mikhail/llama.cpp-mtp-turboquant
1
0
Fork 0
Code Issues Pull Requests Actions 34 Packages Projects Releases Wiki Activity

cuda : support non-contiguous i32 to i32 copy (#17326)

Browse Source 
* support non-contiguous i32 to i32 copy

* add tests

* rename cpy_flt to cpy_scalar and reindent params
This commit is contained in:
Sigbjørn Skjæret
2025-11-23 11:13:34 +01:00
committed by GitHub
parent bc809e9c53
commit 96ac5a2329
4 changed files with 92 additions and 49 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/test-backend-ops.cpp
+3 -1
View File
@@ -6953,9 +6953,11 @@ static std::vector<std::unique_ptr<test_case>> make_test_cases_eval() {
test_cases.emplace_back(new test_cpy(GGML_TYPE_F16, GGML_TYPE_F16, {256, 4, 1, 1}, {0, 0, 0, 0}, {0, 0, 0, 0}, true));
test_cases.emplace_back(new test_cpy(GGML_TYPE_F32, GGML_TYPE_F32, {256, 4, 1, 1}, {0, 0, 0, 0}, {0, 0, 0, 0}, true));
test_cases.emplace_back(new test_cpy(GGML_TYPE_BF16, GGML_TYPE_BF16, {256, 4, 1, 1}, {0, 0, 0, 0}, {0, 0, 0, 0}, true));
test_cases.emplace_back(new test_cpy(GGML_TYPE_I32, GGML_TYPE_I32, {256, 4, 1, 1}, {0, 0, 0, 0}, {0, 0, 0, 0}, true));
test_cases.emplace_back(new test_cpy(GGML_TYPE_I32, GGML_TYPE_I32, {256, 1, 4, 1}, {1, 2, 0, 3}, {0, 0, 0, 0}));
test_cases.emplace_back(new test_cpy(GGML_TYPE_F32, GGML_TYPE_F32, {256, 1, 4, 1}, {1, 2, 0, 3}, {0, 0, 0, 0}));
for (ggml_type type_dst : { GGML_TYPE_F32, GGML_TYPE_F16, GGML_TYPE_BF16 }) {
for (ggml_type type_dst : { GGML_TYPE_F32, GGML_TYPE_I32, GGML_TYPE_F16, GGML_TYPE_BF16 }) {
for (bool use_view_slice : { true, false }) {
for (std::array<int64_t, 4> ne : std::initializer_list<std::array<int64_t, 4>>{ {2, 1, 1, 1}, {2, 1, 3, 5},
{2, 3, 5, 7}, {1, 4, 4, 1}, {1, 8, 17, 1}, {10, 10, 10, 1} }) {