x64: Sink constant loads into xmm instructions (#5880)

A number of places in the x64 backend make use of 128-bit constants for
various wasm SIMD-related instructions although most of them currently
use the `x64_xmm_load_const` helper to load the constant into a
register. Almost all xmm instructions, however, enable using a memory
operand which means that these loads can be folded into instructions to
help reduce register pressure. Automatic conversions were added for a
`VCodeConstant` into an `XmmMem` value and then explicit loads were all
removed in favor of forwarding the `XmmMem` value directly to the
underlying instruction. Note that some instances of `x64_xmm_load_const`
remain since they're used in contexts where load sinking won't work
(e.g. they're the first operand, not the second for non-commutative
instructions).

cranelift/filetests/filetests/isa/x64/fcvt.clif

@@ -304,10 +304,8 @@ block0(v0: i32x4):
 ;   pushq   %rbp
 ;   movq    %rsp, %rbp
 ; block0:
-;   movdqu  const(0), %xmm2
-;   unpcklps %xmm0, %xmm2, %xmm0
-;   movdqu  const(1), %xmm6
-;   subpd   %xmm0, %xmm6, %xmm0
+;   unpcklps %xmm0, const(0), %xmm0
+;   subpd   %xmm0, const(1), %xmm0
 ;   movq    %rbp, %rsp
 ;   popq    %rbp
 ;   ret
@@ -317,14 +315,16 @@ block0(v0: i32x4):
 ;   pushq %rbp
 ;   movq %rsp, %rbp
 ; block1: ; offset 0x4
-;   movdqu 0x14(%rip), %xmm2
-;   unpcklps %xmm2, %xmm0
-;   movdqu 0x19(%rip), %xmm6
-;   subpd %xmm6, %xmm0
+;   unpcklps 0x15(%rip), %xmm0
+;   subpd 0x1d(%rip), %xmm0
 ;   movq %rbp, %rsp
 ;   popq %rbp
 ;   retq
 ;   addb %al, (%rax)
+;   addb %al, (%rax)
+;   addb %al, (%rax)
+;   addb %al, (%rax)
+;   addb %al, (%rax)
 ;   xorb %al, (%rbx)
 ;   addb %dh, (%rax)
 ;   addb %al, (%r8)

cranelift/filetests/filetests/isa/x64/float-avx.clif

@@ -566,10 +566,9 @@ block0(v0: f64x2):
 ;   movq    %rsp, %rbp
 ; block0:
 ;   vcmppd  $0 %xmm0, %xmm0, %xmm2
-;   movupd  const(0), %xmm4
-;   vandps  %xmm2, %xmm4, %xmm6
-;   vminpd  %xmm0, %xmm6, %xmm8
-;   vcvttpd2dq %xmm8, %xmm0
+;   vandps  %xmm2, const(0), %xmm4
+;   vminpd  %xmm0, %xmm4, %xmm6
+;   vcvttpd2dq %xmm6, %xmm0
 ;   movq    %rbp, %rsp
 ;   popq    %rbp
 ;   ret
@@ -580,19 +579,13 @@ block0(v0: f64x2):
 ;   movq %rsp, %rbp
 ; block1: ; offset 0x4
 ;   vcmpeqpd %xmm0, %xmm0, %xmm2
-;   movupd 0x1f(%rip), %xmm4
-;   vandps %xmm4, %xmm2, %xmm6
-;   vminpd %xmm6, %xmm0, %xmm8
-;   vcvttpd2dq %xmm8, %xmm0
+;   vandps 0xf(%rip), %xmm2, %xmm4
+;   vminpd %xmm4, %xmm0, %xmm6
+;   vcvttpd2dq %xmm6, %xmm0
 ;   movq %rbp, %rsp
 ;   popq %rbp
 ;   retq
 ;   addb %al, (%rax)
 ;   addb %al, (%rax)
-;   addb %al, (%rax)
-;   addb %al, (%rax)
-;   addb %al, (%rax)
-;   addb %al, (%rax)
-;   addb %al, (%rax)
-;   addb %al, %al
+;   sarb $0xff, %bh
 

