All pull requests

Pull requests

• [graphExecution] warm up, time capture separately, and average 32 replays

  #360·zjin-lcf opened Sep 23, 2026Collaborator·

  ·

• [cmake] compile C++-in-.c sources as CXX and fix SYCL GPU flags

  #359·zjin-lcf opened Sep 22, 2026Collaborator·

  ·
  3

• [pingpong] fail fast unless MPI is GPU-aware

  #357·zjin-lcf opened Sep 22, 2026Collaborator·

  ·
  1

  1

• [mace] add CUDA, HIP, SYCL, and OpenMP HydraGNN MACE pipeline benchmarks

  #356·zjin-lcf opened Sep 21, 2026Collaborator·

  ·
  3

• [rmsnorm-quant] add CUDA, HIP, and SYCL fused RMSNorm+FP8 quantization benchmarks

  #355·zjin-lcf opened Sep 21, 2026Collaborator·

  ·
  1

• Use C++ compiler when file includes <chrono> (only for 'omp' variant)

  #354·tob2 opened Sep 21, 2026Contributor·

  ·
  2

• [bitonic-sort] reduce kernel launches and verify results against std::sort

  #340·zjin-lcf opened Sep 14, 2026Collaborator·

  ·

• [pathfinder] fix the launch geometry and verify results against a host checker

  #339·zjin-lcf opened Sep 14, 2026Collaborator·

  ·

• [segment-reduce2] add CUDA, HIP, and SYCL segmented reduction benchmarks

  #337·zjin-lcf opened Sep 13, 2026Collaborator·

  ·
  5

• [kda] add fused recurrent Kimi Delta Attention benchmarks

  #336·zjin-lcf opened Sep 12, 2026Collaborator·

  ·
  1

• [trimul] add CUDA, HIP, and SYCL benchmarks

  #335·zjin-lcf opened Sep 9, 2026Collaborator·

  ·
  5

• Add OpenAI SwiGLU benchmarks

  #334·zjin-lcf opened Sep 7, 2026Collaborator·

  ·
  8

• [svt] add sparse volume traversal benchmark

  #325·zjin-lcf opened Aug 22, 2026Collaborator·

  ·
  21

• [gsplat4d] add a physics-aware 4D Gaussian splatting benchmark

  #324·zjin-lcf opened Aug 22, 2026Collaborator·

  ·
  11

• [schur-complement] validate the arguments and report allocation and launch failures

  #316·zjin-lcf opened Aug 9, 2026Collaborator·

  ·
  6

• Add OpenMP versions of axpby, bincount, dwconv, pointerchase, and sundials

  #310·zjin-lcf opened Aug 7, 2026Collaborator·

  ·
  12

  

• [hadamard-hip] Use type-dependant tolerance for validation

  #277·maarquitos14 opened Jul 6, 2026Contributor·

  ·
  3

• Adding a verification system to verify benchmark results

  #111·thomasRoglin opened Jul 3, 2025Contributor·

  ·

• Exit with error code 1 when verification fails

  #76·linehill opened Feb 19, 2024·

  ·
  2