sleepy/llama.cpp
1
0
Fork 0
Code Issues 11 Pull Requests Actions 108 Packages Projects Releases Wiki Activity

CUDA : Fix CUB's argsort when nrows % block_size == 0 CCCL < 3.1 (#21181)

Browse Source 
* CUDA: Fix CUB's argsort when nrows % block_size == 0 CCCL < 3.1

We wrongly calculated offset_grid as `ceildiv(nrows, block_size)`,
while it must be `ceildiv(nrows + 1, block_size)`. As a consequence, we
had uninitialized values in `offset_iterator[nrows]` for the case when
`nrows % block_size == 0`.

Fixes #21162

* Reduce nrows in test case to 256, don't need 768
This commit is contained in:
Oliver Simons
2026-03-30 16:20:00 +02:00
committed by GitHub
parent cad2d3884c
commit 64ac9ab66a
2 changed files with 5 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/test-backend-ops.cpp
+1
View File
@@ -8424,6 +8424,7 @@ static std::vector<std::unique_ptr<test_case>> make_test_cases_eval() {
test_cases.emplace_back(new test_argsort(GGML_TYPE_F32, {1023, 2, 1, 3}, order));
test_cases.emplace_back(new test_argsort(GGML_TYPE_F32, {1024, 2, 1, 3}, order));
test_cases.emplace_back(new test_argsort(GGML_TYPE_F32, {1025, 2, 1, 3}, order));
test_cases.emplace_back(new test_argsort(GGML_TYPE_F32, {1025, 256, 1, 1}, order)); // test ceildiv in CUDA's CUB's DeviceSegmentedSort
test_cases.emplace_back(new test_argsort(GGML_TYPE_F32, {2047, 2, 1, 3}, order));
test_cases.emplace_back(new test_argsort(GGML_TYPE_F32, {2048, 2, 1, 3}, order));
test_cases.emplace_back(new test_argsort(GGML_TYPE_F32, {2049, 2, 1, 3}, order));