cranelift/filetests/filetests/isa/x64/narrowing.clif

@@ -63,13 +63,12 @@ block0(v0: f64x2):
 ;   pushq   %rbp
 ;   movq    %rsp, %rbp
 ; block0:
-;   movdqa  %xmm0, %xmm4
-;   cmppd   $0, %xmm4, %xmm0, %xmm4
-;   movupd  const(0), %xmm5
-;   andps   %xmm4, %xmm5, %xmm4
-;   movdqa  %xmm0, %xmm8
-;   minpd   %xmm8, %xmm4, %xmm8
-;   cvttpd2dq %xmm8, %xmm0
+;   movdqa  %xmm0, %xmm3
+;   cmppd   $0, %xmm3, %xmm0, %xmm3
+;   andps   %xmm3, const(0), %xmm3
+;   movdqa  %xmm0, %xmm6
+;   minpd   %xmm6, %xmm3, %xmm6
+;   cvttpd2dq %xmm6, %xmm0
 ;   movq    %rbp, %rsp
 ;   popq    %rbp
 ;   ret
@@ -79,20 +78,22 @@ block0(v0: f64x2):
 ;   pushq %rbp
 ;   movq %rsp, %rbp
 ; block1: ; offset 0x4
-;   movdqa %xmm0, %xmm4
-;   cmpeqpd %xmm0, %xmm4
-;   movupd 0x1b(%rip), %xmm5
-;   andps %xmm5, %xmm4
-;   movdqa %xmm0, %xmm8
-;   minpd %xmm4, %xmm8
-;   cvttpd2dq %xmm8, %xmm0
+;   movdqa %xmm0, %xmm3
+;   cmpeqpd %xmm0, %xmm3
+;   andps 0x1c(%rip), %xmm3
+;   movdqa %xmm0, %xmm6
+;   minpd %xmm3, %xmm6
+;   cvttpd2dq %xmm6, %xmm0
 ;   movq %rbp, %rsp
 ;   popq %rbp
 ;   retq
 ;   addb %al, (%rax)
 ;   addb %al, (%rax)
 ;   addb %al, (%rax)
-;   sarb $0xff, %bh
+;   addb %al, (%rax)
+;   addb %al, (%rax)
+;   addb %al, (%rax)
+;   addb %al, %al
 
 function %f4(i16x8, i16x8) -> i8x16 {
 block0(v0: i16x8, v1: i16x8):

cranelift/filetests/filetests/isa/x64/shuffle-avx512.clif

@@ -55,12 +55,11 @@ block0(v0: i8x16, v1: i8x16):
 ;   pushq   %rbp
 ;   movq    %rsp, %rbp
 ; block0:
-;   movdqa  %xmm0, %xmm7
-;   movdqu  const(1), %xmm0
-;   movdqu  const(0), %xmm6
-;   movdqa  %xmm7, %xmm9
-;   vpermi2b %xmm1, %xmm9, %xmm6, %xmm6
-;   andps   %xmm0, %xmm6, %xmm0
+;   movdqa  %xmm0, %xmm6
+;   movdqu  const(0), %xmm0
+;   movdqa  %xmm6, %xmm7
+;   vpermi2b %xmm1, %xmm7, %xmm0, %xmm0
+;   andps   %xmm0, const(1), %xmm0
 ;   movq    %rbp, %rsp
 ;   popq    %rbp
 ;   ret
@@ -70,12 +69,11 @@ block0(v0: i8x16, v1: i8x16):
 ;   pushq %rbp
 ;   movq %rsp, %rbp
 ; block1: ; offset 0x4
-;   movdqa %xmm0, %xmm7
-;   movdqu 0x30(%rip), %xmm0
-;   movdqu 0x18(%rip), %xmm6
-;   movdqa %xmm7, %xmm9
-;   vpermi2b %xmm1, %xmm9, %xmm6
-;   andps %xmm6, %xmm0
+;   movdqa %xmm0, %xmm6
+;   movdqu 0x20(%rip), %xmm0
+;   movdqa %xmm6, %xmm7
+;   vpermi2b %xmm1, %xmm7, %xmm0
+;   andps 0x1f(%rip), %xmm0
 ;   movq %rbp, %rsp
 ;   popq %rbp
 ;   retq
@@ -89,7 +87,9 @@ block0(v0: i8x16, v1: i8x16):
 ;   addb %al, (%rax)
 ;   addb %al, (%rax)
 ;   addb %al, (%rax)
-;   cmpb $0xff, %bh
+;   addb %al, (%rax)
+;   addb %al, (%rax)
+;   addb %al, -1(%rax)
 
