cmd/compile: intrinsify math/bits.Len on riscv64

For riscv64/rva22u64 and above, we can intrinsify math/bits.Len using the
CLZ/CLZW machine instructions.

On a StarFive VisionFive 2 with GORISCV64=rva22u64:

                 │   clz.b.1   │               clz.b.2               │
                 │   sec/op    │   sec/op     vs base                │
LeadingZeros-4     28.89n ± 0%   12.08n ± 0%  -58.19% (p=0.000 n=10)
LeadingZeros8-4    18.79n ± 0%   14.76n ± 0%  -21.45% (p=0.000 n=10)
LeadingZeros16-4   25.27n ± 0%   14.76n ± 0%  -41.59% (p=0.000 n=10)
LeadingZeros32-4   25.12n ± 0%   12.08n ± 0%  -51.92% (p=0.000 n=10)
LeadingZeros64-4   25.89n ± 0%   12.08n ± 0%  -53.35% (p=0.000 n=10)
geomean            24.55n        13.09n       -46.70%

Change-Id: I0dda684713dbdf5336af393f5ccbdae861c4f694
Reviewed-on: https://go-review.googlesource.com/c/go/+/652321
Reviewed-by: David Chase <drchase@google.com>
Reviewed-by: Meng Zhuo <mengzhuo1203@gmail.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Mark Ryan <markdryan@rivosinc.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>

7 files changed, 142 insertions, 1 deletions

diff --git a/src/cmd/compile/internal/riscv64/ssa.go b/src/cmd/compile/internal/riscv64/ssa.go
index 4392081f6e..952a2050a0 100644
--- a/src/cmd/compile/internal/riscv64/ssa.go
+++ b/src/cmd/compile/internal/riscv64/ssa.go
@@ -419,7 +419,7 @@ func ssaGenValue(s *ssagen.State, v *ssa.Value) {
 		ssa.OpRISCV64FMVSX, ssa.OpRISCV64FMVDX,
 		ssa.OpRISCV64FCVTSW, ssa.OpRISCV64FCVTSL, ssa.OpRISCV64FCVTWS, ssa.OpRISCV64FCVTLS,
 		ssa.OpRISCV64FCVTDW, ssa.OpRISCV64FCVTDL, ssa.OpRISCV64FCVTWD, ssa.OpRISCV64FCVTLD, ssa.OpRISCV64FCVTDS, ssa.OpRISCV64FCVTSD,
-		ssa.OpRISCV64NOT, ssa.OpRISCV64NEG, ssa.OpRISCV64NEGW, ssa.OpRISCV64CTZ, ssa.OpRISCV64CTZW:
+		ssa.OpRISCV64NOT, ssa.OpRISCV64NEG, ssa.OpRISCV64NEGW, ssa.OpRISCV64CLZ, ssa.OpRISCV64CLZW, ssa.OpRISCV64CTZ, ssa.OpRISCV64CTZW:
 		p := s.Prog(v.Op.Asm())
 		p.From.Type = obj.TYPE_REG
 		p.From.Reg = v.Args[0].Reg()
diff --git a/src/cmd/compile/internal/ssa/_gen/RISCV64.rules b/src/cmd/compile/internal/ssa/_gen/RISCV64.rules
index 016eb53f04..96b9b11cf9 100644
--- a/src/cmd/compile/internal/ssa/_gen/RISCV64.rules
+++ b/src/cmd/compile/internal/ssa/_gen/RISCV64.rules
@@ -225,6 +225,12 @@
 (Ctz16 x) => (CTZW (ORI <typ.UInt32> [1<<16] x))
 (Ctz8  x) => (CTZW (ORI <typ.UInt32> [1<<8]  x))
 
