sync : ggml#3127

Merged

ggerganov merged 13 commits intomasterfrom

sync-ggml-25-05-07-2

May 7, 2025

Commits on May 7, 2025

vulkan: Additional type support for unary, binary, and copy (llama/13266)

jeffbolznv
authored and
ggerganov
committed
CUDA: fix race condition in MMQ ids_dst (llama/13294)

JohannesGaessler
authored and
ggerganov
committed
CUDA: fix race condition in MMQ stream-k fixup (llama/13299)

JohannesGaessler
authored and
ggerganov
committed
ggml : activate s390x simd for Q3_K (llama/13301)

taronaeo
authored and
ggerganov
committed
rpc : use backend registry, support dl backends (llama/13304)

slaren
authored and
ggerganov
committed
SYCL: Disable mul_mat kernels for noncontiguous tensor b (llama/13308)

qnixsynapse
authored and
ggerganov
committed
CUDA: fix logic for clearing padding with -ngl 0 (llama/13320)

JohannesGaessler
authored and
ggerganov
committed
CUDA: fix --split-mode row for MMQ (llama/13323)

JohannesGaessler
authored and
ggerganov
committed
CUDA: fix bad asserts for partial offload (llama/13337)

JohannesGaessler
authored and
ggerganov
committed
SYCL: Disable reorder optimize by default and stop setting tensor extras when optimize is disabled (llama/13254)

qnixsynapse
authored and
ggerganov
committed
CUDA: mix virt/real CUDA archs for GGML_NATIVE=OFF (llama/13135)

JohannesGaessler
authored and
ggerganov
committed
cuda : remove nrows_x in mul_mat_q_process_tile (llama/13325)

yeahdongcn
authored and
ggerganov
committed
sync : ggml
ggerganov
committed