Skip to content

Instantly share code, notes, and snippets.

@AmosLewis
Created March 17, 2025 01:17
Show Gist options
  • Save AmosLewis/688b70aaf0295793448f5b0c70704887 to your computer and use it in GitHub Desktop.
Save AmosLewis/688b70aaf0295793448f5b0c70704887 to your computer and use it in GitHub Desktop.
# downdload the issue1.sh by the following wget
# wget https://gist.githubusercontent.com/AmosLewis/00fdb4e9a96f29c188828e3ff4ea29ef/raw/8377ffd2f53e58e12a9adae7b92b3d5a7f35d98b/bisect-issue1.sh
(bisect.venv) ➜ bisect git:(main) ✗ python ./bisect_packages.py \
--good-ref=00e88733e6b8c8cdb351d4516509f56daebdf604 \
--bad-ref=4451b8ba42b1249eb3ed1b5031a46f51733984ed \
--test-script=/sharedfile/attn/bisect/issue1.sh
Welcome to bisect_packages.py!
------------------------------------------------------------------
--------- Configuration ------------------------------------------
------------------------------------------------------------------
Searching range : '00e88733e6b8c8cdb351d4516509f56daebdf604' - '4451b8ba42b1249eb3ed1b5031a46f51733984ed'
Using working directory : '/home/chi/.iree/bisect'
Using test script : '/sharedfile/attn/bisect/issue1.sh'
Current platform is 'Linux-6.8.0-52-generic-x86_64-with-glibc2.35', platform.system is 'Linux'.
Current Python version is '3.11.1 (main, Oct 7 2024, 06:16:08) [GCC 11.4.0]'. This script requires 3.11.
Found gh at '/usr/bin/gh'.
------------------------------------------------------------------
------------------------------------------------------------------
--------- Running git bisect -------------------------------------
------------------------------------------------------------------
Bisecting: 3 revisions left to test after this (roughly 2 steps)
[3bfd3628d03587d25fe7f5de126cbb672fd0d71f] No `extern "C"` main functions (#20223)
running '/home/chi/.iree/bisect/bisect_run_script.sh'
++ git rev-parse BISECT_HEAD
+ REF_HASH=3bfd3628d03587d25fe7f5de126cbb672fd0d71f
+ python /home/chi/src/iree/build_tools/pkgci/bisect/../setup_venv.py /home/chi/.iree/bisect/3bfd3628d03587d25fe7f5de126cbb672fd0d71f/.venv --artifact-path=/home/chi/.iree/bisect/3bfd3628d03587d25fe7f5de126cbb672fd0d71f --fetch-git-ref=3bfd3628d03587d25fe7f5de126cbb672fd0d71f
Finding workflow run for ref: 3bfd3628d03587d25fe7f5de126cbb672fd0d71f
Using normalized ref: 3bfd3628d03587d25fe7f5de126cbb672fd0d71f
Running command to list workflow runs:
gh api -H Accept: application/vnd.github+json -H X-GitHub-Api-Version: 2022-11-28 /repos/iree-org/iree/actions/workflows/pkgci.yml/runs?head_sha=3bfd3628d03587d25fe7f5de126cbb672fd0d71f
Found workflow run: https://github.com/iree-org/iree/actions/runs/13818437894
Package iree-base-compiler not found in cache. Fetching from linux_x86_64_release_packages...
Fetching artifacts for workflow run: 13818437894
Found artifacts:
linux_x86_64_release_packages: /repos/iree-org/iree/actions/artifacts/2739977017/zip
Downloading artifact /repos/iree-org/iree/actions/artifacts/2739977017/zip
Extracting /home/chi/.iree/bisect/3bfd3628d03587d25fe7f5de126cbb672fd0d71f/linux_x86_64_release_packages.zip
Installing wheels: [(PosixPath('/home/chi/.iree/bisect/3bfd3628d03587d25fe7f5de126cbb672fd0d71f'), 'iree-base-compiler'), (PosixPath('/home/chi/.iree/bisect/3bfd3628d03587d25fe7f5de126cbb672fd0d71f'), 'iree-base-runtime')]
Creating venv at /home/chi/.iree/bisect/3bfd3628d03587d25fe7f5de126cbb672fd0d71f/.venv
Running command: /home/chi/.iree/bisect/3bfd3628d03587d25fe7f5de126cbb672fd0d71f/.venv/bin/python -m pip install --no-deps --no-index -f /home/chi/.iree/bisect/3bfd3628d03587d25fe7f5de126cbb672fd0d71f --force-reinstall iree-base-compiler
Looking in links: /home/chi/.iree/bisect/3bfd3628d03587d25fe7f5de126cbb672fd0d71f
Processing /home/chi/.iree/bisect/3bfd3628d03587d25fe7f5de126cbb672fd0d71f/iree_base_compiler-3.3.0.dev0+3bfd3628d03587d25fe7f5de126cbb672fd0d71f-cp311-cp311-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
Installing collected packages: iree-base-compiler
Successfully installed iree-base-compiler-3.3.0.dev0+3bfd3628d03587d25fe7f5de126cbb672fd0d71f
Running command: /home/chi/.iree/bisect/3bfd3628d03587d25fe7f5de126cbb672fd0d71f/.venv/bin/python -m pip install --no-deps --no-index -f /home/chi/.iree/bisect/3bfd3628d03587d25fe7f5de126cbb672fd0d71f --force-reinstall iree-base-runtime
Looking in links: /home/chi/.iree/bisect/3bfd3628d03587d25fe7f5de126cbb672fd0d71f
Processing /home/chi/.iree/bisect/3bfd3628d03587d25fe7f5de126cbb672fd0d71f/iree_base_runtime-3.3.0.dev0+3bfd3628d03587d25fe7f5de126cbb672fd0d71f-cp311-cp311-manylinux_2_28_x86_64.whl
Installing collected packages: iree-base-runtime
Successfully installed iree-base-runtime-3.3.0.dev0+3bfd3628d03587d25fe7f5de126cbb672fd0d71f
venv setup complete at '/home/chi/.iree/bisect/3bfd3628d03587d25fe7f5de126cbb672fd0d71f/.venv'. Activate it with
source /home/chi/.iree/bisect/3bfd3628d03587d25fe7f5de126cbb672fd0d71f/.venv/bin/activate
+ PATH=/home/chi/.iree/bisect/3bfd3628d03587d25fe7f5de126cbb672fd0d71f/.venv/bin:/usr/lib/git-core:/usr/lib/git-core:/sharedfile/bisect.venv/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin
+ set +e
+ iree-compile /sharedfile/attn/128/fp8_attn.mlir --iree-hip-target=gfx942 -o=/sharedfile/attn/128/fp8_attn.vmfb --iree-hal-target-device=hip --iree-dispatch-creation-enable-aggressive-fusion=true --iree-global-opt-propagate-transposes=true --iree-opt-aggressively-propagate-transposes=true --iree-opt-data-tiling=false '--iree-preprocessing-pass-pipeline=builtin.module(util.func(iree-preprocessing-generalize-linalg-matmul-experimental))' --iree-hal-indirect-command-buffers=true --iree-stream-resource-memory-model=discrete --iree-hal-memoization=true --iree-opt-strip-assertions
+ ROCR_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
+ iree-benchmark-module --hip_use_streams=true --module=/sharedfile/attn/128/fp8_attn.vmfb --parameters=model=/sharedfile/attn/fp8_attn.irpa --device=hip://4 --function=prefill_bs4 --input=4x128xi64=@/sharedfile/128/prefill/prefill_token_ids_4x128xi64.bin --input=4xi64=@/sharedfile/128/prefill/prefill_seq_lens_4xi64.bin --input=4x4xi64=@/sharedfile/128/prefill/prefill_seq_block_ids_4x4xi64.bin --input=261x2097152xf8E4M3FNUZ=@/sharedfile/128/prefill/prefill_cache_state_261x2097152xf8E4M3FNUZ.bin --benchmark_repetitions=3
2025-03-16T18:15:20-07:00
Running /home/chi/.iree/bisect/3bfd3628d03587d25fe7f5de126cbb672fd0d71f/.venv/lib/python3.11/site-packages/iree/_runtime_libs/iree-benchmark-module
Run on (96 X 3810.79 MHz CPU s)
CPU Caches:
L1 Data 32 KiB (x96)
L1 Instruction 32 KiB (x96)
L2 Unified 1024 KiB (x96)
L3 Unified 32768 KiB (x16)
