Skip to content

Add lit tests; shows initial bfloat16 support for tt.dot #253

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

Merged
abrown merged 2 commits into from
Jan 29, 2026
Merged

Add lit tests; shows initial bfloat16 support for tt.dot #253

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
80000fe
Add lit tests; shows initial `bfloat16` support
abrown Jan 24, 2026
d250c84
review: add --canonicalize pass
abrown Jan 28, 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
41 changes: 41 additions & 0 deletions test/Conversion/dot.mlir
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,41 @@
// RUN: triton-opt %s -split-input-file --convert-triton-cpu-to-llvm --canonicalize | FileCheck %s

#blocked = #ttg.blocked<{sizePerThread = [2, 2], threadsPerWarp = [1, 1], warpsPerCTA = [1, 1], order = [1, 0]}>
module attributes {"ttg.num-ctas" = 1 : i32, "ttg.num-warps" = 1 : i32, ttg.shared = 0 : i32, ttg.target = "cpu", "ttg.threads-per-warp" = 1 : i32} {
tt.func public @kernel(
%a: tensor<2x4xf16, #ttg.dot_op<{opIdx = 0, parent = #blocked}>>,
%b: tensor<4x2xf16, #ttg.dot_op<{opIdx = 1, parent = #blocked}>>,
%c: tensor<2x2xf32, #blocked>) -> tensor<2x2xf32, #blocked> attributes {noinline = false} {
%d = tt.dot %a, %b, %c :
tensor<2x4xf16, #ttg.dot_op<{opIdx = 0, parent = #blocked}>> *
tensor<4x2xf16, #ttg.dot_op<{opIdx = 1, parent = #blocked}>> ->
tensor<2x2xf32, #blocked>
// 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
tt.return %d : tensor<2x2xf32, #blocked>
}
}

// -----

#blocked = #ttg.blocked<{sizePerThread = [2, 2], threadsPerWarp = [1, 1], warpsPerCTA = [1, 1], order = [1, 0]}>
module attributes {"ttg.num-ctas" = 1 : i32, "ttg.num-warps" = 1 : i32, ttg.shared = 0 : i32, ttg.target = "cpu", "ttg.threads-per-warp" = 1 : i32} {
tt.func public @kernel(
%a: tensor<2x4xbf16, #ttg.dot_op<{opIdx = 0, parent = #blocked}>>,
%b: tensor<4x2xbf16, #ttg.dot_op<{opIdx = 1, parent = #blocked}>>,
%c: tensor<2x2xf32, #blocked>) -> tensor<2x2xf32, #blocked> attributes {noinline = false} {
%d = tt.dot %a, %b, %c :
tensor<2x4xbf16, #ttg.dot_op<{opIdx = 0, parent = #blocked}>> *
tensor<4x2xbf16, #ttg.dot_op<{opIdx = 1, parent = #blocked}>> ->
tensor<2x2xf32, #blocked>
// 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
tt.return %d : tensor<2x2xf32, #blocked>
}
}