 function %f3(i8x16, i8x16) -> i8x16 {
 block0(v0: i8x16, v1: i8x16):

cranelift/filetests/filetests/isa/x64/simd-arith-avx.clif

@@ -574,10 +574,9 @@ block0(v0: i16x8, v1: i16x8):
 ;   pushq   %rbp
 ;   movq    %rsp, %rbp
 ; block0:
-;   movdqu  const(0), %xmm3
-;   vpmulhrsw %xmm0, %xmm1, %xmm5
-;   vpcmpeqw %xmm3, %xmm5, %xmm7
-;   vpxor   %xmm5, %xmm7, %xmm0
+;   vpmulhrsw %xmm0, %xmm1, %xmm3
+;   vpcmpeqw %xmm3, const(0), %xmm5
+;   vpxor   %xmm3, %xmm5, %xmm0
 ;   movq    %rbp, %rsp
 ;   popq    %rbp
 ;   ret
@@ -587,14 +586,15 @@ block0(v0: i16x8, v1: i16x8):
 ;   pushq %rbp
 ;   movq %rsp, %rbp
 ; block1: ; offset 0x4
-;   movdqu 0x14(%rip), %xmm3
-;   vpmulhrsw %xmm1, %xmm0, %xmm5
-;   vpcmpeqw %xmm5, %xmm3, %xmm7
-;   vpxor %xmm7, %xmm5, %xmm0
+;   vpmulhrsw %xmm1, %xmm0, %xmm3
+;   vpcmpeqw 0xf(%rip), %xmm3, %xmm5
+;   vpxor %xmm5, %xmm3, %xmm0
 ;   movq %rbp, %rsp
 ;   popq %rbp
 ;   retq
 ;   addb %al, (%rax)
+;   addb %al, (%rax)
+;   addb %al, (%rax)
 ;   addb %al, -0x7fff8000(%rax)
 ;   addb %al, -0x7fff8000(%rax)
 
