Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Feature: Added LIGHTNING_INDEXER support for Deepseek V4 ops on Vulkan Backend (#27453)

Browse Source 
* vulkan: add LIGHTNING_INDEXER op

* vulkan: updated lightning_indexer.comp and ggml-vulkan.cpp with 128-lane dot-product reduction moved from a shared-memory tree to subgroupAdd.

* vulkan: cleanup; Skip bounds checks

* vulkan: cleanup FA_K_ONLY

* Revert "vulkan: cleanup FA_K_ONLY"

This reverts commit fdcbdd91511945d6878d9070b5447e4a34dce010.

* vulkan: restore interleaved K/V buffer ordering

* vulkan: Remove FA_K_ONLY

* vulkan: Revert flash_attn_dequant

* vulkan: Revert tests in backend-ops.cpp
This commit is contained in:
Shobhit
2026-08-27 15:34:42 +02:00
committed by GitHub
parent 1a946ec745
commit cb300598d5
5 changed files with 365 additions and 54 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp
+6
View File
@@ -1069,6 +1069,12 @@ void process_shaders() {
string_to_spv("gated_linear_attn_f32", "gla.comp", merge_maps(base_dict, {{"A_TYPE", "float"}}));
// Compile IQ4_NL support in so its shared LUT is available when K uses it.
// K quant type is selected at runtime via the FaTypeK spec constant.
std::map<std::string, std::string> li_dict = {{"FLOAT_TYPE", "float"}, {"FLOAT_TYPEV4", "vec4"}, {"DATA_A_IQ4_NL", "1"}};
string_to_spv("lightning_indexer_f32", "lightning_indexer.comp", li_dict);
string_to_spv("lightning_indexer_subgroup_f32", "lightning_indexer.comp", merge_maps(li_dict, {{"USE_SUBGROUP_ADD", "1"}}));
string_to_spv("rwkv_wkv7_f32", "wkv7.comp", merge_maps(base_dict, {{"A_TYPE", "float"}}));
string_to_spv("gated_delta_net_f32", "gated_delta_net.comp", merge_maps(base_dict, {{"FLOAT_TYPE", "float"}, {"USE_SUBGROUP_ADD", "1"}, {"USE_SUBGROUP_CLUSTERED", "1"}}));