Load Average: 3.52, 2.79, 2.94
***WARNING*** CPU scaling is enabled, the benchmark real time measurements may be noisy and will incur extra overhead.
-------------------------------------------------------------------------------------------------------
Benchmark Time CPU Iterations UserCounters...
-------------------------------------------------------------------------------------------------------
BM_prefill_bs4/process_time/real_time 29.9 ms 30.7 ms 24 items_per_second=33.4553/s
BM_prefill_bs4/process_time/real_time 29.9 ms 30.8 ms 24 items_per_second=33.4601/s
BM_prefill_bs4/process_time/real_time 29.8 ms 30.7 ms 24 items_per_second=33.5245/s
BM_prefill_bs4/process_time/real_time_mean 29.9 ms 30.7 ms 3 items_per_second=33.48/s
BM_prefill_bs4/process_time/real_time_median 29.9 ms 30.7 ms 3 items_per_second=33.4601/s
BM_prefill_bs4/process_time/real_time_stddev 0.034 ms 0.062 ms 3 items_per_second=0.0386615/s
BM_prefill_bs4/process_time/real_time_cv 0.12 % 0.20 % 3 items_per_second=0.12%
(bisect.venv) ➜ bisect git:(main) ✗ python ./bisect_packages.py \
--good-ref=00e88733e6b8c8cdb351d4516509f56daebdf604 \
--bad-ref=3bfd3628d03587d25fe7f5de126cbb672fd0d71f \
--test-script=/sharedfile/attn/bisect/issue1.sh
Welcome to bisect_packages.py!
------------------------------------------------------------------
--------- Configuration ------------------------------------------
------------------------------------------------------------------
Searching range : '00e88733e6b8c8cdb351d4516509f56daebdf604' - '3bfd3628d03587d25fe7f5de126cbb672fd0d71f'
Using working directory : '/home/chi/.iree/bisect'
Using test script : '/sharedfile/attn/bisect/issue1.sh'
Current platform is 'Linux-6.8.0-52-generic-x86_64-with-glibc2.35', platform.system is 'Linux'.
Current Python version is '3.11.1 (main, Oct 7 2024, 06:16:08) [GCC 11.4.0]'. This script requires 3.11.
Found gh at '/usr/bin/gh'.
------------------------------------------------------------------
------------------------------------------------------------------
--------- Running git bisect -------------------------------------
------------------------------------------------------------------
Bisecting: 0 revisions left to test after this (roughly 1 step)
[db5b69aab08593e215e4c999854fa633bcedb346] [AMDGPU] Do not rewrite or approximate math functions on ROCm (#20222)
running '/home/chi/.iree/bisect/bisect_run_script.sh'
++ git rev-parse BISECT_HEAD
+ REF_HASH=db5b69aab08593e215e4c999854fa633bcedb346
+ python /home/chi/src/iree/build_tools/pkgci/bisect/../setup_venv.py /home/chi/.iree/bisect/db5b69aab08593e215e4c999854fa633bcedb346/.venv --artifact-path=/home/chi/.iree/bisect/db5b69aab08593e215e4c999854fa633bcedb346 --fetch-git-ref=db5b69aab08593e215e4c999854fa633bcedb346
Finding workflow run for ref: db5b69aab08593e215e4c999854fa633bcedb346
Using normalized ref: db5b69aab08593e215e4c999854fa633bcedb346
Running command to list workflow runs:
gh api -H Accept: application/vnd.github+json -H X-GitHub-Api-Version: 2022-11-28 /repos/iree-org/iree/actions/workflows/pkgci.yml/runs?head_sha=db5b69aab08593e215e4c999854fa633bcedb346
Found workflow run: https://github.com/iree-org/iree/actions/runs/13816514809
Package iree-base-compiler not found in cache. Fetching from linux_x86_64_release_packages...
