Skip to content

A lightweight WebAssembly runtime that is fast, secure, and standards-compliant

x64: Fix typo in `splat` lowering (#12986)

Alex Crichton37ef87db5078
diff --git a/cranelift/codegen/src/isa/x64/lower.isle b/cranelift/codegen/src/isa/x64/lower.isle
index 9819ef8c87..ac904f6944 100644
--- a/cranelift/codegen/src/isa/x64/lower.isle
+++ b/cranelift/codegen/src/isa/x64/lower.isle
@@ -4876,7 +4876,7 @@
;; on AVX2 to leverage that specific instruction for this operation.
(rule 0 (lower (has_type $F32X4 (splat _ src)))
(let ((tmp Xmm src))
- (x64_shufps src src 0)))
+ (x64_shufps tmp tmp 0)))
(rule 1 (lower (has_type $F32X4 (splat _ src)))
(if-let true (use_avx2))
(x64_vbroadcastss src))
diff --git a/cranelift/filetests/filetests/isa/aarch64/nan-canonicalization.clif b/cranelift/filetests/filetests/isa/aarch64/nan-canonicalization.clif
index 619b245e1a..af76af4e75 100644
--- a/cranelift/filetests/filetests/isa/aarch64/nan-canonicalization.clif
+++ b/cranelift/filetests/filetests/isa/aarch64/nan-canonicalization.clif
@@ -15,7 +15,7 @@ block0(v0: f32x4, v1: f32x4):
; addps %xmm1, %xmm0
; movl $0x7fc00000, %edi
; movd %edi, %xmm7
-; shufps $0x0, (%rip), %xmm7
+; shufps $0x0, %xmm7, %xmm7
; movdqa %xmm0, %xmm1
; cmpunordps %xmm0, %xmm1
; movdqa %xmm0, %xmm2
@@ -35,7 +35,7 @@ block0(v0: f32x4, v1: f32x4):
; addps %xmm1, %xmm0
; movl $0x7fc00000, %edi
; movd %edi, %xmm7
-; shufps $0, 0x28(%rip), %xmm7
+; shufps $0, %xmm7, %xmm7
; movdqa %xmm0, %xmm1
; cmpunordps %xmm0, %xmm1
; movdqa %xmm0, %xmm2
@@ -46,16 +46,6 @@ block0(v0: f32x4, v1: f32x4):
; movq %rbp, %rsp
; popq %rbp
; retq
-; addb %al, (%rax)
-; addb %al, (%rax)
-; addb %al, (%rax)
-; addb %al, (%rax)
-; sarb $0, (%rdi)
-; addb %al, (%rax)
-; addb %al, (%rax)
-; addb %al, (%rax)
-; addb %al, (%rax)
-; addb %al, (%rax)
function %f1(f64, f64) -> f64 {
block0(v0: f64, v1: f64):
diff --git a/cranelift/filetests/filetests/isa/x64/nan-canonicalization-sse41.clif b/cranelift/filetests/filetests/isa/x64/nan-canonicalization-sse41.clif
index 830008b45e..98fee9677b 100644
--- a/cranelift/filetests/filetests/isa/x64/nan-canonicalization-sse41.clif
+++ b/cranelift/filetests/filetests/isa/x64/nan-canonicalization-sse41.clif
@@ -15,7 +15,7 @@ block0(v0: f32x4, v1: f32x4):
; addps %xmm1, %xmm0
; movl $0x7fc00000, %edi
; movd %edi, %xmm7
-; shufps $0x0, (%rip), %xmm7
+; shufps $0x0, %xmm7, %xmm7
; movdqa %xmm0, %xmm1
; cmpunordps %xmm0, %xmm1
; movdqa %xmm0, %xmm2
@@ -35,7 +35,7 @@ block0(v0: f32x4, v1: f32x4):
; addps %xmm1, %xmm0
; movl $0x7fc00000, %edi
; movd %edi, %xmm7
-; shufps $0, 0x28(%rip), %xmm7
+; shufps $0, %xmm7, %xmm7
; movdqa %xmm0, %xmm1
; cmpunordps %xmm0, %xmm1
; movdqa %xmm0, %xmm2
@@ -46,16 +46,6 @@ block0(v0: f32x4, v1: f32x4):
; movq %rbp, %rsp
; popq %rbp
; retq
-; addb %al, (%rax)
-; addb %al, (%rax)
-; addb %al, (%rax)
-; addb %al, (%rax)
-; sarb $0, (%rdi)
-; addb %al, (%rax)
-; addb %al, (%rax)
-; addb %al, (%rax)
-; addb %al, (%rax)
-; addb %al, (%rax)
function %f1(f64, f64) -> f64 {
block0(v0: f64, v1: f64):
diff --git a/cranelift/filetests/filetests/isa/x64/nan-canonicalization.clif b/cranelift/filetests/filetests/isa/x64/nan-canonicalization.clif
index a9f1cd3735..d54c978aea 100644
--- a/cranelift/filetests/filetests/isa/x64/nan-canonicalization.clif
+++ b/cranelift/filetests/filetests/isa/x64/nan-canonicalization.clif
@@ -16,7 +16,7 @@ block0(v0: f32x4, v1: f32x4):
; movdqa %xmm0, %xmm1
; movl $0x7fc00000, %r9d
; movd %r9d, %xmm5
-; shufps $0x0, (%rip), %xmm5
+; shufps $0x0, %xmm5, %xmm5
; cmpunordps %xmm1, %xmm0
; andps %xmm0, %xmm5
; andnps %xmm1, %xmm0
@@ -34,7 +34,7 @@ block0(v0: f32x4, v1: f32x4):
; movdqa %xmm0, %xmm1
; movl $0x7fc00000, %r9d
; movd %r9d, %xmm5
-; shufps $0, 0x12(%rip), %xmm5
+; shufps $0, %xmm5, %xmm5
; cmpunordps %xmm1, %xmm0
; andps %xmm0, %xmm5
; andnps %xmm1, %xmm0
@@ -42,13 +42,6 @@ block0(v0: f32x4, v1: f32x4):
; movq %rbp, %rsp
; popq %rbp
; retq
-; addb %al, (%rax)
-; sarb $0, (%rdi)
-; addb %al, (%rax)
-; addb %al, (%rax)
-; addb %al, (%rax)
-; addb %al, (%rax)
-; addb %al, (%rax)
function %f1(f64, f64) -> f64 {
block0(v0: f64, v1: f64):