util_common.cc:520] (HLO module jit_less): Executable fingerprint:35c4cec8e92e8cc8dca7839e591ad5b5ff3a2bc33a8d416713a1cbe8bfea7e09
I1204 10:24:07.702941 3001 isa_program_util_common.cc:524] (HLO module jit_less): Executable fingerprint (including data segments):b6dfb81f2124f37247c236beaed32223900393979bd055a7cfc141ffa1b764d0
I1204 10:24:07.702943 3001 isa_program_util_common.cc:527] (HLO module jit_less): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:07.703271 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 37.57942ms
I1204 10:24:07.703302 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:07.742782 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.6337ms
I1204 10:24:07.744478 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:07.745146 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:07.747344 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:07.756413 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 12.04026ms
I1204 10:24:07.765929 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 9.38566ms
I1204 10:24:07.765986 3001 isa_program_util_common.cc:520] (HLO module jit_add): Executable fingerprint:21bda17af4c81842aba49d87f9675c4e48e9986c436edbd6526e3c621eb0c021
I1204 10:24:07.765995 3001 isa_program_util_common.cc:524] (HLO module jit_add): Executable fingerprint (including data segments):fac29855c403755254c9a8b3b4250e9ec8fe78578b05c922bc3264f0ac2fde88
I1204 10:24:07.765998 3001 isa_program_util_common.cc:527] (HLO module jit_add): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:07.766253 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 36.48125075ms
I1204 10:24:07.766263 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:07.804306 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.06081075ms
I1204 10:24:07.805975 3013 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:07.806638 3013 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:07.808536 3013 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:07.816847 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 10.97382925ms
I1204 10:24:07.826109 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 9.12722ms
I1204 10:24:07.826163 3001 isa_program_util_common.cc:520] (HLO module jit_select_n): Executable fingerprint:1b4af806359ca11a9aeedcab09cec4115d359228b512823f789d069d0a412454
I1204 10:24:07.826166 3001 isa_program_util_common.cc:524] (HLO module jit_select_n): Executable fingerprint (including data segments):101c5260488115e2b7ec811bcd3a241daf5827b1cc7bf76ea72a4ccf9fc17426
I1204 10:24:07.826169 3001 isa_program_util_common.cc:527] (HLO module jit_select_n): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:07.826488 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 34.63204925ms
I1204 10:24:07.826497 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:07.859756 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.80148ms
I1204 10:24:07.861334 3011 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:07.861979 3011 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:07.863265 3011 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:07.870639 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 9.3902ms
I1204 10:24:07.880040 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 9.20199075ms
I1204 10:24:07.880098 3001 isa_program_util_common.cc:520] (HLO module jit__broadcast_arrays): Executable fingerprint:2465fb80b7e81e394d71acdce5964d521e7f17722fe790d47e342a72d0ee4162
I1204 10:24:07.880101 3001 isa_program_util_common.cc:524] (HLO module jit__broadcast_arrays): Executable fingerprint (including data segments):0d74f996d24d804ac425f2da854d60c51a95459ed55395522afa0de4027e8577
I1204 10:24:07.880104 3001 isa_program_util_common.cc:527] (HLO module jit__broadcast_arrays): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:07.880331 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 32.42347075ms
I1204 10:24:07.880340 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:07.917548 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.01868ms
I1204 10:24:07.919250 3011 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:07.919918 3011 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:07.921684 3011 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:07.930692 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 11.53534ms
I1204 10:24:07.939883 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 9.04954ms
I1204 10:24:07.939945 3001 isa_program_util_common.cc:520] (HLO module jit_broadcast_in_dim): Executable fingerprint:f8a8a1bcc14437ec76793f8d974adac664e3c5f5e676f34082c65624e495eac5
I1204 10:24:07.939949 3001 isa_program_util_common.cc:524] (HLO module jit_broadcast_in_dim): Executable fingerprint (including data segments):b7175bd87ce618f7322eb6a7c85b4a92b395ad6e0edb5ac68d5ee005fd07f01b
I1204 10:24:07.939951 3001 isa_program_util_common.cc:527] (HLO module jit_broadcast_in_dim): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:07.940227 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 35.26784925ms
I1204 10:24:07.940236 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:07.982562 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 8.3753ms
I1204 10:24:07.984325 3011 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:07.985043 3011 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:07.987974 3011 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:08.037487 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 53.22724ms
I1204 10:24:08.051976 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 14.30334925ms
I1204 10:24:08.052051 3001 isa_program_util_common.cc:520] (HLO module jit_gather): Executable fingerprint:91a4f1e38fe2930e32f7554c86d1ccb8bdc143c2455ee497403fe3af4ca8390b
I1204 10:24:08.052055 3001 isa_program_util_common.cc:524] (HLO module jit_gather): Executable fingerprint (including data segments):1b2d850f36448dc61eeba97c481df78ff369ec0b2e8ddc0c53d6f5c61e8117ca
I1204 10:24:08.052057 3001 isa_program_util_common.cc:527] (HLO module jit_gather): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:08.052371 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 86.20678ms
I1204 10:24:08.052380 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:08.089410 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.40278925ms
I1204 10:24:08.090979 3011 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:08.091713 3011 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:08.093177 3011 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:08.391095 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 300.19586925ms
I1204 10:24:08.432636 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 41.34028925ms
I1204 10:24:08.432801 3001 isa_program_util_common.cc:520] (HLO module jit_broadcast_in_dim): Executable fingerprint:dadf519e3f3ce8a15e826a434604227832fc40b271dc7cd685920ce2c30356c0
I1204 10:24:08.432811 3001 isa_program_util_common.cc:524] (HLO module jit_broadcast_in_dim): Executable fingerprint (including data segments):d99f5fd8b80d2133e146bc589b06abcc15e698cf20e4b2248513546d60dbcece
I1204 10:24:08.432815 3001 isa_program_util_common.cc:527] (HLO module jit_broadcast_in_dim): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:08.433049 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 355.22389075ms
I1204 10:24:08.433058 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:08.469591 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.04092ms
I1204 10:24:08.471194 3011 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:08.471954 3011 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:08.473584 3011 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:08.751385 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 280.29228ms
I1204 10:24:08.793764 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 42.17331075ms
I1204 10:24:08.793931 3001 isa_program_util_common.cc:520] (HLO module jit_reshape): Executable fingerprint:93f0bf5ff194e054767462707b33532c9e9884ea846b1ab6ccc0fcccbc667641
I1204 10:24:08.793935 3001 isa_program_util_common.cc:524] (HLO module jit_reshape): Executable fingerprint (including data segments):045c7a3c18f9c9e4b3dbfef8af436d6848187f565a379e99203ecade80cd7615
I1204 10:24:08.793937 3001 isa_program_util_common.cc:527] (HLO module jit_reshape): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:08.794206 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 336.81788ms
