Repository navigation
[rocRoller] Add unique IDs to rrsuite - #1592
Merged
Merged
Conversation
Contributor
Generated Documentation |
Contributor
Contributor
Code Coverage Report for gfx942Summary
This PR adds/edits 78 newly uncovered lines. Artifacts
Commit Hashes |
jaopaulolc
reviewed
Sep 16, 2025
Contributor
Performance Report for gfx12Results@@ Significant (p-val <0.05) Performance Diffs @@
====================================================================================================
- 0.50% | p=7.0009e-03
| CodeGen(instCount: 40000, instructions: comments)| CodeGen() | CodeGen(instCount: 40000, instructions: comments)
- 1.70% | p=2.5347e-02
| CodeGen(instCount: 40000, instructions: complex_mi_with_coop)| CodeGen() | CodeGen(instCount: 40000, instructions: complex_mi_with_coop)
+ 1.67% | p=1.7451e-03
| CodeGen(instCount: 40000, instructions: simple_mi)| CodeGen() | CodeGen(instCount: 40000, instructions: simple_mi)
Full table of results
|
Contributor
Performance Report for gfx942Results@@ Significant (p-val <0.05) Performance Diffs @@
====================================================================================================
- 1.46% | p=2.5347e-02
| 3. FloatsGEMM(M: 7680, N: 8448, K: 8448, alpha: 2, beta: 0.5, types: {'type_A': 'half', 'type_B': 'half', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'N', 'trans_B': 'T', 'scale_A': 'None', 'scaleType_A': 'None', 'scale_B': 'None', 'scaleType_B': 'None', 'scaleBlockSize': -1, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: -1, workgroupMappingValue: -1)| GEMM(mac_m: 128, mac_n: 128, mac_k: 32, wave_m: 32, wave_n: 32, wave_k: 8, wave_b: 1, workgroup_size_x: 128, workgroup_size_y: 2, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: True, betaInFma: True, direct2LDS_A: False, direct2LDS_B: False, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: False, prefetchInFlight: 2, prefetchLDSFactor: 0, prefetchMixMemOps: False, loadLDSScale_A: False, loadLDSScale_B: False, swizzleScale: False, prefetchScale: False, streamK: True, numWGs: 304, streamKTwoTile: False, architecture: {'ArchString': 'gfx942', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=7680, N=8448, K=8448, alpha=2, beta=0.5, types={'type_A': 'half', 'type_B': 'half', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'N', 'trans_B': 'T', 'scale_A': 'None', 'scaleType_A': 'None', 'scale_B': 'None', 'scaleType_B': 'None', 'scaleBlockSize': -1, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=-1, workgroupMappingValue=-1)GEMMSolution(mac_m=128, mac_n=128, mac_k=32, wave_m=32, wave_n=32, wave_k=8, wave_b=1, workgroup_size_x=128, workgroup_size_y=2, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=True, betaInFma=True, direct2LDS_A=False, direct2LDS_B=False, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=False, prefetchInFlight=2, prefetchLDSFactor=0, prefetchMixMemOps=False, loadLDSScale_A=False, loadLDSScale_B=False, swizzleScale=False, prefetchScale=False, streamK=True, numWGs=304, streamKTwoTile=False, architecture={'ArchString': 'gfx942', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
- 4.65% | p=2.5347e-02
| 3. FloatsGEMM(M: 7680, N: 8448, K: 8448, alpha: 2, beta: 0.5, types: {'type_A': 'half', 'type_B': 'half', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'N', 'trans_B': 'T', 'scale_A': 'None', 'scaleType_A': 'None', 'scale_B': 'None', 'scaleType_B': 'None', 'scaleBlockSize': -1, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: -1, workgroupMappingValue: -1)| GEMM(mac_m: 128, mac_n: 256, mac_k: 16, wave_m: 32, wave_n: 32, wave_k: 8, wave_b: 1, workgroup_size_x: 128, workgroup_size_y: 2, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: False, betaInFma: True, direct2LDS_A: False, direct2LDS_B: False, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: True, prefetchInFlight: 2, prefetchLDSFactor: 0, prefetchMixMemOps: False, loadLDSScale_A: False, loadLDSScale_B: False, swizzleScale: False, prefetchScale: False, streamK: False, numWGs: 0, streamKTwoTile: False, architecture: {'ArchString': 'gfx942', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=7680, N=8448, K=8448, alpha=2, beta=0.5, types={'type_A': 'half', 'type_B': 'half', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'N', 'trans_B': 'T', 'scale_A': 'None', 'scaleType_A': 'None', 'scale_B': 'None', 'scaleType_B': 'None', 'scaleBlockSize': -1, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=-1, workgroupMappingValue=-1)GEMMSolution(mac_m=128, mac_n=256, mac_k=16, wave_m=32, wave_n=32, wave_k=8, wave_b=1, workgroup_size_x=128, workgroup_size_y=2, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=False, betaInFma=True, direct2LDS_A=False, direct2LDS_B=False, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=True, prefetchInFlight=2, prefetchLDSFactor=0, prefetchMixMemOps=False, loadLDSScale_A=False, loadLDSScale_B=False, swizzleScale=False, prefetchScale=False, streamK=False, numWGs=0, streamKTwoTile=False, architecture={'ArchString': 'gfx942', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
+ 0.71% | p=2.5347e-02
| 3. FloatsGEMM(M: 7680, N: 8448, K: 8448, alpha: 2, beta: 0.5, types: {'type_A': 'half', 'type_B': 'half', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'N', 'trans_B': 'T', 'scale_A': 'None', 'scaleType_A': 'None', 'scale_B': 'None', 'scaleType_B': 'None', 'scaleBlockSize': -1, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: -1, workgroupMappingValue: -1)| GEMM(mac_m: 128, mac_n: 64, mac_k: 32, wave_m: 32, wave_n: 32, wave_k: 8, wave_b: 1, workgroup_size_x: 128, workgroup_size_y: 2, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: True, betaInFma: True, direct2LDS_A: False, direct2LDS_B: False, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: False, prefetchInFlight: 2, prefetchLDSFactor: 0, prefetchMixMemOps: False, loadLDSScale_A: False, loadLDSScale_B: False, swizzleScale: False, prefetchScale: False, streamK: True, numWGs: 304, streamKTwoTile: True, architecture: {'ArchString': 'gfx942', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=7680, N=8448, K=8448, alpha=2, beta=0.5, types={'type_A': 'half', 'type_B': 'half', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'N', 'trans_B': 'T', 'scale_A': 'None', 'scaleType_A': 'None', 'scale_B': 'None', 'scaleType_B': 'None', 'scaleBlockSize': -1, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=-1, workgroupMappingValue=-1)GEMMSolution(mac_m=128, mac_n=64, mac_k=32, wave_m=32, wave_n=32, wave_k=8, wave_b=1, workgroup_size_x=128, workgroup_size_y=2, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=True, betaInFma=True, direct2LDS_A=False, direct2LDS_B=False, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=False, prefetchInFlight=2, prefetchLDSFactor=0, prefetchMixMemOps=False, loadLDSScale_A=False, loadLDSScale_B=False, swizzleScale=False, prefetchScale=False, streamK=True, numWGs=304, streamKTwoTile=True, architecture={'ArchString': 'gfx942', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
- 1.92% | p=2.5347e-02
| 3. FloatsGEMM(M: 7680, N: 8448, K: 8448, alpha: 2, beta: 0.5, types: {'type_A': 'half', 'type_B': 'half', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'N', 'trans_B': 'T', 'scale_A': 'None', 'scaleType_A': 'None', 'scale_B': 'None', 'scaleType_B': 'None', 'scaleBlockSize': -1, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: -1, workgroupMappingValue: -1)| GEMM(mac_m: 64, mac_n: 128, mac_k: 64, wave_m: 32, wave_n: 32, wave_k: 8, wave_b: 1, workgroup_size_x: 128, workgroup_size_y: 2, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: True, betaInFma: True, direct2LDS_A: False, direct2LDS_B: False, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: False, prefetchInFlight: 2, prefetchLDSFactor: 0, prefetchMixMemOps: False, loadLDSScale_A: False, loadLDSScale_B: False, swizzleScale: False, prefetchScale: False, streamK: True, numWGs: 304, streamKTwoTile: False, architecture: {'ArchString': 'gfx942', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=7680, N=8448, K=8448, alpha=2, beta=0.5, types={'type_A': 'half', 'type_B': 'half', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'N', 'trans_B': 'T', 'scale_A': 'None', 'scaleType_A': 'None', 'scale_B': 'None', 'scaleType_B': 'None', 'scaleBlockSize': -1, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=-1, workgroupMappingValue=-1)GEMMSolution(mac_m=64, mac_n=128, mac_k=64, wave_m=32, wave_n=32, wave_k=8, wave_b=1, workgroup_size_x=128, workgroup_size_y=2, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=True, betaInFma=True, direct2LDS_A=False, direct2LDS_B=False, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=False, prefetchInFlight=2, prefetchLDSFactor=0, prefetchMixMemOps=False, loadLDSScale_A=False, loadLDSScale_B=False, swizzleScale=False, prefetchScale=False, streamK=True, numWGs=304, streamKTwoTile=False, architecture={'ArchString': 'gfx942', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
+ 0.91% | p=2.5347e-02
| 3. FloatsGEMM(M: 7680, N: 8448, K: 8448, alpha: 2, beta: 0.5, types: {'type_A': 'half', 'type_B': 'half', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'N', 'trans_B': 'T', 'scale_A': 'None', 'scaleType_A': 'None', 'scale_B': 'None', 'scaleType_B': 'None', 'scaleBlockSize': -1, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: -1, workgroupMappingValue: -1)| GEMM(mac_m: 64, mac_n: 256, mac_k: 16, wave_m: 32, wave_n: 32, wave_k: 8, wave_b: 1, workgroup_size_x: 128, workgroup_size_y: 2, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: True, betaInFma: True, direct2LDS_A: False, direct2LDS_B: False, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: False, prefetchInFlight: 2, prefetchLDSFactor: 0, prefetchMixMemOps: False, loadLDSScale_A: False, loadLDSScale_B: False, swizzleScale: False, prefetchScale: False, streamK: True, numWGs: 304, streamKTwoTile: False, architecture: {'ArchString': 'gfx942', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=7680, N=8448, K=8448, alpha=2, beta=0.5, types={'type_A': 'half', 'type_B': 'half', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'N', 'trans_B': 'T', 'scale_A': 'None', 'scaleType_A': 'None', 'scale_B': 'None', 'scaleType_B': 'None', 'scaleBlockSize': -1, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=-1, workgroupMappingValue=-1)GEMMSolution(mac_m=64, mac_n=256, mac_k=16, wave_m=32, wave_n=32, wave_k=8, wave_b=1, workgroup_size_x=128, workgroup_size_y=2, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=True, betaInFma=True, direct2LDS_A=False, direct2LDS_B=False, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=False, prefetchInFlight=2, prefetchLDSFactor=0, prefetchMixMemOps=False, loadLDSScale_A=False, loadLDSScale_B=False, swizzleScale=False, prefetchScale=False, streamK=True, numWGs=304, streamKTwoTile=False, architecture={'ArchString': 'gfx942', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
- 2.13% | p=1.7451e-03
| 3. FloatsGEMM(M: 8448, N: 8448, K: 128, alpha: 2, beta: 0.5, types: {'type_A': 'half', 'type_B': 'half', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'N', 'trans_B': 'T', 'scale_A': 'None', 'scaleType_A': 'None', 'scale_B': 'None', 'scaleType_B': 'None', 'scaleBlockSize': -1, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: -1, workgroupMappingValue: -1)| GEMM(mac_m: 128, mac_n: 256, mac_k: 16, wave_m: 32, wave_n: 32, wave_k: 8, wave_b: 1, workgroup_size_x: 128, workgroup_size_y: 2, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: True, betaInFma: True, direct2LDS_A: False, direct2LDS_B: False, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: True, prefetchInFlight: 2, prefetchLDSFactor: 0, prefetchMixMemOps: False, loadLDSScale_A: False, loadLDSScale_B: False, swizzleScale: False, prefetchScale: False, streamK: False, numWGs: 0, streamKTwoTile: False, architecture: {'ArchString': 'gfx942', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=8448, N=8448, K=128, alpha=2, beta=0.5, types={'type_A': 'half', 'type_B': 'half', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'N', 'trans_B': 'T', 'scale_A': 'None', 'scaleType_A': 'None', 'scale_B': 'None', 'scaleType_B': 'None', 'scaleBlockSize': -1, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=-1, workgroupMappingValue=-1)GEMMSolution(mac_m=128, mac_n=256, mac_k=16, wave_m=32, wave_n=32, wave_k=8, wave_b=1, workgroup_size_x=128, workgroup_size_y=2, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=True, betaInFma=True, direct2LDS_A=False, direct2LDS_B=False, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=True, prefetchInFlight=2, prefetchLDSFactor=0, prefetchMixMemOps=False, loadLDSScale_A=False, loadLDSScale_B=False, swizzleScale=False, prefetchScale=False, streamK=False, numWGs=0, streamKTwoTile=False, architecture={'ArchString': 'gfx942', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
- 1.04% | p=2.5347e-02
