nikola/llama_cpp
1
0
Fork 0
Code Issues Pull Requests Actions 11 Packages Projects Releases Wiki Activity

hexagon: ssm-conv fix for large prompts (#23307)

Browse Source 
* hexagon: remove gathers and better handling of vtcm in ssm-conv

* hexagon: relax ssm-conv gating requirements

* hexagon: add new prefill ssm-conv backend test

* hexagon: remove trailing white space

* hex-rope: uninline rope_cache_init, otherwise it breaks after rebaseing with SSM_CONV changes

---------

Co-authored-by: Max Krasnyansky <maxk@qti.qualcomm.com>
This commit is contained in:
Todor Boinovski
2026-05-20 22:14:13 -07:00
committed by GitHub
co-authored by Max Krasnyansky
parent ce02093fdd
commit 0be84685bd
4 changed files with 252 additions and 158 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/test-backend-ops.cpp
+1
View File
@@ -9337,6 +9337,7 @@ static std::vector<std::unique_ptr<test_case>> make_test_cases_perf() {
// Examples from granite-4.0-h-1b/ggml-model-Q8_0.gguf
test_cases.emplace_back(new test_ssm_conv(GGML_TYPE_F32, {515, 3328, 1, 1}, {4, 3328, 1, 1})); // prefill
test_cases.emplace_back(new test_ssm_conv(GGML_TYPE_F32, {937, 8192, 1, 1}, {4, 8192, 1, 1})); // prefill
test_cases.emplace_back(new test_ssm_conv(GGML_TYPE_F32, {4, 3328, 1, 1}, {4, 3328, 1, 1})); // generate
test_cases.emplace_back(new test_ssm_conv_bias_silu(GGML_TYPE_F32, {515, 3328, 1, 1}, {4, 3328, 1, 1}, true)); // prefill
test_cases.emplace_back(new test_ssm_conv_bias_silu(GGML_TYPE_F32, {4, 3328, 1, 1}, {4, 3328, 1, 1}, true)); // generate