Piotr Wilkin (ilintar) 16378d93f9 CUDA/HIP: Flash Attention tuning (gfx1201) (#28102) ...

* HIP: enable mma FA for head size 256 on RDNA4, tune configs

Assisted-by: Claude
Assisted-by: Codex

* HIP: prefer whole-tile FA grids over stream-k on AMD WMMA

Assisted-by: Claude
Assisted-by: Codex

* revise stream_k logic

* revise kernel selection logic

---------

Co-authored-by: Johannes Gäßler <johannesg@5d6.de>

2026-09-11 09:58:20 +02:00

..

ggml-blas

llama: add default load-mode auto, which avoids mmap on iGPUs (#26081)

2026-08-11 09:20:46 +03:00

ggml-cann

ggml: add SWIGLU_CLAMP (#27930)

2026-08-30 23:00:02 +08:00

ggml-cpu

ggml-cpu(s390x): add repack support for q4_0 (#28667)

2026-09-10 09:51:17 +03:00

ggml-cuda

CUDA/HIP: Flash Attention tuning (gfx1201) (#28102)

2026-09-11 09:58:20 +02:00

ggml-et

ggml: add SWIGLU_CLAMP (#27930)

2026-08-30 23:00:02 +08:00

ggml-hexagon

hexagon: rope updates (#28628)

2026-09-09 08:40:24 -07:00

ggml-hip

CUDA: replace GGML_FA_ALL_QUANTS with GGML_FA_QUANTS, more control over what is compiled (#28079)

2026-09-09 12:50:08 +02:00

ggml-metal

metal : fix idle threads in mul_mv_iq3_xxs for ne00 < 1024 (#28086)

2026-09-08 15:54:42 +03:00

ggml-musa

CUDA: replace GGML_FA_ALL_QUANTS with GGML_FA_QUANTS, more control over what is compiled (#28079)

2026-09-09 12:50:08 +02:00

ggml-opencl

opencl: add A8 Q4_0 mm binary kernel support (#28268)

2026-09-10 11:25:40 -07:00

ggml-openvino

vulkan: support type-aligned GET_ROWS (#28253)

2026-09-07 12:22:10 +02:00

ggml-rpc

sync : ggml (#28379)

2026-09-04 14:39:19 +03:00

ggml-sycl

Add IQ type handling for MoE (#28476)

2026-09-09 09:25:41 +08:00

ggml-virtgpu

ggml: allow passing alloc dependencies in graph_optimize (#27301)

2026-08-30 11:34:20 +08:00

ggml-vulkan

vulkan: fix data race and OOB access in argsort(large) (#28705)

2026-09-11 08:44:13 +03:00

ggml-webgpu

webgpu: format the GET_ROWS case block (#28542)

2026-09-07 15:55:14 +02:00

ggml-zdnn

llama: add default load-mode auto, which avoids mmap on iGPUs (#26081)

2026-08-11 09:20:46 +03:00

ggml-zendnn

CI: Use LLVM's OpenMP over MSVC_DEBUG_non_redist on Windows (#26678)

2026-08-20 15:42:26 +02:00

CMakeLists.txt

ggml : replace compile definitions with version.h.in (#28364)

2026-09-04 10:28:23 +02:00

ggml-alloc.c

ggml : fix ggml_clamp (#27644)

2026-08-24 10:43:04 +03:00

ggml-backend-dl.cpp

hexagon: enable offloading to Hexagon on Windows on Snapdragon (#19150)

2026-01-29 12:33:21 -08:00

ggml-backend-dl.h

hexagon: enable offloading to Hexagon on Windows on Snapdragon (#19150)

2026-01-29 12:33:21 -08:00

ggml-backend-impl.h

sync : ggml (#28379)

2026-09-04 14:39:19 +03:00

ggml-backend-meta.cpp

ggml-meta: propagate buffer usage and call init on the new tensors (#27586)

2026-08-26 08:27:51 +03:00

ggml-backend-reg.cpp

ggml : don't crash when backend search path can't be read (#28271)

2026-09-04 10:24:06 +03:00

ggml-backend.cpp

ggml: allow backend inputs to not create another split (#28387)

2026-09-07 09:10:40 +03:00

ggml-common.h

AVX2: Speed up large batch size prompt processing of IQ models (#27402)

2026-08-31 14:33:50 -04:00

ggml-feats.h

ggml : fix arm builds, unused var (#26991)

2026-08-13 07:57:24 +03:00

ggml-impl.h

ggml : update ggml_prec specification (#26675)

2026-09-08 09:06:24 +03:00

ggml-opt.cpp

fix: free ctx_copy in ggml_opt_free to plug per-training-session leak (#21592)

2026-04-08 17:40:15 +02:00

ggml-quants.c

Add Q2_0 quantization: type definition and CPU backend (#24448)

2026-07-07 12:05:47 -07:00

ggml-quants.h

Add Q2_0 quantization: type definition and CPU backend (#24448)

2026-07-07 12:05:47 -07:00

ggml-threading.cpp

ggml : build backends as libraries (#10256)

2024-11-14 18:04:35 +01:00

ggml-threading.h

remove CMAKE_WINDOWS_EXPORT_ALL_SYMBOLS (#10797)

2024-12-12 19:02:49 +01:00

ggml-version.h.in

ggml : replace compile definitions with version.h.in (#28364)

2026-09-04 10:28:23 +02:00

ggml.c

ggml : update ggml_prec specification (#26675)

2026-09-08 09:06:24 +03:00

ggml.cpp

ggml : Print backtrace on uncaught C++ exceptions (ggml/1232)

2025-06-01 13:43:57 +03:00

gguf.cpp

llama : add missing headers (#28566)

2026-09-08 12:59:53 +02:00