| 3. FloatsGEMM(M: 8448, N: 8448, K: 128, alpha: 2, beta: 0.5, types: {'type_A': 'half', 'type_B': 'half', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'N', 'trans_B': 'T', 'scale_A': 'None', 'scaleType_A': 'None', 'scale_B': 'None', 'scaleType_B': 'None', 'scaleBlockSize': -1, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: -1, workgroupMappingValue: -1)| GEMM(mac_m: 128, mac_n: 256, mac_k: 16, wave_m: 32, wave_n: 32, wave_k: 8, wave_b: 1, workgroup_size_x: 128, workgroup_size_y: 2, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: True, betaInFma: True, direct2LDS_A: False, direct2LDS_B: False, scheduler: Sequential, schedulerCost: LinearWeighted, prefetch: True, prefetchInFlight: 2, prefetchLDSFactor: 0, prefetchMixMemOps: False, loadLDSScale_A: False, loadLDSScale_B: False, swizzleScale: False, prefetchScale: False, streamK: False, numWGs: 0, streamKTwoTile: False, architecture: {'ArchString': 'gfx942', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=8448, N=8448, K=128, alpha=2, beta=0.5, types={'type_A': 'half', 'type_B': 'half', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'N', 'trans_B': 'T', 'scale_A': 'None', 'scaleType_A': 'None', 'scale_B': 'None', 'scaleType_B': 'None', 'scaleBlockSize': -1, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=-1, workgroupMappingValue=-1)GEMMSolution(mac_m=128, mac_n=256, mac_k=16, wave_m=32, wave_n=32, wave_k=8, wave_b=1, workgroup_size_x=128, workgroup_size_y=2, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=True, betaInFma=True, direct2LDS_A=False, direct2LDS_B=False, scheduler='Sequential', schedulerCost='LinearWeighted', prefetch=True, prefetchInFlight=2, prefetchLDSFactor=0, prefetchMixMemOps=False, loadLDSScale_A=False, loadLDSScale_B=False, swizzleScale=False, prefetchScale=False, streamK=False, numWGs=0, streamKTwoTile=False, architecture={'ArchString': 'gfx942', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
+ 16.95% | p=5.6994e-05
| CodeGen(instCount: 40000, instructions: comments)| CodeGen() | CodeGen(instCount: 40000, instructions: comments)
+ 10.83% | p=5.6994e-05
| CodeGen(instCount: 40000, instructions: complex_mi_with_coop)| CodeGen() | CodeGen(instCount: 40000, instructions: complex_mi_with_coop)
+ 17.40% | p=5.6994e-05
| CodeGen(instCount: 40000, instructions: simple_mi)| CodeGen() | CodeGen(instCount: 40000, instructions: simple_mi)
Full table of results
|
Contributor
Performance Report for gfx950-rocrollerResults@@ Significant (p-val <0.05) Performance Diffs @@
====================================================================================================
+ 0.41% | p=1.3906e-02
| 3. FloatsGEMM(M: 4096, N: 4096, K: 32768, alpha: 2, beta: 0, types: {'type_A': 'fp4', 'type_B': 'fp4', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: -1, workgroupMappingValue: -1)| GEMM(mac_m: 256, mac_n: 256, mac_k: 128, wave_m: 32, wave_n: 32, wave_k: 64, wave_b: 1, workgroup_size_x: 128, workgroup_size_y: 2, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: False, betaInFma: True, direct2LDS_A: True, direct2LDS_B: True, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: True, prefetchInFlight: 4, prefetchLDSFactor: 1, prefetchMixMemOps: True, loadLDSScale_A: False, loadLDSScale_B: False, swizzleScale: True, prefetchScale: True, streamK: False, numWGs: 0, streamKTwoTile: False, architecture: {'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=4096, N=4096, K=32768, alpha=2, beta=0, types={'type_A': 'fp4', 'type_B': 'fp4', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=-1, workgroupMappingValue=-1)GEMMSolution(mac_m=256, mac_n=256, mac_k=128, wave_m=32, wave_n=32, wave_k=64, wave_b=1, workgroup_size_x=128, workgroup_size_y=2, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=False, betaInFma=True, direct2LDS_A=True, direct2LDS_B=True, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=True, prefetchInFlight=4, prefetchLDSFactor=1, prefetchMixMemOps=True, loadLDSScale_A=False, loadLDSScale_B=False, swizzleScale=True, prefetchScale=True, streamK=False, numWGs=0, streamKTwoTile=False, architecture={'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
+ 0.37% | p=3.5562e-02
| 3. FloatsGEMM(M: 4096, N: 4096, K: 32768, alpha: 2, beta: 0, types: {'type_A': 'fp4', 'type_B': 'fp4', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: 0, workgroupMappingValue: -1)| GEMM(mac_m: 256, mac_n: 256, mac_k: 128, wave_m: 16, wave_n: 16, wave_k: 128, wave_b: 1, workgroup_size_x: 128, workgroup_size_y: 2, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: False, betaInFma: True, direct2LDS_A: True, direct2LDS_B: True, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: True, prefetchInFlight: 4, prefetchLDSFactor: 1, prefetchMixMemOps: True, loadLDSScale_A: False, loadLDSScale_B: False, swizzleScale: True, prefetchScale: True, streamK: False, numWGs: 0, streamKTwoTile: False, architecture: {'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=4096, N=4096, K=32768, alpha=2, beta=0, types={'type_A': 'fp4', 'type_B': 'fp4', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=0, workgroupMappingValue=-1)GEMMSolution(mac_m=256, mac_n=256, mac_k=128, wave_m=16, wave_n=16, wave_k=128, wave_b=1, workgroup_size_x=128, workgroup_size_y=2, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=False, betaInFma=True, direct2LDS_A=True, direct2LDS_B=True, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=True, prefetchInFlight=4, prefetchLDSFactor=1, prefetchMixMemOps=True, loadLDSScale_A=False, loadLDSScale_B=False, swizzleScale=True, prefetchScale=True, streamK=False, numWGs=0, streamKTwoTile=False, architecture={'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
+ 0.62% | p=3.8822e-05
| 3. FloatsGEMM(M: 4096, N: 4096, K: 32768, alpha: 2, beta: 0, types: {'type_A': 'fp4', 'type_B': 'fp4', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: 0, workgroupMappingValue: -1)| GEMM(mac_m: 256, mac_n: 256, mac_k: 128, wave_m: 32, wave_n: 32, wave_k: 64, wave_b: 1, workgroup_size_x: 128, workgroup_size_y: 2, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: False, betaInFma: True, direct2LDS_A: True, direct2LDS_B: True, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: True, prefetchInFlight: 4, prefetchLDSFactor: 1, prefetchMixMemOps: True, loadLDSScale_A: False, loadLDSScale_B: False, swizzleScale: True, prefetchScale: True, streamK: False, numWGs: 0, streamKTwoTile: False, architecture: {'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=4096, N=4096, K=32768, alpha=2, beta=0, types={'type_A': 'fp4', 'type_B': 'fp4', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=0, workgroupMappingValue=-1)GEMMSolution(mac_m=256, mac_n=256, mac_k=128, wave_m=32, wave_n=32, wave_k=64, wave_b=1, workgroup_size_x=128, workgroup_size_y=2, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=False, betaInFma=True, direct2LDS_A=True, direct2LDS_B=True, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=True, prefetchInFlight=4, prefetchLDSFactor=1, prefetchMixMemOps=True, loadLDSScale_A=False, loadLDSScale_B=False, swizzleScale=True, prefetchScale=True, streamK=False, numWGs=0, streamKTwoTile=False, architecture={'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
+ 0.45% | p=6.7676e-04
