Optimize Vulkan REPEAT performance · SkutteOleg/stable-diffusion.cpp@4b26f15

Commit

Optimize Vulkan REPEAT performance

Co-Authored-By: 0cc4m <[email protected]>

SkutteOleg and 0cc4m committed Aug 23, 2024

1 parent e278f68 commit 4b26f15

+18 −0		include/ggml.h
+1 −1		scripts/sync-llama.last
+1 −1		scripts/sync-whisper.last
+8 −0		src/ggml-cuda.cu
+56 −0		src/ggml-cuda/unary.cu
+6 −0		src/ggml-cuda/unary.cuh
+37 −2		src/ggml-metal.m
+14 −0		src/ggml-metal.metal
+35 −1		src/ggml-rpc.cpp
+92 −100		src/ggml-vulkan.cpp
+247 −34		src/ggml.c
+15 −0		src/vulkan-shaders/cos.comp
+24 −0		src/vulkan-shaders/repeat.comp
+15 −0		src/vulkan-shaders/sin.comp
+12 −0		src/vulkan-shaders/vulkan-shaders-gen.cpp
+54 −0		tests/test-backend-ops.cpp