Attila T. Áfra
Attila T. Áfra
`enable_language(HIP)` in CMake 3.22.3 and 3.23.0-rc3 fails when linking the test program using ROCm 5.0.2 on Ubuntu 20.04: ``` -- The HIP compiler identification is Clang 14.0.0 -- Detecting HIP...
ISPC can be used to implement very efficient convolution / machine learning kernels but unfortunately it's not possible to use AI acceleration instruction sets like VNNI. Unlike AMX, adding support...
Having support for type template arguments is very useful but it's often insufficient. In many cases, non-type arguments, e.g. int or enum, are also very important. A typical use case...
ISPC 1.21.x generates illegal instruction(s) when building Intel Open Image Denoise on macOS for ARM64. This is a regression: prior versions, e.g. ISPC 1.20.0, generate correct code. Steps to reproduce...
After upgrading to Xcode 15, linking ISPC objects results in the following linker error for each ISPC object: > ld: warning: no platform load command found in 'oidn/build/devices/cpu/local__/color.dev.o', assuming: macOS...
XDL kernels are instantiated with various different BlockSize/MPerBlock/NPerBlock/etc. template parameters in the library, so it's easy to pick a good set of parameters for a particular convolution. But unfortunately the...