| 3. FloatsGEMM(M: 4096, N: 4096, K: 32768, alpha: 2, beta: 0, types: {'type_A': 'fp4', 'type_B': 'fp4', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': True, 'scaleShuffleTileA': [64, 4, 2], 'scaleShuffleTileB': [64, 4, 2]}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: 0, workgroupMappingValue: -1)| GEMM(mac_m: 256, mac_n: 256, mac_k: 128, wave_m: 32, wave_n: 32, wave_k: 64, wave_b: 1, workgroup_size_x: 128, workgroup_size_y: 2, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: False, betaInFma: True, direct2LDS_A: True, direct2LDS_B: True, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: True, prefetchInFlight: 4, prefetchLDSFactor: 1, prefetchMixMemOps: True, loadLDSScale_A: False, loadLDSScale_B: False, swizzleScale: True, prefetchScale: True, streamK: False, numWGs: 0, streamKTwoTile: False, architecture: {'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=4096, N=4096, K=32768, alpha=2, beta=0, types={'type_A': 'fp4', 'type_B': 'fp4', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': True, 'scaleShuffleTileA': [64, 4, 2], 'scaleShuffleTileB': [64, 4, 2]}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=0, workgroupMappingValue=-1)GEMMSolution(mac_m=256, mac_n=256, mac_k=128, wave_m=32, wave_n=32, wave_k=64, wave_b=1, workgroup_size_x=128, workgroup_size_y=2, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=False, betaInFma=True, direct2LDS_A=True, direct2LDS_B=True, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=True, prefetchInFlight=4, prefetchLDSFactor=1, prefetchMixMemOps=True, loadLDSScale_A=False, loadLDSScale_B=False, swizzleScale=True, prefetchScale=True, streamK=False, numWGs=0, streamKTwoTile=False, architecture={'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
+ 0.53% | p=2.9851e-31
| 3. FloatsGEMM(M: 4096, N: 4096, K: 32768, alpha: 2, beta: 0, types: {'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'None', 'scaleType_A': 'None', 'scale_B': 'None', 'scaleType_B': 'None', 'scaleBlockSize': -1, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: -1, workgroupMappingValue: -1)| GEMM(mac_m: 128, mac_n: 128, mac_k: 128, wave_m: 16, wave_n: 16, wave_k: 128, wave_b: 1, workgroup_size_x: 128, workgroup_size_y: 2, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: False, betaInFma: True, direct2LDS_A: True, direct2LDS_B: True, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: True, prefetchInFlight: 4, prefetchLDSFactor: 1, prefetchMixMemOps: True, loadLDSScale_A: False, loadLDSScale_B: False, swizzleScale: False, prefetchScale: False, streamK: False, numWGs: 0, streamKTwoTile: False, architecture: {'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=4096, N=4096, K=32768, alpha=2, beta=0, types={'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'None', 'scaleType_A': 'None', 'scale_B': 'None', 'scaleType_B': 'None', 'scaleBlockSize': -1, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=-1, workgroupMappingValue=-1)GEMMSolution(mac_m=128, mac_n=128, mac_k=128, wave_m=16, wave_n=16, wave_k=128, wave_b=1, workgroup_size_x=128, workgroup_size_y=2, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=False, betaInFma=True, direct2LDS_A=True, direct2LDS_B=True, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=True, prefetchInFlight=4, prefetchLDSFactor=1, prefetchMixMemOps=True, loadLDSScale_A=False, loadLDSScale_B=False, swizzleScale=False, prefetchScale=False, streamK=False, numWGs=0, streamKTwoTile=False, architecture={'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
+ 0.19% | p=3.1952e-05
| 3. FloatsGEMM(M: 4096, N: 4096, K: 32768, alpha: 2, beta: 0, types: {'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'None', 'scaleType_A': 'None', 'scale_B': 'None', 'scaleType_B': 'None', 'scaleBlockSize': -1, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: 1, workgroupMappingValue: -1)| GEMM(mac_m: 128, mac_n: 128, mac_k: 128, wave_m: 16, wave_n: 16, wave_k: 128, wave_b: 1, workgroup_size_x: 128, workgroup_size_y: 2, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: False, betaInFma: True, direct2LDS_A: True, direct2LDS_B: True, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: True, prefetchInFlight: 4, prefetchLDSFactor: 1, prefetchMixMemOps: True, loadLDSScale_A: False, loadLDSScale_B: False, swizzleScale: False, prefetchScale: False, streamK: False, numWGs: 0, streamKTwoTile: False, architecture: {'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=4096, N=4096, K=32768, alpha=2, beta=0, types={'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'None', 'scaleType_A': 'None', 'scale_B': 'None', 'scaleType_B': 'None', 'scaleBlockSize': -1, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=1, workgroupMappingValue=-1)GEMMSolution(mac_m=128, mac_n=128, mac_k=128, wave_m=16, wave_n=16, wave_k=128, wave_b=1, workgroup_size_x=128, workgroup_size_y=2, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=False, betaInFma=True, direct2LDS_A=True, direct2LDS_B=True, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=True, prefetchInFlight=4, prefetchLDSFactor=1, prefetchMixMemOps=True, loadLDSScale_A=False, loadLDSScale_B=False, swizzleScale=False, prefetchScale=False, streamK=False, numWGs=0, streamKTwoTile=False, architecture={'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
+ 0.26% | p=1.2004e-04
| 3. FloatsGEMM(M: 4096, N: 4096, K: 32768, alpha: 2, beta: 0, types: {'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: -1, workgroupMappingValue: -1)| GEMM(mac_m: 128, mac_n: 128, mac_k: 128, wave_m: 16, wave_n: 16, wave_k: 128, wave_b: 1, workgroup_size_x: 128, workgroup_size_y: 2, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: False, betaInFma: True, direct2LDS_A: True, direct2LDS_B: True, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: True, prefetchInFlight: 4, prefetchLDSFactor: 1, prefetchMixMemOps: True, loadLDSScale_A: False, loadLDSScale_B: False, swizzleScale: True, prefetchScale: True, streamK: False, numWGs: 0, streamKTwoTile: False, architecture: {'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=4096, N=4096, K=32768, alpha=2, beta=0, types={'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=-1, workgroupMappingValue=-1)GEMMSolution(mac_m=128, mac_n=128, mac_k=128, wave_m=16, wave_n=16, wave_k=128, wave_b=1, workgroup_size_x=128, workgroup_size_y=2, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=False, betaInFma=True, direct2LDS_A=True, direct2LDS_B=True, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=True, prefetchInFlight=4, prefetchLDSFactor=1, prefetchMixMemOps=True, loadLDSScale_A=False, loadLDSScale_B=False, swizzleScale=True, prefetchScale=True, streamK=False, numWGs=0, streamKTwoTile=False, architecture={'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
+ 0.59% | p=1.1015e-19
