x64: Lower bitcast, fabs, and fneg in ISLE (#4729)

* Add tests for bitcast

* Migrate bitcast to ISLE

* Add tests for fabs

* Lower fabs in ISLE

* Add tests for fneg

* Lower fneg in ISLE

cranelift/filetests/filetests/isa/x64/bitcast.clif

@@ -0,0 +1,59 @@
+test compile precise-output
+target x86_64
+
+function %f1(f32) -> i32 {
+block0(v0: f32):
+  v1 = bitcast.i32 v0
+  return v1
+}
+
+;   pushq   %rbp
+;   movq    %rsp, %rbp
+; block0:
+;   movd    %xmm0, %eax
+;   movq    %rbp, %rsp
+;   popq    %rbp
+;   ret
+
+function %f2(i32) -> f32 {
+block0(v0: i32):
+  v1 = bitcast.f32 v0
+  return v1
+}
+
+;   pushq   %rbp
+;   movq    %rsp, %rbp
+; block0:
+;   movd    %edi, %xmm0
+;   movq    %rbp, %rsp
+;   popq    %rbp
+;   ret
+
+function %f3(f64) -> i64 {
+block0(v0: f64):
+  v1 = bitcast.i64 v0
+  return v1
+}
+
+;   pushq   %rbp
+;   movq    %rsp, %rbp
+; block0:
+;   movq    %xmm0, %rax
+;   movq    %rbp, %rsp
+;   popq    %rbp
+;   ret
+
+function %f4(i64) -> f64 {
+block0(v0: i64):
+  v1 = bitcast.f64 v0
+  return v1
+}
+
+;   pushq   %rbp
+;   movq    %rsp, %rbp
+; block0:
+;   movq    %rdi, %xmm0
+;   movq    %rbp, %rsp
+;   popq    %rbp
+;   ret
+

cranelift/filetests/filetests/isa/x64/fabs.clif

@@ -0,0 +1,67 @@
+test compile precise-output
+target x86_64
+
+function %f1(f32) -> f32 {
+block0(v0: f32):
+  v1 = fabs v0
+  return v1
+}
+
+;   pushq   %rbp
+;   movq    %rsp, %rbp
+; block0:
+;   movl    $2147483647, %ecx
+;   movd    %ecx, %xmm5
+;   andps   %xmm0, %xmm5, %xmm0
+;   movq    %rbp, %rsp
+;   popq    %rbp
+;   ret
+
+function %f2(f64) -> f64 {
+block0(v0: f64):
+  v1 = fabs v0
+  return v1
+}
+
+;   pushq   %rbp
+;   movq    %rsp, %rbp
+; block0:
+;   movabsq $9223372036854775807, %rcx
+;   movq    %rcx, %xmm5
+;   andpd   %xmm0, %xmm5, %xmm0
+;   movq    %rbp, %rsp
+;   popq    %rbp
+;   ret
+
+function %f3(f32x4) -> f32x4 {
+block0(v0: f32x4):
+  v1 = fabs v0
+  return v1
+}
+
+;   pushq   %rbp
+;   movq    %rsp, %rbp
+; block0:
+;   pcmpeqd %xmm4, %xmm4, %xmm4
+;   psrld   %xmm4, $1, %xmm4
+;   andps   %xmm0, %xmm4, %xmm0
+;   movq    %rbp, %rsp
+;   popq    %rbp
+;   ret
+
+function %f4(f64x2) -> f64x2 {
+block0(v0: f64x2):
+  v1 = fabs v0
+  return v1
+}
+
+;   pushq   %rbp
+;   movq    %rsp, %rbp
+; block0:
+;   pcmpeqd %xmm4, %xmm4, %xmm4
+;   psrlq   %xmm4, $1, %xmm4
+;   andpd   %xmm0, %xmm4, %xmm0
+;   movq    %rbp, %rsp
+;   popq    %rbp
+;   ret
+

cranelift/filetests/filetests/isa/x64/floating-point.clif

@@ -10,11 +10,9 @@ block0(v0: f64):
 ;   pushq   %rbp
 ;   movq    %rsp, %rbp
 ; block0:
-;   movdqa  %xmm0, %xmm5
-;   movabsq $9223372036854775807, %rdx
-;   movq    %rdx, %xmm0
-;   movdqa  %xmm5, %xmm7
-;   andpd   %xmm0, %xmm7, %xmm0
+;   movabsq $9223372036854775807, %rcx
+;   movq    %rcx, %xmm5
+;   andpd   %xmm0, %xmm5, %xmm0
 ;   movq    %rbp, %rsp
 ;   popq    %rbp
 ;   ret
@@ -29,10 +27,10 @@ block0(v0: i64):
 ;   pushq   %rbp
 ;   movq    %rsp, %rbp
 ; block0:
-;   movsd   0(%rdi), %xmm5
-;   movabsq $9223372036854775807, %r8
-;   movq    %r8, %xmm0
-;   andpd   %xmm0, %xmm5, %xmm0
+;   movsd   0(%rdi), %xmm0
+;   movabsq $9223372036854775807, %rdx
+;   movq    %rdx, %xmm6
+;   andpd   %xmm0, %xmm6, %xmm0
 ;   movq    %rbp, %rsp
 ;   popq    %rbp
 ;   ret

cranelift/filetests/filetests/isa/x64/fneg.clif

@@ -0,0 +1,67 @@
+test compile precise-output
+target x86_64
+
+function %f1(f32) -> f32 {
+block0(v0: f32):
+  v1 = fneg v0
+  return v1
+}
+
+;   pushq   %rbp
+;   movq    %rsp, %rbp
+; block0:
+;   movl    $-2147483648, %ecx
+;   movd    %ecx, %xmm5
+;   xorps   %xmm0, %xmm5, %xmm0
+;   movq    %rbp, %rsp
+;   popq    %rbp
+;   ret
+
+function %f2(f64) -> f64 {
+block0(v0: f64):
+  v1 = fneg v0
+  return v1
+}
+
+;   pushq   %rbp
+;   movq    %rsp, %rbp
+; block0:
+;   movabsq $-9223372036854775808, %rcx
+;   movq    %rcx, %xmm5
+;   xorpd   %xmm0, %xmm5, %xmm0
+;   movq    %rbp, %rsp
+;   popq    %rbp
+;   ret
+
+function %f3(f32x4) -> f32x4 {
+block0(v0: f32x4):
+  v1 = fneg v0
+  return v1
+}
+
+;   pushq   %rbp
+;   movq    %rsp, %rbp
+; block0:
+;   pcmpeqd %xmm4, %xmm4, %xmm4
+;   pslld   %xmm4, $31, %xmm4
+;   xorps   %xmm0, %xmm4, %xmm0
+;   movq    %rbp, %rsp
+;   popq    %rbp
+;   ret
+
+function %f4(f64x2) -> f64x2 {
+block0(v0: f64x2):
+  v1 = fneg v0
+  return v1
+}
+
+;   pushq   %rbp
+;   movq    %rsp, %rbp
+; block0:
+;   pcmpeqd %xmm4, %xmm4, %xmm4
+;   psllq   %xmm4, $63, %xmm4
+;   xorpd   %xmm0, %xmm4, %xmm0
+;   movq    %rbp, %rsp
+;   popq    %rbp
+;   ret
+