I1204 10:24:08.794215 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:08.832925 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 6.49966925ms
I1204 10:24:08.834599 3011 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:08.835339 3011 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:08.837824 3011 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:08.932672 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 98.16356ms
I1204 10:24:08.951140 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 18.27956925ms
I1204 10:24:08.951226 3001 isa_program_util_common.cc:520] (HLO module jit_gather): Executable fingerprint:8beaf31fd0fda9cac6393da71a693fc5b57e6b339b258df3ccff46defc156b23
I1204 10:24:08.951230 3001 isa_program_util_common.cc:524] (HLO module jit_gather): Executable fingerprint (including data segments):2086ffc58556f347fba83b06291df39d22bba91783ce439d655bf9a53dc20183
I1204 10:24:08.951232 3001 isa_program_util_common.cc:527] (HLO module jit_gather): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:08.951503 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 132.03358925ms
I1204 10:24:08.951512 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:08.986318 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.29924ms
I1204 10:24:08.987828 3011 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:08.988562 3011 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:08.989997 3011 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:09.066099 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 78.34234ms
I1204 10:24:09.083158 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 16.8761ms
I1204 10:24:09.083243 3001 isa_program_util_common.cc:520] (HLO module jit_broadcast_in_dim): Executable fingerprint:fe76b306c340dbf0e8e502008db10114d2a0b2dafefb28e94ab75c0607603fa8
I1204 10:24:09.083247 3001 isa_program_util_common.cc:524] (HLO module jit_broadcast_in_dim): Executable fingerprint (including data segments):fc85e8d990116f1f23866e1bcb44c4deedc0f444209f0c14a9fafed248e815ef
I1204 10:24:09.083249 3001 isa_program_util_common.cc:527] (HLO module jit_broadcast_in_dim): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:09.083464 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 108.62946ms
I1204 10:24:09.083473 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:09.118160 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.30952925ms
I1204 10:24:09.119695 3011 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:09.120413 3011 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:09.121860 3011 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:09.198679 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 79.05794ms
I1204 10:24:09.215937 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 17.06797075ms
I1204 10:24:09.216020 3001 isa_program_util_common.cc:520] (HLO module jit__squeeze): Executable fingerprint:fe76b306c340dbf0e8e502008db10114d2a0b2dafefb28e94ab75c0607603fa8
I1204 10:24:09.216025 3001 isa_program_util_common.cc:524] (HLO module jit__squeeze): Executable fingerprint (including data segments):fc85e8d990116f1f23866e1bcb44c4deedc0f444209f0c14a9fafed248e815ef
I1204 10:24:09.216028 3001 isa_program_util_common.cc:527] (HLO module jit__squeeze): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:09.216259 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 109.56363075ms
I1204 10:24:09.216268 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:09.259612 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 12.35553075ms
I1204 10:24:09.261366 3011 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:09.262093 3011 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:09.266827 3011 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
W1204 10:24:09.404907 3282 internals-datagram-socket.cc:411] udp sendto fd=196, dest=172.20.0.120:9530: Network is unreachable
I1204 10:24:09.683378 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 422.11768925ms
I1204 10:24:09.728742 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 45.15858ms
I1204 10:24:09.728960 3001 isa_program_util_common.cc:520] (HLO module jit_scatter): Executable fingerprint:127a8cb58b1fce4a0e06b39684f59fa3b6415b1f03fc55bdf155378e4dd66b08
I1204 10:24:09.728966 3001 isa_program_util_common.cc:524] (HLO module jit_scatter): Executable fingerprint (including data segments):2223e81fc364f82c15b95ceca2b2d40f7f50ecb59482446b3adb1e47cb7c3c53
I1204 10:24:09.728968 3001 isa_program_util_common.cc:527] (HLO module jit_scatter): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:09.729255 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 489.12913075ms
I1204 10:24:09.729264 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:09.768151 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.28874925ms
I1204 10:24:09.769665 3012 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:09.770379 3012 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:09.771833 3012 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:10.069383 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 299.80384ms
I1204 10:24:10.110948 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 41.35747075ms
I1204 10:24:10.111113 3001 isa_program_util_common.cc:520] (HLO module jit__squeeze): Executable fingerprint:dadf519e3f3ce8a15e826a434604227832fc40b271dc7cd685920ce2c30356c0
I1204 10:24:10.111117 3001 isa_program_util_common.cc:524] (HLO module jit__squeeze): Executable fingerprint (including data segments):d99f5fd8b80d2133e146bc589b06abcc15e698cf20e4b2248513546d60dbcece
I1204 10:24:10.111119 3001 isa_program_util_common.cc:527] (HLO module jit__squeeze): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:10.111362 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 354.5039ms
I1204 10:24:10.111371 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:10.150945 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 8.40388925ms
I1204 10:24:10.152606 3012 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:10.153357 3012 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:10.156717 3012 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:10.336744 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 184.22502ms
I1204 10:24:10.358740 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 21.79052ms
I1204 10:24:10.358873 3001 isa_program_util_common.cc:520] (HLO module jit_scatter): Executable fingerprint:6d3ac6c46f4f337ffdc4ba7dae298b536be5d0fb1d6b51c2c39d562ea3e0518f
I1204 10:24:10.358877 3001 isa_program_util_common.cc:524] (HLO module jit_scatter): Executable fingerprint (including data segments):1b6aa04a7480e389c6127b58a5cdf8fc3a2c546fe3b55884558823a590b849b9
I1204 10:24:10.358880 3001 isa_program_util_common.cc:527] (HLO module jit_scatter): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:10.359187 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 223.7638ms
I1204 10:24:10.359196 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:10.396907 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.68640925ms
I1204 10:24:10.398476 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:10.399223 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:10.400830 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:10.447079 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 48.69206ms
I1204 10:24:10.459907 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 12.66885075ms
I1204 10:24:10.459977 3001 isa_program_util_common.cc:520] (HLO module jit_reshape): Executable fingerprint:0fa9c13821fe40aeccbce7bea23962f7c57fefbc43e6851346d8a132771e8064
I1204 10:24:10.459980 3001 isa_program_util_common.cc:524] (HLO module jit_reshape): Executable fingerprint (including data segments):2ae23ec471bbe139b6a0bd7871202910cd1df232be21b4f16df82dbb524b69c1
I1204 10:24:10.459982 3001 isa_program_util_common.cc:527] (HLO module jit_reshape): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:10.460239 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 75.17482ms
I1204 10:24:10.460250 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:10.495562 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.53339075ms
I1204 10:24:10.497056 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:10.497800 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:10.499306 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:10.560145 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 63.15349075ms
I1204 10:24:10.575483 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 15.1599ms