+// Bit length (note that these will only be emitted for rva22u64 and above).
+(BitLen64 <t> x) => (SUB (MOVDconst [64]) (CLZ  <t> x))
+(BitLen32 <t> x) => (SUB (MOVDconst [32]) (CLZW <t> x))
+(BitLen16 x) => (BitLen64 (ZeroExt16to64 x))
+(BitLen8  x) => (BitLen64 (ZeroExt8to64 x))
+
 (Less64  ...) => (SLT  ...)
 (Less32  x y) => (SLT  (SignExt32to64 x) (SignExt32to64 y))
 (Less16  x y) => (SLT  (SignExt16to64 x) (SignExt16to64 y))
diff --git a/src/cmd/compile/internal/ssa/_gen/RISCV64Ops.go b/src/cmd/compile/internal/ssa/_gen/RISCV64Ops.go
index 85e9e47e82..cc2302ff37 100644
--- a/src/cmd/compile/internal/ssa/_gen/RISCV64Ops.go
+++ b/src/cmd/compile/internal/ssa/_gen/RISCV64Ops.go
@@ -229,6 +229,8 @@ func init() {
 		{name: "AND", argLength: 2, reg: gp21, asm: "AND", commutative: true},   // arg0 & arg1
 		{name: "ANDN", argLength: 2, reg: gp21, asm: "ANDN"},                    // ^arg0 & arg1
 		{name: "ANDI", argLength: 1, reg: gp11, asm: "ANDI", aux: "Int64"},      // arg0 & auxint
+		{name: "CLZ", argLength: 1, reg: gp11, asm: "CLZ"},                      // count leading zeros
+		{name: "CLZW", argLength: 1, reg: gp11, asm: "CLZW"},                    // count leading zeros of least significant word
 		{name: "CTZ", argLength: 1, reg: gp11, asm: "CTZ"},                      // count trailing zeros
 		{name: "CTZW", argLength: 1, reg: gp11, asm: "CTZW"},                    // count trailing zeros of least significant word
 		{name: "NOT", argLength: 1, reg: gp11, asm: "NOT"},                      // ^arg0
diff --git a/src/cmd/compile/internal/ssa/opGen.go b/src/cmd/compile/internal/ssa/opGen.go
index 3fd5b310ac..0ae88ab246 100644
--- a/src/cmd/compile/internal/ssa/opGen.go
+++ b/src/cmd/compile/internal/ssa/opGen.go
@@ -2523,6 +2523,8 @@ const (
 	OpRISCV64AND
 	OpRISCV64ANDN
 	OpRISCV64ANDI
+	OpRISCV64CLZ
+	OpRISCV64CLZW
 	OpRISCV64CTZ
 	OpRISCV64CTZW
 	OpRISCV64NOT
@@ -34005,6 +34007,32 @@ var opcodeTable = [...]opInfo{
 		},
 	},
 	{
+		name:   "CLZ",
+		argLen: 1,
+		asm:    riscv.ACLZ,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{0, 1006632944}, // X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X15 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X28 X29 X30
+			},
+			outputs: []outputInfo{
+				{0, 1006632944}, // X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X15 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X28 X29 X30
+			},
+		},
+	},
+	{
+		name:   "CLZW",
+		argLen: 1,
+		asm:    riscv.ACLZW,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{0, 1006632944}, // X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X15 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X28 X29 X30
+			},
+			outputs: []outputInfo{
+				{0, 1006632944}, // X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X15 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X28 X29 X30
+			},
+		},
+	},
+	{
 		name:   "CTZ",
 		argLen: 1,
 		asm:    riscv.ACTZ,
