cmd/compile: don't depend on arch-dependent conversions in the compiler

Leave those constant foldings for runtime, similar to how we do it
for NaN generation.

These are the only instances I could find in cmd/compile/..., using

objdump -d ../pkg/tool/darwin_arm64/compile| egrep "(fcvtz|>:)" | grep -B1 fcvt

(There are instances in other places, like runtime and reflect, but I don't
think those places would affect compiler output.)

Change-Id: I4113fe4570115e4765825cf442cb1fde97cf2f27
Reviewed-on: https://go-review.googlesource.com/c/go/+/711281
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: David Chase <drchase@google.com>
Reviewed-by: Keith Randall <khr@google.com>

Author: randall77

src/cmd/compile/internal/ssa/_gen/generic.rules

@@ -50,10 +50,10 @@
 (Cvt32to64F  (Const32  [c])) => (Const64F [float64(c)])
 (Cvt64to32F  (Const64  [c])) => (Const32F [float32(c)])
 (Cvt64to64F  (Const64  [c])) => (Const64F [float64(c)])
-(Cvt32Fto32  (Const32F [c])) => (Const32  [int32(c)])
-(Cvt32Fto64  (Const32F [c])) => (Const64  [int64(c)])
-(Cvt64Fto32  (Const64F [c])) => (Const32  [int32(c)])
-(Cvt64Fto64  (Const64F [c])) => (Const64  [int64(c)])
+(Cvt32Fto32  (Const32F [c])) && c >= -1<<31 && c < 1<<31 => (Const32 [int32(c)])
+(Cvt32Fto64  (Const32F [c])) && c >= -1<<63 && c < 1<<63 => (Const64 [int64(c)])
+(Cvt64Fto32  (Const64F [c])) && c >= -1<<31 && c < 1<<31 => (Const32 [int32(c)])
+(Cvt64Fto64  (Const64F [c])) && c >= -1<<63 && c < 1<<63 => (Const64 [int64(c)])
 (Round32F x:(Const32F)) => x
 (Round64F x:(Const64F)) => x
 (CvtBoolToUint8 (ConstBool [false])) => (Const8 [0])

src/cmd/compile/internal/ssa/rewritegeneric.go

@@ -330,3 +330,48 @@ func nanGenerate32() float32 {
 	// amd64/v3:"VFMADD231SS"
 	return z0 + z1
 }
+
+func outOfBoundsConv(i32 *[2]int32, u32 *[2]uint32, i64 *[2]int64, u64 *[2]uint64) {
+	// arm64: "FCVTZSDW"
+	// amd64: "CVTTSD2SL", "CVTSD2SS"
+	i32[0] = int32(two40())
+	// arm64: "FCVTZSDW"
+	// amd64: "CVTTSD2SL", "CVTSD2SS"
+	i32[1] = int32(-two40())
+	// arm64: "FCVTZSDW"
+	// amd64: "CVTTSD2SL", "CVTSD2SS"
+	u32[0] = uint32(two41())
+	// on arm64, this uses an explicit <0 comparison, so it constant folds.
+	// on amd64, this uses an explicit <0 comparison, so it constant folds.
+	// amd64: "MOVL\t[$]0,"
+	u32[1] = uint32(minus1())
+	// arm64: "FCVTZSD"
+	// amd64: "CVTTSD2SQ"
+	i64[0] = int64(two80())
+	// arm64: "FCVTZSD"
+	// amd64: "CVTTSD2SQ"
+	i64[1] = int64(-two80())
+	// arm64: "FCVTZUD"
+	// amd64: "CVTTSD2SQ"
+	u64[0] = uint64(two81())
+	// arm64: "FCVTZUD"
+	// on amd64, this uses an explicit <0 comparison, so it constant folds.
+	// amd64: "MOVQ\t[$]0,"
+	u64[1] = uint64(minus1())
+}
+
+func two40() float64 {
+	return 1 << 40
+}
+func two41() float64 {
+	return 1 << 41
+}
+func two80() float64 {
+	return 1 << 80
+}
+func two81() float64 {
+	return 1 << 81
+}
+func minus1() float64 {
+	return -1
+}