I1204 10:24:10.575565 3001 isa_program_util_common.cc:520] (HLO module jit_reshape): Executable fingerprint:8ec17c91ee59da3a7bcff0625a1c068bbeeceeb89b7ed3fa86548020d1113c8c
I1204 10:24:10.575569 3001 isa_program_util_common.cc:524] (HLO module jit_reshape): Executable fingerprint (including data segments):b55fb09519c02513df5949591fe337ddf91097b136b0862b635ecdee1e7181a9
I1204 10:24:10.575571 3001 isa_program_util_common.cc:527] (HLO module jit_reshape): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:10.575845 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 91.9601ms
I1204 10:24:10.575856 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:10.616830 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 7.28434925ms
I1204 10:24:10.618568 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:10.619280 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:10.622055 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:10.703691 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 85.22548925ms
I1204 10:24:10.719636 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 15.74820925ms
I1204 10:24:10.719718 3001 isa_program_util_common.cc:520] (HLO module jit_gather): Executable fingerprint:6b3c669dac3b5630f3a249c9f074632e84382955f172cb3dfe487f3fda115588
I1204 10:24:10.719728 3001 isa_program_util_common.cc:524] (HLO module jit_gather): Executable fingerprint (including data segments):af06618962ff24a8573aa3350e2b7395eeb7491983819140a9acf446b461fc5b
I1204 10:24:10.719732 3001 isa_program_util_common.cc:527] (HLO module jit_gather): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:10.720035 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 117.61948ms
I1204 10:24:10.720048 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:10.755313 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.3432ms
I1204 10:24:10.756870 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:10.757580 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:10.759063 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:10.785219 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 28.43457075ms
I1204 10:24:10.795784 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 10.42936ms
I1204 10:24:10.795847 3001 isa_program_util_common.cc:520] (HLO module jit_broadcast_in_dim): Executable fingerprint:3cc34f75fba85f5e37bc74aa982c756b619d17ddc9fe3732fd10fc6809757243
I1204 10:24:10.795850 3001 isa_program_util_common.cc:524] (HLO module jit_broadcast_in_dim): Executable fingerprint (including data segments):111b2d780445da494f6e7f228a1a420325d62651fe23697a9ebce509fefbbb47
I1204 10:24:10.795852 3001 isa_program_util_common.cc:527] (HLO module jit_broadcast_in_dim): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:10.796081 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 52.19656925ms
I1204 10:24:10.796093 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:10.834375 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.49098ms
I1204 10:24:10.835912 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:10.836624 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:10.838399 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:10.994726 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 158.91313075ms
I1204 10:24:11.020450 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 25.53064ms
I1204 10:24:11.020566 3001 isa_program_util_common.cc:520] (HLO module jit_broadcast_in_dim): Executable fingerprint:54c7f201df5b6e0bd83ab59de84131a750272295a1a5e7721840b4f816b306b6
I1204 10:24:11.020569 3001 isa_program_util_common.cc:524] (HLO module jit_broadcast_in_dim): Executable fingerprint (including data segments):ccd7679af9fb09fe5bddedfccc11273870ebb364ffec14d1ed4cf8038f8aaf08
I1204 10:24:11.020572 3001 isa_program_util_common.cc:527] (HLO module jit_broadcast_in_dim): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:11.020863 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 198.06322ms
I1204 10:24:11.020875 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:11.057511 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.70206925ms
I1204 10:24:11.059064 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:11.059804 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:11.061366 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:11.265798 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 206.81915075ms
I1204 10:24:11.283274 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 17.27885075ms
I1204 10:24:11.283368 3001 isa_program_util_common.cc:520] (HLO module jit_reshape): Executable fingerprint:5d3c7a5fcc094dce77baf68803cc1a61870de57a8eb9706aeff67184a275188a
I1204 10:24:11.283372 3001 isa_program_util_common.cc:524] (HLO module jit_reshape): Executable fingerprint (including data segments):2c3ffd3bd1a6b9ba0877ed395ed27f33a737d28ca2018bd0f721e688273b2f4f
I1204 10:24:11.283374 3001 isa_program_util_common.cc:527] (HLO module jit_reshape): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:11.283667 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 238.00422ms
I1204 10:24:11.283680 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:11.322892 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 6.09921075ms
I1204 10:24:11.324461 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:11.325190 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:11.328068 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:12.039827 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 715.47062ms
I1204 10:24:12.129687 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 89.63754925ms
I1204 10:24:12.129958 3001 isa_program_util_common.cc:520] (HLO module jit__einsum): Executable fingerprint:1e2a83653ca0e783c367d03408f2d1b6636df034401f264ab0ea3e059888f3ed
I1204 10:24:12.129963 3001 isa_program_util_common.cc:524] (HLO module jit__einsum): Executable fingerprint (including data segments):e2c11c8e33bdd1ffa3094deaf15ff4f6b22edc1b6ea0e99dbbfdadb5ae9d4a07
I1204 10:24:12.129965 3001 isa_program_util_common.cc:527] (HLO module jit__einsum): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:12.130250 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 820.66515075ms
I1204 10:24:12.130262 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:12.172857 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.37034ms
I1204 10:24:12.174408 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:12.175132 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:12.177265 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:12.247063 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 72.73875075ms
I1204 10:24:12.261831 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 14.58415075ms
I1204 10:24:12.261900 3001 isa_program_util_common.cc:520] (HLO module jit_multiply): Executable fingerprint:c2ba4e054775f897a3b5f83db7563f38aa4f014a989dbfc6340b5e58d72d9b8b
I1204 10:24:12.261909 3001 isa_program_util_common.cc:524] (HLO module jit_multiply): Executable fingerprint (including data segments):cdc9a36c64f0e9b76c68930be184accf9d00a042603e3f477053349d2e7aa040
I1204 10:24:12.261913 3001 isa_program_util_common.cc:527] (HLO module jit_multiply): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:12.262207 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 101.88746ms
I1204 10:24:12.262218 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:12.297934 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.27498ms
I1204 10:24:12.299539 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:12.300269 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:12.301595 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:12.332475 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 33.02322ms
I1204 10:24:12.343372 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 10.75876ms
I1204 10:24:12.343440 3001 isa_program_util_common.cc:520] (HLO module jit_iota): Executable fingerprint:953c81bf90d60e085630b75eeba8fbfa2ef9da5b77c6e552699f0971f2c26016
I1204 10:24:12.343444 3001 isa_program_util_common.cc:524] (HLO module jit_iota): Executable fingerprint (including data segments):1a8414389e416f736ae7cebca254f63783d8196c50d54a70f0dba4f2d0103209
I1204 10:24:12.343469 3001 isa_program_util_common.cc:527] (HLO module jit_iota): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:12.343710 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 57.15702ms
I1204 10:24:12.343720 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:12.379703 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.2341ms
