mirror of
https://git.datalinker.icu/vllm-project/vllm.git
synced 2025-12-15 07:45:01 +08:00
[Kernel][Moe Configs] Add more tuned triton configs for ExpertsInt8 and FP8 (#25858)
Signed-off-by: asafg <39553475+Josephasafg@users.noreply.github.com>
This commit is contained in:
parent
bb6d43047e
commit
35fe398c7c
@ -0,0 +1,146 @@
|
|||||||
|
{
|
||||||
|
"1": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 32,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"2": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 32,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"4": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 32,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"8": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 32,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"16": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"24": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 32,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"32": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 32,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"48": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"64": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"96": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"128": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"256": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"512": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"1024": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"1536": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"2048": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"3072": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"4096": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
}
|
||||||
|
}
|
||||||
@ -1,218 +1,146 @@
|
|||||||
{
|
{
|
||||||
"1": {
|
"1": {
|
||||||
"BLOCK_SIZE_M": 16,
|
"BLOCK_SIZE_M": 16,
|
||||||
"BLOCK_SIZE_N": 32,
|
"BLOCK_SIZE_N": 64,
|
||||||
"BLOCK_SIZE_K": 256,
|
"BLOCK_SIZE_K": 256,
|
||||||
"GROUP_SIZE_M": 32,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 4,
|
"num_warps": 4,
|
||||||
"num_stages": 3
|
"num_stages": 5
|
||||||
},
|
},
|
||||||
"2": {
|
"2": {
|
||||||
"BLOCK_SIZE_M": 16,
|
"BLOCK_SIZE_M": 16,
|
||||||
"BLOCK_SIZE_N": 64,
|
"BLOCK_SIZE_N": 64,
|
||||||
"BLOCK_SIZE_K": 128,
|
"BLOCK_SIZE_K": 256,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 5
|
"num_stages": 5
|
||||||
},
|
},
|
||||||
"4": {
|
"4": {
|
||||||
"BLOCK_SIZE_M": 16,
|
"BLOCK_SIZE_M": 16,
|
||||||
"BLOCK_SIZE_N": 64,
|
"BLOCK_SIZE_N": 64,
|
||||||
"BLOCK_SIZE_K": 128,
|
"BLOCK_SIZE_K": 256,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 5
|
"num_stages": 5
|
||||||
},
|
},
|
||||||
"8": {
|
"8": {
|
||||||
"BLOCK_SIZE_M": 16,
|
"BLOCK_SIZE_M": 16,
|
||||||
"BLOCK_SIZE_N": 64,
|
"BLOCK_SIZE_N": 64,
|
||||||
"BLOCK_SIZE_K": 128,
|
"BLOCK_SIZE_K": 256,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 4
|
"num_stages": 5
|
||||||
},
|
},
|
||||||
"16": {
|
"16": {
|
||||||
"BLOCK_SIZE_M": 16,
|
"BLOCK_SIZE_M": 16,
|
||||||
"BLOCK_SIZE_N": 64,
|
"BLOCK_SIZE_N": 64,
|
||||||
"BLOCK_SIZE_K": 128,
|
"BLOCK_SIZE_K": 256,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 5
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"24": {
|
"24": {
|
||||||
"BLOCK_SIZE_M": 32,
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"32": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
"BLOCK_SIZE_N": 64,
|
"BLOCK_SIZE_N": 64,
|
||||||
"BLOCK_SIZE_K": 256,
|
"BLOCK_SIZE_K": 256,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 8,
|
"num_warps": 8,
|
||||||
"num_stages": 3
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"32": {
|
|
||||||
"BLOCK_SIZE_M": 16,
|
|
||||||
"BLOCK_SIZE_N": 32,
|
|
||||||
"BLOCK_SIZE_K": 256,
|
|
||||||
"GROUP_SIZE_M": 1,
|
|
||||||
"num_warps": 4,
|
|
||||||
"num_stages": 5
|
|
||||||
},
|
|
||||||
"48": {
|
"48": {
|
||||||
"BLOCK_SIZE_M": 64,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 32,
|
"BLOCK_SIZE_N": 64,
|
||||||
"BLOCK_SIZE_K": 256,
|
"BLOCK_SIZE_K": 128,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 4,
|
"num_warps": 4,
|
||||||
"num_stages": 3
|
"num_stages": 4
|
||||||
},
|
},
|
||||||
"64": {
|
"64": {
|
||||||
"BLOCK_SIZE_M": 64,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 32,
|
"BLOCK_SIZE_N": 64,
|
||||||
"BLOCK_SIZE_K": 256,
|
"BLOCK_SIZE_K": 128,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 4,
|
"num_warps": 4,
|
||||||
"num_stages": 3
|
"num_stages": 4
|
||||||
},
|
},
|
||||||
"96": {
|
"96": {
|
||||||
"BLOCK_SIZE_M": 64,
|
"BLOCK_SIZE_M": 64,
|
||||||
"BLOCK_SIZE_N": 64,
|
"BLOCK_SIZE_N": 64,
|
||||||
"BLOCK_SIZE_K": 256,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 4,
|
"num_warps": 4,
|
||||||
"num_stages": 2
|
"num_stages": 5
|
||||||
},
|
},
|
||||||
"128": {
|
"128": {
|
||||||
"BLOCK_SIZE_M": 64,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 64,
|
"BLOCK_SIZE_N": 64,
|
||||||
"BLOCK_SIZE_K": 256,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 16,
|
||||||
"num_warps": 4,
|
"num_warps": 4,
|
||||||
"num_stages": 2
|