Fetching artifacts for workflow run: 13816514809
Found artifacts:
linux_x86_64_release_packages: /repos/iree-org/iree/actions/artifacts/2739270803/zip
Downloading artifact /repos/iree-org/iree/actions/artifacts/2739270803/zip
Extracting /home/chi/.iree/bisect/db5b69aab08593e215e4c999854fa633bcedb346/linux_x86_64_release_packages.zip
Installing wheels: [(PosixPath('/home/chi/.iree/bisect/db5b69aab08593e215e4c999854fa633bcedb346'), 'iree-base-compiler'), (PosixPath('/home/chi/.iree/bisect/db5b69aab08593e215e4c999854fa633bcedb346'), 'iree-base-runtime')]
Creating venv at /home/chi/.iree/bisect/db5b69aab08593e215e4c999854fa633bcedb346/.venv
Running command: /home/chi/.iree/bisect/db5b69aab08593e215e4c999854fa633bcedb346/.venv/bin/python -m pip install --no-deps --no-index -f /home/chi/.iree/bisect/db5b69aab08593e215e4c999854fa633bcedb346 --force-reinstall iree-base-compiler
Looking in links: /home/chi/.iree/bisect/db5b69aab08593e215e4c999854fa633bcedb346
Processing /home/chi/.iree/bisect/db5b69aab08593e215e4c999854fa633bcedb346/iree_base_compiler-3.3.0.dev0+db5b69aab08593e215e4c999854fa633bcedb346-cp311-cp311-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
Installing collected packages: iree-base-compiler
Successfully installed iree-base-compiler-3.3.0.dev0+db5b69aab08593e215e4c999854fa633bcedb346
Running command: /home/chi/.iree/bisect/db5b69aab08593e215e4c999854fa633bcedb346/.venv/bin/python -m pip install --no-deps --no-index -f /home/chi/.iree/bisect/db5b69aab08593e215e4c999854fa633bcedb346 --force-reinstall iree-base-runtime
Looking in links: /home/chi/.iree/bisect/db5b69aab08593e215e4c999854fa633bcedb346
Processing /home/chi/.iree/bisect/db5b69aab08593e215e4c999854fa633bcedb346/iree_base_runtime-3.3.0.dev0+db5b69aab08593e215e4c999854fa633bcedb346-cp311-cp311-manylinux_2_28_x86_64.whl
Installing collected packages: iree-base-runtime
Successfully installed iree-base-runtime-3.3.0.dev0+db5b69aab08593e215e4c999854fa633bcedb346
venv setup complete at '/home/chi/.iree/bisect/db5b69aab08593e215e4c999854fa633bcedb346/.venv'. Activate it with
source /home/chi/.iree/bisect/db5b69aab08593e215e4c999854fa633bcedb346/.venv/bin/activate
+ PATH=/home/chi/.iree/bisect/db5b69aab08593e215e4c999854fa633bcedb346/.venv/bin:/usr/lib/git-core:/usr/lib/git-core:/sharedfile/bisect.venv/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin
+ set +e
+ iree-compile /sharedfile/attn/128/fp8_attn.mlir --iree-hip-target=gfx942 -o=/sharedfile/attn/128/fp8_attn.vmfb --iree-hal-target-device=hip --iree-dispatch-creation-enable-aggressive-fusion=true --iree-global-opt-propagate-transposes=true --iree-opt-aggressively-propagate-transposes=true --iree-opt-data-tiling=false '--iree-preprocessing-pass-pipeline=builtin.module(util.func(iree-preprocessing-generalize-linalg-matmul-experimental))' --iree-hal-indirect-command-buffers=true --iree-stream-resource-memory-model=discrete --iree-hal-memoization=true --iree-opt-strip-assertions
