Search results

HIP: adjust RDNA3.5 MMQ kernel selction logic

ggmlchanges relating to the ggml tensor library for machine learningchanges relating to the ggml tensor library for machine learning

Nvidia GPUIssues specific to Nvidia GPUsIssues specific to Nvidia GPUs

Status: Open (in progress).

ggml-org/llama.cppnumber 18666

#18666 In ggml-org/llama.cpp;

· JohannesGaessler opened on Jan 7, 2026

cuda : check src shapes for CUDA graphs

ggmlchanges relating to the ggml tensor library for machine learningchanges relating to the ggml tensor library for machine learning

Nvidia GPUIssues specific to Nvidia GPUsIssues specific to Nvidia GPUs

Status: Draft (not ready).

ggml-org/llama.cppnumber 18561

#18561 In ggml-org/llama.cpp;

· ggerganov opened on Jan 3, 2026

ggml : add ggml_build_forward_select

Apple Metalhttps://en.wikipedia.org/wiki/Metal_(API)https://en.wikipedia.org/wiki/Metal_(API)

Ascend NPUissues specific to Ascend NPUsissues specific to Ascend NPUs

ggmlchanges relating to the ggml tensor library for machine learningchanges relating to the ggml tensor library for machine learning

IBM zDNNissues specific to IBM zDNN Acceleratorissues specific to IBM zDNN Accelerator

modelModel specificModel specific

Nvidia GPUIssues specific to Nvidia GPUsIssues specific to Nvidia GPUs

OpenCLIssues specific to the OpenCL backendIssues specific to the OpenCL backend

SYCLhttps://en.wikipedia.org/wiki/SYCL - GPU programming languagehttps://en.wikipedia.org/wiki/SYCL - GPU programming language

VulkanIssues specific to the Vulkan backendIssues specific to the Vulkan backend

Status: Open (in progress).

ggml-org/llama.cppnumber 18550

#18550 In ggml-org/llama.cpp;

· ggerganov opened on Jan 2, 2026

CUDA: cache intermediate tensors

ggmlchanges relating to the ggml tensor library for machine learningchanges relating to the ggml tensor library for machine learning

Nvidia GPUIssues specific to Nvidia GPUsIssues specific to Nvidia GPUs

Status: Open (in progress).

ggml-org/llama.cppnumber 18538

#18538 In ggml-org/llama.cpp;

· am17an opened on Jan 2, 2026

ggml: add env var GGML_OP_OFFLOAD_MIN_BATCH

Apple Metalhttps://en.wikipedia.org/wiki/Metal_(API)https://en.wikipedia.org/wiki/Metal_(API)

Ascend NPUissues specific to Ascend NPUsissues specific to Ascend NPUs

ggmlchanges relating to the ggml tensor library for machine learningchanges relating to the ggml tensor library for machine learning

Nvidia GPUIssues specific to Nvidia GPUsIssues specific to Nvidia GPUs

SYCLhttps://en.wikipedia.org/wiki/SYCL - GPU programming languagehttps://en.wikipedia.org/wiki/SYCL - GPU programming language

VulkanIssues specific to the Vulkan backendIssues specific to the Vulkan backend

Status: Open (in progress).

ggml-org/llama.cppnumber 18535

#18535 In ggml-org/llama.cpp;

· DocShotgun opened on Jan 2, 2026

HIP: add fattn-mma-f16 for RDNA4

ggmlchanges relating to the ggml tensor library for machine learningchanges relating to the ggml tensor library for machine learning

Nvidia GPUIssues specific to Nvidia GPUsIssues specific to Nvidia GPUs

Status: Open (in progress).

ggml-org/llama.cppnumber 18481

#18481 In ggml-org/llama.cpp;

· zhang-hui-yulo opened on Dec 30, 2025

CUDA: fix 0.0f/0.0f for FA fixup

ggmlchanges relating to the ggml tensor library for machine learningchanges relating to the ggml tensor library for machine learning

Nvidia GPUIssues specific to Nvidia GPUsIssues specific to Nvidia GPUs

Status: Open (in progress).

ggml-org/llama.cppnumber 18472

#18472 In ggml-org/llama.cpp;

· JohannesGaessler opened on Dec 29, 2025

Compile bug: CUDA MUL_MAT crash on Blackwell (sm_121) - nvcc O3 optimization bug

bug-unconfirmed

CUDARelated to the CUDA backendRelated to the CUDA backend

Nvidia GPUIssues specific to Nvidia GPUsIssues specific to Nvidia GPUs

Status: Open.

#18331 In ggml-org/llama.cpp;

· AmesianX opened on Dec 23, 2025

KYLIN: fix compile error for cuda backend

ggmlchanges relating to the ggml tensor library for machine learningchanges relating to the ggml tensor library for machine learning

Nvidia GPUIssues specific to Nvidia GPUsIssues specific to Nvidia GPUs

Status: Open (in progress).

ggml-org/llama.cppnumber 18275

#18275 In ggml-org/llama.cpp;

· lizhenneng opened on Dec 22, 2025

Eval bug: Major performance drop since b7406

bug-unconfirmed

CUDARelated to the CUDA backendRelated to the CUDA backend

Nvidia GPUIssues specific to Nvidia GPUsIssues specific to Nvidia GPUs

performanceSpeed related topicsSpeed related topics

regressionA regression introduced in a new build (something that was previously working correctly)A regression introduced in a new build (something that was previously working correctly)

windowsIssues specific to WindowsIssues specific to Windows

Status: Open.

#18258 In ggml-org/llama.cpp;

· Kaspur2012 opened on Dec 21, 2025

ggml-cuda: Delta-Net linear attention for Qwen3-Next

ggmlchanges relating to the ggml tensor library for machine learningchanges relating to the ggml tensor library for machine learning

modelModel specificModel specific

Nvidia GPUIssues specific to Nvidia GPUsIssues specific to Nvidia GPUs

testingEverything test relatedEverything test related

Status: Open (in progress).

ggml-org/llama.cppnumber 18102

#18102 In ggml-org/llama.cpp;

· hauhaut opened on Dec 16, 2025

CUDA: Improve performance via less synchronizations between token

ggmlchanges relating to the ggml tensor library for machine learningchanges relating to the ggml tensor library for machine learning

Nvidia GPUIssues specific to Nvidia GPUsIssues specific to Nvidia GPUs

Status: Draft (not ready).

ggml-org/llama.cppnumber 17795

#17795 In ggml-org/llama.cpp;

· aendk opened on Dec 5, 2025