ggml/include at master - ggml - Gitea: Git with a cup of tea

mirrors/ggml

0 Fork 0

mirror of https://github.com/ggml-org/ggml.git synced 2026-10-28 16:37:16 +08:00

Files

T

History

1f344416ec cuda, hip: GDN chunked kernel (llama/29353)

* cuda, hip: GDN chunked kernel

Ported-from: https://github.com/tetherto/qvac-fabric-llm.cpp/pull/257
Source-commit: 53f0027d8070715fe9ad2131f3201bb0001d4eed
Co-authored-by: Aman Gupta <aman.gupta@tether.io>

* fix older builds

* review fixes

* cuda: enable GDN MMA on tested Ampere and Ada GPUs

Enable SM86 and SM89 after RTX 3090 and RTX 4090 correctness and prefill benchmarks. Keep other untested architectures on the existing fallback.

Assisted-by: Codex

* cuda : use refactored swizzle API in GDN MMA

* ggml : GDN only guarantees the final state when n_tokens > K

* cuda : add short GDN MMA phase comments

---------

Co-authored-by: gianni-cor <gianfranco.cordella@tether.io>
Co-authored-by: Aman Gupta <aman.gupta@tether.io>

2026-10-15 11:52:11 +03:00

ggml-alloc.h

ggml : add alloc_buffer_n to buffer type interface (llama/23671)

2026-10-04 22:46:07 +03:00

ggml-backend.h

llama : report compute graph reallocations during inference (llama/30329)

2026-10-15 11:52:05 +03:00

ggml-blas.h

ggml : build backends as libraries (llama/10256)

2024-11-15 22:51:53 +02:00

ggml-cann.h

docs : Minor cleanups (llama/19252)