"num_stages": 4
|
||||||
},
|
},
|
||||||
"256": {
|
"256": {
|
||||||
"BLOCK_SIZE_M": 256,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 64,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 128,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 64,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 3
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"512": {
|
"512": {
|
||||||
"BLOCK_SIZE_M": 256,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 128,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 128,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 3
|
"num_stages": 4
|
||||||
},
|
},
|
||||||
"1024": {
|
"1024": {
|
||||||
"BLOCK_SIZE_M": 256,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 128,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 128,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 64,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 3
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"1536": {
|
"1536": {
|
||||||
"BLOCK_SIZE_M": 256,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 128,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 128,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 32,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 3
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"2048": {
|
"2048": {
|
||||||
"BLOCK_SIZE_M": 256,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 128,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 128,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 16,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 3
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"3072": {
|
"3072": {
|
||||||
"BLOCK_SIZE_M": 256,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 128,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 128,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 16,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 3
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"4096": {
|
"4096": {
|
||||||
"BLOCK_SIZE_M": 256,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 128,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 128,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 16,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 3
|
|
||||||
},
|
|
||||||
"5120": {
|
|
||||||
"BLOCK_SIZE_M": 256,
|
|
||||||
"BLOCK_SIZE_N": 128,
|
|
||||||
"BLOCK_SIZE_K": 128,
|
|
||||||
"GROUP_SIZE_M": 1,
|
|
||||||
"num_warps": 8,
|
|
||||||
"num_stages": 3
|
|
||||||
},
|
|
||||||
"9216": {
|
|
||||||
"BLOCK_SIZE_M": 256,
|
|
||||||
"BLOCK_SIZE_N": 128,
|
|
||||||
"BLOCK_SIZE_K": 128,
|
|
||||||
"GROUP_SIZE_M": 1,
|
|
||||||
"num_warps": 8,
|
|
||||||
"num_stages": 3
|
|
||||||
},
|
|
||||||
"13312": {
|
|
||||||
"BLOCK_SIZE_M": 256,
|
|
||||||
"BLOCK_SIZE_N": 128,
|
|
||||||
"BLOCK_SIZE_K": 128,
|
|
||||||
"GROUP_SIZE_M": 1,
|
|
||||||
"num_warps": 8,
|
|
||||||
"num_stages": 3
|
|
||||||
},
|
|
||||||
"17408": {
|
|
||||||
"BLOCK_SIZE_M": 256,
|
|
||||||
"BLOCK_SIZE_N": 128,
|
|
||||||
"BLOCK_SIZE_K": 128,
|
|
||||||
"GROUP_SIZE_M": 1,
|
|
||||||
"num_warps": 8,
|
|
||||||
"num_stages": 3
|
|
||||||
},
|
|
||||||
"25600": {
|
|
||||||
"BLOCK_SIZE_M": 256,
|
|
||||||
"BLOCK_SIZE_N": 128,
|
|
||||||
"BLOCK_SIZE_K": 128,
|
|
||||||
"GROUP_SIZE_M": 1,
|
|
||||||
"num_warps": 8,
|
|
||||||
"num_stages": 3
|
|
||||||
},
|
|
||||||
"33792": {
|
|
||||||
"BLOCK_SIZE_M": 256,
|
|
||||||
"BLOCK_SIZE_N": 128,
|
|
||||||
"BLOCK_SIZE_K": 128,
|
|
||||||
"GROUP_SIZE_M": 1,
|
|
||||||
"num_warps": 8,
|
|
||||||
"num_stages": 3
|
|
||||||
},
|
|
||||||
"41984": {
|
|
||||||
"BLOCK_SIZE_M": 256,
|
|
||||||
"BLOCK_SIZE_N": 128,
|
|
||||||
"BLOCK_SIZE_K": 128,
|
|
||||||
"GROUP_SIZE_M": 1,
|
|
||||||
"num_warps": 8,
|
|
||||||
"num_stages": 3
|
|
||||||
},
|
|
||||||
"50176": {
|
|
||||||
"BLOCK_SIZE_M": 256,
|
|
||||||
"BLOCK_SIZE_N": 128,
|
|
||||||
"BLOCK_SIZE_K": 128,
|
|
||||||
"GROUP_SIZE_M": 1,
|
|
||||||
"num_warps": 8,
|
|
||||||
"num_stages": 3
|
|
||||||
},
|
|
||||||
"58368": {
|
|
||||||
"BLOCK_SIZE_M": 256,
|
|
||||||
"BLOCK_SIZE_N": 128,
|
|
||||||
"BLOCK_SIZE_K": 128,
|
|
||||||
"GROUP_SIZE_M": 1,
|
|
||||||
"num_warps": 8,
|
|
||||||
"num_stages": 3
|
"num_stages": 3
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
@ -5,7 +5,7 @@
|
|||||||
"BLOCK_SIZE_K": 256,
|
"BLOCK_SIZE_K": 256,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 4,
|
"num_warps": 4,
|
||||||
"num_stages": 4
|
"num_stages": 5
|
||||||
},
|
},
|
||||||
"2": {
|
"2": {
|
||||||
"BLOCK_SIZE_M": 16,
|
"BLOCK_SIZE_M": 16,
|
||||||
@ -13,7 +13,7 @@
|
|||||||
"BLOCK_SIZE_K": 256,
|
"BLOCK_SIZE_K": 256,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 4,
|
"num_warps": 4,
|
||||||
"num_stages": 3
|
"num_stages": 5
|
||||||
},
|
},
|
||||||
"4": {
|
"4": {
|
||||||
"BLOCK_SIZE_M": 16,
|
"BLOCK_SIZE_M": 16,
|
||||||
@ -21,7 +21,7 @@
|
|||||||
"BLOCK_SIZE_K": 256,
|
"BLOCK_SIZE_K": 256,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 4,
|
"num_warps": 4,
|
||||||
"num_stages": 3
|
"num_stages": 5
|
||||||
},
|
},
|
||||||
"8": {
|
"8": {
|
||||||
"BLOCK_SIZE_M": 16,
