Skip to content

Use llvm.fma for tt.dot lowering #254

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

Open
abrown wants to merge 4 commits into
base: main
Choose a base branch
from
Open

Use llvm.fma for tt.dot lowering #254

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
8cee28e
Use `llvm.fma` for `tt.dot` lowering
abrown Jan 26, 2026
23c84ab
Apply clang formatting
abrown Jan 26, 2026
f17f266
review: remove fastmath flags
abrown Jan 28, 2026
df6cae7
Update lit tests
abrown Jan 29, 2026
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
3 changes: 1 addition & 2 deletions cpu/lib/TritonCPUToLLVM/DotOpToLLVM.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -49,8 +49,7 @@ class GenericFMAVectorMultiplier : public triton::gpu::FMAVectorMultiplier {
}

// Multiply and accumulate.
auto mul = LLVM::FMulOp::create(builder, loc, tgtTy, aElem, bElem);
accum = LLVM::FAddOp::create(builder, loc, tgtTy, accum, mul);
accum = LLVM::FMAOp::create(builder, loc, tgtTy, aElem, bElem, accum);
}
return accum;
}
Expand Down
6 changes: 2 additions & 4 deletions test/Conversion/dot.mlir
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,8 +13,7 @@ module attributes {"ttg.num-ctas" = 1 : i32, "ttg.num-warps" = 1 : i32, ttg.shar
// COM: We should see a bunch of repetitions of this pattern:
// CHECK: [[A:%.*]] = llvm.fpext {{%.*}} : f16 to f32
// CHECK: [[B:%.*]] = llvm.fpext {{%.*}} : f16 to f32
// CHECK: [[MUL:%.*]] = llvm.fmul [[A]], [[B]] : f32
// CHECK: {{%.*}} = llvm.fadd {{%.*}}, [[MUL]] : f32
// CHECK: [[MUL:%.*]] = llvm.intr.fma([[A]], [[B]], {{%.*}}) : (f32, f32, f32) -> f32
tt.return %d : tensor<2x2xf32, #blocked>
}
}
Expand All @@ -34,8 +33,7 @@ module attributes {"ttg.num-ctas" = 1 : i32, "ttg.num-warps" = 1 : i32, ttg.shar
// COM: We should see a bunch of repetitions of this pattern:
// CHECK: [[A:%.*]] = llvm.fpext {{%.*}} : bf16 to f32
// CHECK: [[B:%.*]] = llvm.fpext {{%.*}} : bf16 to f32
// CHECK: [[MUL:%.*]] = llvm.fmul [[A]], [[B]] : f32
// CHECK: {{%.*}} = llvm.fadd {{%.*}}, [[MUL]] : f32
// CHECK: [[MUL:%.*]] = llvm.intr.fma([[A]], [[B]], {{%.*}}) : (f32, f32, f32) -> f32
tt.return %d : tensor<2x2xf32, #blocked>
}
}