@@ -671,10 +671,8 @@ block0(v0: i32x4):
 ;   pushq   %rbp
 ;   movq    %rsp, %rbp
 ; block0:
-;   movdqu  const(0), %xmm2
-;   vunpcklps %xmm0, %xmm2, %xmm4
-;   movdqu  const(1), %xmm6
-;   vsubpd  %xmm4, %xmm6, %xmm0
+;   vunpcklps %xmm0, const(0), %xmm2
+;   vsubpd  %xmm2, const(1), %xmm0
 ;   movq    %rbp, %rsp
 ;   popq    %rbp
 ;   ret
@@ -684,10 +682,8 @@ block0(v0: i32x4):
 ;   pushq %rbp
 ;   movq %rsp, %rbp
 ; block1: ; offset 0x4
-;   movdqu 0x24(%rip), %xmm2
-;   vunpcklps %xmm2, %xmm0, %xmm4
-;   movdqu 0x28(%rip), %xmm6
-;   vsubpd %xmm6, %xmm4, %xmm0
+;   vunpcklps 0x14(%rip), %xmm0, %xmm2
+;   vsubpd 0x1c(%rip), %xmm2, %xmm0
 ;   movq %rbp, %rsp
 ;   popq %rbp
 ;   retq
@@ -695,10 +691,6 @@ block0(v0: i32x4):
 ;   addb %al, (%rax)
 ;   addb %al, (%rax)
 ;   addb %al, (%rax)
-;   addb %al, (%rax)
-;   addb %al, (%rax)
-;   addb %al, (%rax)
-;   addb %al, (%rax)
 ;   addb %dh, (%rax)
 ;   addb %al, (%r8)
 ;   xorb %al, (%rbx)
@@ -1283,8 +1275,7 @@ block0(v0: i16x8):
 ;   pushq   %rbp
 ;   movq    %rsp, %rbp
 ; block0:
-;   movdqu  const(0), %xmm2
-;   vpmaddwd %xmm0, %xmm2, %xmm0
+;   vpmaddwd %xmm0, const(0), %xmm0
 ;   movq    %rbp, %rsp
 ;   popq    %rbp
 ;   ret
@@ -1294,8 +1285,7 @@ block0(v0: i16x8):
 ;   pushq %rbp
 ;   movq %rsp, %rbp
 ; block1: ; offset 0x4
-;   movdqu 0x14(%rip), %xmm2
-;   vpmaddwd %xmm2, %xmm0, %xmm0
+;   vpmaddwd 0x14(%rip), %xmm0, %xmm0
 ;   movq %rbp, %rsp
 ;   popq %rbp
 ;   retq
@@ -1304,6 +1294,8 @@ block0(v0: i16x8):
 ;   addb %al, (%rax)
 ;   addb %al, (%rax)
 ;   addb %al, (%rax)
+;   addb %al, (%rax)
+;   addb %al, (%rax)
 ;   addb %al, (%rcx)
 ;   addb %al, (%rcx)
 ;   addb %al, (%rcx)
@@ -1357,12 +1349,10 @@ block0(v0: f64x2):
 ; block0:
 ;   xorpd   %xmm2, %xmm2, %xmm2
 ;   vmaxpd  %xmm0, %xmm2, %xmm4
-;   movupd  const(0), %xmm6
-;   vminpd  %xmm4, %xmm6, %xmm8
-;   vroundpd $3, %xmm8, %xmm10
-;   movupd  const(1), %xmm12
-;   vaddpd  %xmm10, %xmm12, %xmm14
-;   vshufps $136 %xmm14, %xmm2, %xmm0
+;   vminpd  %xmm4, const(0), %xmm6
+;   vroundpd $3, %xmm6, %xmm8
+;   vaddpd  %xmm8, const(1), %xmm10
+;   vshufps $136 %xmm10, %xmm2, %xmm0
 ;   movq    %rbp, %rsp
 ;   popq    %rbp
 ;   ret
@@ -1374,22 +1364,17 @@ block0(v0: f64x2):
 ; block1: ; offset 0x4
 ;   xorpd %xmm2, %xmm2
 ;   vmaxpd %xmm2, %xmm0, %xmm4
-;   movupd 0x2c(%rip), %xmm6
-;   vminpd %xmm6, %xmm4, %xmm8
-;   vroundpd $3, %xmm8, %xmm10
-;   movupd 0x29(%rip), %xmm12
-;   vaddpd %xmm12, %xmm10, %xmm14
-;   vshufps $0x88, %xmm2, %xmm14, %xmm0
+;   vminpd 0x1c(%rip), %xmm4, %xmm6
+;   vroundpd $3, %xmm6, %xmm8
+;   vaddpd 0x1e(%rip), %xmm8, %xmm10
+;   vshufps $0x88, %xmm2, %xmm10, %xmm0
 ;   movq %rbp, %rsp
 ;   popq %rbp
 ;   retq
 ;   addb %al, (%rax)
 ;   addb %al, (%rax)
 ;   addb %al, (%rax)
-;   addb %al, (%rax)
-;   addb %al, (%rax)
-;   addb %al, (%rax)
-;   loopne 0x43
+;   loopne 0x33
 
 function %i8x16_shl(i8x16, i32) -> i8x16 {
 block0(v0: i8x16, v1: i32):

