forked from ggml-org/llama.cpp
-
Notifications
You must be signed in to change notification settings - Fork 90
All issues
Issue creation is restricted in this repository
- #54 · khosravipasha opened
on Jul 10, 2026
Issues
is:issue state:open
is:issue state:open
Search results
Metal: no TQ1_0 kernels, and supports_op claims the type anyway
enhancementNew feature or requestNew feature or requestStatus: Open.#141 In PrismML-Eng/llama.cpp;CUDA: TQ1_0 prefill has no MMQ tile loader and falls back to dequantize plus cuBLAS
enhancementNew feature or requestNew feature or requestStatus: Open.#140 In PrismML-Eng/llama.cpp;CUDA: TQ1_0 mat-vec is load-bound by the per-32-element chunk contract
enhancementNew feature or requestNew feature or requestStatus: Open.#139 In PrismML-Eng/llama.cpp;CUDA: native TQ1_0 support (MMVQ decode, MMQ prefill)
enhancementNew feature or requestNew feature or requestStatus: Open.#138 In PrismML-Eng/llama.cpp;- Status: Open.#112 In PrismML-Eng/llama.cpp;
- Status: Open.#110 In PrismML-Eng/llama.cpp;
- Status: Open.#108 In PrismML-Eng/llama.cpp;
- Status: Open.#101 In PrismML-Eng/llama.cpp;
- Status: Open.#99 In PrismML-Eng/llama.cpp;
- Status: Open.#89 In PrismML-Eng/llama.cpp;
- Status: Open.#87 In PrismML-Eng/llama.cpp;
- Status: Open.#85 In PrismML-Eng/llama.cpp;