All pull requests

Pull requests

• Bugfix: Update mpirun options for Open MPI 5 compatibility

  #896·WenqingLan1 opened Nov 9, 2026Contributor·

  ·
  2

• Benchmarks: Refine GPU stream performance with autotune and preconfig

  #895·WenqingLan1 opened Nov 4, 2026Contributor·

  ·
  6

• Benchmarks: Support compact core ranges in CPU STREAM

  micro-benchmarks
  #894·pdr (dpower4) opened Nov 4, 2026Contributor·

  ·
  3

• Bugfix: Preserve Configuration-Specific PyTorch Traces

  #893·Shenghsun Cho (shcho) opened Oct 28, 2026Contributor·

  ·
  12

• Bump the github-actions group with 10 updates

  dependenciesgithub_actions
  #892·dependabot[bot] opened Oct 28, 2026Bot·

  ·
  4

• Benchmarks - Add official-checkpoint V2-Lite FSDP training

  #887·gusui-msft opened Oct 19, 2026Contributor·

  ·
  26

• Benchmarks - Add official-checkpoint V2-Lite generative inference

  #886·gusui-msft opened Oct 19, 2026Contributor·

  ·
  19

• Benchmarks: Micro benchmark - Add Blackwell and Rubin CUTLASS kernels to gemm-flops

  #883·Hongtao Zhang (polarG) opened Oct 15, 2026Contributor·

  ·
  11

• Benchmarks - Add experimental MXFP4 hipBLASLt GEMM

  #882·gusui-msft opened Oct 15, 2026Contributor·5 tasks·

  ·
  5

• Benchmarks: Preserve cuBLASLt precision labels with autotuning

  #881·gusui-msft opened Oct 14, 2026Contributor·

  ·
  2

• Analyzer: Keep aggregate and baseline metric keys consistent

  #880·gusui-msft opened Oct 14, 2026Contributor·

  ·
  2

• Analyzer: Fix existing-baseline merge return contracts

  #879·gusui-msft opened Oct 14, 2026Contributor·

  ·
  2

• Runner: Preserve monitor samples across result files

  #878·gusui-msft opened Oct 14, 2026Contributor·

  ·
  2

• Executor: Support numeric list arguments

  #877·gusui-msft opened Oct 14, 2026Contributor·

  ·
  2

• Benchmarks: Parse singleton and mixed NUMA CPU lists

  #876·gusui-msft opened Oct 14, 2026Contributor·

  ·
  2

• Analyzer: Fix inclusive comparisons in Excel reports

  #875·gusui-msft opened Oct 14, 2026Contributor·

  ·
  2

• Devcontainer: Remove retired pip in-tree-build feature flag

  #874·gusui-msft opened Oct 14, 2026Contributor·

  ·
  2

• Benchmarks: Stop inference when model creation fails

  #873·gusui-msft opened Oct 14, 2026Contributor·

  ·
  2

• Benchmarks: Honor CNN class count during model creation

  #872·gusui-msft opened Oct 14, 2026Contributor·

  ·
  2

• Runner - Use double-dash mpirun options for Open MPI 5

  #871·Nguyen Vuong (normanvuong) opened Oct 12, 2026·

  ·
  2

• Benchmarks - Add ROCm TransferBench mem-bw backend and fix PyTorch DDP store init

  #868·Nguyen Vuong (normanvuong) opened Oct 9, 2026·

  ·
  14

• Dockerfile - Add ROCm 10.1 dockerfile with MI455X (gfx1250) support

  #867·Nguyen Vuong (normanvuong) opened Oct 9, 2026·

  ·
  26

• Dockerfile - Add ROCm7.2 dockerfile

  #855·Nguyen Vuong (normanvuong) opened Sep 23, 2026·2 tasks done·

  ·
  34

• Benchmarks: Decouple cuDNN workload and legacy execution setup

  #853·gusui-msft opened Sep 16, 2026Contributor·

  ·
  11

• Benchmarks: Fix cuDNN storage/compute types and FP16 initialization

  #852·gusui-msft opened Sep 10, 2026Contributor·

  ·
  4