cmd/compile: combine more 32 bit shift and mask operations on ppc64

Combine (AND m (SRWconst x)) or (SRWconst (AND m x)) when mask m is
and the shift value produce constant which can be encoded into an
RLWINM instruction.

Combine (CLRLSLDI (SRWconst x)) if the combining of the underling rotate
masks produces a constant which can be encoded into RLWINM.

Likewise for (SLDconst (SRWconst x)) and (CLRLSDI (RLWINM x)).

Combine rotate word + and operations which can be encoded as a single
RLWINM/RLWNM instruction.

The most notable performance improvements arise from the crypto
benchmarks below (GOARCH=power8 on a ppc64le/linux):

pkg:golang.org/x/crypto/blowfish goos:linux goarch:ppc64le
ExpandKeyWithSalt                               52.2µs ± 0%    47.5µs ± 0%  -8.88%
ExpandKey                                       44.4µs ± 0%    40.3µs ± 0%  -9.15%

pkg:golang.org/x/crypto/ssh/internal/bcrypt_pbkdf goos:linux goarch:ppc64le
Key                                             57.6ms ± 0%    52.3ms ± 0%  -9.13%

pkg:golang.org/x/crypto/bcrypt goos:linux goarch:ppc64le
Equal                                           90.9ms ± 0%    82.6ms ± 0%  -9.13%
DefaultCost                                     91.0ms ± 0%    82.7ms ± 0%  -9.12%

Change-Id: I59a0ca29face38f4ab46e37124c32906f216c4ce
Reviewed-on: https://go-review.googlesource.com/c/go/+/260798
Run-TryBot: Carlos Eduardo Seo <carlos.seo@linaro.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Lynn Boger <laboger@linux.vnet.ibm.com>
Reviewed-by: Carlos Eduardo Seo <carlos.seo@linaro.com>
Trust: Lynn Boger <laboger@linux.vnet.ibm.com>

2 files changed, 118 insertions, 21 deletions

diff --git a/test/codegen/rotate.go b/test/codegen/rotate.go
index ce24b57877..0c8b030970 100644
--- a/test/codegen/rotate.go
+++ b/test/codegen/rotate.go
@@ -6,6 +6,8 @@
 