|
"BLOCK_SIZE_M": 16,
|
||||||
@ -29,7 +29,7 @@
|
|||||||
"BLOCK_SIZE_K": 256,
|
"BLOCK_SIZE_K": 256,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 4,
|
"num_warps": 4,
|
||||||
"num_stages": 3
|
"num_stages": 5
|
||||||
},
|
},
|
||||||
"16": {
|
"16": {
|
||||||
"BLOCK_SIZE_M": 16,
|
"BLOCK_SIZE_M": 16,
|
||||||
@ -37,52 +37,52 @@
|
|||||||
"BLOCK_SIZE_K": 256,
|
"BLOCK_SIZE_K": 256,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 4,
|
"num_warps": 4,
|
||||||
"num_stages": 5
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"24": {
|
"24": {
|
||||||
"BLOCK_SIZE_M": 32,
|
"BLOCK_SIZE_M": 64,
|
||||||
"BLOCK_SIZE_N": 64,
|
"BLOCK_SIZE_N": 64,
|
||||||
"BLOCK_SIZE_K": 256,
|
"BLOCK_SIZE_K": 256,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 16,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 4
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"32": {
|
"32": {
|
||||||
"BLOCK_SIZE_M": 64,
|
"BLOCK_SIZE_M": 64,
|
||||||
"BLOCK_SIZE_N": 64,
|
"BLOCK_SIZE_N": 64,
|
||||||
"BLOCK_SIZE_K": 128,
|
"BLOCK_SIZE_K": 256,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 4,
|
"num_warps": 8,
|
||||||
"num_stages": 5
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"48": {
|
"48": {
|
||||||
"BLOCK_SIZE_M": 64,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 64,
|
"BLOCK_SIZE_N": 64,
|
||||||
"BLOCK_SIZE_K": 128,
|
"BLOCK_SIZE_K": 128,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 4,
|
"num_warps": 4,
|
||||||
"num_stages": 5
|
|
||||||
},
|
|
||||||
"64": {
|
|
||||||
"BLOCK_SIZE_M": 64,
|
|
||||||
"BLOCK_SIZE_N": 64,
|
|
||||||
"BLOCK_SIZE_K": 128,
|
|
||||||
"GROUP_SIZE_M": 1,
|
|
||||||
"num_warps": 4,
|
|
||||||
"num_stages": 5
|
|
||||||
},
|
|
||||||
"96": {
|
|
||||||
"BLOCK_SIZE_M": 64,
|
|
||||||
"BLOCK_SIZE_N": 128,
|
|
||||||
"BLOCK_SIZE_K": 128,
|
|
||||||
"GROUP_SIZE_M": 16,
|
|
||||||
"num_warps": 4,
|
|
||||||
"num_stages": 4
|
"num_stages": 4
|
||||||
},
|
},
|
||||||
"128": {
|
"64": {
|
||||||
"BLOCK_SIZE_M": 64,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 128,
|
"BLOCK_SIZE_N": 64,
|
||||||
"BLOCK_SIZE_K": 128,
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"96": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"128": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 16,
|
"GROUP_SIZE_M": 16,
|
||||||
"num_warps": 4,
|
"num_warps": 4,
|
||||||
"num_stages": 4
|
"num_stages": 4
|
||||||
@ -91,57 +91,57 @@
|
|||||||
"BLOCK_SIZE_M": 128,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 128,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 64,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 64,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 5
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"512": {
|
"512": {
|
||||||
"BLOCK_SIZE_M": 128,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 256,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 64,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 16,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 4
|
"num_stages": 4
|
||||||
},
|
},
|
||||||
"1024": {
|
"1024": {
|
||||||
"BLOCK_SIZE_M": 128,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 256,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 64,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 32,
|
"GROUP_SIZE_M": 64,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 4
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"1536": {
|
"1536": {
|
||||||
"BLOCK_SIZE_M": 128,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 256,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 64,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 32,
|
"GROUP_SIZE_M": 32,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 4
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"2048": {
|
"2048": {
|
||||||
"BLOCK_SIZE_M": 128,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 256,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 64,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 16,
|
"GROUP_SIZE_M": 16,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 3
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"3072": {
|
"3072": {
|
||||||
"BLOCK_SIZE_M": 128,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 256,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 64,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 64,
|
"GROUP_SIZE_M": 16,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 4
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"4096": {
|
"4096": {
|
||||||
"BLOCK_SIZE_M": 128,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 256,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 64,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 32,
|
"GROUP_SIZE_M": 16,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 4
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"5120": {
|
"5120": {
|
||||||
"BLOCK_SIZE_M": 128,
|
"BLOCK_SIZE_M": 128,
|
||||||
|
|||||||
@ -0,0 +1,146 @@
|
|||||||
|
{
|
||||||
|
