oneDNN icon indicating copy to clipboard operation
oneDNN copied to clipboard

xe: softmax: restore missing inf_as_zero functionality

Open pv-pterab-s opened this issue 7 months ago • 3 comments

PLEASE CODE REVIEW, BUT TESTING WAITING ON CI

This PR addresses the issue that PR #2525 was missing inf_as_zero functionality in the OpenCL softmax kernels (MFDNN-13724).

pv-pterab-s avatar May 22 '25 20:05 pv-pterab-s

make test disable os_win disable test_device_cpu enable test_device_gpu disable build_cpu_runtime_omp disable build_cpu_runtime_sycl disable build_cpu_runtime_tbb enable build_graph disable compiler_icx-previous disable compiler_gnu9 disable compiler_clang14 disable compiler_vs2022 disable build_gpu_runtime_sycl disable benchdnn_all enable benchdnn_softmax enable benchdnn_graph enable arch_gpu_xe-hpc enable arch_gpu_xe-hpg-atsm enable arch_gpu_xe-hpg-dg2 disable arch_gpu_xe-lp disable arch_gpu_xe-lpg disable arch_gpu_xe-lpg+ enable arch_gpu_xe2-hpg-bmg disable arch_gpu_xe2-lpg

pv-pterab-s avatar May 22 '25 21:05 pv-pterab-s

Still waiting on CI to pass this until getting this merged...

pv-pterab-s avatar Jun 13 '25 17:06 pv-pterab-s

make test disable os_win disable test_device_cpu enable test_device_gpu disable build_cpu_runtime_omp disable build_cpu_runtime_sycl disable build_cpu_runtime_tbb enable build_graph disable compiler_icx-previous disable compiler_gnu9 disable compiler_clang14 disable compiler_vs2022 disable build_gpu_runtime_sycl disable benchdnn_all enable benchdnn_softmax enable benchdnn_graph enable arch_gpu_xe-hpc enable arch_gpu_xe-hpg-atsm enable arch_gpu_xe-hpg-dg2 disable arch_gpu_xe-lp disable arch_gpu_xe-lpg disable arch_gpu_xe-lpg+ enable arch_gpu_xe2-hpg-bmg disable arch_gpu_xe2-lpg

pv-pterab-s avatar Jun 16 '25 13:06 pv-pterab-s

make test disable test_device_cpu enable test_device_gpu disable build_cpu_runtime_omp disable build_cpu_runtime_sycl disable build_cpu_runtime_tbb enable build_graph disable benchdnn_all enable benchdnn_softmax enable benchdnn_graph enable arch_gpu_xe-hpc enable arch_gpu_xe-hpg-atsm enable arch_gpu_xe-hpg-dg2 disable arch_gpu_xe-lp disable arch_gpu_xe-lpg disable arch_gpu_xe-lpg+ enable arch_gpu_xe2-hpg-bmg disable arch_gpu_xe2-lpg

umar456 avatar Jun 17 '25 17:06 umar456