BA8F0D39
BA8F0D39
@9prady9  https://www.khronos.org/opencl/ Khronos Group says that OpenCL kernels can run be compiled to the intermediate language SPIR-V. SPIR-V can run on the Vulkan Compute runtime....
For k < 256 and input array > 100000000, arrayfire crashes with CUDA illegal instruction. Seems topk is unstable
It doesn't seem to insert the values into the array correctly for u64 ``` let A_dims = arrayfire::Dim4::new(&[6,1,1,1]); let mut A = arrayfire::randu::(A_dims); let bool_cpu: Vec = vec![true, true, false,...
@9prady9 Ok, I'll close this
@syurkevi Did you manage to replicate it?
@syurkevi Some functions already take in the index as input af_where (af_array *idx, const af_array in) Can't you detect if idx is 32bit or 64bit and use the appropriate kernel?...
Could we also update the half 1.5.0 to half 2.2.0 in arrayfire-rust? to get new features
@moyang RTX 3090 has native FP16 support in tensor cores https://www.nvidia.com/content/PDF/nvidia-ampere-ga-102-gpu-architecture-whitepaper-v2.pdf 512 FP16 FMA per SM 128 FP16 FMA per Tensor core RTX 3090 has 82 SM and 328 Tensor...
Moonlight Sway Wayland on the client Sunshine 1.60 on the windows host When I plug a mouse into the host there is a cursor When I remove the mouse from...
@arthuryuan1987 I am on Arch Linux Kernel 6.2rc8 with A770 16 GB. Not on Windows. Not on WSL Allocating less than 8GB on A770 16GB works. Allocating more than 8GB...