Search results

Quantized Matmul with Higher precision src (BF16/F32)

question

Status: Open.

#4481 In uxlfoundation/oneDNN;

· SriAlavandar opened on Dec 23, 2025

AArch64: f16 matmul performance degradation

sightingSuspicious library behavior. Should be promoted to a bug when confirmedSuspicious library behavior. Should be promoted to a bug when confirmed

Status: Open.

#4476 In uxlfoundation/oneDNN;

· alvoron opened on Dec 22, 2025

Using oneDNN v3.6 with the maximum ISA set to avx, and the attribute post-op configured as eltwise_relu, the actual execution runs on jit:sse41.

sightingSuspicious library behavior. Should be promoted to a bug when confirmedSuspicious library behavior. Should be promoted to a bug when confirmed

Status: Open.

#4462 In uxlfoundation/oneDNN;

· Li-Theone opened on Dec 15, 2025

Running examples with DPC++ SYCL GPU-only

platform:gpu-intelCodeowner: @oneapi-src/onednn-gpu-intelCodeowner: @oneapi-src/onednn-gpu-intel

question

Status: Open.

#4456 In uxlfoundation/oneDNN;

· frenchwr opened on Dec 12, 2025

Feasibility of Adding a New Backend Library to oneDNN

platform:cpu-x64Intel64/AMD64 processors. Codeowner: @oneapi-src/onednn-cpu-x64Intel64/AMD64 processors. Codeowner: @oneapi-src/onednn-cpu-x64

question

Status: Open.

#4446 In uxlfoundation/oneDNN;

· avinashcpandey opened on Dec 10, 2025

brgemm_matmul segfaults with multiple threads when broadcasting dims

bugA confirmed library bugA confirmed library bug

platform:cpu-aarch64Codeowner: @oneapi-src/onednn-cpu-aarch64Codeowner: @oneapi-src/onednn-cpu-aarch64

platform:cpu-x64Intel64/AMD64 processors. Codeowner: @oneapi-src/onednn-cpu-x64Intel64/AMD64 processors. Codeowner: @oneapi-src/onednn-cpu-x64

Status: Open.

#4396 In uxlfoundation/oneDNN;

· Sqvid opened on Nov 28, 2025

[Graph API] Is compiled_partition thread-safe for concurrent calls on same instance?

question

Status: Open.

#4381 In uxlfoundation/oneDNN;

· lxq2t opened on Nov 25, 2025

dnnl_lstm_backward_primitive_desc_create() can be called with parameter prop_kind=dnnl_forward_training

bugA confirmed library bugA confirmed library bug

Status: Open.

#4306 In uxlfoundation/oneDNN;

· lacak-sk opened on Nov 10, 2025

Support for long integers (aka int64, s64, long)?

enhancementA feature or an optimization requestA feature or an optimization request

help wanted

Status: Open.

#4299 In uxlfoundation/oneDNN;

· blueberry opened on Nov 7, 2025

[Doc] Update documentation on "flags" parameter in dnnl_lstm_backward_primitive_desc_create()

documentationA request to change/fix/improve the documentation. Codeowner: @oneapi-src/onednn-docA request to change/fix/improve the documentation. Codeowner: @oneapi-src/onednn-doc

Status: Open.

#4216 In uxlfoundation/oneDNN;

· lacak-sk opened on Oct 27, 2025

How to query for memory descriptors for weights_layer_md and weights_iter_md in RNN LSTM from primitive descriptor

question

Status: Open.

#4215 In uxlfoundation/oneDNN;

· lacak-sk opened on Oct 27, 2025

BWDW JIT 256 error reproducible through gtests, but not through benchdnn

component:testsCodeowner: @oneapi-src/onednn-archCodeowner: @oneapi-src/onednn-arch

help wanted

platform:cpu-aarch64Codeowner: @oneapi-src/onednn-cpu-aarch64Codeowner: @oneapi-src/onednn-cpu-aarch64

sightingSuspicious library behavior. Should be promoted to a bug when confirmedSuspicious library behavior. Should be promoted to a bug when confirmed

Status: Open.

#4124 In uxlfoundation/oneDNN;

· gassan-arm opened on Oct 13, 2025