[Kernels] Improve H200 Fused MoE Config (#28992)

Signed-off-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>
This commit is contained in:
Robert Shaw 2025-11-19 14:23:54 -05:00 committed by GitHub
parent 3319a493fc
commit fe69f331f8
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194

View File

@ -1,11 +1,11 @@
{ {
"1": { "1": {
"BLOCK_SIZE_M": 16, "BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64, "BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 128, "BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1, "GROUP_SIZE_M": 16,
"num_warps": 4, "num_warps": 4,
"num_stages": 5 "num_stages": 4
}, },
"2": { "2": {
"BLOCK_SIZE_M": 16, "BLOCK_SIZE_M": 16,
@ -13,82 +13,82 @@
"BLOCK_SIZE_K": 128, "BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 16, "GROUP_SIZE_M": 16,
"num_warps": 4, "num_warps": 4,
"num_stages": 3 "num_stages": 4
}, },
"4": { "4": {
"BLOCK_SIZE_M": 64, "BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64, "BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 128, "BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 64, "GROUP_SIZE_M": 64,
"num_warps": 4, "num_warps": 4,
"num_stages": 4 "num_stages": 4
}, },
"8": { "8": {
"BLOCK_SIZE_M": 64, "BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 16,
"num_warps": 4,
"num_stages": 4
},
"16": {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 16,
"num_warps": 4,
"num_stages": 3
},
"24": {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 16,
"num_warps": 4,
"num_stages": 4
},
"32": {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 16,
"num_warps": 4,
"num_stages": 5
},
"48": {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 128, "BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 128, "BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 64, "GROUP_SIZE_M": 64,
"num_warps": 4, "num_warps": 4,
"num_stages": 3 "num_stages": 3
}, },
"16": {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 256,
"BLOCK_SIZE_K": 64,
"GROUP_SIZE_M": 32,
"num_warps": 4,
"num_stages": 3
},
"24": {
"BLOCK_SIZE_M": 64,
"BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 16,
"num_warps": 4,
"num_stages": 3
},
"32": {
"BLOCK_SIZE_M": 64,
"BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1,
"num_warps": 4,
"num_stages": 3
},
"48": {
"BLOCK_SIZE_M": 64,
"BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 32,
"num_warps": 4,
"num_stages": 3
},
"64": { "64": {
"BLOCK_SIZE_M": 64, "BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 128, "BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 128, "BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 32, "GROUP_SIZE_M": 64,
"num_warps": 4, "num_warps": 4,
"num_stages": 3 "num_stages": 3
}, },
"96": { "96": {
"BLOCK_SIZE_M": 64, "BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 32,
"num_warps": 4,
"num_stages": 3
},
"128": {
"BLOCK_SIZE_M": 64,
"BLOCK_SIZE_N": 128, "BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 128, "BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 16, "GROUP_SIZE_M": 16,
"num_warps": 4, "num_warps": 4,
"num_stages": 3 "num_stages": 3
}, },
"128": {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 32,
"num_warps": 4,
"num_stages": 3
},
"256": { "256": {
"BLOCK_SIZE_M": 64, "BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 128, "BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 128, "BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 32, "GROUP_SIZE_M": 32,
@ -96,10 +96,10 @@
"num_stages": 3 "num_stages": 3
}, },
"512": { "512": {
"BLOCK_SIZE_M": 64, "BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 128, "BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 128, "BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 32, "GROUP_SIZE_M": 16,
"num_warps": 4, "num_warps": 4,
"num_stages": 3 "num_stages": 3
}, },
@ -109,7 +109,7 @@
"BLOCK_SIZE_K": 128, "BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 32, "GROUP_SIZE_M": 32,
"num_warps": 4, "num_warps": 4,
"num_stages": 3 "num_stages": 4
}, },
"1536": { "1536": {
"BLOCK_SIZE_M": 64, "BLOCK_SIZE_M": 64,
@ -117,21 +117,21 @@
"BLOCK_SIZE_K": 128, "BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 32, "GROUP_SIZE_M": 32,
"num_warps": 4, "num_warps": 4,
"num_stages": 3 "num_stages": 4
}, },
"2048": { "2048": {
"BLOCK_SIZE_M": 64, "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": 32,
"num_warps": 4, "num_warps": 4,
"num_stages": 3 "num_stages": 4
}, },
"3072": { "3072": {
"BLOCK_SIZE_M": 128, "BLOCK_SIZE_M": 64,
"BLOCK_SIZE_N": 64, "BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 128, "BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 16, "GROUP_SIZE_M": 32,
"num_warps": 4, "num_warps": 4,
"num_stages": 4 "num_stages": 4
}, },
@ -139,7 +139,7 @@
"BLOCK_SIZE_M": 64, "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": 16,
"num_warps": 4, "num_warps": 4,
"num_stages": 3 "num_stages": 3
} }