"1": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"2": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"4": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"8": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"16": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"24": {
|
||||||
|
"BLOCK_SIZE_M": 32,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"32": {
|
||||||
|
"BLOCK_SIZE_M": 32,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"48": {
|
||||||
|
"BLOCK_SIZE_M": 32,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"64": {
|
||||||
|
"BLOCK_SIZE_M": 32,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"96": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"128": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"256": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"512": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"1024": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"1536": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"2048": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 64,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"3072": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"4096": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
}
|
||||||
|
}
|
||||||
@ -0,0 +1,146 @@
|
|||||||
|
{
|
||||||
|
"1": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"2": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"4": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"8": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"16": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"24": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 32,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"32": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"48": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"64": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"96": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"128": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"256": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"512": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"1024": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"1536": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"2048": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 64,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"3072": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"4096": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
}
|
||||||
|
}
|
||||||
@ -0,0 +1,146 @@
|
|||||||
|
{
|
||||||
|
"1": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 32,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"2": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 32,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"4": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 32,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"8": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 32,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"16": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"24": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"32": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"48": {
|
||||||
|
"BLOCK_SIZE_M": 32,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"64": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"96": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"128": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"256": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"512": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"1024": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"1536": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"2048": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"3072": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"4096": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
}
|
||||||
|
}
|
||||||
@ -0,0 +1,146 @@
|
|||||||
|
{
|
||||||
|
"1": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 32,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"2": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"4": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 32,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"8": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"16": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 64,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"24": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"32": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"48": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"64": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"96": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"128": {
|
||||||
|
"BLOCK_SIZE_M": 32,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"256": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"512": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"1024": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 64,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"1536": {
|
||||||
|
"BLOCK_SIZE_M": 256,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"2048": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"3072": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"4096": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