| 3. FloatsGEMM(M: 4096, N: 4096, K: 32768, alpha: 2, beta: 0, types: {'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: -1, workgroupMappingValue: -1)| GEMM(mac_m: 128, mac_n: 128, mac_k: 128, wave_m: 32, wave_n: 32, wave_k: 64, wave_b: 1, workgroup_size_x: 128, workgroup_size_y: 2, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: False, betaInFma: True, direct2LDS_A: True, direct2LDS_B: True, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: True, prefetchInFlight: 4, prefetchLDSFactor: 1, prefetchMixMemOps: True, loadLDSScale_A: False, loadLDSScale_B: False, swizzleScale: True, prefetchScale: True, streamK: False, numWGs: 0, streamKTwoTile: False, architecture: {'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=4096, N=4096, K=32768, alpha=2, beta=0, types={'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=-1, workgroupMappingValue=-1)GEMMSolution(mac_m=128, mac_n=128, mac_k=128, wave_m=32, wave_n=32, wave_k=64, wave_b=1, workgroup_size_x=128, workgroup_size_y=2, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=False, betaInFma=True, direct2LDS_A=True, direct2LDS_B=True, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=True, prefetchInFlight=4, prefetchLDSFactor=1, prefetchMixMemOps=True, loadLDSScale_A=False, loadLDSScale_B=False, swizzleScale=True, prefetchScale=True, streamK=False, numWGs=0, streamKTwoTile=False, architecture={'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
+ 0.30% | p=1.4393e-04
| 3. FloatsGEMM(M: 4096, N: 4096, K: 32768, alpha: 2, beta: 0, types: {'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: 1, workgroupMappingValue: -1)| GEMM(mac_m: 128, mac_n: 128, mac_k: 128, wave_m: 16, wave_n: 16, wave_k: 128, wave_b: 1, workgroup_size_x: 128, workgroup_size_y: 2, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: False, betaInFma: True, direct2LDS_A: True, direct2LDS_B: True, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: True, prefetchInFlight: 4, prefetchLDSFactor: 1, prefetchMixMemOps: True, loadLDSScale_A: False, loadLDSScale_B: False, swizzleScale: True, prefetchScale: True, streamK: False, numWGs: 0, streamKTwoTile: False, architecture: {'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=4096, N=4096, K=32768, alpha=2, beta=0, types={'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=1, workgroupMappingValue=-1)GEMMSolution(mac_m=128, mac_n=128, mac_k=128, wave_m=16, wave_n=16, wave_k=128, wave_b=1, workgroup_size_x=128, workgroup_size_y=2, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=False, betaInFma=True, direct2LDS_A=True, direct2LDS_B=True, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=True, prefetchInFlight=4, prefetchLDSFactor=1, prefetchMixMemOps=True, loadLDSScale_A=False, loadLDSScale_B=False, swizzleScale=True, prefetchScale=True, streamK=False, numWGs=0, streamKTwoTile=False, architecture={'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
+ 0.17% | p=4.8407e-03
| 3. FloatsGEMM(M: 4096, N: 4096, K: 32768, alpha: 2, beta: 0, types: {'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: 1, workgroupMappingValue: -1)| GEMM(mac_m: 128, mac_n: 128, mac_k: 128, wave_m: 32, wave_n: 32, wave_k: 64, wave_b: 1, workgroup_size_x: 128, workgroup_size_y: 2, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: False, betaInFma: True, direct2LDS_A: True, direct2LDS_B: True, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: True, prefetchInFlight: 4, prefetchLDSFactor: 1, prefetchMixMemOps: True, loadLDSScale_A: False, loadLDSScale_B: False, swizzleScale: True, prefetchScale: True, streamK: False, numWGs: 0, streamKTwoTile: False, architecture: {'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=4096, N=4096, K=32768, alpha=2, beta=0, types={'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=1, workgroupMappingValue=-1)GEMMSolution(mac_m=128, mac_n=128, mac_k=128, wave_m=32, wave_n=32, wave_k=64, wave_b=1, workgroup_size_x=128, workgroup_size_y=2, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=False, betaInFma=True, direct2LDS_A=True, direct2LDS_B=True, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=True, prefetchInFlight=4, prefetchLDSFactor=1, prefetchMixMemOps=True, loadLDSScale_A=False, loadLDSScale_B=False, swizzleScale=True, prefetchScale=True, streamK=False, numWGs=0, streamKTwoTile=False, architecture={'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
+ 0.15% | p=2.2561e-02
| 3. FloatsGEMM(M: 4096, N: 4096, K: 32768, alpha: 2, beta: 0, types: {'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': True, 'scaleShuffleTileA': [64, 4, 2], 'scaleShuffleTileB': [64, 4, 2]}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: 1, workgroupMappingValue: -1)| GEMM(mac_m: 128, mac_n: 128, mac_k: 128, wave_m: 32, wave_n: 32, wave_k: 64, wave_b: 1, workgroup_size_x: 128, workgroup_size_y: 2, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: False, betaInFma: True, direct2LDS_A: True, direct2LDS_B: True, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: True, prefetchInFlight: 4, prefetchLDSFactor: 1, prefetchMixMemOps: True, loadLDSScale_A: False, loadLDSScale_B: False, swizzleScale: True, prefetchScale: True, streamK: False, numWGs: 0, streamKTwoTile: False, architecture: {'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=4096, N=4096, K=32768, alpha=2, beta=0, types={'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': True, 'scaleShuffleTileA': [64, 4, 2], 'scaleShuffleTileB': [64, 4, 2]}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=1, workgroupMappingValue=-1)GEMMSolution(mac_m=128, mac_n=128, mac_k=128, wave_m=32, wave_n=32, wave_k=64, wave_b=1, workgroup_size_x=128, workgroup_size_y=2, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=False, betaInFma=True, direct2LDS_A=True, direct2LDS_B=True, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=True, prefetchInFlight=4, prefetchLDSFactor=1, prefetchMixMemOps=True, loadLDSScale_A=False, loadLDSScale_B=False, swizzleScale=True, prefetchScale=True, streamK=False, numWGs=0, streamKTwoTile=False, architecture={'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
+ 0.19% | p=4.8407e-03
| 3. FloatsGEMM(M: 4096, N: 4096, K: 32768, alpha: 2, beta: 0, types: {'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': True, 'scaleShuffleTileA': [64, 4, 4], 'scaleShuffleTileB': [64, 4, 4]}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: 1, workgroupMappingValue: -1)| GEMM(mac_m: 128, mac_n: 128, mac_k: 128, wave_m: 16, wave_n: 16, wave_k: 128, wave_b: 1, workgroup_size_x: 128, workgroup_size_y: 2, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: False, betaInFma: True, direct2LDS_A: True, direct2LDS_B: True, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: True, prefetchInFlight: 4, prefetchLDSFactor: 1, prefetchMixMemOps: True, loadLDSScale_A: False, loadLDSScale_B: False, swizzleScale: True, prefetchScale: True, streamK: False, numWGs: 0, streamKTwoTile: False, architecture: {'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=4096, N=4096, K=32768, alpha=2, beta=0, types={'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': True, 'scaleShuffleTileA': [64, 4, 4], 'scaleShuffleTileB': [64, 4, 4]}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=1, workgroupMappingValue=-1)GEMMSolution(mac_m=128, mac_n=128, mac_k=128, wave_m=16, wave_n=16, wave_k=128, wave_b=1, workgroup_size_x=128, workgroup_size_y=2, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=False, betaInFma=True, direct2LDS_A=True, direct2LDS_B=True, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=True, prefetchInFlight=4, prefetchLDSFactor=1, prefetchMixMemOps=True, loadLDSScale_A=False, loadLDSScale_B=False, swizzleScale=True, prefetchScale=True, streamK=False, numWGs=0, streamKTwoTile=False, architecture={'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