I1204 10:24:12.381273 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:12.381995 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:12.384045 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:12.455360 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 74.17484925ms
I1204 10:24:12.470106 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 14.55595075ms
I1204 10:24:12.470177 3001 isa_program_util_common.cc:520] (HLO module jit_add): Executable fingerprint:76a1327eef3e679c7f756b5deda66309944bdcbe71a4fa143fe081d03d4d0213
I1204 10:24:12.470181 3001 isa_program_util_common.cc:524] (HLO module jit_add): Executable fingerprint (including data segments):e7471b6d3ea71b147a43b515666a2c775a2c1e87be2f2f14d2c7437b78d2388c
I1204 10:24:12.470183 3001 isa_program_util_common.cc:527] (HLO module jit_add): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:12.470492 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 103.15022925ms
I1204 10:24:12.470504 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:12.505844 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.25343075ms
I1204 10:24:12.507509 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:12.508237 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:12.509608 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:12.540087 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 32.66718925ms
I1204 10:24:12.551474 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 11.25020925ms
I1204 10:24:12.551541 3001 isa_program_util_common.cc:520] (HLO module jit_iota): Executable fingerprint:124895c743d142b901ea11d5636bd1982c210de975477d0ed495f146cfa5bde7
I1204 10:24:12.551545 3001 isa_program_util_common.cc:524] (HLO module jit_iota): Executable fingerprint (including data segments):1f537d31bf440b93042e36b6630c45a35e946777b2a4a1452699100715bc46e8
I1204 10:24:12.551548 3001 isa_program_util_common.cc:527] (HLO module jit_iota): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:12.551802 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 57.30631075ms
I1204 10:24:12.551812 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:12.587668 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.85654ms
I1204 10:24:12.589340 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:12.589932 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:12.591323 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
W1204 10:24:12.708716 3300 internals-datagram-socket.cc:411] udp sendto fd=198, dest=172.20.0.120:9530: Network is unreachable
I1204 10:24:12.790433 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 201.19838ms
I1204 10:24:12.810221 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 19.61684925ms
I1204 10:24:12.810308 3001 isa_program_util_common.cc:520] (HLO module jit_less): Executable fingerprint:374d5e344ead83f8acac9ef32090d6ed44e9ff7fd22ee87c2ea8e5a7a9a84c08
I1204 10:24:12.810312 3001 isa_program_util_common.cc:524] (HLO module jit_less): Executable fingerprint (including data segments):2cde952cad1aecfb40e2c9ffecbf6684b6bf40caf193609cadf6b4d50ace077a
I1204 10:24:12.810314 3001 isa_program_util_common.cc:527] (HLO module jit_less): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:12.810610 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 234.86576ms
I1204 10:24:12.810622 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:12.849870 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.76829075ms
I1204 10:24:12.851464 3011 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:12.852172 3011 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:12.854930 3011 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:12.942167 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 90.80925075ms
I1204 10:24:12.957901 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 15.57877075ms
I1204 10:24:12.957980 3001 isa_program_util_common.cc:520] (HLO module jit__where): Executable fingerprint:724231b3ce4a751c22774867640c14ba5620067cb89bc00911ffc3c659fe9ee1
I1204 10:24:12.957989 3001 isa_program_util_common.cc:524] (HLO module jit__where): Executable fingerprint (including data segments):f0cea53588cff34fdb71507e29bfa9e4abbd29d83e911045682cb3bf9e0c5758
I1204 10:24:12.957993 3001 isa_program_util_common.cc:527] (HLO module jit__where): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:12.958282 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 121.30085075ms
I1204 10:24:12.958291 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:12.996202 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.70864ms
I1204 10:24:12.997703 3011 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:12.998416 3011 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:13.000068 3011 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:13.129875 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 132.24655075ms
I1204 10:24:13.146695 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 16.61266ms
I1204 10:24:13.146774 3001 isa_program_util_common.cc:520] (HLO module jit_add): Executable fingerprint:c19426e0dba97c93f905edfe788d8a2a71c4a012f9136890b60daca22270bbd7
I1204 10:24:13.146778 3001 isa_program_util_common.cc:524] (HLO module jit_add): Executable fingerprint (including data segments):a602c3029d728774718e550866ced72127001a96826add4bc1f1ff69f408040c
I1204 10:24:13.146780 3001 isa_program_util_common.cc:527] (HLO module jit_add): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:13.147032 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 162.90784925ms
I1204 10:24:13.147040 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:13.181613 2807 deepsea_compiler_base.cc:7855] XLA::TPU running hlo passes for 10 instructions, module: jit__reduce_max
I1204 10:24:13.181652 2807 deepsea_compiler_base.cc:7910] Initial HLO module: jit__reduce_max instructions: 10 fingerprint: 9838537573041770476-4035828241379739704-17078542282357815393-1242137874306953380/5
I1204 10:24:13.181919 2807 deepsea_compiler_base.cc:7982] HLO optimizing module: jit__reduce_max instructions: 10
I1204 10:24:13.181925 2807 deepsea_compiler_base.cc:7997] XLA::TPU HLO optimization
I1204 10:24:13.187625 2807 deepsea_compiler_base.cc:6963] XLA::TPU HLO PostOptimizationPipeline
I1204 10:24:13.189556 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 7.90428ms
I1204 10:24:13.191547 3011 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:13.192241 3011 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:13.194695 3011 window_config_assignment_prod.cc:30] Retrieving backend configs from FDO profiles.
I1204 10:24:13.194989 3011 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:13.295071 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 103.63671075ms
I1204 10:24:13.303079 3001 deepsea_compiler_base.cc:3021] final program bundle count: 723 note this count does not reflect cycles spent executing delays.
I1204 10:24:13.309537 3001 deepsea_compiler_base.cc:3021] final program bundle count: 176 note this count does not reflect cycles spent executing delays.
I1204 10:24:13.310792 3001 deepsea_compiler_base.cc:3507] Program divided into 2 overlays without HLO functions (73.0K).
I1204 10:24:13.310943 3001 deepsea_compiler_base.cc:3675] XLA::TPU module name: jit__reduce_max
I1204 10:24:13.310948 3001 deepsea_compiler_base.cc:3677] XLA::TPU program HBM usage: 1.07M / 31.25G
I1204 10:24:13.310952 3001 deepsea_compiler_base.cc:3724] XLA::TPU program VMEM usage: 2.02M / 128.00M
I1204 10:24:13.310963 3001 deepsea_compiler_base.cc:3735] Total hbm usage >= 263.07M:
I1204 10:24:13.310965 3001 deepsea_compiler_base.cc:3735] reserved 260.00M
I1204 10:24:13.310967 3001 deepsea_compiler_base.cc:3735] program 1.07M
I1204 10:24:13.310969 3001 deepsea_compiler_base.cc:3735] arguments 2.00M
I1204 10:24:13.310975 3001 deepsea_compiler_base.cc:3735]
I1204 10:24:13.310978 3001 deepsea_compiler_base.cc:3735] Output size 16.0K; shares 0B with arguments.
I1204 10:24:13.310980 3001 deepsea_compiler_base.cc:3735]
I1204 10:24:13.310982 3001 deepsea_compiler_base.cc:3735] Total host usage >= 0B:
I1204 10:24:13.310983 3001 deepsea_compiler_base.cc:3735] reserved 0B
I1204 10:24:13.310993 3001 deepsea_compiler_base.cc:3735] program unknown size
I1204 10:24:13.310994 3001 deepsea_compiler_base.cc:3735] arguments 0B
I1204 10:24:13.310996 3001 deepsea_compiler_base.cc:3735]
I1204 10:24:13.310998 3001 deepsea_compiler_base.cc:3735] Output size 0B; shares 0B with arguments.