}
|
||||||
|
}
|
||||||
@ -0,0 +1,146 @@
|
|||||||
|
{
|
||||||
|
"1": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 32,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"2": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 32,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"4": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"8": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 64,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"16": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"24": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"32": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"48": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"64": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"96": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 32,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"128": {
|
||||||
|
"BLOCK_SIZE_M": 32,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"256": {
|
||||||
|
"BLOCK_SIZE_M": 32,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 64,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"512": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"1024": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"1536": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"2048": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"3072": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"4096": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
}
|
||||||
|
}
|
||||||
@ -0,0 +1,146 @@
|
|||||||
|
{
|
||||||
|
"1": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"2": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 64,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"4": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 64,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"8": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 32,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"16": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"24": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"32": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"48": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"64": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"96": {
|
||||||
|
"BLOCK_SIZE_M": 32,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"128": {
|
||||||
|
"BLOCK_SIZE_M": 32,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"256": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"512": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"1024": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"1536": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"2048": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"3072": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"4096": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 4
|
||||||
|
}
|
||||||
|
}
|
||||||
@ -0,0 +1,146 @@
|
|||||||
|
{
|
||||||
|
"1": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"2": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 64,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"4": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 64,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"8": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"16": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"24": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 64,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"32": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"48": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"64": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"96": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"128": {
|
||||||
|
"BLOCK_SIZE_M": 32,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"256": {
|
||||||
|
"BLOCK_SIZE_M": 32,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"512": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 64,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"1024": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"1536": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"2048": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"3072": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"4096": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
}
|
||||||
|
}
|
||||||
@ -1,29 +1,5 @@
|
|||||||
{
|
{
|
||||||
"1": {
|
"1": {
|
||||||
"BLOCK_SIZE_M": 16,
|
|
||||||
"BLOCK_SIZE_N": 32,
|
|
||||||
"BLOCK_SIZE_K": 256,
|
|
||||||
"GROUP_SIZE_M": 1,
|
|
||||||
"num_warps": 4,
|
|
||||||
"num_stages": 5
|
|
||||||
},
|
|
||||||
"2": {
|
|
||||||
"BLOCK_SIZE_M": 16,
|
|
||||||
"BLOCK_SIZE_N": 32,
|
|
||||||
"BLOCK_SIZE_K": 256,
|
|
||||||
"GROUP_SIZE_M": 1,
|
|
||||||
"num_warps": 4,
|
|
||||||
"num_stages": 4
|
|
||||||
},
|
|
||||||
"4": {
|
|
||||||
"BLOCK_SIZE_M": 16,
|
|
||||||
"BLOCK_SIZE_N": 64,
|
|
||||||
"BLOCK_SIZE_K": 256,
|
|
||||||
"GROUP_SIZE_M": 1,
|
|
||||||
"num_warps": 8,
|
|
||||||
"num_stages": 4
|
|
||||||
},
|
|
||||||
"8": {
|
|
||||||
"BLOCK_SIZE_M": 16,
|
"BLOCK_SIZE_M": 16,
|
||||||
"BLOCK_SIZE_N": 32,
