public_git
/
whisper.cpp
mirror of https://github.com/ggml-org/whisper.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
whisper.cpp/ggml/src
History
Jeff Bolz cea22b3075 vulkan: For coopmat2 FA, use fp16 accumulators for the final result (llama/19376) 
The cpu and cuda backends use fp16 for the VKQ accumulator type, this change
does the same for vulkan. This helps particularly with large head sizes which
are very register-limited.

I tried this for the coopmat1 path and it slowed down a bit. I didn't try for
scalar.

I applied the softmax bias that the cuda backend uses to avoid overflow,
although I was not able to reproduce the original bug without it.
 		2026-02-08 09:29:10 +02:00
..
ggml-blas ggml : add ggml_build_forward_select (llama/18550) 2026-01-30 15:56:40 +02:00
ggml-cann docs : Minor cleanups (llama/19252) 2026-02-08 09:29:10 +02:00
ggml-cpu ggml-cpu: use LUT for converting e8->f32 scales on x86 (llama/19288) 2026-02-08 09:29:10 +02:00
ggml-cuda cuda : cuda graphs now compare all node params (llama/19383) 2026-02-08 09:29:10 +02:00
ggml-hexagon ggml-hexagon: flash-attention and reduce-sum optimizations (llama/19141) 2026-02-08 09:29:10 +02:00
ggml-hip HIP: add mmf for CDNA (llama/18896) 2026-01-30 15:56:40 +02:00
ggml-metal metal : skip loading all-zero mask (llama/19337) 2026-02-08 09:29:10 +02:00
ggml-musa CUDA: faster tile FA, add oob checks, more HSs (llama/16492) 2025-10-15 09:29:17 +03:00
ggml-opencl opencl: refactor some ops, concat, repeat, tanh and scale (llama/19226) 2026-02-08 09:29:10 +02:00
ggml-rpc rpc : use unordered_map::reserve and emplace (llama/18513) 2026-01-14 09:11:59 +02:00
ggml-sycl Remove support for Nvidia & AMD GPU, because the oneAPI plugin for Nvidia & AMD GPU is unavailable: download/installation channels are out of work. (llama/19246) 2026-02-08 09:29:10 +02:00
ggml-virtgpu ggml-virtgpu: make the code thread safe (llama/19204) 2026-02-08 09:29:10 +02:00
ggml-vulkan vulkan: For coopmat2 FA, use fp16 accumulators for the final result (llama/19376) 2026-02-08 09:29:10 +02:00
ggml-webgpu Remove pipeline cache mutexes (llama/19195) 2026-02-08 09:29:10 +02:00
ggml-zdnn ggml-zdnn : mark zDNN buffers as non-host (llama/18967) 2026-01-30 15:56:40 +02:00
ggml-zendnn ggml-zendnn : resolve ZenDNN backend cross-module symbol dependency (llama/19159) 2026-01-30 15:56:40 +02:00
CMakeLists.txt hexagon: enable offloading to Hexagon on Windows on Snapdragon (llama/19150) 2026-01-30 15:56:40 +02:00
ggml-alloc.c llama: automatically set parameters not set by the user in such a way that maximizes GPU utilization (llama/16653) 2025-12-18 08:20:56 +02:00
ggml-backend-dl.cpp hexagon: enable offloading to Hexagon on Windows on Snapdragon (llama/19150) 2026-01-30 15:56:40 +02:00
ggml-backend-dl.h hexagon: enable offloading to Hexagon on Windows on Snapdragon (llama/19150) 2026-01-30 15:56:40 +02:00
ggml-backend-impl.h llama: use host memory if device reports 0 memory (llama/18587) 2026-01-14 09:11:59 +02:00
ggml-backend-reg.cpp hexagon: enable offloading to Hexagon on Windows on Snapdragon (llama/19150) 2026-01-30 15:56:40 +02:00
ggml-backend.cpp ggml-backend: fix async set/get fallback sync (llama/19179) 2026-02-08 09:29:10 +02:00
ggml-common.h llama : add gpt-oss (llama/15091) 2025-08-18 20:30:45 +03:00
ggml-impl.h ggml : add ggml_build_forward_select (llama/18550) 2026-01-30 15:56:40 +02:00
ggml-opt.cpp finetune: SGD optimizer, more CLI args (llama/13873) 2025-08-18 20:30:45 +03:00
ggml-quants.c ggml : fix uninitialized is_on_grid in quantize_row_iq3_xxs_impl (llama/15928) 2025-09-29 15:18:09 +03:00
ggml-quants.h llama : add gpt-oss (llama/15091) 2025-08-18 20:30:45 +03:00
ggml-threading.cpp
…
ggml-threading.h
…
ggml.c ggml: added cleanups in ggml_quantize_free (llama/19278) 2026-02-08 09:29:10 +02:00
ggml.cpp
…
gguf.cpp
…