+ ROCR_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
+ iree-benchmark-module --hip_use_streams=true --module=/sharedfile/attn/128/fp8_attn.vmfb --parameters=model=/sharedfile/attn/fp8_attn.irpa --device=hip://4 --function=prefill_bs4 --input=4x128xi64=@/sharedfile/128/prefill/prefill_token_ids_4x128xi64.bin --input=4xi64=@/sharedfile/128/prefill/prefill_seq_lens_4xi64.bin --input=4x4xi64=@/sharedfile/128/prefill/prefill_seq_block_ids_4x4xi64.bin --input=261x2097152xf8E4M3FNUZ=@/sharedfile/128/prefill/prefill_cache_state_261x2097152xf8E4M3FNUZ.bin --benchmark_repetitions=3
2025-03-16T18:18:24-07:00
Running /home/chi/.iree/bisect/db5b69aab08593e215e4c999854fa633bcedb346/.venv/lib/python3.11/site-packages/iree/_runtime_libs/iree-benchmark-module
Run on (96 X 3810.79 MHz CPU s)
CPU Caches:
L1 Data 32 KiB (x96)
L1 Instruction 32 KiB (x96)
L2 Unified 1024 KiB (x96)
L3 Unified 32768 KiB (x16)
Load Average: 4.70, 3.19, 3.04
***WARNING*** CPU scaling is enabled, the benchmark real time measurements may be noisy and will incur extra overhead.
-------------------------------------------------------------------------------------------------------
Benchmark Time CPU Iterations UserCounters...
-------------------------------------------------------------------------------------------------------
BM_prefill_bs4/process_time/real_time 29.7 ms 30.4 ms 24 items_per_second=33.6578/s
BM_prefill_bs4/process_time/real_time 29.8 ms 30.4 ms 24 items_per_second=33.612/s
BM_prefill_bs4/process_time/real_time 29.8 ms 30.3 ms 24 items_per_second=33.5952/s
BM_prefill_bs4/process_time/real_time_mean 29.7 ms 30.4 ms 3 items_per_second=33.6217/s
BM_prefill_bs4/process_time/real_time_median 29.8 ms 30.4 ms 3 items_per_second=33.612/s
BM_prefill_bs4/process_time/real_time_stddev 0.029 ms 0.032 ms 3 items_per_second=0.0324114/s
BM_prefill_bs4/process_time/real_time_cv 0.10 % 0.11 % 3 items_per_second=0.10%
(bisect.venv) ➜ bisect git:(main) ✗ python ./bisect_packages.py \
--good-ref=00e88733e6b8c8cdb351d4516509f56daebdf604 \
--bad-ref=db5b69aab08593e215e4c999854fa633bcedb346 \
--test-script=/sharedfile/attn/bisect/issue1.sh
Welcome to bisect_packages.py!
------------------------------------------------------------------
--------- Configuration ------------------------------------------
------------------------------------------------------------------
Searching range : '00e88733e6b8c8cdb351d4516509f56daebdf604' - 'db5b69aab08593e215e4c999854fa633bcedb346'
Using working directory : '/home/chi/.iree/bisect'
Using test script : '/sharedfile/attn/bisect/issue1.sh'
Current platform is 'Linux-6.8.0-52-generic-x86_64-with-glibc2.35', platform.system is 'Linux'.
Current Python version is '3.11.1 (main, Oct 7 2024, 06:16:08) [GCC 11.4.0]'. This script requires 3.11.
Found gh at '/usr/bin/gh'.
------------------------------------------------------------------
------------------------------------------------------------------
--------- Running git bisect -------------------------------------
------------------------------------------------------------------
Bisecting: 0 revisions left to test after this (roughly 0 steps)
[9d693cb25223d0520007d97614412e5d7a9f3606] [Codegen][Tuner] improve verifier for the default attribute (#20173)
running '/home/chi/.iree/bisect/bisect_run_script.sh'
++ git rev-parse BISECT_HEAD
+ REF_HASH=9d693cb25223d0520007d97614412e5d7a9f3606
+ python /home/chi/src/iree/build_tools/pkgci/bisect/../setup_venv.py /home/chi/.iree/bisect/9d693cb25223d0520007d97614412e5d7a9f3606/.venv --artifact-path=/home/chi/.iree/bisect/9d693cb25223d0520007d97614412e5d7a9f3606 --fetch-git-ref=9d693cb25223d0520007d97614412e5d7a9f3606
Finding workflow run for ref: 9d693cb25223d0520007d97614412e5d7a9f3606
Using normalized ref: 9d693cb25223d0520007d97614412e5d7a9f3606
Running command to list workflow runs:
gh api -H Accept: application/vnd.github+json -H X-GitHub-Api-Version: 2022-11-28 /repos/iree-org/iree/actions/workflows/pkgci.yml/runs?head_sha=9d693cb25223d0520007d97614412e5d7a9f3606
Found workflow run: https://github.com/iree-org/iree/actions/runs/13815044240
Package iree-base-compiler not found in cache. Fetching from linux_x86_64_release_packages...
Fetching artifacts for workflow run: 13815044240
Found artifacts:
linux_x86_64_release_packages: /repos/iree-org/iree/actions/artifacts/2738691432/zip
Downloading artifact /repos/iree-org/iree/actions/artifacts/2738691432/zip
Extracting /home/chi/.iree/bisect/9d693cb25223d0520007d97614412e5d7a9f3606/linux_x86_64_release_packages.zip
Installing wheels: [(PosixPath('/home/chi/.iree/bisect/9d693cb25223d0520007d97614412e5d7a9f3606'), 'iree-base-compiler'), (PosixPath('/home/chi/.iree/bisect/9d693cb25223d0520007d97614412e5d7a9f3606'), 'iree-base-runtime')]
Creating venv at /home/chi/.iree/bisect/9d693cb25223d0520007d97614412e5d7a9f3606/.venv
Running command: /home/chi/.iree/bisect/9d693cb25223d0520007d97614412e5d7a9f3606/.venv/bin/python -m pip install --no-deps --no-index -f /home/chi/.iree/bisect/9d693cb25223d0520007d97614412e5d7a9f3606 --force-reinstall iree-base-compiler
Looking in links: /home/chi/.iree/bisect/9d693cb25223d0520007d97614412e5d7a9f3606
Processing /home/chi/.iree/bisect/9d693cb25223d0520007d97614412e5d7a9f3606/iree_base_compiler-3.3.0.dev0+9d693cb25223d0520007d97614412e5d7a9f3606-cp311-cp311-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
Installing collected packages: iree-base-compiler
Successfully installed iree-base-compiler-3.3.0.dev0+9d693cb25223d0520007d97614412e5d7a9f3606
Running command: /home/chi/.iree/bisect/9d693cb25223d0520007d97614412e5d7a9f3606/.venv/bin/python -m pip install --no-deps --no-index -f /home/chi/.iree/bisect/9d693cb25223d0520007d97614412e5d7a9f3606 --force-reinstall iree-base-runtime
Looking in links: /home/chi/.iree/bisect/9d693cb25223d0520007d97614412e5d7a9f3606
Processing /home/chi/.iree/bisect/9d693cb25223d0520007d97614412e5d7a9f3606/iree_base_runtime-3.3.0.dev0+9d693cb25223d0520007d97614412e5d7a9f3606-cp311-cp311-manylinux_2_28_x86_64.whl
Installing collected packages: iree-base-runtime
Successfully installed iree-base-runtime-3.3.0.dev0+9d693cb25223d0520007d97614412e5d7a9f3606
venv setup complete at '/home/chi/.iree/bisect/9d693cb25223d0520007d97614412e5d7a9f3606/.venv'. Activate it with
source /home/chi/.iree/bisect/9d693cb25223d0520007d97614412e5d7a9f3606/.venv/bin/activate
+ PATH=/home/chi/.iree/bisect/9d693cb25223d0520007d97614412e5d7a9f3606/.venv/bin:/usr/lib/git-core:/usr/lib/git-core:/sharedfile/bisect.venv/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin
+ set +e
+ iree-compile /sharedfile/attn/128/fp8_attn.mlir --iree-hip-target=gfx942 -o=/sharedfile/attn/128/fp8_attn.vmfb --iree-hal-target-device=hip --iree-dispatch-creation-enable-aggressive-fusion=true --iree-global-opt-propagate-transposes=true --iree-opt-aggressively-propagate-transposes=true --iree-opt-data-tiling=false '--iree-preprocessing-pass-pipeline=builtin.module(util.func(iree-preprocessing-generalize-linalg-matmul-experimental))' --iree-hal-indirect-command-buffers=true --iree-stream-resource-memory-model=discrete --iree-hal-memoization=true --iree-opt-strip-assertions
+ ROCR_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
+ iree-benchmark-module --hip_use_streams=true --module=/sharedfile/attn/128/fp8_attn.vmfb --parameters=model=/sharedfile/attn/fp8_attn.irpa --device=hip://4 --function=prefill_bs4 --input=4x128xi64=@/sharedfile/128/prefill/prefill_token_ids_4x128xi64.bin --input=4xi64=@/sharedfile/128/prefill/prefill_seq_lens_4xi64.bin --input=4x4xi64=@/sharedfile/128/prefill/prefill_seq_block_ids_4x4xi64.bin --input=261x2097152xf8E4M3FNUZ=@/sharedfile/128/prefill/prefill_cache_state_261x2097152xf8E4M3FNUZ.bin --benchmark_repetitions=3
2025-03-16T18:19:47-07:00
Running /home/chi/.iree/bisect/9d693cb25223d0520007d97614412e5d7a9f3606/.venv/lib/python3.11/site-packages/iree/_runtime_libs/iree-benchmark-module
Run on (96 X 3810.79 MHz CPU s)
CPU Caches:
L1 Data 32 KiB (x96)
L1 Instruction 32 KiB (x96)
L2 Unified 1024 KiB (x96)
L3 Unified 32768 KiB (x16)
Load Average: 3.79, 3.15, 3.04
***WARNING*** CPU scaling is enabled, the benchmark real time measurements may be noisy and will incur extra overhead.
-------------------------------------------------------------------------------------------------------
Benchmark Time CPU Iterations UserCounters...
-------------------------------------------------------------------------------------------------------
BM_prefill_bs4/process_time/real_time 27.5 ms 28.1 ms 26 items_per_second=36.4008/s
BM_prefill_bs4/process_time/real_time 27.5 ms 28.1 ms 26 items_per_second=36.3567/s
BM_prefill_bs4/process_time/real_time 27.5 ms 28.1 ms 26 items_per_second=36.3624/s
BM_prefill_bs4/process_time/real_time_mean 27.5 ms 28.1 ms 3 items_per_second=36.3733/s
BM_prefill_bs4/process_time/real_time_median 27.5 ms 28.1 ms 3 items_per_second=36.3624/s
BM_prefill_bs4/process_time/real_time_stddev 0.018 ms 0.003 ms 3 items_per_second=0.023992/s
BM_prefill_bs4/process_time/real_time_cv 0.07 % 0.01 % 3 items_per_second=0.07%
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment