Skip to content

GitLab

Commit fcef1550 authored by Po-Yen, Chen's avatar Po-Yen, Chen
Browse files

Update sheduling pattern for IGLP_OPT_STRATEGY=1

parent 08d12d4f
Show whitespace changes
Inline Side-by-side
Showing with 1 addition and 6 deletions
include/ck/tensor_operation/gpu/grid/gridwise_gemm_pipeline_v2.hpp
View file @ fcef1550
...@@ -110,25 +110,20 @@ struct GridwiseGemmPipeline_v2 ...@@ -110,25 +110,20 @@ struct GridwiseGemmPipeline_v2
__builtin_amdgcn_sched_group_barrier(0x020, 2, 0); // VMEM read __builtin_amdgcn_sched_group_barrier(0x020, 2, 0); // VMEM read
__builtin_amdgcn_sched_group_barrier(0x008, 1, 0); // MFMA __builtin_amdgcn_sched_group_barrier(0x008, 1, 0); // MFMA
__builtin_amdgcn_sched_group_barrier(0x020, 1, 0); // VMEM read
__builtin_amdgcn_sched_group_barrier(0x008, 1, 0); // MFMA __builtin_amdgcn_sched_group_barrier(0x008, 1, 0); // MFMA
__builtin_amdgcn_sched_group_barrier(0x020, 2, 0); // VMEM read __builtin_amdgcn_sched_group_barrier(0x020, 2, 0); // VMEM read
__builtin_amdgcn_sched_group_barrier(0x008, 1, 0); // MFMA __builtin_amdgcn_sched_group_barrier(0x008, 1, 0); // MFMA
__builtin_amdgcn_sched_group_barrier(0x020, 1, 0); // VMEM read
__builtin_amdgcn_sched_group_barrier(0x008, 1, 0); // MFMA __builtin_amdgcn_sched_group_barrier(0x008, 1, 0); // MFMA
__builtin_amdgcn_sched_group_barrier(0x020, 2, 0); // VMEM read __builtin_amdgcn_sched_group_barrier(0x020, 1, 0); // VMEM read
__builtin_amdgcn_sched_group_barrier(0x008, 1, 0); // MFMA __builtin_amdgcn_sched_group_barrier(0x008, 1, 0); // MFMA
__builtin_amdgcn_sched_group_barrier(0x020, 1, 0); // VMEM read
__builtin_amdgcn_sched_group_barrier(0x008, 1, 0); // MFMA __builtin_amdgcn_sched_group_barrier(0x008, 1, 0); // MFMA
__builtin_amdgcn_sched_group_barrier(0x020, 2, 0); // VMEM read
__builtin_amdgcn_sched_group_barrier(0x008, 1, 0); // MFMA __builtin_amdgcn_sched_group_barrier(0x008, 1, 0); // MFMA
__builtin_amdgcn_sched_group_barrier(0x020, 1, 0); // VMEM read
__builtin_amdgcn_sched_group_barrier(0x008, 1, 0); // MFMA __builtin_amdgcn_sched_group_barrier(0x008, 1, 0); // MFMA
#elif IGLP_OPT_STRATEGY == 2 #elif IGLP_OPT_STRATEGY == 2
// 16 MFMAs // 16 MFMAs
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment