Search results

E2E accuracy workflow is broken

ci

tests: e2e

Status: Open.

Bug

#5294 In intel/intel-xpu-backend-for-triton;

· anmyachev opened on Oct 13, 2025

· 

1. [PT 2.10 Upstream] TorchInductor

[CI][Liger-Kernels] Liger kernel tests take >12hrs in CI

ci

tests: ecosystem

Status: Open.

Bug

#5292 In intel/intel-xpu-backend-for-triton;

· Egor-Krivov opened on Oct 13, 2025

· 

5. Ecosystem enabling and performance

[FlexAttention] Performance regression

performance

Status: Open.

Bug

#5290 In intel/intel-xpu-backend-for-triton;

· hoshibara opened on Oct 13, 2025

· 

1. [PT 2.10 Upstream] TorchInductor

[Helion] Functional enabling and perf optimizations

tests: ecosystem

Status: Open.

Epic

#5288 In intel/intel-xpu-backend-for-triton;

· vlad-penkin opened on Oct 10, 2025

· 

3. [Triton] Language and Runtime

Build PTI wheels from sources

ci

profiling

Status: Open.

Task

#5278 In intel/intel-xpu-backend-for-triton;

· anmyachev opened on Oct 9, 2025

· 

3. [Triton] Language and Runtime

[pytorch upstream] softmax bwd on BMG is slow

performance

upstream: pytorch

Status: Open.

Bug

#5275 In intel/intel-xpu-backend-for-triton;

· jianyizh opened on Oct 9, 2025

· 

1. [PT 2.10 Upstream] TorchInductor

[pytorch upstream] Layernorm bwd has register spill on BMG

dependencies: pytorch

performance

upstream: pytorch

Status: Open.

Bug

#5274 In intel/intel-xpu-backend-for-triton;

· jianyizh opened on Oct 9, 2025

· 

4. [Performance] Core

[Helion] FlashAttention kernel containing 3-dim block-ptrs not optimized

codegen: mlir

performance

tests: ecosystem

Status: Open.

Bug

#5272 In intel/intel-xpu-backend-for-triton;

· etiotto opened on Oct 8, 2025

· 

5. Ecosystem enabling and performance

Reland upstream commit <code>60605d8</code>

tests: ut

upstream: triton

Status: Open.

Task

#5269 In intel/intel-xpu-backend-for-triton;

· whitneywhtsang opened on Oct 8, 2025

· 

3. [Triton] Language and Runtime

[Gluon] Enable GEMM benchmark, reach perf parity with Triton implementation

benchmarks

codegen: gemm

reporting

Status: Open.

Feature

#5263 In intel/intel-xpu-backend-for-triton;

· vlad-penkin opened on Oct 6, 2025

· 

3. [Triton] Language and Runtime

[VLLM] Improve performance of MOE GEMM kernel

benchmarks

performance

tests: ecosystem

Status: Open.

Task

#5258 In intel/intel-xpu-backend-for-triton;

· Egor-Krivov opened on Oct 6, 2025

· 

5. Ecosystem enabling and performance

[VLLM] Add benchmark for unified attention

ci

performance

tests: ecosystem

Status: Open.

Task

#5257 In intel/intel-xpu-backend-for-triton;

· Egor-Krivov opened on Oct 6, 2025

· 

5. Ecosystem enabling and performance