compile: prefer an AND instead of SHR+SHL instructions

On modern 64bit CPUs a SHR, SHL or AND instruction take 1 cycle to execute.
A pair of shifts that operate on the same register will take 2 cycles
and needs to wait for the input register value to be available.

Large constants used to mask the high bits of a register with an AND
instruction can not be encoded as an immediate in the AND instruction
on amd64 and therefore need to be loaded into a register with a MOV
instruction.

However that MOV instruction is not dependent on the output register and
on many CPUs does not compete with the AND or shift instructions for
execution ports.

Using a pair of shifts to mask high bits instead of an AND to mask high
bits of a register has a shorter encoding and uses one less general
purpose register but is slower due to taking one clock cycle longer
if there is no register pressure that would make the AND variant need to
generate a spill.

For example the instructions emitted for (x & 1 << 63) before this CL are:
48c1ea3f                SHRQ $0x3f, DX
48c1e23f                SHLQ $0x3f, DX

after this CL the instructions are the same as GCC and LLVM use:
48b80000000000000080    MOVQ $0x8000000000000000, AX
4821d0                  ANDQ DX, AX

Some platforms such as arm64 already have SSA optimization rules to fuse
two shift instructions back into an AND.

Removing the general rule to rewrite AND to SHR+SHL speeds up this benchmark:

var GlobalU uint

func BenchmarkAndHighBits(b *testing.B) {
	x := uint(0)
	for i := 0; i < b.N; i++ {
		x &= 1 << 63
	}
	GlobalU = x
}

amd64/darwin on Intel(R) Core(TM) i7-3520M CPU @ 2.90GHz:
name           old time/op  new time/op  delta
AndHighBits-4  0.61ns ± 6%  0.42ns ± 6%  -31.42%  (p=0.000 n=25+25):

Updates #33826
Updates #32781

Change-Id: I862d3587446410c447b9a7265196b57f85358633
Reviewed-on: https://go-review.googlesource.com/c/go/+/191780
Run-TryBot: Martin Möhrmann <moehrmann@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>

src/cmd/compile/internal/ssa/gen/ARM64.rules

@@ -1863,9 +1863,8 @@
 (XORshiftLL <t> [c] (UBFX [bfc] x) x2) && c < 32 && t.Size() == 4 && bfc == armBFAuxInt(32-c, c)
 	-> (EXTRWconst [32-c] x2 x)
 
-// Generic rules rewrite certain AND to a pair of shifts.
-// However, on ARM64 the bitmask can fit into an instruction.
-// Rewrite it back to AND.
+// Rewrite special pairs of shifts to AND.
+// On ARM64 the bitmask can fit into an instruction.
 (SRLconst [c] (SLLconst [c] x)) && 0 < c && c < 64 -> (ANDconst [1<<uint(64-c)-1] x) // mask out high bits
 (SLLconst [c] (SRLconst [c] x)) && 0 < c && c < 64 -> (ANDconst [^(1<<uint(c)-1)] x) // mask out low bits
 

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

@@ -542,14 +542,6 @@
 (Slicemask (Const64 [x])) && x > 0 -> (Const64 [-1])
 (Slicemask (Const64 [0]))          -> (Const64 [0])
 
-// Rewrite AND of consts as shifts if possible, slightly faster for 64 bit operands
-// leading zeros can be shifted left, then right
-(And64 <t> (Const64 [y]) x) && nlz(y) + nto(y) == 64 && nto(y) >= 32
-  -> (Rsh64Ux64 (Lsh64x64 <t> x (Const64 <t> [nlz(y)])) (Const64 <t> [nlz(y)]))
-// trailing zeros can be shifted right, then left
-(And64 <t> (Const64 [y]) x) && nlo(y) + ntz(y) == 64 && ntz(y) >= 32
-  -> (Lsh64x64 (Rsh64Ux64 <t> x (Const64 <t> [ntz(y)])) (Const64 <t> [ntz(y)]))
-
 // simplifications often used for lengths.  e.g. len(s[i:i+5])==5
 (Sub(64|32|16|8) (Add(64|32|16|8) x y) x) -> y
 (Sub(64|32|16|8) (Add(64|32|16|8) x y) y) -> x

test/codegen/math.go

@@ -81,7 +81,7 @@ func abs32(x float32) float32 {
 
 // Check that it's using integer registers
 func copysign(a, b, c float64) {
-	// amd64:"BTRQ\t[$]63","SHRQ\t[$]63","SHLQ\t[$]63","ORQ"
+	// amd64:"BTRQ\t[$]63","ANDQ","ORQ"
 	// s390x:"CPSDR",-"MOVD"         (no integer load/store)
 	// ppc64:"FCPSGN"
 	// ppc64le:"FCPSGN"
@@ -100,7 +100,7 @@ func copysign(a, b, c float64) {
 	// s390x:"LNDFR\t",-"MOVD\t"     (no integer load/store)
 	sink64[2] = math.Float64frombits(math.Float64bits(a) | 1<<63)
 
-	// amd64:-"SHLQ\t[$]1",-"SHRQ\t[$]1","SHRQ\t[$]63","SHLQ\t[$]63","ORQ"
+	// amd64:"ANDQ","ORQ"
 	// s390x:"CPSDR\t",-"MOVD\t"     (no integer load/store)
 	// ppc64:"FCPSGN"
 	// ppc64le:"FCPSGN"