diff options
| author | Sam Parker <sam.parker@arm.com> | 2025-10-13 16:50:53 +0100 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2025-10-13 16:50:53 +0100 |
| commit | 1820102167a5ace14a5c1f79d11d5eb4cce93001 (patch) | |
| tree | 4e5f2e07f3cb292261291a1714ce2f2a2e34f500 /cross-project-tests | |
| parent | 095cad6add16df3f6273f5b24293e48a08e3230e (diff) | |
Wasm fmuladd relaxed (#163177)
Reland #161355, after fixing up the cross-projects-tests for the wasm
simd intrinsics.
Original commit message:
Lower v4f32 and v2f64 fmuladd calls to relaxed_madd instructions.
If we have FP16, then lower v8f16 fmuladds to FMA.
I've introduced an ISD node for fmuladd to maintain the rounding
ambiguity through legalization / combine / isel.
Diffstat (limited to 'cross-project-tests')
| -rw-r--r-- | cross-project-tests/intrinsic-header-tests/wasm_simd128.c | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/cross-project-tests/intrinsic-header-tests/wasm_simd128.c b/cross-project-tests/intrinsic-header-tests/wasm_simd128.c index b601d90cfcc9..0f9ef39e4ca6 100644 --- a/cross-project-tests/intrinsic-header-tests/wasm_simd128.c +++ b/cross-project-tests/intrinsic-header-tests/wasm_simd128.c @@ -1511,13 +1511,13 @@ v128_t test_f16x8_convert_u16x8(v128_t a) { } // CHECK-LABEL: test_f16x8_relaxed_madd: -// CHECK: f16x8.relaxed_madd{{$}} +// CHECK: f16x8.madd{{$}} v128_t test_f16x8_relaxed_madd(v128_t a, v128_t b, v128_t c) { return wasm_f16x8_relaxed_madd(a, b, c); } // CHECK-LABEL: test_f16x8_relaxed_nmadd: -// CHECK: f16x8.relaxed_nmadd{{$}} +// CHECK: f16x8.nmadd{{$}} v128_t test_f16x8_relaxed_nmadd(v128_t a, v128_t b, v128_t c) { return wasm_f16x8_relaxed_nmadd(a, b, c); } |