I1204 10:24:13.311000 3001 deepsea_compiler_base.cc:3735]
I1204 10:24:13.311011 3001 deepsea_compiler_base.cc:3739] Program sflag requirement 216B:
I1204 10:24:13.311013 3001 deepsea_compiler_base.cc:3739] reserved 204B
I1204 10:24:13.311015 3001 deepsea_compiler_base.cc:3739] global 4B
I1204 10:24:13.311016 3001 deepsea_compiler_base.cc:3739] scoped 8B
I1204 10:24:13.311018 3001 deepsea_compiler_base.cc:3739] Program vmem requirement 2.02M:
I1204 10:24:13.311020 3001 deepsea_compiler_base.cc:3739] scoped 2.00M
I1204 10:24:13.311021 3001 deepsea_compiler_base.cc:3739] HLO temp 16.0K (0.0% utilization: Unpadded (0B) Padded (0B), 100.0% fragmentation (16.0K))
I1204 10:24:13.311023 3001 deepsea_compiler_base.cc:3739] Program smem requirement 56B:
I1204 10:24:13.311025 3001 deepsea_compiler_base.cc:3739] global 16B
I1204 10:24:13.311026 3001 deepsea_compiler_base.cc:3739] scoped 40B
I1204 10:24:13.311028 3001 deepsea_compiler_base.cc:3739] Program hbm requirement 1.07M:
I1204 10:24:13.311030 3001 deepsea_compiler_base.cc:3739] global 1.00M
I1204 10:24:13.311032 3001 deepsea_compiler_base.cc:3739] overlays 73.0K
I1204 10:24:13.311033 3001 deepsea_compiler_base.cc:3739] Program host requirement 0B:
I1204 10:24:13.311035 3001 deepsea_compiler_base.cc:3747] XLA::TPU program SMEM usage: 1.9K / 1.00M (2 parameters)
I1204 10:24:13.311050 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 15.79510925ms
I1204 10:24:13.311120 3001 isa_program_util_common.cc:520] (HLO module jit__reduce_max): Executable fingerprint:13e244b03004b8be1c52f064e789776e61759efc4b04f154ab5f67a822d67a55
I1204 10:24:13.311123 3001 isa_program_util_common.cc:524] (HLO module jit__reduce_max): Executable fingerprint (including data segments):43c4308e3955828ee92197772c16729ca1b3549d85a4aeb5eb893877f2644563
I1204 10:24:13.311125 3001 isa_program_util_common.cc:527] (HLO module jit__reduce_max): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:13.311462 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 137.23496925ms
I1204 10:24:13.311476 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:13.349250 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.63238ms
I1204 10:24:13.350888 3011 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:13.351645 3011 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:13.353627 3011 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:13.436705 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 85.9106ms
I1204 10:24:13.453437 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 16.52125075ms
I1204 10:24:13.453515 3001 isa_program_util_common.cc:520] (HLO module jit_subtract): Executable fingerprint:01a39bf18d6e51f34fdba481cdb427ac08b7ae7cf4effb012dc222122e8fcdf8
I1204 10:24:13.453519 3001 isa_program_util_common.cc:524] (HLO module jit_subtract): Executable fingerprint (including data segments):532052e4d86cd287a31df0a8c1abdd3e695560f8bd2ba8b46826849015e13341
I1204 10:24:13.453522 3001 isa_program_util_common.cc:527] (HLO module jit_subtract): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:13.453823 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 117.28888ms
I1204 10:24:13.453831 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:13.489537 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.4445ms
I1204 10:24:13.491065 3011 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:13.491802 3011 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:13.493332 3011 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:13.566283 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 75.32514925ms
I1204 10:24:13.582524 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 16.05657075ms
I1204 10:24:13.582601 3001 isa_program_util_common.cc:520] (HLO module jit_exp): Executable fingerprint:c5daf2d0d5a6270c70bd436c71b9156135ecd9543b13ca5753132b9fc60920bd
I1204 10:24:13.582604 3001 isa_program_util_common.cc:524] (HLO module jit_exp): Executable fingerprint (including data segments):00fa6f58f1cd9f6fad23c63d1ee994be8221a918e48d512f15224cf8e5f3c0fd
I1204 10:24:13.582607 3001 isa_program_util_common.cc:527] (HLO module jit_exp): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:13.582860 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 104.89333075ms
I1204 10:24:13.582880 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:13.621199 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 6.11844925ms
I1204 10:24:13.622792 3011 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:13.623537 3011 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:13.625871 3011 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:13.727042 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 104.34654ms
I1204 10:24:13.742422 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 15.18594ms
I1204 10:24:13.742503 3001 isa_program_util_common.cc:520] (HLO module jit__reduce_sum): Executable fingerprint:72a4cab48f947bb54f80ee672bb0a6440fe952008647e7e9ca2259e16b30ec2f
I1204 10:24:13.742512 3001 isa_program_util_common.cc:524] (HLO module jit__reduce_sum): Executable fingerprint (including data segments):5b444b238184fd1dc5d55a13b98259ab8eae77c5551dbd193b88e73cbcdbccaf
I1204 10:24:13.742515 3001 isa_program_util_common.cc:527] (HLO module jit__reduce_sum): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:13.742813 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 134.87355075ms
I1204 10:24:13.742827 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:13.780221 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.76988ms
I1204 10:24:13.781800 3011 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:13.782534 3011 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:13.784585 3011 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:13.876969 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 95.25487075ms
I1204 10:24:13.894913 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 17.74242925ms
I1204 10:24:13.895002 3001 isa_program_util_common.cc:520] (HLO module jit_true_divide): Executable fingerprint:ff9904ae39fcc12ef2dd73b6636b51b5bfe1a8e4cbf7127b771be271c333d85d
I1204 10:24:13.895005 3001 isa_program_util_common.cc:524] (HLO module jit_true_divide): Executable fingerprint (including data segments):21724743b99762795531ba97a73256c7b5a86f51c7281ddc6623dd836279dbb1
I1204 10:24:13.895008 3001 isa_program_util_common.cc:527] (HLO module jit_true_divide): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:13.895314 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 127.98997075ms
I1204 10:24:13.895324 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:13.935304 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 6.71144ms
I1204 10:24:13.936812 3011 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:13.937575 3011 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:13.940714 3011 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:14.618348 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 681.64998ms
I1204 10:24:14.698112 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 79.57217075ms
I1204 10:24:14.698390 3001 isa_program_util_common.cc:520] (HLO module jit__einsum): Executable fingerprint:80e1e8a6e7ed67a4361e84632d2ebcdcc308c10465e2db6d08f23900c0c2ff96
I1204 10:24:14.698394 3001 isa_program_util_common.cc:524] (HLO module jit__einsum): Executable fingerprint (including data segments):5897583000512a756f3806f1bb4e828534e6f35c59f974fa960b28392b10a283
I1204 10:24:14.698396 3001 isa_program_util_common.cc:527] (HLO module jit__einsum): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:14.698687 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 777.27814ms
I1204 10:24:14.698697 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:14.741119 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 6.08246925ms
I1204 10:24:14.742709 3012 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:14.743427 3012 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:14.745900 3012 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:14.887141 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 144.54119075ms
I1204 10:24:14.903478 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 16.14842925ms
I1204 10:24:14.903556 3001 isa_program_util_common.cc:520] (HLO module jit_concatenate): Executable fingerprint:c6cc7737f8a76bb6514ad80633955cc75426bf8eb5f29845c13e8fbf4bcb55ea
I1204 10:24:14.903560 3001 isa_program_util_common.cc:524] (HLO module jit_concatenate): Executable fingerprint (including data segments):115a1755bef16585ea3a04042928d040094d3b6810c5ec5f47fa74210acbb26d
I1204 10:24:14.903562 3001 isa_program_util_common.cc:527] (HLO module jit_concatenate): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:14.903827 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 175.91756ms
I1204 10:24:14.903836 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
W1204 10:24:16.012545 3334 internals-datagram-socket.cc:411] udp sendto fd=198, dest=172.20.0.120:9530: Network is unreachable
I1204 10:24:16.700968 2807 deepsea_compiler_base.cc:7855] XLA::TPU running hlo passes for 35 instructions, module: jit_ragged_paged_attention
I1204 10:24:16.701014 2807 deepsea_compiler_base.cc:7910] Initial HLO module: jit_ragged_paged_attention instructions: 35 fingerprint: 12458230163957485823-17445055428001010404-17078542282357815393-1242137874306953380/5
I1204 10:24:16.701480 2807 deepsea_compiler_base.cc:7982] HLO optimizing module: jit_ragged_paged_attention instructions: 28
I1204 10:24:16.701487 2807 deepsea_compiler_base.cc:7997] XLA::TPU HLO optimization
I1204 10:24:16.714531 2807 deepsea_compiler_base.cc:6963] XLA::TPU HLO PostOptimizationPipeline
I1204 10:24:16.718116 2807 tpu_plumb_vmem_info.cc:552] PursueNoOptimizationBaselineScavengingVmem is kicking in.
