]> git.djapps.eu Git - pkg/ggml/sources/whisper.cpp/commit
Vulkan: Add DP4A MMQ and Q8_1 quantization shader (llama/12135)
author0cc4m <redacted>
Mon, 31 Mar 2025 12:37:01 +0000 (14:37 +0200)
committerGeorgi Gerganov <redacted>
Wed, 2 Apr 2025 12:51:57 +0000 (15:51 +0300)
commit0810f025477dde29c338a9c7f9e9ceefc71eb62f
tree188576fc20431a0f33ed579c029e7fa5e66d533b
parent8c13c78f9d1b9fbe5dd7997261bc37eac8f9ca85
Vulkan: Add DP4A MMQ and Q8_1 quantization shader (llama/12135)

* Vulkan: Add DP4A MMQ and Q8_1 quantization shader

* Add q4_0 x q8_1 matrix matrix multiplication support

* Vulkan: Add int8 coopmat MMQ support

* Vulkan: Add q4_1, q5_0 and q5_1 quants, improve integer dot code

* Add GL_EXT_integer_dot_product check

* Remove ggml changes, fix mmq pipeline picker

* Remove ggml changes, restore Intel coopmat behaviour

* Fix glsl compile attempt when integer vec dot is not supported

* Remove redundant code, use non-saturating integer dot, enable all matmul sizes for mmq

* Remove redundant comment

* Fix integer dot check

* Fix compile issue with unsupported int dot glslc

* Update Windows build Vulkan SDK version
ggml/src/ggml-vulkan/CMakeLists.txt
ggml/src/ggml-vulkan/ggml-vulkan.cpp
ggml/src/ggml-vulkan/vulkan-shaders/mul_mm.comp
ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq.comp [new file with mode: 0644]
ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq_funcs.comp [new file with mode: 0644]
ggml/src/ggml-vulkan/vulkan-shaders/quantize_q8_1.comp [new file with mode: 0644]
ggml/src/ggml-vulkan/vulkan-shaders/test_integer_dot_support.comp [new file with mode: 0644]
ggml/src/ggml-vulkan/vulkan-shaders/types.comp
ggml/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp