Repository navigation

All pull requests

Labels (10)Milestones (0)New pull request

Open

(35)

Closed

(168)

hip: fuse Gemma scalar and GELU multiplications on RDNA3.5
#209liangliangchang opened Nov 13, 2026
hip: fuse Gemma vision RoPE and enable D512 WMMA attention
#208liangliangchang opened Nov 13, 2026
mtmd: cast vision K/V before attention permutation
#207liangliangchang opened Nov 13, 2026
hip: reduce tensor traffic and tune small RoPE launches
#206liangliangchang opened Nov 13, 2026
hip: RDNA3.5 MMDQ prefill and quantized FFN fusions
#205liangliangchang opened Nov 13, 2026
hip: add dedicated prefill and vision attention with mask classification
#204liangliangchang opened Nov 12, 2026
Sync master with upstream (2026-11-12)
#202liangliangchang opened Nov 12, 2026
ggml-cuda: fuse FP16/BF16 BLAS producers and outputs on RDNA3.5
#201liangliangchang opened Nov 12, 2026
ggml-hrx: extract Q8 prefill and GPU repeat optimizations
#199AaronStGeorge opened Nov 12, 2026Collaborator
ggml-hrx: add native Q2_K dequantization
#197AaronStGeorge opened Nov 11, 2026Collaborator
hrx: prefer row64 weights for dense Qwen3
#196AaronStGeorge opened Nov 11, 2026Collaborator
hip: accelerate D64/D72/D80 vision attention with WMMA
#195liangliangchang opened Nov 11, 2026
mtmd: batch Qwen2.5-VL windows and avoid extra copies
#192liangliangchang opened Nov 10, 2026
ggml-hrx: allow caller-owned RMSNorm loop unrolling
#184AaronStGeorge opened Nov 8, 2026Collaborator
[hrx/gfx1151] DSV4.0 Flash optimizations
#170jammm opened Nov 1, 2026
Adapt HRX kernel corpus to typed vector splats
#169AaronStGeorge opened Oct 31, 2026Collaborator
Support Qwen3.8 vision and text graphs on HRX
#168AaronStGeorge opened Oct 31, 2026Collaborator
2
[hrx-llama] update compatability floor
#166AaronStGeorge opened Oct 30, 2026Collaborator
llama-server: reset roofline profiler after warmup
#159amd-callumm opened Oct 28, 2026
ggml-hrx: update upstream and enable Flint through 32K context
#158AaronStGeorge opened Oct 27, 2026Collaborator1 task
rdna35 prefill optimizations other than mmq
#129liangliangchang opened Oct 1, 2026
ggml-hrx: fix empty views and improve long-context decode
#114jammm opened Sep 23, 2026
Fix hang on exit and build issues on Windows
#105pschlan-amd opened Sep 18, 2026
cuda/hip: optional wave64 for the quantized mat-vec kernels on RDNA
#89mgehre-amd opened Aug 17, 2026Collaborator
q4_K mat-vec tuning for gfx1151
#84mgehre-amd opened Aug 13, 2026Collaborator

Pull requests