René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"[BUG] Matmul with fp16 resulting in some NaN values","articleBody":"The OpenCL backend test for ConvolveStrided in convolve.cpp fails for half (fp16) data type due to NaN appearing in the array which is a result of the matmul operation done upstream. The library used for matrix multiplication is clblast which may have this [related issue](https://github.com/CNugteren/CLBlast/issues/561).\n\nDescription\n===========\n* Built Arrayfire commit eaa49caced87c6eb21d612b1d546ab3061d30a73\n* Which backend is experiencing this issue? OpenCL\n* Do you have a workaround?: Forcing internatlly cast to float before the matrix operation resolves it\n* Can the bug be reproduced reliably on your system? Yes\n\nReproducible Code and/or Steps\n------------------------------\nRun test_convolve_opencl. Reproducible with CPU, Nvidia, and Intel GPUs","author":{"url":"https://github.com/edwinsolisf","@type":"Person","name":"edwinsolisf"},"datePublished":"2025-06-30T00:44:54.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":1},"url":"https://github.com/3674/arrayfire/issues/3674"}