diff options
author | Johannes Gäßler <johannesg@5d6.de> | 2023-08-09 09:42:34 +0200 |
---|---|---|
committer | GitHub <noreply@github.com> | 2023-08-09 09:42:34 +0200 |
commit | 25d43e0eb578b6e73046d9d6644a3a14d460600d (patch) | |
tree | fddb8e9a044ce7eda09024e345a871cdada4cac8 /README.md | |
parent | f5bfea0580e417f99850d5456ca541d871a3e48c (diff) |
CUDA: tuned mul_mat_q kernels (#2546)
Diffstat (limited to 'README.md')
-rw-r--r-- | README.md | 1 |
1 files changed, 0 insertions, 1 deletions
@@ -406,7 +406,6 @@ Building the program with BLAS support may lead to some performance improvements ---> | Option | Legal values | Default | Description | |-------------------------|------------------------|---------|-------------| - | LLAMA_CUDA_MMQ_Y | Positive integer >= 32 | 64 | Tile size in y direction when using the custom CUDA kernels for prompt processing. Higher values can be faster depending on the amount of shared memory available. Power of 2 heavily recommended. | | LLAMA_CUDA_FORCE_DMMV | Boolean | false | Force the use of dequantization + matrix vector multiplication kernels instead of using kernels that do matrix vector multiplication on quantized data. By default the decision is made based on compute capability (MMVQ for 6.1/Pascal/GTX 1000 or higher). Does not affect k-quants. | | LLAMA_CUDA_DMMV_X | Positive integer >= 32 | 32 | Number of values in x direction processed by the CUDA dequantization + matrix vector multiplication kernel per iteration. Increasing this value can improve performance on fast GPUs. Power of 2 heavily recommended. Does not affect k-quants. | | LLAMA_CUDA_MMV_Y | Positive integer | 1 | Block size in y direction for the CUDA mul mat vec kernels. Increasing this value can improve performance on fast GPUs. Power of 2 recommended. Does not affect k-quants. | |