+ 1.77% | p=1.2736e-51
| 3. FloatsGEMM(M: 4096, N: 4096, K: 32768, alpha: 2, beta: 0.5, types: {'type_A': 'fp4', 'type_B': 'fp4', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: -1, workgroupMappingValue: -1)| GEMM(mac_m: 256, mac_n: 256, mac_k: 128, wave_m: 32, wave_n: 32, wave_k: 64, wave_b: 1, workgroup_size_x: 128, workgroup_size_y: 2, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: True, betaInFma: True, direct2LDS_A: False, direct2LDS_B: False, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: True, prefetchInFlight: 2, prefetchLDSFactor: 2, prefetchMixMemOps: False, loadLDSScale_A: True, loadLDSScale_B: True, swizzleScale: False, prefetchScale: False, streamK: False, numWGs: 0, streamKTwoTile: False, architecture: {'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=4096, N=4096, K=32768, alpha=2, beta=0.5, types={'type_A': 'fp4', 'type_B': 'fp4', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=-1, workgroupMappingValue=-1)GEMMSolution(mac_m=256, mac_n=256, mac_k=128, wave_m=32, wave_n=32, wave_k=64, wave_b=1, workgroup_size_x=128, workgroup_size_y=2, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=True, betaInFma=True, direct2LDS_A=False, direct2LDS_B=False, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=True, prefetchInFlight=2, prefetchLDSFactor=2, prefetchMixMemOps=False, loadLDSScale_A=True, loadLDSScale_B=True, swizzleScale=False, prefetchScale=False, streamK=False, numWGs=0, streamKTwoTile=False, architecture={'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
+ 0.33% | p=1.7451e-03
| 3. FloatsGEMM(M: 4096, N: 4096, K: 32768, alpha: 2, beta: 0.5, types: {'type_A': 'fp4', 'type_B': 'fp4', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: -1, workgroupMappingValue: -1)| GEMM(mac_m: 256, mac_n: 256, mac_k: 128, wave_m: 32, wave_n: 32, wave_k: 64, wave_b: 1, workgroup_size_x: 128, workgroup_size_y: 2, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: True, betaInFma: True, direct2LDS_A: True, direct2LDS_B: True, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: True, prefetchInFlight: 2, prefetchLDSFactor: 1, prefetchMixMemOps: False, loadLDSScale_A: True, loadLDSScale_B: True, swizzleScale: False, prefetchScale: False, streamK: False, numWGs: 0, streamKTwoTile: False, architecture: {'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=4096, N=4096, K=32768, alpha=2, beta=0.5, types={'type_A': 'fp4', 'type_B': 'fp4', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=-1, workgroupMappingValue=-1)GEMMSolution(mac_m=256, mac_n=256, mac_k=128, wave_m=32, wave_n=32, wave_k=64, wave_b=1, workgroup_size_x=128, workgroup_size_y=2, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=True, betaInFma=True, direct2LDS_A=True, direct2LDS_B=True, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=True, prefetchInFlight=2, prefetchLDSFactor=1, prefetchMixMemOps=False, loadLDSScale_A=True, loadLDSScale_B=True, swizzleScale=False, prefetchScale=False, streamK=False, numWGs=0, streamKTwoTile=False, architecture={'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
+ 0.32% | p=4.8404e-03
| 3. FloatsGEMM(M: 4096, N: 4096, K: 32768, alpha: 2, beta: 0.5, types: {'type_A': 'fp4', 'type_B': 'fp4', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: 0, workgroupMappingValue: -1)| GEMM(mac_m: 256, mac_n: 256, mac_k: 128, wave_m: 32, wave_n: 32, wave_k: 64, wave_b: 1, workgroup_size_x: 128, workgroup_size_y: 2, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: True, betaInFma: True, direct2LDS_A: True, direct2LDS_B: True, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: True, prefetchInFlight: 2, prefetchLDSFactor: 1, prefetchMixMemOps: False, loadLDSScale_A: True, loadLDSScale_B: True, swizzleScale: False, prefetchScale: False, streamK: False, numWGs: 0, streamKTwoTile: False, architecture: {'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=4096, N=4096, K=32768, alpha=2, beta=0.5, types={'type_A': 'fp4', 'type_B': 'fp4', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=0, workgroupMappingValue=-1)GEMMSolution(mac_m=256, mac_n=256, mac_k=128, wave_m=32, wave_n=32, wave_k=64, wave_b=1, workgroup_size_x=128, workgroup_size_y=2, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=True, betaInFma=True, direct2LDS_A=True, direct2LDS_B=True, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=True, prefetchInFlight=2, prefetchLDSFactor=1, prefetchMixMemOps=False, loadLDSScale_A=True, loadLDSScale_B=True, swizzleScale=False, prefetchScale=False, streamK=False, numWGs=0, streamKTwoTile=False, architecture={'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
+ 0.59% | p=1.3014e-16
| 3. FloatsGEMM(M: 4096, N: 4096, K: 32768, alpha: 2, beta: 0.5, types: {'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: -1, workgroupMappingValue: -1)| GEMM(mac_m: 128, mac_n: 256, mac_k: 128, wave_m: 16, wave_n: 16, wave_k: 128, wave_b: 1, workgroup_size_x: 128, workgroup_size_y: 2, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: True, betaInFma: True, direct2LDS_A: False, direct2LDS_B: False, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: True, prefetchInFlight: 2, prefetchLDSFactor: 2, prefetchMixMemOps: False, loadLDSScale_A: True, loadLDSScale_B: True, swizzleScale: False, prefetchScale: False, streamK: False, numWGs: 0, streamKTwoTile: False, architecture: {'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=4096, N=4096, K=32768, alpha=2, beta=0.5, types={'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=-1, workgroupMappingValue=-1)GEMMSolution(mac_m=128, mac_n=256, mac_k=128, wave_m=16, wave_n=16, wave_k=128, wave_b=1, workgroup_size_x=128, workgroup_size_y=2, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=True, betaInFma=True, direct2LDS_A=False, direct2LDS_B=False, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=True, prefetchInFlight=2, prefetchLDSFactor=2, prefetchMixMemOps=False, loadLDSScale_A=True, loadLDSScale_B=True, swizzleScale=False, prefetchScale=False, streamK=False, numWGs=0, streamKTwoTile=False, architecture={'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
+ 1.22% | p=1.1300e-84
| 3. FloatsGEMM(M: 4096, N: 4096, K: 32768, alpha: 2, beta: 0.5, types: {'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: -1, workgroupMappingValue: -1)| GEMM(mac_m: 256, mac_n: 128, mac_k: 128, wave_m: 16, wave_n: 16, wave_k: 128, wave_b: 1, workgroup_size_x: 128, workgroup_size_y: 2, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: True, betaInFma: True, direct2LDS_A: False, direct2LDS_B: False, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: True, prefetchInFlight: 2, prefetchLDSFactor: 2, prefetchMixMemOps: False, loadLDSScale_A: True, loadLDSScale_B: True, swizzleScale: False, prefetchScale: False, streamK: False, numWGs: 0, streamKTwoTile: False, architecture: {'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=4096, N=4096, K=32768, alpha=2, beta=0.5, types={'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=-1, workgroupMappingValue=-1)GEMMSolution(mac_m=256, mac_n=128, mac_k=128, wave_m=16, wave_n=16, wave_k=128, wave_b=1, workgroup_size_x=128, workgroup_size_y=2, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=True, betaInFma=True, direct2LDS_A=False, direct2LDS_B=False, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=True, prefetchInFlight=2, prefetchLDSFactor=2, prefetchMixMemOps=False, loadLDSScale_A=True, loadLDSScale_B=True, swizzleScale=False, prefetchScale=False, streamK=False, numWGs=0, streamKTwoTile=False, architecture={'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
- 0.08% | p=5.7413e-04
| 3. FloatsGEMM(M: 4096, N: 4096, K: 32768, alpha: 2, beta: 0.5, types: {'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: -1, workgroupMappingValue: -1)| GEMM(mac_m: 256, mac_n: 256, mac_k: 128, wave_m: 32, wave_n: 32, wave_k: 64, wave_b: 1, workgroup_size_x: 128, workgroup_size_y: 2, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: True, betaInFma: True, direct2LDS_A: True, direct2LDS_B: True, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: True, prefetchInFlight: 2, prefetchLDSFactor: 1, prefetchMixMemOps: False, loadLDSScale_A: True, loadLDSScale_B: True, swizzleScale: False, prefetchScale: False, streamK: False, numWGs: 0, streamKTwoTile: False, architecture: {'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=4096, N=4096, K=32768, alpha=2, beta=0.5, types={'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=-1, workgroupMappingValue=-1)GEMMSolution(mac_m=256, mac_n=256, mac_k=128, wave_m=32, wave_n=32, wave_k=64, wave_b=1, workgroup_size_x=128, workgroup_size_y=2, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=True, betaInFma=True, direct2LDS_A=True, direct2LDS_B=True, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=True, prefetchInFlight=2, prefetchLDSFactor=1, prefetchMixMemOps=False, loadLDSScale_A=True, loadLDSScale_B=True, swizzleScale=False, prefetchScale=False, streamK=False, numWGs=0, streamKTwoTile=False, architecture={'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
+ 0.10% | p=5.0770e-06
| 3. FloatsGEMM(M: 4096, N: 4096, K: 32768, alpha: 2, beta: 0.5, types: {'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: 0, workgroupMappingValue: -1)| GEMM(mac_m: 256, mac_n: 256, mac_k: 128, wave_m: 32, wave_n: 32, wave_k: 64, wave_b: 1, workgroup_size_x: 128, workgroup_size_y: 2, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: True, betaInFma: True, direct2LDS_A: True, direct2LDS_B: True, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: True, prefetchInFlight: 2, prefetchLDSFactor: 1, prefetchMixMemOps: False, loadLDSScale_A: True, loadLDSScale_B: True, swizzleScale: False, prefetchScale: False, streamK: False, numWGs: 0, streamKTwoTile: False, architecture: {'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=4096, N=4096, K=32768, alpha=2, beta=0.5, types={'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'half', 'type_D': 'half', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'N', 'scale_A': 'Separate', 'scaleType_A': 'E8M0', 'scale_B': 'Separate', 'scaleType_B': 'E8M0', 'scaleBlockSize': 32, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=0, workgroupMappingValue=-1)GEMMSolution(mac_m=256, mac_n=256, mac_k=128, wave_m=32, wave_n=32, wave_k=64, wave_b=1, workgroup_size_x=128, workgroup_size_y=2, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=True, betaInFma=True, direct2LDS_A=True, direct2LDS_B=True, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=True, prefetchInFlight=2, prefetchLDSFactor=1, prefetchMixMemOps=False, loadLDSScale_A=True, loadLDSScale_B=True, swizzleScale=False, prefetchScale=False, streamK=False, numWGs=0, streamKTwoTile=False, architecture={'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
+ 2.73% | p=5.6994e-05
| 3. FloatsGEMM(M: 4096, N: 4096, K: 8192, alpha: 2, beta: 0.5, types: {'type_A': 'bf8', 'type_B': 'bf8', 'type_C': 'float', 'type_D': 'float', 'type_acc': 'float', 'trans_A': 'N', 'trans_B': 'T', 'scale_A': 'None', 'scaleType_A': 'None', 'scale_B': 'None', 'scaleType_B': 'None', 'scaleBlockSize': -1, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: -1, workgroupMappingValue: -1)| GEMM(mac_m: 64, mac_n: 64, mac_k: 128, wave_m: 16, wave_n: 16, wave_k: 128, wave_b: 1, workgroup_size_x: 256, workgroup_size_y: 1, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: True, betaInFma: True, direct2LDS_A: False, direct2LDS_B: False, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: True, prefetchInFlight: 2, prefetchLDSFactor: 0, prefetchMixMemOps: False, loadLDSScale_A: False, loadLDSScale_B: False, swizzleScale: False, prefetchScale: False, streamK: False, numWGs: 0, streamKTwoTile: False, architecture: {'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=4096, N=4096, K=8192, alpha=2, beta=0.5, types={'type_A': 'bf8', 'type_B': 'bf8', 'type_C': 'float', 'type_D': 'float', 'type_acc': 'float', 'trans_A': 'N', 'trans_B': 'T', 'scale_A': 'None', 'scaleType_A': 'None', 'scale_B': 'None', 'scaleType_B': 'None', 'scaleBlockSize': -1, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=-1, workgroupMappingValue=-1)GEMMSolution(mac_m=64, mac_n=64, mac_k=128, wave_m=16, wave_n=16, wave_k=128, wave_b=1, workgroup_size_x=256, workgroup_size_y=1, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=True, betaInFma=True, direct2LDS_A=False, direct2LDS_B=False, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=True, prefetchInFlight=2, prefetchLDSFactor=0, prefetchMixMemOps=False, loadLDSScale_A=False, loadLDSScale_B=False, swizzleScale=False, prefetchScale=False, streamK=False, numWGs=0, streamKTwoTile=False, architecture={'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
+ 1.98% | p=5.6994e-05
| 3. FloatsGEMM(M: 4096, N: 4096, K: 8192, alpha: 2, beta: 0.5, types: {'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'float', 'type_D': 'float', 'type_acc': 'float', 'trans_A': 'N', 'trans_B': 'N', 'scale_A': 'None', 'scaleType_A': 'None', 'scale_B': 'None', 'scaleType_B': 'None', 'scaleBlockSize': -1, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: -1, workgroupMappingValue: -1)| GEMM(mac_m: 128, mac_n: 128, mac_k: 64, wave_m: 32, wave_n: 32, wave_k: 64, wave_b: 1, workgroup_size_x: 256, workgroup_size_y: 1, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: True, betaInFma: True, direct2LDS_A: False, direct2LDS_B: False, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: True, prefetchInFlight: 2, prefetchLDSFactor: 0, prefetchMixMemOps: False, loadLDSScale_A: False, loadLDSScale_B: False, swizzleScale: False, prefetchScale: False, streamK: False, numWGs: 0, streamKTwoTile: False, architecture: {'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=4096, N=4096, K=8192, alpha=2, beta=0.5, types={'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'float', 'type_D': 'float', 'type_acc': 'float', 'trans_A': 'N', 'trans_B': 'N', 'scale_A': 'None', 'scaleType_A': 'None', 'scale_B': 'None', 'scaleType_B': 'None', 'scaleBlockSize': -1, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=-1, workgroupMappingValue=-1)GEMMSolution(mac_m=128, mac_n=128, mac_k=64, wave_m=32, wave_n=32, wave_k=64, wave_b=1, workgroup_size_x=256, workgroup_size_y=1, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=True, betaInFma=True, direct2LDS_A=False, direct2LDS_B=False, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=True, prefetchInFlight=2, prefetchLDSFactor=0, prefetchMixMemOps=False, loadLDSScale_A=False, loadLDSScale_B=False, swizzleScale=False, prefetchScale=False, streamK=False, numWGs=0, streamKTwoTile=False, architecture={'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
+ 2.45% | p=1.7451e-03
| 3. FloatsGEMM(M: 4096, N: 4096, K: 8192, alpha: 2, beta: 0.5, types: {'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'float', 'type_D': 'float', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'T', 'scale_A': 'None', 'scaleType_A': 'None', 'scale_B': 'None', 'scaleType_B': 'None', 'scaleBlockSize': -1, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A: 1, scaleValue_B: 1, workgroupMappingDim: -1, workgroupMappingValue: -1)| GEMM(mac_m: 128, mac_n: 128, mac_k: 64, wave_m: 32, wave_n: 32, wave_k: 64, wave_b: 1, workgroup_size_x: 256, workgroup_size_y: 1, workgroupRemapXCC: False, workgroupRemapXCCValue: -1, unroll_x: 0, unroll_y: 0, loadLDS_A: True, loadLDS_B: True, storeLDS_D: True, betaInFma: True, direct2LDS_A: False, direct2LDS_B: False, scheduler: Priority, schedulerCost: LinearWeighted, prefetch: True, prefetchInFlight: 2, prefetchLDSFactor: 0, prefetchMixMemOps: False, loadLDSScale_A: False, loadLDSScale_B: False, swizzleScale: False, prefetchScale: False, streamK: False, numWGs: 0, streamKTwoTile: False, architecture: {'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess: True, version: ) | GEMMProblem(M=4096, N=4096, K=8192, alpha=2, beta=0.5, types={'type_A': 'fp8', 'type_B': 'fp8', 'type_C': 'float', 'type_D': 'float', 'type_acc': 'float', 'trans_A': 'T', 'trans_B': 'T', 'scale_A': 'None', 'scaleType_A': 'None', 'scale_B': 'None', 'scaleType_B': 'None', 'scaleBlockSize': -1, 'scaleSkipPermlane': False, 'scaleShuffleTileA': [], 'scaleShuffleTileB': []}, scaleValue_A=1, scaleValue_B=1, workgroupMappingDim=-1, workgroupMappingValue=-1)GEMMSolution(mac_m=128, mac_n=128, mac_k=64, wave_m=32, wave_n=32, wave_k=64, wave_b=1, workgroup_size_x=256, workgroup_size_y=1, workgroupRemapXCC=False, workgroupRemapXCCValue=-1, unroll_x=0, unroll_y=0, loadLDS_A=True, loadLDS_B=True, storeLDS_D=True, betaInFma=True, direct2LDS_A=False, direct2LDS_B=False, scheduler='Priority', schedulerCost='LinearWeighted', prefetch=True, prefetchInFlight=2, prefetchLDSFactor=0, prefetchMixMemOps=False, loadLDSScale_A=False, loadLDSScale_B=False, swizzleScale=False, prefetchScale=False, streamK=False, numWGs=0, streamKTwoTile=False, architecture={'ArchString': 'gfx950', 'Xnack': False, 'Sramecc': True}, matchMemoryAccess=True, version='')
Full table of results
|
Contributor
There was a problem hiding this comment.
Pull Request Overview
This PR adds unique IDs to benchmarks in the rocRoller performance testing suite to enable filtering and running specific benchmarks by ID. The ID system is based on dataclass fields that participate in comparison operations.
- Implements ID generation using SHA1 hashes of dataclass fields marked for comparison
- Replaces the unused
filterargument withid_filterfor benchmark selection - Improves default value handling and validation in the GEMM client
Reviewed Changes
Copilot reviewed 8 out of 8 changed files in this pull request and generated 2 comments.
Show a summary per file
| File | Description |
|---|---|
| shared/rocroller/scripts/lib/rrperf/utils.py | Adds get_dataclass_id function for generating SHA1-based IDs from dataclass fields |
| shared/rocroller/scripts/lib/rrperf/run.py | Replaces filter with id_filter parameter and adds ID display in output |
| shared/rocroller/scripts/lib/rrperf/problems.py | Adds id properties to dataclasses and marks output fields as non-comparable |
| shared/rocroller/scripts/lib/rrperf/ls.py | Updates to display benchmark IDs and use improved terminal size handling |
| shared/rocroller/scripts/lib/rrperf/generate.py | Adds validation for missing suite specification |
| shared/rocroller/scripts/lib/rrperf/autoperf.py | Updates parameter name from filter to id_filter |
| shared/rocroller/scripts/lib/rrperf/args.py | Adds id_filter argument definition |
| shared/rocroller/client/src/gemm.cpp | Improves default value handling and adds validation for macro/wave tile sizes |
Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.
jaopaulolc
previously approved these changes
Sep 17, 2025
memmett
reviewed
Sep 23, 2025
jaopaulolc
approved these changes
Sep 29, 2025
assistant-librarian Bot
pushed a commit
to ROCm/rocRoller
that referenced
this pull request
Sep 29, 2025
[rocRoller] Add unique IDs to rrsuite - Adds IDs to benchmarks in rrsuites.py based on the run parameters to facilitate running specific failures in a run. - Fixes already_run check in run.py by adding compare=False to output log path and other similar variables - Improves default value handling in the gemm.cpp - Adds check for invalid mac/wave configuration in the GEMM client
qiangpan2
pushed a commit
to qiangpan2/rocm-libraries
that referenced
this pull request
May 29, 2026
* topk_softmax * remove some file * fix atomix linear_offset * address various comment, and change sfc get_index api to static(tuple) [ROCm/composable_kernel commit: b098b71]
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Motivation
Add IDs to benchmarks in
rrsuites.pybased on the run parameters to facilitate running specific failures in a run.Also adds a few improvements to rrperf and the GEMM client.
Technical Details
filterargument withid_filterto implement idsrun.pyby addingcompare=Falseto output log path and other similar variablesgemm.cppTest Plan
Current test suites can be run to check the new ids in each benchmark, and the flag
--id-filteris used to select a benchmark.Submission Checklist