mirror of
https://github.com/wassname/vllm.git
synced 2026-07-03 02:01:46 +08:00
[Kernel][ROCm][AMD] fused_moe Triton configs v2 for mi300X (#5932)
This commit is contained in:
+118
-46
@@ -1,128 +1,200 @@
|
||||
{
|
||||
"1": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"BLOCK_SIZE_N": 32,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_stages": 0
|
||||
"num_warps": 2,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"2": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_N": 16,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_stages": 0
|
||||
"num_warps": 2,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"4": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_N": 32,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"GROUP_SIZE_M": 64,
|
||||
"num_stages": 1
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 2,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"8": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_N": 16,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"GROUP_SIZE_M": 32,
|
||||
"num_stages": 1
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 1,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"16": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_N": 16,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"GROUP_SIZE_M": 8,
|
||||
"num_stages": 1
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 4,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"24": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"GROUP_SIZE_M": 64,
|
||||
"num_stages": 1
|
||||
"BLOCK_SIZE_N": 32,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 1,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"32": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"GROUP_SIZE_M": 8,
|
||||
"num_stages": 1
|
||||
"BLOCK_SIZE_N": 16,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 4,
|
||||
"num_warps": 2,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"48": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_N": 16,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 8,
|
||||
"num_stages": 0
|
||||
"GROUP_SIZE_M": 4,
|
||||
"num_warps": 2,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"64": {
|
||||
"BLOCK_SIZE_M": 64,
|
||||
"BLOCK_SIZE_M": 32,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 8,
|
||||
"num_stages": 0
|
||||
"GROUP_SIZE_M": 4,
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"96": {
|
||||
"BLOCK_SIZE_M": 32,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_N": 32,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 16,
|
||||
"num_stages": 0
|
||||
"GROUP_SIZE_M": 4,
|
||||
"num_warps": 4,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"128": {
|
||||
"BLOCK_SIZE_M": 64,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 8,
|
||||
"num_stages": 0
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 4,
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"256": {
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 8,
|
||||
"num_stages": 0
|
||||
"GROUP_SIZE_M": 4,
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"512": {
|
||||
"BLOCK_SIZE_M": 256,
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 8,
|
||||
"num_stages": 0
|
||||
"GROUP_SIZE_M": 4,
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"1024": {
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_stages": 0
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 32,
|
||||
"kpack": 2
|
||||
},
|
||||
"1536": {
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_stages": 0
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"2048": {
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 256,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_stages": 0
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"3072": {
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 256,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_stages": 0
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"4096": {
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_stages": 0
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
}
|
||||
}
|
||||
|
||||
Regular → Executable
+136
-46
@@ -1,110 +1,200 @@
|
||||
{
|
||||
"1": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 64
|
||||
"BLOCK_SIZE_N": 32,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 2,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"2": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 32,
|
||||
"GROUP_SIZE_M": 32
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 4,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"4": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 32,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 8
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 4,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"8": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_N": 16,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"GROUP_SIZE_M": 1
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 2,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"16": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_M": 64,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"GROUP_SIZE_M": 1
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"24": {
|
||||
"BLOCK_SIZE_M": 32,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 1
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 4,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"32": {
|
||||
"BLOCK_SIZE_M": 64,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 8
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 16,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"GROUP_SIZE_M": 4,
|
||||
"num_warps": 2,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"48": {
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 8
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"64": {
|
||||
"BLOCK_SIZE_M": 64,
|
||||
"BLOCK_SIZE_N": 32,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 1
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 2,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"96": {
|
||||
"BLOCK_SIZE_M": 32,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 8
|
||||
"BLOCK_SIZE_N": 32,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"GROUP_SIZE_M": 4,
|
||||
"num_warps": 4,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"128": {
|
||||
"BLOCK_SIZE_M": 64,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 32
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 4,
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"256": {
|
||||
"BLOCK_SIZE_M": 32,
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1
|
||||
"GROUP_SIZE_M": 4,
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"512": {
|
||||
"BLOCK_SIZE_M": 64,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 1
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"1024": {
|
||||
"BLOCK_SIZE_M": 64,
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"1536": {
|
||||
"BLOCK_SIZE_M": 64,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 1
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"2048": {
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"3072": {
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"4096": {
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
}
|
||||
}
|
||||
|
||||
+125
-53
@@ -1,128 +1,200 @@
|
||||
{
|
||||
"1": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 8,
|
||||
"num_stages": 0
|
||||
"BLOCK_SIZE_N": 16,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 2,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"2": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_N": 16,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_stages": 0
|
||||
"num_warps": 2,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"4": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 32,
|
||||
"GROUP_SIZE_M": 32,
|
||||
"num_stages": 1
|
||||
"BLOCK_SIZE_N": 32,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 2,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"8": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 32,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"GROUP_SIZE_M": 8,
|
||||
"num_stages": 1
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 2,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"16": {
|
||||
"BLOCK_SIZE_M": 32,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 16,
|
||||
"num_stages": 1
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 32,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 2,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"24": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"GROUP_SIZE_M": 8,
|
||||
"num_stages": 1
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 4,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"32": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 256,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 16,
|
||||
"num_stages": 0
|
||||
"BLOCK_SIZE_N": 16,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"GROUP_SIZE_M": 4,
|
||||
"num_warps": 2,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"48": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_N": 32,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"GROUP_SIZE_M": 16,
|
||||
"num_stages": 1
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 2,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"64": {
|
||||
"BLOCK_SIZE_M": 64,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 32,
|
||||
"num_stages": 0
|
||||
"BLOCK_SIZE_M": 32,
|
||||
"BLOCK_SIZE_N": 32,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"GROUP_SIZE_M": 4,
|
||||
"num_warps": 4,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"96": {
|
||||
"BLOCK_SIZE_M": 32,
|
||||
"BLOCK_SIZE_N": 32,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 16,
|
||||
"num_stages": 0
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 4,
|
||||
"num_warps": 4,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"128": {
|
||||
"BLOCK_SIZE_M": 64,
|
||||
"BLOCK_SIZE_N": 256,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 8,
|
||||
"num_stages": 0
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 4,
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"256": {
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 8,
|
||||
"num_stages": 0
|
||||
"GROUP_SIZE_M": 4,
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"512": {
|
||||
"BLOCK_SIZE_M": 64,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_stages": 0
|
||||
},
|
||||
"1024": {
|
||||
"BLOCK_SIZE_M": 64,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_stages": 0
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 32,
|
||||
"kpack": 2
|
||||
},
|
||||
"1024": {
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"1536": {
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_stages": 0
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"2048": {
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_stages": 0
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"3072": {
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_stages": 0
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"4096": {
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_stages": 0
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
}
|
||||
}
|
||||
|
||||
+127
-55
@@ -1,128 +1,200 @@
|
||||
{
|
||||
"1": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_N": 16,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_stages": 0
|
||||
"num_warps": 2,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"2": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 32,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_stages": 1
|
||||
"num_warps": 4,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"4": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"GROUP_SIZE_M": 32,
|
||||
"num_stages": 1
|
||||
"BLOCK_SIZE_N": 32,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 4,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"8": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_N": 32,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"GROUP_SIZE_M": 8,
|
||||
"num_stages": 1
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 2,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"16": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_N": 16,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"GROUP_SIZE_M": 8,
|
||||
"num_stages": 1
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 4,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"24": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"GROUP_SIZE_M": 8,
|
||||
"num_stages": 1
|
||||
"BLOCK_SIZE_M": 32,
|
||||
"BLOCK_SIZE_N": 32,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"32": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"GROUP_SIZE_M": 16,
|
||||
"num_stages": 0
|
||||
"BLOCK_SIZE_N": 32,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 4,
|
||||
"num_warps": 2,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"48": {
|
||||
"BLOCK_SIZE_M": 16,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"GROUP_SIZE_M": 16,
|
||||
"num_stages": 0
|
||||
"BLOCK_SIZE_N": 32,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 4,
|
||||
"num_warps": 2,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"64": {
|
||||
"BLOCK_SIZE_M": 32,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 256,
|
||||
"GROUP_SIZE_M": 8,
|
||||
"num_stages": 1
|
||||
"BLOCK_SIZE_N": 32,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 4,
|
||||
"num_warps": 4,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"96": {
|
||||
"BLOCK_SIZE_M": 32,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_N": 32,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 8,
|
||||
"num_stages": 0
|
||||
"GROUP_SIZE_M": 4,
|
||||
"num_warps": 4,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"128": {
|
||||
"BLOCK_SIZE_M": 64,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 128,
|
||||
"GROUP_SIZE_M": 8,
|
||||
"num_stages": 0
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 4,
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"256": {
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 64,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 8,
|
||||
"num_stages": 0
|
||||
},
|
||||
"512": {
|
||||
"BLOCK_SIZE_M": 256,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 8,
|
||||
"num_stages": 0
|
||||
"GROUP_SIZE_M": 4,
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 32,
|
||||
"kpack": 2
|
||||
},
|
||||
"512": {
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"1024": {
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_stages": 0
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"1536": {
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_stages": 0
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"2048": {
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_stages": 0
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
},
|
||||
"3072": {
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_stages": 0
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 2
|
||||
},
|
||||
"4096": {
|
||||
"BLOCK_SIZE_M": 256,
|
||||
"BLOCK_SIZE_N": 256,
|
||||
"BLOCK_SIZE_K": 32,
|
||||
"BLOCK_SIZE_M": 128,
|
||||
"BLOCK_SIZE_N": 128,
|
||||
"BLOCK_SIZE_K": 64,
|
||||
"GROUP_SIZE_M": 1,
|
||||
"num_stages": 0
|
||||
"num_warps": 8,
|
||||
"num_stages": 0,
|
||||
"waves_per_eu": 0,
|
||||
"matrix_instr_nonkdim": 16,
|
||||
"kpack": 1
|
||||
}
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user