leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

Add couple configs into generator.py for mixed input MM (#1350)

Browse Source 
* Add couple configs into generator.py for mixed input MM

* change one unit test name; reenable 128x32 in the profiler

* Added U8/BF16 tests.

---------

Co-authored-by: Haicheng Wu <haichengw@nvidia.com>
Co-authored-by: Haicheng Wu <57973641+hwu36@users.noreply.github.com>
This commit is contained in:
Aleksandar Samardžić
2024-08-16 06:59:29 +02:00
committed by GitHub
parent 865be73a97
commit 3f084f7f3c
21 changed files with 1506 additions and 80 deletions
Download Patch File Download Diff File Expand all files Collapse all files

102
tools/library/src/reference/gemm_fp_mixed_input.cu
View File

@@ -50,23 +50,13 @@ void initialize_gemm_reference_operations_fp_mixed_input(Manifest &manifest) {
make_gemm_real_canonical_layouts<
int8_t,
half_t,
half_t,
half_t,
half_t
float,
float
>(manifest);
make_gemm_real_canonical_layouts<
uint8_t,
half_t,
half_t,
half_t,
half_t
>(manifest);
make_gemm_real_canonical_layouts<
uint8_t,
half_t,
half_t,
float,
float
>(manifest);
@@ -75,6 +65,33 @@ void initialize_gemm_reference_operations_fp_mixed_input(Manifest &manifest) {
int8_t,
half_t,
half_t,
float
>(manifest);
make_gemm_real_canonical_layouts<
uint8_t,
half_t,
half_t,
float
>(manifest);
make_gemm_real_canonical_layouts<
int8_t,
half_t,
half_t,
half_t
>(manifest);
make_gemm_real_canonical_layouts<
uint8_t,
half_t,
half_t,
half_t
>(manifest);
make_gemm_real_canonical_layouts<
half_t,
int8_t,
float,
float
>(manifest);
@@ -82,7 +99,6 @@ void initialize_gemm_reference_operations_fp_mixed_input(Manifest &manifest) {
make_gemm_real_canonical_layouts<
half_t,
uint8_t,
half_t,
float,
float
>(manifest);
@@ -91,14 +107,40 @@ void initialize_gemm_reference_operations_fp_mixed_input(Manifest &manifest) {
half_t,
int8_t,
half_t,
float,
half_t
>(manifest);
make_gemm_real_canonical_layouts<
half_t,
uint8_t,
half_t,
half_t
>(manifest);
make_gemm_real_canonical_layouts<
half_t,
int8_t,
half_t,
float
>(manifest);
make_gemm_real_canonical_layouts<
half_t,
uint8_t,
half_t,
float
>(manifest);
// bfloat16_t mixed with 8-bit integer input
make_gemm_real_canonical_layouts<
uint8_t,
int8_t,
bfloat16_t,
float,
float
>(manifest);
make_gemm_real_canonical_layouts<
uint8_t,
bfloat16_t,
float,
float
@@ -107,15 +149,20 @@ void initialize_gemm_reference_operations_fp_mixed_input(Manifest &manifest) {
make_gemm_real_canonical_layouts<
int8_t,
bfloat16_t,
float,
float,
bfloat16_t,
float
>(manifest);
make_gemm_real_canonical_layouts<
uint8_t,
bfloat16_t,
bfloat16_t,
float
>(manifest);
make_gemm_real_canonical_layouts<
bfloat16_t,
int8_t,
bfloat16_t,
bfloat16_t,
float,
float
>(manifest);
@@ -124,15 +171,6 @@ void initialize_gemm_reference_operations_fp_mixed_input(Manifest &manifest) {
bfloat16_t,
uint8_t,
float,
float,
float
>(manifest);
make_gemm_real_canonical_layouts<
bfloat16_t,
uint8_t,
bfloat16_t,
float,
float
>(manifest);
@@ -140,7 +178,13 @@ void initialize_gemm_reference_operations_fp_mixed_input(Manifest &manifest) {
bfloat16_t,
int8_t,
bfloat16_t,
float,
float
>(manifest);
make_gemm_real_canonical_layouts<
bfloat16_t,
uint8_t,
bfloat16_t,
float
>(manifest);
}