cranelift/filetests/filetests/isa/x64/simd-lane-access-compile.clif

@@ -17,12 +17,10 @@ block0:
 ;   movq    %rsp, %rbp
 ; block0:
 ;   movdqu  const(3), %xmm0
-;   movdqu  const(2), %xmm4
-;   movdqu  const(0), %xmm2
-;   pshufb  %xmm0, %xmm2, %xmm0
-;   movdqu  const(1), %xmm6
-;   pshufb  %xmm4, %xmm6, %xmm4
-;   por     %xmm0, %xmm4, %xmm0
+;   movdqu  const(2), %xmm2
+;   pshufb  %xmm0, const(0), %xmm0
+;   pshufb  %xmm2, const(1), %xmm2
+;   por     %xmm0, %xmm2, %xmm0
 ;   movq    %rbp, %rsp
 ;   popq    %rbp
 ;   ret
@@ -32,13 +30,11 @@ block0:
 ;   pushq %rbp
 ;   movq %rsp, %rbp
 ; block1: ; offset 0x4
-;   movdqu 0x64(%rip), %xmm0
-;   movdqu 0x4c(%rip), %xmm4
-;   movdqu 0x24(%rip), %xmm2
-;   pshufb %xmm2, %xmm0
-;   movdqu 0x27(%rip), %xmm6
-;   pshufb %xmm6, %xmm4
-;   por %xmm4, %xmm0
+;   movdqu 0x54(%rip), %xmm0
+;   movdqu 0x3c(%rip), %xmm2
+;   pshufb 0x13(%rip), %xmm0
+;   pshufb 0x1a(%rip), %xmm2
+;   por %xmm2, %xmm0
 ;   movq %rbp, %rsp
 ;   popq %rbp
 ;   retq
@@ -50,10 +46,6 @@ block0:
 ;   addb %al, (%rax)
 ;   addb %al, (%rax)
 ;   addb %al, (%rax)
-;   addb %al, (%rax)
-;   addb %al, (%rax)
-;   addb %al, (%rax)
-;   addb %al, (%rax)
 ;   addb $0x80, -0x7f7f7f80(%rax)
 ;   addb $0x80, -0x7f7f7f80(%rax)
 ;   addb $0, 0x101(%rax)
@@ -84,8 +76,7 @@ block0:
 ;   movq    %rsp, %rbp
 ; block0:
 ;   movdqu  const(1), %xmm0
-;   movdqu  const(0), %xmm1
-;   pshufb  %xmm0, %xmm1, %xmm0
+;   pshufb  %xmm0, const(0), %xmm0
 ;   movq    %rbp, %rsp
 ;   popq    %rbp
 ;   ret
@@ -96,8 +87,7 @@ block0:
 ;   movq %rsp, %rbp
 ; block1: ; offset 0x4
 ;   movdqu 0x24(%rip), %xmm0
-;   movdqu 0xc(%rip), %xmm1
-;   pshufb %xmm1, %xmm0
+;   pshufb 0xb(%rip), %xmm0
 ;   movq %rbp, %rsp
 ;   popq %rbp
 ;   retq
@@ -109,6 +99,8 @@ block0:
 ;   addb %al, (%rax)
 ;   addb %al, (%rax)
 ;   addb %al, (%rax)
+;   addb %al, (%rax)
+;   addb %al, (%rax)
 ;   addb %al, (%rcx, %rax)
 ;   addb %al, (%rax)
 ;   addb %al, (%rax)
@@ -131,10 +123,9 @@ block0:
 ;   movq    %rsp, %rbp
 ; block0:
 ;   movdqu  const(1), %xmm0
-;   movdqu  const(1), %xmm2
-;   movdqu  const(0), %xmm3
-;   paddusb %xmm2, %xmm3, %xmm2
-;   pshufb  %xmm0, %xmm2, %xmm0
+;   movdqu  const(1), %xmm1
+;   paddusb %xmm1, const(0), %xmm1
+;   pshufb  %xmm0, %xmm1, %xmm0
 ;   movq    %rbp, %rsp
 ;   popq    %rbp
 ;   ret
@@ -145,16 +136,17 @@ block0:
 ;   movq %rsp, %rbp
 ; block1: ; offset 0x4
 ;   movdqu 0x34(%rip), %xmm0
-;   movdqu 0x2c(%rip), %xmm2
-;   movdqu 0x14(%rip), %xmm3
-;   paddusb %xmm3, %xmm2
-;   pshufb %xmm2, %xmm0
+;   movdqu 0x2c(%rip), %xmm1
+;   paddusb 0x14(%rip), %xmm1
+;   pshufb %xmm1, %xmm0
 ;   movq %rbp, %rsp
 ;   popq %rbp
 ;   retq
 ;   addb %al, (%rax)
 ;   addb %al, (%rax)
 ;   addb %al, (%rax)
+;   addb %al, (%rax)
+;   addb %al, (%rax)
 ;   jo 0xa2
 ;   jo 0xa4
 ;   jo 0xa6

