]> git.djapps.eu Git - pkg/ggml/sources/whisper.cpp/commit
vulkan: Support GGML_TYPE_NVFP4 (llama/21455)
authorJeff Bolz <redacted>
Tue, 14 Apr 2026 09:34:23 +0000 (11:34 +0200)
committerGeorgi Gerganov <redacted>
Thu, 30 Apr 2026 08:29:07 +0000 (11:29 +0300)
commitcdeaa341742c4d558d7020079ef0e282803511a6
tree1424cd00c6f5c08087ddc298fecd77bfb9369d0c
parent0f99a47177a887b3771c876fcd5b8de4711c9fe3
vulkan: Support GGML_TYPE_NVFP4 (llama/21455)

This adds nvfp4 support for get_rows, dequant, and mul_mat(_id). For
mul_mat, it does not add support for the dp4/q8_1 path, it's all via
fp16/fp32.
ggml/src/ggml-vulkan/ggml-vulkan.cpp
ggml/src/ggml-vulkan/vulkan-shaders/copy_from_quant.comp
ggml/src/ggml-vulkan/vulkan-shaders/dequant_funcs.glsl
ggml/src/ggml-vulkan/vulkan-shaders/dequant_funcs_cm2.glsl
ggml/src/ggml-vulkan/vulkan-shaders/dequant_nvfp4.comp [new file with mode: 0644]
ggml/src/ggml-vulkan/vulkan-shaders/mul_mm_funcs.glsl
ggml/src/ggml-vulkan/vulkan-shaders/types.glsl
ggml/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp