Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[BACKEND] Fix transpose optimization missed during refactor #5226

Merged
merged 1 commit into from
Nov 22, 2024
Merged

[BACKEND] Fix transpose optimization missed during refactor #5226

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 2 additions & 2 deletions lib/Dialect/TritonGPU/Transforms/OptimizeDotOperands.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -326,13 +326,13 @@ class FuseTransHopper : public OpRewritePattern<LocalAllocOp> {
return failure();

// Match outerCvt(trans(innerCvt(x))).
auto trans = allocOp.getSrc().getDefiningOp<MemDescTransOp>();
auto trans = allocOp.getSrc().getDefiningOp<TransOp>();
if (!trans || trans.getOrder() != ArrayRef<int32_t>({1, 0}))
return failure();

MemDescType allocType = allocOp.getType();
auto allocEncoding = cast<SharedEncodingAttr>(allocType.getEncoding());
MemDescType srcTy = trans.getSrc().getType();
RankedTensorType srcTy = trans.getSrc().getType();

// MMAv3 with transpose only supports f16 and bf16. Fall back to MMAv3
// without transpose for other data types.)
Expand Down
20 changes: 20 additions & 0 deletions test/TritonGPU/dot-operands.mlir
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -256,3 +256,23 @@ module attributes {"triton_gpu.target" = "cuda:90", "triton_gpu.num-ctas" = 1 :
tt.return %r : tensor<128x64xf32, #mma>
}
}

// -----

#blocked = #triton_gpu.blocked<{sizePerThread = [16, 1], threadsPerWarp = [32, 1], warpsPerCTA = [4, 1], order = [1, 0]}>
#blocked1 = #triton_gpu.blocked<{sizePerThread = [1, 16], threadsPerWarp = [1, 32], warpsPerCTA = [1, 4], order = [0, 1]}>
#mma = #triton_gpu.nvidia_mma<{versionMajor = 3, versionMinor = 0, warpsPerCTA = [4, 1], instrShape = [16, 64, 16]}>
#shared = #triton_gpu.shared<{vec = 8, perPhase = 1, maxPhase = 8, order = [1, 0], hasLeadingOffset = true}>
#shared1 = #triton_gpu.shared<{vec = 8, perPhase = 1, maxPhase = 8, order = [1, 0], hasLeadingOffset = true}>
module attributes {"triton_gpu.target" = "cuda:90", "triton_gpu.num-ctas" = 1 : i32, "triton_gpu.num-warps" = 4 : i32, "triton_gpu.threads-per-warp" = 32 : i32} {
// CHECK-LABEL: mma_reorder_transpose
// CHECK: triton_gpu.local_alloc
// CHECK: triton_gpu.memdesc_trans
// CHECK: triton_nvidia_gpu.warp_group_dot
tt.func @mma_reorder_transpose(%t: tensor<64x128xf16, #blocked1>, %dotb: !triton_gpu.memdesc<64x64xf16, #shared>, %dotc: tensor<128x64xf32, #mma>) -> tensor<128x64xf32, #mma>{
%a = tt.trans %t {order = array<i32: 1, 0>} : tensor<64x128xf16, #blocked1> -> tensor<128x64xf16, #blocked>
%dota = triton_gpu.local_alloc %a: (tensor<128x64xf16, #blocked>) -> !triton_gpu.memdesc<128x64xf16, #shared1>
%r = triton_nvidia_gpu.warp_group_dot %dota, %dotb, %dotc : !triton_gpu.memdesc<128x64xf16, #shared1> * !triton_gpu.memdesc<64x64xf16, #shared> -> tensor<128x64xf32, #mma>
tt.return %r : tensor<128x64xf32, #mma>
}
}
Loading