All pull requests

Pull requests

• [ReduceVariableLiveness] Count nested-region pressure in the in-loop sink gate

  #8392·oelayan7 opened Oct 12, 2026Contributor·

  ·

• [TritonIntelGPUToLLVM] Only use a predicated load with a constant default

  #8391·slawblauciak opened Oct 12, 2026Contributor·

  ·
  1

• [SGLANG] Update pin to 7953751bfa83239e1dad6dd3a45864db8ce28021

  #8390·wdziurdz opened Oct 12, 2026Contributor·

  ·
  1

• Merge OpenAI Triton commit b575c0c

  #8388·wdziurdz opened Oct 12, 2026Contributor·

  ·
  1

  

• [SGLANG] Update sgl-kernel-xpu pin to ada8a03

  #8387·wdziurdz opened Oct 12, 2026Contributor·

  ·

• [vLLM] Remove quantized-only fused_moe tests from the vllm-tdesc suite

  #8382·dudilester opened Oct 12, 2026Contributor·

  ·
  1

• [vLLM] Keep int8 accumulation in the fused_moe tdesc patch

  #8381·dudilester opened Oct 12, 2026Contributor·

  ·
  1

• [Benchmarks] Fix B60 addmatrix OOM: allocate the do_bench flush buffer first, upcast bf16 on the host in assert_close

  run-benchmarks
  #8379·oelayan7 opened Oct 11, 2026Contributor·

  ·
  1

  5

  

• [TritonIntelGPUToLLVM] Avoid predicated load/store for uniform address with divergent mask

  #8377·anmyachev opened Oct 10, 2026Contributor·

  ·
  1

  3

• [SymbolicBounds] Add tripCountAtLeast

  #8371·etiotto opened Oct 9, 2026Contributor·

  ·
  

• [SGLANG] Unskip test_extend_attention_sliding_window

  #8367·wdziurdz opened Oct 9, 2026Contributor·

  ·
  1

• [SGLANG] Unskip test_all_kernels_triton_matches_torch

  #8364·wdziurdz opened Oct 9, 2026Contributor·

  ·
  1

  1

• [Benchmarks] Use oneDNN's reduced-cache k order in the stream-K benchmark

  run-benchmarks
  #8363·januszjah opened Oct 8, 2026Contributor·

  ·
  1

• Ua offline tuning configs

  #8359·sarah12121212 opened Oct 8, 2026Contributor·2/2·

  ·

• [vLLM] Extend CI suites to cover more vLLM Triton kernels

  #8349·quinnlp opened Oct 8, 2026Contributor·

  ·
  1

• [TritonIntelGPU] Lower evict_first loads to default cache control (#8235)

  #8345·aditzhaki opened Oct 8, 2026Contributor·

  ·
  5

• [TritonIntelGPUToLLVM] Fix i1 add/maxsi/minsi sub-group ops

  #8344·aditzhaki opened Oct 8, 2026Contributor·

  ·
  1

• [CRI] Use 512 GRFs in benchmarks that request 256

  #8335·whitneywhtsang opened Oct 7, 2026Contributor·

  ·

• [RemoveMasks] Use the symbolic bounds prover to remove masks

  #8333·etiotto opened Oct 7, 2026Contributor·

  ·
  1

  

• [Analysis] Add a symbolic bounds prover

  #8332·etiotto opened Oct 7, 2026Contributor·

  ·
  1

  32

  

• [XPU] Accept grf_mode='160', '192' and '512' on CRI only

  #8330·whitneywhtsang opened Oct 7, 2026Contributor·

  ·
  1

• Request explicit GRF modes through SPV_INTEL_maximum_registers

  #8322·whitneywhtsang opened Oct 7, 2026Contributor·

  ·
  1

• [INTEL] Adapt to the upstream fp fusion default change

  run-benchmarks
  #8314·wdziurdz opened Oct 7, 2026Contributor·

  ·
  1

  3

• [SGLANG] Run test_fp4_indexer.py where sgl-kernel-xpu is installed

  #8306·wdziurdz opened Oct 6, 2026Contributor·

  ·

• [vLLM] Update pin

  #8300·quinnlp opened Oct 5, 2026Contributor·

  ·
  1