 package codegen
 
+import "math/bits"
+
 // ------------------- //
 //    const rotates    //
 // ------------------- //
@@ -166,3 +168,46 @@ func f32(x uint32) uint32 {
 	// amd64:"ROLL\t[$]7"
 	return rot32nc(x, 7)
 }
+
+// --------------------------------------- //
+//    Combined Rotate + Masking operations //
+// --------------------------------------- //
+
+func checkMaskedRotate32(a []uint32, r int) {
+	i := 0
+
+	// ppc64le: "RLWNM\t[$]16, R[0-9]+, [$]16711680, R[0-9]+"
+	// ppc64: "RLWNM\t[$]16, R[0-9]+, [$]16711680, R[0-9]+"
+	a[i] = bits.RotateLeft32(a[i], 16) & 0xFF0000
+	i++
+	// ppc64le: "RLWNM\t[$]16, R[0-9]+, [$]16711680, R[0-9]+"
+	// ppc64: "RLWNM\t[$]16, R[0-9]+, [$]16711680, R[0-9]+"
+	a[i] = bits.RotateLeft32(a[i]&0xFF, 16)
+	i++
+	// ppc64le: "RLWNM\t[$]4, R[0-9]+, [$]4080, R[0-9]+"
+	// ppc64: "RLWNM\t[$]4, R[0-9]+, [$]4080, R[0-9]+"
+	a[i] = bits.RotateLeft32(a[i], 4) & 0xFF0
+	i++
+	// ppc64le: "RLWNM\t[$]16, R[0-9]+, [$]255, R[0-9]+"
+	// ppc64: "RLWNM\t[$]16, R[0-9]+, [$]255, R[0-9]+"
+	a[i] = bits.RotateLeft32(a[i]&0xFF0000, 16)
+	i++
+
+	// ppc64le: "RLWNM\tR[0-9]+, R[0-9]+, [$]16711680, R[0-9]+"
+	// ppc64: "RLWNM\tR[0-9]+, R[0-9]+, [$]16711680, R[0-9]+"
+	a[i] = bits.RotateLeft32(a[i], r) & 0xFF0000
+	i++
+	// ppc64le: "RLWNM\tR[0-9]+, R[0-9]+, [$]65280, R[0-9]+"
+	// ppc64: "RLWNM\tR[0-9]+, R[0-9]+, [$]65280, R[0-9]+"
+	a[i] = bits.RotateLeft32(a[3], r) & 0xFF00
+	i++
+
+	// ppc64le: "RLWNM\tR[0-9]+, R[0-9]+, [$]4293922815, R[0-9]+"
+	// ppc64: "RLWNM\tR[0-9]+, R[0-9]+, [$]4293922815, R[0-9]+"
+	a[i] = bits.RotateLeft32(a[3], r) & 0xFFF00FFF
+	i++
+	// ppc64le: "RLWNM\t[$]4, R[0-9]+, [$]4293922815, R[0-9]+"
+	// ppc64: "RLWNM\t[$]4, R[0-9]+, [$]4293922815, R[0-9]+"
+	a[i] = bits.RotateLeft32(a[3], 4) & 0xFFF00FFF
+	i++
+}
diff --git a/test/codegen/shift.go b/test/codegen/shift.go
index bbfc85ffbb..a45f27c9cf 100644
--- a/test/codegen/shift.go
+++ b/test/codegen/shift.go
@@ -156,29 +156,29 @@ func checkUnneededTrunc(tab *[100000]uint32, d uint64, v uint32, h uint16, b byt
 	// ppc64:-".*RLWINM",-".*RLDICR",".*CLRLSLDI"
 	f := tab[byte(v)^b]
 	// ppc64le:-".*RLWINM",-".*RLDICR",".*CLRLSLDI"
-        // ppc64:-".*RLWINM",-".*RLDICR",".*CLRLSLDI"
+	// ppc64:-".*RLWINM",-".*RLDICR",".*CLRLSLDI"
 	f += tab[byte(v)&b]
 	// ppc64le:-".*RLWINM",-".*RLDICR",".*CLRLSLDI"
-        // ppc64:-".*RLWINM",-".*RLDICR",".*CLRLSLDI"
+	// ppc64:-".*RLWINM",-".*RLDICR",".*CLRLSLDI"
 	f += tab[byte(v)|b]
 	// ppc64le:-".*RLWINM",-".*RLDICR",".*CLRLSLDI"
-        // ppc64:-".*RLWINM",-".*RLDICR",".*CLRLSLDI"
+	// ppc64:-".*RLWINM",-".*RLDICR",".*CLRLSLDI"
 	f += tab[uint16(v)&h]
 	// ppc64le:-".*RLWINM",-".*RLDICR",".*CLRLSLDI"
-        // ppc64:-".*RLWINM",-".*RLDICR",".*CLRLSLDI"
+	// ppc64:-".*RLWINM",-".*RLDICR",".*CLRLSLDI"
 	f += tab[uint16(v)^h]
 	// ppc64le:-".*RLWINM",-".*RLDICR",".*CLRLSLDI"
-        // ppc64:-".*RLWINM",-".*RLDICR",".*CLRLSLDI"
+	// ppc64:-".*RLWINM",-".*RLDICR",".*CLRLSLDI"
 	f += tab[uint16(v)|h]
 	// ppc64le:-".*AND",-"RLDICR",".*CLRLSLDI"
 	// ppc64:-".*AND",-"RLDICR",".*CLRLSLDI"
 	f += tab[v&0xff]
 	// ppc64le:-".*AND",".*CLRLSLWI"
-        // ppc64:-".*AND",".*CLRLSLWI"
-        f += 2*uint32(uint16(d))
+	// ppc64:-".*AND",".*CLRLSLWI"
+	f += 2 * uint32(uint16(d))
 	// ppc64le:-".*AND",-"RLDICR",".*CLRLSLDI"
 	// ppc64:-".*AND",-"RLDICR",".*CLRLSLDI"
-	g := 2*uint64(uint32(d))
+	g := 2 * uint64(uint32(d))
 	return f, g
 }
 
@@ -186,10 +186,10 @@ func checkCombinedShifts(v8 uint8, v16 uint16, v32 uint32, x32 int32, v64 uint64
 
 	// ppc64le:-"AND","CLRLSLWI"
 	// ppc64:-"AND","CLRLSLWI"
-	f := (v8 &0xF) << 2
+	f := (v8 & 0xF) << 2
 	// ppc64le:"CLRLSLWI"
 	// ppc64:"CLRLSLWI"
-	f += byte(v16)<<3
+	f += byte(v16) << 3
 	// ppc64le:-"AND","CLRLSLWI"
 	// ppc64:-"AND","CLRLSLWI"
 	g := (v16 & 0xFF) << 3
@@ -207,29 +207,81 @@ func checkCombinedShifts(v8 uint8, v16 uint16, v32 uint32, x32 int32, v64 uint64
 	i += (v64 & 0xFFFF00) << 10
 	// ppc64le/power9:-"SLD","EXTSWSLI"
 	// ppc64/power9:-"SLD","EXTSWSLI"
-	j := int64(x32+32)*8
+	j := int64(x32+32) * 8
 	return f, g, h, i, j
 }
 
 func checkWidenAfterShift(v int64, u uint64) (int64, uint64) {
 
 	// ppc64le:-".*MOVW"
-	f := int32(v>>32)
+	f := int32(v >> 32)
 	// ppc64le:".*MOVW"
-	f += int32(v>>31)
+	f += int32(v >> 31)
 	// ppc64le:-".*MOVH"
-	g := int16(v>>48)
+	g := int16(v >> 48)
 	// ppc64le:".*MOVH"
-	g += int16(v>>30)
+	g += int16(v >> 30)
 	// ppc64le:-".*MOVH"
-	g += int16(f>>16)
+	g += int16(f >> 16)
 	// ppc64le:-".*MOVB"
-	h := int8(v>>56)
+	h := int8(v >> 56)
 	// ppc64le:".*MOVB"
-	h += int8(v>>28)
+	h += int8(v >> 28)
 	// ppc64le:-".*MOVB"
-	h += int8(f>>24)
+	h += int8(f >> 24)
 	// ppc64le:".*MOVB"
-	h += int8(f>>16)
-	return int64(h),uint64(g)
+	h += int8(f >> 16)
+	return int64(h), uint64(g)
+}
+
+func checkShiftAndMask32(v []uint32) {
+	i := 0
+
+	// ppc64le: "RLWNM\t[$]24, R[0-9]+, [$]1044480, R[0-9]+"
+	// ppc64: "RLWNM\t[$]24, R[0-9]+, [$]1044480, R[0-9]+"
+	v[i] = (v[i] & 0xFF00000) >> 8
+	i++
+	// ppc64le: "RLWNM\t[$]26, R[0-9]+, [$]1020, R[0-9]+"
+	// ppc64: "RLWNM\t[$]26, R[0-9]+, [$]1020, R[0-9]+"
+	v[i] = (v[i] & 0xFF00) >> 6
+	i++
+	// ppc64le: "MOVW\tR0"
+	// ppc64: "MOVW\tR0"
+	v[i] = (v[i] & 0xFF) >> 8
+	i++
+	// ppc64le: "MOVW\tR0"
+	// ppc64: "MOVW\tR0"
+	v[i] = (v[i] & 0xF000000) >> 28
+	i++
+	// ppc64le: "RLWNM\t[$]26, R[0-9]+, [$]255, R[0-9]+"
+	// ppc64: "RLWNM\t[$]26, R[0-9]+, [$]255, R[0-9]+"
+	v[i] = (v[i] >> 6) & 0xFF
+	i++
+	// ppc64le: "RLWNM\t[$]26, R[0-9]+, [$]1044480, R[0-9]+"
+	// ppc64: "RLWNM\t[$]26, R[0-9]+, [$]1044480, R[0-9]+"
+	v[i] = (v[i] >> 6) & 0xFF000
+	i++
+	// ppc64le: "MOVW\tR0"
+	// ppc64: "MOVW\tR0"
+	v[i] = (v[i] >> 20) & 0xFF000
+	i++
+	// ppc64le: "MOVW\tR0"
+	// ppc64: "MOVW\tR0"
+	v[i] = (v[i] >> 24) & 0xFF00
+	i++
+}
+
+func checkMergedShifts32(a [256]uint32, b [256]uint64, u uint32, v uint32) {
+	//ppc64le: -"CLRLSLDI", "RLWNM\t[$]10, R[0-9]+, [$]1020, R[0-9]+"
+	//ppc64: -"CLRLSLDI", "RLWNM\t[$]10, R[0-9]+, [$]1020, R[0-9]+"
+	a[0] = a[uint8(v>>24)]
+	//ppc64le: -"CLRLSLDI", "RLWNM\t[$]11, R[0-9]+, [$]2040, R[0-9]+"
+	//ppc64: -"CLRLSLDI", "RLWNM\t[$]11, R[0-9]+, [$]2040, R[0-9]+"
+	b[0] = b[uint8(v>>24)]
+	//ppc64le: -"CLRLSLDI", "RLWNM\t[$]15, R[0-9]+, [$]2040, R[0-9]+"
+	//ppc64: -"CLRLSLDI", "RLWNM\t[$]15, R[0-9]+, [$]2040, R[0-9]+"
+	b[1] = b[(v>>20)&0xFF]
+	//ppc64le: -"SLD", "RLWNM\t[$]10, R[0-9]+, [$]1016, R[0-9]+"
+	//ppc64: -"SLD", "RLWNM\t[$]10, R[0-9]+, [$]1016, R[0-9]+"
+	b[2] = b[v>>25]
 }