I1204 10:24:16.718531 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 17.51646ms
I1204 10:24:16.720353 3012 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:16.721180 3012 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: T; bytes_for_msa_to_allocate: 134217728
I1204 10:24:16.731235 3012 window_config_assignment_prod.cc:30] Retrieving backend configs from FDO profiles.
I1204 10:24:16.731871 3012 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:19.358905 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 2.63870042925s
I1204 10:24:19.522590 3001 deepsea_compiler_base.cc:3021] final program bundle count: 22,309 note this count does not reflect cycles spent executing delays.
I1204 10:24:19.530313 3001 deepsea_compiler_base.cc:3021] final program bundle count: 176 note this count does not reflect cycles spent executing delays.
I1204 10:24:19.536324 3001 deepsea_compiler_base.cc:3507] Program divided into 4 overlays without HLO functions (1.39M).
I1204 10:24:19.537328 3001 deepsea_compiler_base.cc:3675] XLA::TPU module name: jit_ragged_paged_attention
I1204 10:24:19.537334 3001 deepsea_compiler_base.cc:3677] XLA::TPU program HBM usage: 2.39M / 31.25G
I1204 10:24:19.537338 3001 deepsea_compiler_base.cc:3724] XLA::TPU program VMEM usage: 112.02M / 128.00M
I1204 10:24:19.537350 3001 deepsea_compiler_base.cc:3735] Total hbm usage >= 399.39M:
I1204 10:24:19.537353 3001 deepsea_compiler_base.cc:3735] reserved 260.00M
I1204 10:24:19.537361 3001 deepsea_compiler_base.cc:3735] program 2.39M
I1204 10:24:19.537364 3001 deepsea_compiler_base.cc:3735] arguments 137.00M
I1204 10:24:19.537366 3001 deepsea_compiler_base.cc:3735]
I1204 10:24:19.537368 3001 deepsea_compiler_base.cc:3735] Output size 133.00M; shares 125.00M with arguments.
I1204 10:24:19.537369 3001 deepsea_compiler_base.cc:3735]
I1204 10:24:19.537371 3001 deepsea_compiler_base.cc:3735] Total host usage >= 0B:
I1204 10:24:19.537373 3001 deepsea_compiler_base.cc:3735] reserved 0B
I1204 10:24:19.537374 3001 deepsea_compiler_base.cc:3735] program unknown size
I1204 10:24:19.537376 3001 deepsea_compiler_base.cc:3735] arguments 0B
I1204 10:24:19.537377 3001 deepsea_compiler_base.cc:3735]
I1204 10:24:19.537379 3001 deepsea_compiler_base.cc:3735] Output size 0B; shares 0B with arguments.
I1204 10:24:19.537381 3001 deepsea_compiler_base.cc:3735]
I1204 10:24:19.537394 3001 deepsea_compiler_base.cc:3739] Program sflag requirement 268B:
I1204 10:24:19.537395 3001 deepsea_compiler_base.cc:3739] reserved 204B
I1204 10:24:19.537397 3001 deepsea_compiler_base.cc:3739] global 4B
I1204 10:24:19.537399 3001 deepsea_compiler_base.cc:3739] scoped 36B
I1204 10:24:19.537400 3001 deepsea_compiler_base.cc:3739] HLO temp 24B (100.0% utilization: Unpadded (24B) Padded (24B), 0.0% fragmentation (0B))
I1204 10:24:19.537402 3001 deepsea_compiler_base.cc:3739] Program vmem requirement 112.02M:
I1204 10:24:19.537404 3001 deepsea_compiler_base.cc:3739] HLO temp 112.02M (0.0% utilization: Unpadded (0B) Padded (0B), 100.0% fragmentation (112.02M))
I1204 10:24:19.537405 3001 deepsea_compiler_base.cc:3739] Program smem requirement 4.1K:
I1204 10:24:19.537407 3001 deepsea_compiler_base.cc:3739] global 16B
I1204 10:24:19.537408 3001 deepsea_compiler_base.cc:3739] scoped 4.1K
I1204 10:24:19.537415 3001 deepsea_compiler_base.cc:3739] Program hbm requirement 2.39M:
I1204 10:24:19.537417 3001 deepsea_compiler_base.cc:3739] global 1.00M
I1204 10:24:19.537419 3001 deepsea_compiler_base.cc:3739] overlays 1.39M
I1204 10:24:19.537420 3001 deepsea_compiler_base.cc:3739] Program host requirement 0B:
I1204 10:24:19.537422 3001 deepsea_compiler_base.cc:3747] XLA::TPU program SMEM usage: 6.0K / 1.00M (8 parameters)
I1204 10:24:19.537450 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 178.31876ms
I1204 10:24:19.538312 3001 isa_program_util_common.cc:520] (HLO module jit_ragged_paged_attention): Executable fingerprint:70f9cf606ebf4d450c73c25e73fbf864294821731e2c372d14d2266490981e37
I1204 10:24:19.538318 3001 isa_program_util_common.cc:524] (HLO module jit_ragged_paged_attention): Executable fingerprint (including data segments):8e45925e6aad506d102c37220cc73f0f317d2a579926de6527c8ea261d99907d
I1204 10:24:19.538320 3001 isa_program_util_common.cc:527] (HLO module jit_ragged_paged_attention): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:19.539307 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 2.84582934925s
I1204 10:24:19.539320 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:19.601324 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 7.43866925ms
I1204 10:24:19.603263 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:19.603995 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:19.606530 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:19.942932 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 339.7975ms
I1204 10:24:19.965104 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 21.9642ms
I1204 10:24:19.965216 3001 isa_program_util_common.cc:520] (HLO module jit_gather): Executable fingerprint:6c5ea236a92eeefb38ad37a3c9b5280b1771e51f8183134a71d8f676c03153b2
I1204 10:24:19.965220 3001 isa_program_util_common.cc:524] (HLO module jit_gather): Executable fingerprint (including data segments):6b0d5bd8591988d6fc5dce3f3107ecc731b72db12c58cb1c8bcf259d4b67c0a6
I1204 10:24:19.965223 3001 isa_program_util_common.cc:527] (HLO module jit_gather): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:19.965565 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 378.84300925ms
I1204 10:24:19.965574 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:19.999801 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.3242ms
I1204 10:24:20.001384 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:20.002114 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:20.003609 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
W1204 10:24:20.142355 3300 internals-datagram-socket.cc:411] udp sendto fd=202, dest=172.20.0.120:9530: Network is unreachable
I1204 10:24:20.339088 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 337.79199075ms
I1204 10:24:20.360602 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 21.31435075ms
I1204 10:24:20.360714 3001 isa_program_util_common.cc:520] (HLO module jit_broadcast_in_dim): Executable fingerprint:2f74073077751beb51adb8056977c464315512af5d83b3d43f3f950c1b2b2f49
I1204 10:24:20.360718 3001 isa_program_util_common.cc:524] (HLO module jit_broadcast_in_dim): Executable fingerprint (including data segments):9da123bfa1eb24e5d5470cc9513315cfaf6797678bc0ab7cd1eb21daa3fcdb45
I1204 10:24:20.360720 3001 isa_program_util_common.cc:527] (HLO module jit_broadcast_in_dim): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:20.360951 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 372.49328ms
I1204 10:24:20.360960 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:20.488102 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.84844ms
I1204 10:24:20.491519 3019 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:20.492282 3019 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:20.494080 3019 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:20.574619 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 85.13338ms
I1204 10:24:20.586208 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 11.42826925ms
I1204 10:24:20.586273 3001 isa_program_util_common.cc:520] (HLO module jit_isnan): Executable fingerprint:7f56bfe95cbfbf44893370c340e9bedd4b1884543e08514b26346f41b50e83cb
I1204 10:24:20.586277 3001 isa_program_util_common.cc:524] (HLO module jit_isnan): Executable fingerprint (including data segments):67ecfd8456831dec563306ca03054611ab860c555c1f550fd63d79ac6feee813
I1204 10:24:20.586279 3001 isa_program_util_common.cc:527] (HLO module jit_isnan): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:20.586528 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 110.46099075ms
I1204 10:24:20.586538 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:20.621175 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.36828ms
I1204 10:24:20.622653 3019 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:20.623386 3019 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:20.624928 3019 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:20.673754 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 51.18998ms
I1204 10:24:20.684705 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 10.8085ms
I1204 10:24:20.684771 3001 isa_program_util_common.cc:520] (HLO module jit_bitwise_not): Executable fingerprint:55fd271b85392418c790c615093d952ad4128d9fa22d0167198781cc5a86677d
I1204 10:24:20.684775 3001 isa_program_util_common.cc:524] (HLO module jit_bitwise_not): Executable fingerprint (including data segments):d1f8ba91e5bd39240693eb438e40bca17ba8b189b14886f0b183c92d687582d8
I1204 10:24:20.684778 3001 isa_program_util_common.cc:527] (HLO module jit_bitwise_not): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:20.685023 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 75.31944925ms
I1204 10:24:20.685031 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:20.815698 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.32641075ms
I1204 10:24:20.817157 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:20.817759 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:20.819492 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:20.990678 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 173.58006ms
I1204 10:24:21.013568 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 22.7263ms
I1204 10:24:21.013661 3001 isa_program_util_common.cc:520] (HLO module jit_less): Executable fingerprint:7f8373aab1a876e8bb0a1ae308d3a17bae579100f6706155fdcdfff0bdce1647
I1204 10:24:21.013665 3001 isa_program_util_common.cc:524] (HLO module jit_less): Executable fingerprint (including data segments):11273145bd7b98855bdf94ac305b452bd4bf353ecdcff1d7a68762babaa7baf6
I1204 10:24:21.013667 3001 isa_program_util_common.cc:527] (HLO module jit_less): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:21.013928 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 210.7956ms
I1204 10:24:21.013936 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:21.051028 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.26788ms
I1204 10:24:21.052566 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:21.053158 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:21.054911 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:21.148026 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 95.53605075ms
I1204 10:24:21.164479 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 16.28626ms
I1204 10:24:21.164559 3001 isa_program_util_common.cc:520] (HLO module jit_add): Executable fingerprint:8a2fbb85ca8e849a5f57754164e98bf35d1ec98913832e806f751e05c0380ef8
I1204 10:24:21.164564 3001 isa_program_util_common.cc:524] (HLO module jit_add): Executable fingerprint (including data segments):bbbf01874a981798ab20a416a9c5fe99c532119ec013478bd3193b7a773e6cf1
I1204 10:24:21.164566 3001 isa_program_util_common.cc:527] (HLO module jit_add): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:21.164813 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 126.28128925ms
I1204 10:24:21.164821 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:21.200915 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.73894ms
I1204 10:24:21.202438 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:21.203038 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:21.204543 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:21.249457 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 47.09561075ms
I1204 10:24:21.260744 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 11.1448ms
I1204 10:24:21.260809 3001 isa_program_util_common.cc:520] (HLO module jit_select_n): Executable fingerprint:db681dd213170e3f0b5e61e093a17ecbd0b18c012b2bbbb7b809793c351d4856
I1204 10:24:21.260813 3001 isa_program_util_common.cc:524] (HLO module jit_select_n): Executable fingerprint (including data segments):e6a20bc24cdceb35cfc9c96cf074bdff646dca6da60dc2d71f4926eb3702246a
I1204 10:24:21.260815 3001 isa_program_util_common.cc:527] (HLO module jit_select_n): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:21.261094 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 72.16571075ms
I1204 10:24:21.261106 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:21.313868 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.88317075ms
I1204 10:24:21.315387 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:21.315964 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:21.321009 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:21.405580 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 90.26850925ms
I1204 10:24:21.433215 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 27.48106ms
I1204 10:24:21.433357 3001 isa_program_util_common.cc:520] (HLO module jit__broadcast_arrays): Executable fingerprint:dc0ecd4edb30246eff4cf9d09f5455c6649d4d8746e7d4995dc7ede0c262403d
I1204 10:24:21.433361 3001 isa_program_util_common.cc:524] (HLO module jit__broadcast_arrays): Executable fingerprint (including data segments):1b1a0da88b3bd5c2444ed1101f36494ddbd89d6ebb8cb58b40f15184188f795e
I1204 10:24:21.433363 3001 isa_program_util_common.cc:527] (HLO module jit__broadcast_arrays): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:21.433635 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 133.3609ms
I1204 10:24:21.433651 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:21.473691 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.61714ms
I1204 10:24:21.475164 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:21.475744 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:21.477076 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:21.568695 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 93.59439075ms
I1204 10:24:21.585619 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 16.74317075ms
I1204 10:24:21.585706 3001 isa_program_util_common.cc:520] (HLO module jit_broadcast_in_dim): Executable fingerprint:b5615183d2cec80ffab92c68a399d07d2801d79980df58b608c44f5475d8da84
I1204 10:24:21.585710 3001 isa_program_util_common.cc:524] (HLO module jit_broadcast_in_dim): Executable fingerprint (including data segments):2493c3e686b621290edb3ec0298bd872ab4cb5fd0bf53a3109e3c99be404be9b
I1204 10:24:21.585712 3001 isa_program_util_common.cc:527] (HLO module jit_broadcast_in_dim): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:21.585966 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 124.08213075ms
I1204 10:24:21.585978 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:21.624508 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 5.17226925ms
I1204 10:24:21.625987 3014 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:21.626565 3014 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:21.628281 3014 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:21.706234 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 80.31232925ms
I1204 10:24:21.721457 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 15.06439075ms
I1204 10:24:21.721542 3001 isa_program_util_common.cc:520] (HLO module jit_concatenate): Executable fingerprint:05d7e8b2983b92f1cee7f51f5d81f4de3ea7da229f80b2980146b450fa68c929
I1204 10:24:21.721545 3001 isa_program_util_common.cc:524] (HLO module jit_concatenate): Executable fingerprint (including data segments):9d35f17ea035020d376ef8b4598191423dca19a6e8bb7a4067783fd43ec713cb
I1204 10:24:21.721547 3001 isa_program_util_common.cc:527] (HLO module jit_concatenate): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:21.721837 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 109.77726ms
I1204 10:24:21.721845 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:21.759393 3001 tpu_layout_assignment.cc:3619] Ran 2 additional passes of layout assignment to assign all layouts.
