Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Add dp4a for CC < 610 #2372

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Add dp4a for CC < 610 #2372

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 4 additions & 1 deletion .vscode/settings.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,5 +7,8 @@
"candle-pyo3"
],
"python.testing.unittestEnabled": false,
"python.testing.pytestEnabled": true
"python.testing.pytestEnabled": true,
"files.associations": {
"cstdint": "cpp"
}
}
13 changes: 13 additions & 0 deletions candle-kernels/src/quantized.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -82,6 +82,19 @@ static __device__ __forceinline__ int get_int_from_uint8_aligned(const uint8_t *
#define CC_RDNA2 (CC_OFFSET_AMD + 1030)
#define CC_RDNA3 (CC_OFFSET_AMD + 1100)

#if __CUDA_ARCH__ < MIN_CC_DP4A
// Manually perform 8-bit dot product
__device__ int dp4a(int a, int b, int c) {
int result = c;
for (int i = 0; i < 4; ++i) {
int8_t a_byte = (a >> (i * 8)) & 0xFF;
int8_t b_byte = (b >> (i * 8)) & 0xFF;
result += a_byte * b_byte;
}
return result;
}
#endif

#define MMQ_X_Q4_0_RDNA2 64
#define MMQ_Y_Q4_0_RDNA2 128
#define NWARPS_Q4_0_RDNA2 8
Expand Down
Loading