cranelift/filetests/filetests/isa/x64/simd-pairwise-add.clif

@@ -55,8 +55,7 @@ block0(v0: i16x8):
 ;   pushq   %rbp
 ;   movq    %rsp, %rbp
 ; block0:
-;   movdqu  const(0), %xmm2
-;   pmaddwd %xmm0, %xmm2, %xmm0
+;   pmaddwd %xmm0, const(0), %xmm0
 ;   movq    %rbp, %rsp
 ;   popq    %rbp
 ;   ret
@@ -66,8 +65,7 @@ block0(v0: i16x8):
 ;   pushq %rbp
 ;   movq %rsp, %rbp
 ; block1: ; offset 0x4
-;   movdqu 0x14(%rip), %xmm2
-;   pmaddwd %xmm2, %xmm0
+;   pmaddwd 0x14(%rip), %xmm0
 ;   movq %rbp, %rsp
 ;   popq %rbp
 ;   retq
@@ -76,6 +74,8 @@ block0(v0: i16x8):
 ;   addb %al, (%rax)
 ;   addb %al, (%rax)
 ;   addb %al, (%rax)
+;   addb %al, (%rax)
+;   addb %al, (%rax)
 ;   addb %al, (%rcx)
 ;   addb %al, (%rcx)
 ;   addb %al, (%rcx)
@@ -97,8 +97,7 @@ block0(v0: i8x16):
 ;   pushq   %rbp
 ;   movq    %rsp, %rbp
 ; block0:
-;   movdqu  const(0), %xmm2
-;   pmaddubsw %xmm0, %xmm2, %xmm0
+;   pmaddubsw %xmm0, const(0), %xmm0
 ;   movq    %rbp, %rsp
 ;   popq    %rbp
 ;   ret
@@ -108,8 +107,7 @@ block0(v0: i8x16):
 ;   pushq %rbp
 ;   movq %rsp, %rbp
 ; block1: ; offset 0x4
-;   movdqu 0x14(%rip), %xmm2
-;   pmaddubsw %xmm2, %xmm0
+;   pmaddubsw 0x13(%rip), %xmm0
 ;   movq %rbp, %rsp
 ;   popq %rbp
 ;   retq
@@ -118,6 +116,8 @@ block0(v0: i8x16):
 ;   addb %al, (%rax)
 ;   addb %al, (%rax)
 ;   addb %al, (%rax)
+;   addb %al, (%rax)
+;   addb %al, (%rax)
 ;   addl %eax, (%rcx)
 ;   addl %eax, (%rcx)
 ;   addl %eax, (%rcx)
@@ -139,12 +139,9 @@ block0(v0: i16x8):
 ;   pushq   %rbp
 ;   movq    %rsp, %rbp
 ; block0:
-;   movdqu  const(0), %xmm2
-;   pxor    %xmm0, %xmm2, %xmm0
-;   movdqu  const(1), %xmm6
-;   pmaddwd %xmm0, %xmm6, %xmm0
-;   movdqu  const(2), %xmm10
-;   paddd   %xmm0, %xmm10, %xmm0
+;   pxor    %xmm0, const(0), %xmm0
+;   pmaddwd %xmm0, const(1), %xmm0
+;   paddd   %xmm0, const(2), %xmm0
 ;   movq    %rbp, %rsp
 ;   popq    %rbp
 ;   ret
@@ -154,16 +151,20 @@ block0(v0: i16x8):
 ;   pushq %rbp
 ;   movq %rsp, %rbp
 ; block1: ; offset 0x4
-;   movdqu 0x24(%rip), %xmm2
-;   pxor %xmm2, %xmm0
-;   movdqu 0x28(%rip), %xmm6
-;   pmaddwd %xmm6, %xmm0
-;   movdqu 0x2b(%rip), %xmm10
-;   paddd %xmm10, %xmm0
+;   pxor 0x24(%rip), %xmm0
+;   pmaddwd 0x2c(%rip), %xmm0
+;   paddd 0x34(%rip), %xmm0
 ;   movq %rbp, %rsp
 ;   popq %rbp
 ;   retq
 ;   addb %al, (%rax)
+;   addb %al, (%rax)
+;   addb %al, (%rax)
+;   addb %al, (%rax)
+;   addb %al, (%rax)
+;   addb %al, (%rax)
+;   addb %al, (%rax)
+;   addb %al, (%rax)
 ;   addb $0x80, (%rax)
 ;   addb %al, -0x7fff8000(%rax)
 ;   addb %al, -0x7fff8000(%rax)

cranelift/filetests/filetests/isa/x64/sqmul_round_sat.clif

@@ -11,9 +11,9 @@ block0(v0: i16x8, v1: i16x8):
 ;   pushq   %rbp
 ;   movq    %rsp, %rbp
 ; block0:
-;   movdqu  const(0), %xmm5
 ;   pmulhrsw %xmm0, %xmm1, %xmm0
-;   pcmpeqw %xmm5, %xmm0, %xmm5
+;   movdqa  %xmm0, %xmm5
+;   pcmpeqw %xmm5, const(0), %xmm5
 ;   pxor    %xmm0, %xmm5, %xmm0
 ;   movq    %rbp, %rsp
 ;   popq    %rbp
@@ -24,9 +24,9 @@ block0(v0: i16x8, v1: i16x8):
 ;   pushq %rbp
 ;   movq %rsp, %rbp
 ; block1: ; offset 0x4
-;   movdqu 0x14(%rip), %xmm5
 ;   pmulhrsw %xmm1, %xmm0
-;   pcmpeqw %xmm0, %xmm5
+;   movdqa %xmm0, %xmm5
+;   pcmpeqw 0xb(%rip), %xmm5
 ;   pxor %xmm5, %xmm0
 ;   movq %rbp, %rsp
 ;   popq %rbp

cranelift/filetests/filetests/isa/x64/uunarrow.clif

@@ -14,13 +14,11 @@ block0(v0: f64x2):
 ;   movq    %rsp, %rbp
 ; block0:
 ;   xorpd   %xmm2, %xmm2, %xmm2
-;   movdqa  %xmm0, %xmm6
-;   maxpd   %xmm6, %xmm2, %xmm6
-;   movupd  const(0), %xmm7
-;   minpd   %xmm6, %xmm7, %xmm6
-;   roundpd $3, %xmm6, %xmm0
-;   movupd  const(1), %xmm12
-;   addpd   %xmm0, %xmm12, %xmm0
+;   movdqa  %xmm0, %xmm5
+;   maxpd   %xmm5, %xmm2, %xmm5
+;   minpd   %xmm5, const(0), %xmm5
+;   roundpd $3, %xmm5, %xmm0
+;   addpd   %xmm0, const(1), %xmm0
 ;   shufps  $136, %xmm0, %xmm2, %xmm0
 ;   movq    %rbp, %rsp
 ;   popq    %rbp
@@ -32,20 +30,15 @@ block0(v0: f64x2):
 ;   movq %rsp, %rbp
 ; block1: ; offset 0x4
 ;   xorpd %xmm2, %xmm2
-;   movdqa %xmm0, %xmm6
-;   maxpd %xmm2, %xmm6
-;   movupd 0x28(%rip), %xmm7
-;   minpd %xmm7, %xmm6
-;   roundpd $3, %xmm6, %xmm0
-;   movupd 0x25(%rip), %xmm12
-;   addpd %xmm12, %xmm0
+;   movdqa %xmm0, %xmm5
+;   maxpd %xmm2, %xmm5
+;   minpd 0x18(%rip), %xmm5
+;   roundpd $3, %xmm5, %xmm0
+;   addpd 0x1a(%rip), %xmm0
 ;   shufps $0x88, %xmm2, %xmm0
 ;   movq %rbp, %rsp
 ;   popq %rbp
 ;   retq
 ;   addb %al, (%rax)
-;   addb %al, (%rax)
-;   addb %al, (%rax)
-;   addb %al, (%rax)
 ;   addb %ah, %al