I1204 10:24:21.769842 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 15.88224ms
I1204 10:24:21.771580 3016 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:21.772316 3016 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:21.777654 3016 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:22.681154 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 909.66365075ms
I1204 10:24:22.789704 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 108.36098925ms
I1204 10:24:22.790047 3001 isa_program_util_common.cc:520] (HLO module jit_gather): Executable fingerprint:dc5bad97ca4a34700ad9e4e47973cc2a128f8ab6a57be3c5b2622a9921ca1887
I1204 10:24:22.790052 3001 isa_program_util_common.cc:524] (HLO module jit_gather): Executable fingerprint (including data segments):a72e359c29c3431955971c9d4118f4d20cb967097f1ff0cd60a2b5c0753ad3e5
I1204 10:24:22.790055 3001 isa_program_util_common.cc:527] (HLO module jit_gather): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:22.790332 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 1.04361014s
I1204 10:24:22.790342 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
I1204 10:24:22.835401 2807 deepsea_compiler_base.cc:8071] HLO_PASSES stage duration: 4.54852ms
I1204 10:24:22.837046 3010 tpu_chip_config.cc:329] Resolved chip config alias default->megachip for ghostlite,
I1204 10:24:22.837752 3010 memory_space_assignment_util.cc:803] MSA Vmem breakdown: default scoped vmem (from flag): 33554432; does_msa_directly_allocate_scoped_vmem: F; bytes_for_msa_to_allocate: 100663296
I1204 10:24:22.839272 3010 llo_log_recorder.cc:101] The input HBM logging buffer size is 1048576 and the adjusted size after granule alignment is 1048576
I1204 10:24:22.913619 3001 deepsea_compiler_base.cc:11495] BACKEND_PASSES stage duration: 76.66551075ms
I1204 10:24:22.929331 3001 deepsea_compiler_base.cc:3298] CODE_GENERATION stage duration: 15.50991075ms
I1204 10:24:22.929439 3001 isa_program_util_common.cc:520] (HLO module jit_broadcast_in_dim): Executable fingerprint:dc7cb3c8af537e99426fb2f6800a291d3a9033f3bd1fbe98d769e86aceaa4247
I1204 10:24:22.929443 3001 isa_program_util_common.cc:524] (HLO module jit_broadcast_in_dim): Executable fingerprint (including data segments):0ba606c436b137a24447811a8be258e25d2317a50b0edb6017a2b5b59c1ca475
I1204 10:24:22.929445 3001 isa_program_util_common.cc:527] (HLO module jit_broadcast_in_dim): Host transfer fingerprint:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
I1204 10:24:22.929691 2807 deepsea_compiler_base.cc:8504] END_TO_END stage duration: 106.21578925ms
I1204 10:24:22.929701 2807 deepsea_compiler_base.cc:8505] Note that END_TO_END duration also includes time outside of HLO_PASSES, BACKEND_PASSES, and CODE_GENERATION
[ OK ] RaggedPagedAttentionKernelTest.test_ragged_paged_attention_basic0 (dtype=<class 'jax.numpy.float32'>)
----------------------------------------------------------------------
Ran 1 test in 46.069s
OK
I1204 10:24:23.460033 2807 tpunetd_client.cc:363] Session master stopping the session...
I1204 10:24:23.460208 3012 session_worker_service.cc:109] Session master notifies the worker to disconnect from session a170413005d7ad56
I1204 10:24:23.460359 2807 tpunetd_client.cc:365] Session master stopped the session.
I1204 10:24:23.460367 2807 tpunetd_client.cc:370] Session manager stopping the session...
I1204 10:24:23.460374 2807 noop_barrier.h:25] NoopBarrier::SyncWithTimeout: PRE_STOP_SESSION_BARRIER
I1204 10:24:23.464139 2807 tpunetd_client.cc:372] Session manager stopped the session.
W1204 10:24:23.476119 2825 firmware_indirect_registers.cc:76] Released last reference with existing Open. Performing implicit close-on-destruction.
I1204 10:24:23.996049 2825 async_driver.cc:1119] [/dev/vfio/131 itmm4:pe2:0] vf_id:0 Driver closed.
I1204 10:24:24.025011 2807 httpserver2.cc:939] Starting server termination.
I1204 10:24:24.025051 2807 httpserver2.cc:1102] Closing 0 connections across 32 shards.
I1204 10:24:24.025091 2807 server.cc:2446] Suspending PSP connection callback and invalidating PSP acceptor.
I1204 10:24:24.025115 2807 httpserver2.cc:957] Closing acceptors.
I1204 10:24:24.025311 2807 httpserver2.cc:966] Unexporting HTTPOverRPC.
I1204 10:24:24.025327 2807 httpserver2.cc:975] Unexporting FrontEndStreamService.
I1204 10:24:24.025340 2807 httpserver2.cc:1003] Server termination completed. Remaining pending operations=0
I1204 10:24:24.025344 2807 httpserver2.cc:472] Terminating server with 0 pending operations.
I1204 10:24:24.025352 2807 httpserver2.cc:475] Server pending operations completed. Unregistering server.
I1204 10:24:24.026261 2807 tpu_inspectz.cc:305] Unregistering handler for /tpuz?id=0
I1204 10:24:24.026534 2807 pending_event_logger.cc:160] Waiting for pending events to become available
I1204 10:24:24.133942 2807 metrics_exporter.cc:458] Metrics are flushed.
I1204 10:24:24.134013 2807 client_instance.cc:527] ClientInstance default EndSessions.
I1204 10:24:24.134095 2807 waiter_thread.cc:3104] [test] EndSession for client id 1764872617910046023 (server [2002:a05:663a:64c8:b0:41d:9192:5b3c]:6297)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment