Issues

• benchdnn matmul silently ignores --kernel_name (always runs AOCL-DLP)

  amd/ZenDNN#36 · Filter by author lwandrebeck opened on Aug 13, 2026

• Cost model assumes a 32 MB L3 on CPUs that have no L3

  amd/ZenDNN#37 · Filter by author lwandrebeck opened on Aug 13, 2026

• auto_tuner offers AOCL-DLP candidates it cannot run, and caches non-computing backends as fastest

  amd/ZenDNN#35 · Filter by author lwandrebeck opened on Aug 13, 2026

• scalar_microkernel treats beta as a flag rather than a scale (wrong results for beta not in {0,1})

  amd/ZenDNN#34 · Filter by author lwandrebeck opened on Aug 13, 2026

• Uncaught dnnl::error terminates the process when oneDNN cannot serve a dtype

  amd/ZenDNN#33 · Filter by author lwandrebeck opened on Aug 13, 2026

• Unguarded AVX-512 dispatch (SIGILL) in three more places besides RMS_NORM (#23)

  amd/ZenDNN#32 · Filter by author lwandrebeck opened on Aug 13, 2026

• Flash SDPA silently returns zeros on any --no-aocldlp build (hardcoded GEMM backend + two LIBXSMM gates)

  amd/ZenDNN#31 · Filter by author lwandrebeck opened on Aug 13, 2026

• Post-op chains silently reordered when the fused op is not first in the chain

  amd/ZenDNN#30 · Filter by author lwandrebeck opened on Aug 13, 2026

• Native matmul hands BF16/INT8 problems to the FP32 kernel (heap corruption, silent wrong results)

  amd/ZenDNN#29 · Filter by author lwandrebeck opened on Aug 13, 2026

• Build with --no-aocldlp fails to link: missing W4A8 stubs

  amd/ZenDNN#28 · Filter by author lwandrebeck opened on Aug 13, 2026

• Build failure with gcc 15.2, error: ‘acc’ may be used uninitialized

  amd/ZenDNN#25 · Filter by author krampenschiesser opened on May 8, 2026

• RMS_NORM dispatcher unconditionally calls AVX-512 kernel, SIGILL on Zen3

  amd/ZenDNN#23 · Filter by author bghimireamd opened on Apr 19, 2026
  1

• link with system library

  amd/ZenDNN#18 · Filter by author rhjdvsgsgks opened on Dec 7, 2025

• Broken (Non-working) Composable Kernel Library Installation

  amd/ZenDNN#16 · Filter by author Qubitium opened on Nov 10, 2025

• ## [Bug] Build fails due to missing source files and linking errors with custom ops in ZenDNN v5.0.2 / main

  amd/ZenDNN#14 · Filter by author kksspoi opened on May 15, 2025
  1

• Support for Consumer Grade CPUs?

  amd/ZenDNN#13 · Filter by author LuciAkirami opened on Mar 6, 2025
  11

• Severe lack of information on how to build or use ZenDNN

  amd/ZenDNN#12 · Filter by author teodorkostov opened on Feb 7, 2025
  2

• Support for 3D convolutions?

  amd/ZenDNN#11 · Filter by author stefanozampini opened on Sep 11, 2024

• Support Pytorch with python v3.11

  amd/ZenDNN#10 · Filter by author ZoeDong opened on Mar 18, 2024

• Clarification on ZenDNN tuning for tensorflow

  amd/ZenDNN#9 · Filter by author sagi-scalers-ai opened on Sep 20, 2023
  5

• Support for Pytorch 2.x?

  amd/ZenDNN#8 · Filter by author obsolete-scai-account opened on May 17, 2023

• Can I build PT_v1.12_ZenDNN_v4.0_source_code to obtain a pytorch with both zendnn cpu capability and cuda GPU capability at the same time?

  amd/ZenDNN#6 · Filter by author chencuber opened on Apr 11, 2023

• DLRM suggest benchmark

  amd/ZenDNN#4 · Filter by author JunxiChhen opened on Jul 12, 2022
  8

  

• Issue while setting up ZenDNN in conda container

  question
  amd/ZenDNN#2 · Filter by author Lavanya-Kampara opened on Sep 23, 2021
  1

• about zenMatMulSplit split strategy

  question
  amd/ZenDNN#1 · Filter by author MichoChan opened on Aug 17, 2021
  1