CUTLASS 3.1 (#915)

Co-authored-by: Aniket Shivam <ashivam@nvidia.com>
This commit is contained in:
ANIKET SHIVAM
2023-04-14 23:19:34 -04:00
committed by GitHub
co-authored by Aniket Shivam
parent 9b8166e3f0
commit d572cc1aab
482 changed files with 37175 additions and 16410 deletions
+25 -2
View File
@@ -267,6 +267,19 @@ cutlass_test_unit_add_executable(
sm90_gemm_tf32_tf32_f32_alignx_tensor_op_f32.cu
)
# Fused epilogue tests
cutlass_test_unit_add_executable(
cutlass_test_unit_gemm_device_tensorop_epilogue_fusion_sm90
BATCH_SOURCES ON
BATCH_SIZE 4
sm90_gemm_f16_f16_f16_tensor_op_f32_tensor_broadcast.cu
sm90_gemm_f32_f32_f32_tensor_op_f32_tensor_broadcast.cu
sm90_gemm_s8_s8_s8_tensor_op_s32_tensor_broadcast.cu
sm90_gemm_f16_f16_f16_tensor_op_f32_cluster_warpspecialized_cooperative_bias_elementwise.cu
sm90_gemm_f16_f16_f16_tensor_op_f32_cluster_warpspecialized_pingpong_bias_elementwise.cu
)
cutlass_test_unit_add_executable(
cutlass_test_unit_gemm_device_tensorop_cluster_multicast_sm90
@@ -276,7 +289,17 @@ cutlass_test_unit_add_executable(
sm90_gemm_f16_f16_f16_tensor_op_f32_cluster_unspecialized.cu
sm90_gemm_f16_f16_f16_tensor_op_f32_cluster_warpspecialized.cu
sm90_gemm_f16_f16_f16_tensor_op_f32_cluster_warpspecialized_persistent.cu
sm90_gemm_f16_f16_f16_tensor_op_f32_cluster_warpspecialized_pingpong.cu
sm90_gemm_f16_f16_f16_tensor_op_f32_cluster_warpspecialized_cooperative.cu
)
cutlass_test_unit_add_executable(
cutlass_test_unit_gemm_device_tensorop_gmma_rs_warpspecialized_sm90
BATCH_SOURCES ON
BATCH_SIZE 4
sm90_gemm_tf32_tf32_f32_tensor_op_f32_gmma_rs_cluster_warpspecialized.cu
)
cutlass_test_unit_add_executable(
@@ -337,6 +360,7 @@ cutlass_test_unit_add_executable(
gemm_s8t_s8n_s32n_tensor_op_s32_sm80.cu
gemm_s8t_s8n_s8n_tensor_op_s32_sm80.cu
gemm_s8t_s8n_s8t_tensor_op_s32_sm80.cu
gemm_s8t_s8n_f16t_tensor_op_s32_sm80.cu
gemm_s4t_s4n_s32n_tensor_op_s32_sm80.cu
gemm_s4t_s4n_s32t_tensor_op_s32_sm80.cu
gemm_s4t_s4n_s4n_tensor_op_s32_sm80.cu
@@ -416,7 +440,6 @@ cutlass_test_unit_add_executable(
gemm_planar_complex_f16_f16_f32_tensor_op_sm75.cu
gemm_planar_complex_f16_f16_f32_tensor_op_sm80.cu
)
cutlass_test_unit_add_executable(
cutlass_test_unit_gemm_device_grouped