|
"BLOCK_SIZE_N": 32,
|
||||||
"BLOCK_SIZE_K": 128,
|
"BLOCK_SIZE_K": 128,
|
||||||
@ -31,44 +7,68 @@
|
|||||||
"num_warps": 4,
|
"num_warps": 4,
|
||||||
"num_stages": 5
|
"num_stages": 5
|
||||||
},
|
},
|
||||||
"16": {
|
"2": {
|
||||||
"BLOCK_SIZE_M": 16,
|
"BLOCK_SIZE_M": 16,
|
||||||
"BLOCK_SIZE_N": 64,
|
"BLOCK_SIZE_N": 64,
|
||||||
"BLOCK_SIZE_K": 256,
|
"BLOCK_SIZE_K": 256,
|
||||||
"GROUP_SIZE_M": 64,
|
"GROUP_SIZE_M": 16,
|
||||||
"num_warps": 4,
|
"num_warps": 4,
|
||||||
"num_stages": 4
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"24": {
|
"4": {
|
||||||
"BLOCK_SIZE_M": 16,
|
"BLOCK_SIZE_M": 16,
|
||||||
"BLOCK_SIZE_N": 64,
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 64,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"8": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 32,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"16": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 32,
|
||||||
"BLOCK_SIZE_K": 256,
|
"BLOCK_SIZE_K": 256,
|
||||||
"GROUP_SIZE_M": 32,
|
"GROUP_SIZE_M": 32,
|
||||||
"num_warps": 4,
|
"num_warps": 4,
|
||||||
"num_stages": 3
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
|
"24": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
"32": {
|
"32": {
|
||||||
"BLOCK_SIZE_M": 16,
|
"BLOCK_SIZE_M": 16,
|
||||||
"BLOCK_SIZE_N": 64,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 256,
|
"BLOCK_SIZE_K": 256,
|
||||||
"GROUP_SIZE_M": 32,
|
"GROUP_SIZE_M": 16,
|
||||||
"num_warps": 4,
|
"num_warps": 4,
|
||||||
"num_stages": 4
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"48": {
|
"48": {
|
||||||
"BLOCK_SIZE_M": 16,
|
"BLOCK_SIZE_M": 16,
|
||||||
"BLOCK_SIZE_N": 64,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 256,
|
"BLOCK_SIZE_K": 256,
|
||||||
"GROUP_SIZE_M": 64,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 4,
|
"num_warps": 8,
|
||||||
"num_stages": 4
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"64": {
|
"64": {
|
||||||
"BLOCK_SIZE_M": 16,
|
"BLOCK_SIZE_M": 16,
|
||||||
"BLOCK_SIZE_N": 64,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 256,
|
"BLOCK_SIZE_K": 256,
|
||||||
"GROUP_SIZE_M": 32,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 4,
|
"num_warps": 8,
|
||||||
"num_stages": 3
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"96": {
|
"96": {
|
||||||
@ -77,22 +77,22 @@
|
|||||||
"BLOCK_SIZE_K": 256,
|
"BLOCK_SIZE_K": 256,
|
||||||
"GROUP_SIZE_M": 32,
|
"GROUP_SIZE_M": 32,
|
||||||
"num_warps": 4,
|
"num_warps": 4,
|
||||||
"num_stages": 4
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"128": {
|
"128": {
|
||||||
"BLOCK_SIZE_M": 32,
|
"BLOCK_SIZE_M": 32,
|
||||||
"BLOCK_SIZE_N": 128,
|
"BLOCK_SIZE_N": 256,
|
||||||
"BLOCK_SIZE_K": 128,
|
"BLOCK_SIZE_K": 256,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 4,
|
"num_warps": 8,
|
||||||
"num_stages": 4
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"256": {
|
"256": {
|
||||||
"BLOCK_SIZE_M": 32,
|
"BLOCK_SIZE_M": 64,
|
||||||
"BLOCK_SIZE_N": 128,
|
"BLOCK_SIZE_N": 256,
|
||||||
"BLOCK_SIZE_K": 128,
|
"BLOCK_SIZE_K": 128,
|
||||||
"GROUP_SIZE_M": 16,
|
"GROUP_SIZE_M": 32,
|
||||||
"num_warps": 4,
|
"num_warps": 8,
|
||||||
"num_stages": 4
|
"num_stages": 4
|
||||||
},
|
},
|
||||||
"512": {
|
"512": {
|
||||||
@ -100,47 +100,47 @@
|
|||||||
"BLOCK_SIZE_N": 128,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 64,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 4
|
"num_stages": 4
|
||||||
},
|
},
|
||||||
"1024": {
|
"1024": {
|
||||||
"BLOCK_SIZE_M": 128,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 128,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 64,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 64,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 5
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"1536": {
|
"1536": {
|
||||||
"BLOCK_SIZE_M": 256,
|
"BLOCK_SIZE_M": 256,
|
||||||
"BLOCK_SIZE_N": 128,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 128,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 16,
|
||||||
"num_warps": 8,
|
"num_warps": 8,
|
||||||
"num_stages": 3
|
"num_stages": 4
|
||||||
},
|
},
|
||||||
"2048": {
|
"2048": {
|
||||||
"BLOCK_SIZE_M": 128,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 128,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 64,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 64,
|
"GROUP_SIZE_M": 16,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 5
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"3072": {
|
"3072": {
|
||||||
"BLOCK_SIZE_M": 128,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 128,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 64,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 64,
|
"GROUP_SIZE_M": 16,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 5
|
"num_stages": 4
|
||||||
},
|
},
|
||||||
"4096": {
|
"4096": {
|
||||||
"BLOCK_SIZE_M": 256,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 128,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 128,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 16,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 3
|
"num_stages": 3
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
@ -0,0 +1,146 @@
|
|||||||
|
{
|
||||||
|
"1": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"2": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"4": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"8": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 64,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"16": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"24": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"32": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"48": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"64": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"96": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"128": {
|
||||||
|
"BLOCK_SIZE_M": 32,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"256": {
|
||||||
|
"BLOCK_SIZE_M": 32,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"512": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 64,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"1024": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"1536": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 64,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"2048": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"3072": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"4096": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
}
|
||||||
|
}
|
||||||
@ -0,0 +1,146 @@
|
|||||||
|
{
|
||||||
|
"1": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 32,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"2": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 32,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 64,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"4": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"8": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 32,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"16": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"24": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"32": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"48": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"64": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"96": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"128": {
|
||||||
|
"BLOCK_SIZE_M": 32,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"256": {
|
||||||
|
"BLOCK_SIZE_M": 32,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 64,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"512": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 64,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"1024": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"1536": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"2048": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"3072": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"4096": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
}
|
||||||
|
}
|
||||||
@ -0,0 +1,146 @@
|
|||||||
|
{
|
||||||
|
"1": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"2": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"4": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 4
|
||||||
|
},
|
||||||
|
"8": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"16": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"24": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"32": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"48": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"64": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"96": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"128": {
|
||||||
|
"BLOCK_SIZE_M": 32,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 64,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"256": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 64,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"512": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 16,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"1024": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 64,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"1536": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 64,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"2048": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"3072": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 64,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
},
|
||||||
|
"4096": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 64,
|
||||||
|
"GROUP_SIZE_M": 32,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2
|
||||||
|
}
|
||||||
|
}
|
||||||
@ -2,7 +2,7 @@
|
|||||||
"1": {
|
"1": {
|
||||||
"BLOCK_SIZE_M": 16,
|
"BLOCK_SIZE_M": 16,
|
||||||
"BLOCK_SIZE_N": 32,
|
"BLOCK_SIZE_N": 32,
|
||||||
"BLOCK_SIZE_K": 256,
|
"BLOCK_SIZE_K": 128,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 4,
|
"num_warps": 4,
|
||||||
"num_stages": 4
|
"num_stages": 4
|
||||||
@ -20,78 +20,78 @@
|
|||||||
"BLOCK_SIZE_N": 64,
|
"BLOCK_SIZE_N": 64,
|
||||||
"BLOCK_SIZE_K": 256,
|
"BLOCK_SIZE_K": 256,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 4,
|
"num_warps": 8,
|
||||||
"num_stages": 4
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"8": {
|
"8": {
|
||||||
"BLOCK_SIZE_M": 16,
|
|
||||||
"BLOCK_SIZE_N": 32,
|
|
||||||
"BLOCK_SIZE_K": 256,
|
|
||||||
"GROUP_SIZE_M": 1,
|
|
||||||
"num_warps": 4,
|
|
||||||
"num_stages": 3
|
|
||||||
},
|
|
||||||
"16": {
|
|
||||||
"BLOCK_SIZE_M": 16,
|
|
||||||
"BLOCK_SIZE_N": 64,
|
|
||||||
"BLOCK_SIZE_K": 256,
|
|
||||||
"GROUP_SIZE_M": 32,
|
|
||||||
"num_warps": 4,
|
|
||||||
"num_stages": 3
|
|
||||||
},
|
|
||||||
"24": {
|
|
||||||
"BLOCK_SIZE_M": 16,
|
|
||||||
"BLOCK_SIZE_N": 32,
|
|
||||||
"BLOCK_SIZE_K": 256,
|
|
||||||
"GROUP_SIZE_M": 64,
|
|
||||||
"num_warps": 4,
|
|
||||||
"num_stages": 3
|
|
||||||
},
|
|
||||||
"32": {
|
|
||||||
"BLOCK_SIZE_M": 16,
|
|
||||||
"BLOCK_SIZE_N": 64,
|
|
||||||
"BLOCK_SIZE_K": 256,
|
|
||||||
"GROUP_SIZE_M": 64,
|
|
||||||
"num_warps": 4,
|
|
||||||
"num_stages": 3
|
|
||||||
},
|
|
||||||
"48": {
|
|
||||||
"BLOCK_SIZE_M": 16,
|
|
||||||
"BLOCK_SIZE_N": 64,
|
|
||||||
"BLOCK_SIZE_K": 256,
|
|
||||||
"GROUP_SIZE_M": 1,
|
|
||||||
"num_warps": 4,
|
|
||||||
"num_stages": 3
|
|
||||||
},
|
|
||||||
"64": {
|
|
||||||
"BLOCK_SIZE_M": 16,
|
"BLOCK_SIZE_M": 16,
|
||||||
"BLOCK_SIZE_N": 64,
|
"BLOCK_SIZE_N": 64,
|
||||||
"BLOCK_SIZE_K": 256,
|
"BLOCK_SIZE_K": 256,
|
||||||
"GROUP_SIZE_M": 16,
|
"GROUP_SIZE_M": 16,
|
||||||
"num_warps": 4,
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"16": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 64,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 64,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 5
|
||||||
|
},
|
||||||
|
"24": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"32": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 256,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"48": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 3
|
||||||
|
},
|
||||||
|
"64": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
"num_stages": 3
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"96": {
|
"96": {
|
||||||
"BLOCK_SIZE_M": 16,
|
"BLOCK_SIZE_M": 16,
|
||||||
"BLOCK_SIZE_N": 64,
|
"BLOCK_SIZE_N": 64,
|
||||||
"BLOCK_SIZE_K": 256,
|
"BLOCK_SIZE_K": 256,
|
||||||
"GROUP_SIZE_M": 32,
|
"GROUP_SIZE_M": 64,
|
||||||
"num_warps": 4,
|
"num_warps": 4,
|
||||||
"num_stages": 4
|
"num_stages": 4
|
||||||
},
|
},
|
||||||
"128": {
|
"128": {
|
||||||
"BLOCK_SIZE_M": 32,
|
"BLOCK_SIZE_M": 32,
|
||||||
"BLOCK_SIZE_N": 128,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 128,
|
"BLOCK_SIZE_K": 256,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 4,
|
"num_warps": 8,
|
||||||
"num_stages": 3
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"256": {
|
"256": {
|
||||||
"BLOCK_SIZE_M": 32,
|
"BLOCK_SIZE_M": 64,
|
||||||
"BLOCK_SIZE_N": 128,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 128,
|
"BLOCK_SIZE_K": 128,
|
||||||
"GROUP_SIZE_M": 64,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 4,
|
"num_warps": 4,
|
||||||
"num_stages": 3
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
@ -100,47 +100,47 @@
|
|||||||
"BLOCK_SIZE_N": 128,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 64,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 3
|
"num_stages": 4
|
||||||
},
|
},
|
||||||
"1024": {
|
"1024": {
|
||||||
"BLOCK_SIZE_M": 128,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 128,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 64,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 64,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 5
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"1536": {
|
"1536": {
|
||||||
"BLOCK_SIZE_M": 256,
|
"BLOCK_SIZE_M": 256,
|
||||||
"BLOCK_SIZE_N": 128,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 128,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 1,
|
||||||
"num_warps": 8,
|
"num_warps": 8,
|
||||||
"num_stages": 3
|
"num_stages": 4
|
||||||
},
|
},
|
||||||
"2048": {
|
"2048": {
|
||||||
"BLOCK_SIZE_M": 128,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 128,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 64,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 64,
|
"GROUP_SIZE_M": 16,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 3
|
"num_stages": 4
|
||||||
},
|
},
|
||||||
"3072": {
|
"3072": {
|
||||||
"BLOCK_SIZE_M": 128,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 128,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 64,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 32,
|
"GROUP_SIZE_M": 16,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 4
|
"num_stages": 3
|
||||||
},
|
},
|
||||||
"4096": {
|
"4096": {
|
||||||
"BLOCK_SIZE_M": 256,
|
"BLOCK_SIZE_M": 128,
|
||||||
"BLOCK_SIZE_N": 128,
|
"BLOCK_SIZE_N": 128,
|
||||||
"BLOCK_SIZE_K": 128,
|
"BLOCK_SIZE_K": 64,
|
||||||
"GROUP_SIZE_M": 1,
|
"GROUP_SIZE_M": 16,
|
||||||
"num_warps": 8,
|
"num_warps": 4,
|
||||||
"num_stages": 3
|
"num_stages": 3
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
Loading…
x
Reference in New Issue
Block a user