llama.cpp
ggml: adds CONV_2D op and direct GEMM Vulkan implementation
#14316

Merged

ggml: adds CONV_2D op and direct GEMM Vulkan implementation #14316

0cc4m merged 10 commits into ggml-org:master from etasnadi:ggml/conv_2d

github-actions added testing

github-actions added Vulkan

github-actions added ggml

jeffbolznv commented on 2025-06-21

jeffbolznv commented on 2025-06-21

ggml/ggml-vulkan/test-backend-ops: adds CONV_2D for Vulkan

720b483e

* Performance fixes: minimized branch divergence, uses collectives to

0715985e

etasnadi force pushed from beff9ca8 to 0715985e 160 days ago

* Apple/Win32 compile errors fixed

7f9b6591

Collectives disabled by default.

a09e8f59

Intel support is disabled as the performance is poor.

aa472583

Conv2d enabled for Intel with disabled collectives, disabled for Apple

a672803d

slaren commented on 2025-07-16

test-backend-ops modifications are reverted

e50711f1

Trailing spaces and missing override fixed.

f8295bfc

Triggering pipeline relaunch.

8ddaf2d5

0cc4m approved these changes on 2025-07-19

Code formatted with .clang-format.

b4acacc2

0cc4m merged a979ca22 into master 153 days ago

Reviewers

0cc4m

slaren

am17an

jeffbolznv

Assignees

No one assigned

Labels

testing Vulkan ggml

Milestone

No milestone