cranelift: Align Scalar and SIMD shift semantics (#4520)

* cranelift: Reorganize test suite

Group some SIMD operations by instruction.

* cranelift: Deduplicate some shift tests

Also, new tests with the mod behaviour

* aarch64: Lower shifts with mod behaviour

* x64: Lower shifts with mod behaviour

* wasmtime: Don't mask SIMD shifts
This commit is contained in:
Afonso Bordado
2022-07-27 18:54:00 +01:00
committed by GitHub
parent e121c209fc
commit 0508932174
15 changed files with 314 additions and 423 deletions

View File

@@ -344,9 +344,10 @@ block0(v0: i8x16):
; block0:
; movz x3, #1
; sub w5, wzr, w3
; dup v7.16b, w5
; ushl v0.16b, v0.16b, v7.16b
; and w5, w3, #7
; sub x7, xzr, x5
; dup v17.16b, w7
; ushl v0.16b, v0.16b, v17.16b
; ret
function %add_i128(i128, i128) -> i128 {
@@ -492,4 +493,3 @@ block0(v0: i64):
; b.vc 8 ; udf
; sdiv x0, x0, x3
; ret