diff --git a/src/cmd/compile/internal/ssa/rewriteRISCV64.go b/src/cmd/compile/internal/ssa/rewriteRISCV64.go
index ab93309680..b2318e711b 100644
--- a/src/cmd/compile/internal/ssa/rewriteRISCV64.go
+++ b/src/cmd/compile/internal/ssa/rewriteRISCV64.go
@@ -102,6 +102,14 @@ func rewriteValueRISCV64(v *Value) bool {
 		return true
 	case OpAvg64u:
 		return rewriteValueRISCV64_OpAvg64u(v)
+	case OpBitLen16:
+		return rewriteValueRISCV64_OpBitLen16(v)
+	case OpBitLen32:
+		return rewriteValueRISCV64_OpBitLen32(v)
+	case OpBitLen64:
+		return rewriteValueRISCV64_OpBitLen64(v)
+	case OpBitLen8:
+		return rewriteValueRISCV64_OpBitLen8(v)
 	case OpClosureCall:
 		v.Op = OpRISCV64CALLclosure
 		return true
@@ -928,6 +936,72 @@ func rewriteValueRISCV64_OpAvg64u(v *Value) bool {
 		return true
 	}
 }
+func rewriteValueRISCV64_OpBitLen16(v *Value) bool {
+	v_0 := v.Args[0]
+	b := v.Block
+	typ := &b.Func.Config.Types
+	// match: (BitLen16 x)
+	// result: (BitLen64 (ZeroExt16to64 x))
+	for {
+		x := v_0
+		v.reset(OpBitLen64)
+		v0 := b.NewValue0(v.Pos, OpZeroExt16to64, typ.UInt64)
+		v0.AddArg(x)
+		v.AddArg(v0)
+		return true
+	}
+}
+func rewriteValueRISCV64_OpBitLen32(v *Value) bool {
+	v_0 := v.Args[0]
+	b := v.Block
+	typ := &b.Func.Config.Types
+	// match: (BitLen32 <t> x)
+	// result: (SUB (MOVDconst [32]) (CLZW <t> x))
+	for {
+		t := v.Type
+		x := v_0
+		v.reset(OpRISCV64SUB)
+		v0 := b.NewValue0(v.Pos, OpRISCV64MOVDconst, typ.UInt64)
+		v0.AuxInt = int64ToAuxInt(32)
+		v1 := b.NewValue0(v.Pos, OpRISCV64CLZW, t)
+		v1.AddArg(x)
+		v.AddArg2(v0, v1)
+		return true
+	}
+}
+func rewriteValueRISCV64_OpBitLen64(v *Value) bool {
+	v_0 := v.Args[0]
+	b := v.Block
+	typ := &b.Func.Config.Types
+	// match: (BitLen64 <t> x)
+	// result: (SUB (MOVDconst [64]) (CLZ <t> x))
+	for {
+		t := v.Type
+		x := v_0
+		v.reset(OpRISCV64SUB)
+		v0 := b.NewValue0(v.Pos, OpRISCV64MOVDconst, typ.UInt64)
+		v0.AuxInt = int64ToAuxInt(64)
+		v1 := b.NewValue0(v.Pos, OpRISCV64CLZ, t)
+		v1.AddArg(x)
+		v.AddArg2(v0, v1)
+		return true
+	}
+}
+func rewriteValueRISCV64_OpBitLen8(v *Value) bool {
+	v_0 := v.Args[0]
+	b := v.Block
+	typ := &b.Func.Config.Types
+	// match: (BitLen8 x)
+	// result: (BitLen64 (ZeroExt8to64 x))
+	for {
+		x := v_0
+		v.reset(OpBitLen64)
+		v0 := b.NewValue0(v.Pos, OpZeroExt8to64, typ.UInt64)
+		v0.AddArg(x)
+		v.AddArg(v0)
+		return true
+	}
+}
 func rewriteValueRISCV64_OpConst16(v *Value) bool {
 	// match: (Const16 [val])
 	// result: (MOVDconst [int64(val)])
diff --git a/src/cmd/compile/internal/ssagen/intrinsics.go b/src/cmd/compile/internal/ssagen/intrinsics.go
index f2b13045eb..eaced0b277 100644
--- a/src/cmd/compile/internal/ssagen/intrinsics.go
+++ b/src/cmd/compile/internal/ssagen/intrinsics.go
@@ -962,6 +962,30 @@ func initIntrinsics(cfg *intrinsicBuildConfig) {
 			return s.newValue1(ssa.OpBitLen8, types.Types[types.TINT], args[0])
 		},
 		sys.AMD64, sys.ARM, sys.ARM64, sys.Loong64, sys.MIPS, sys.PPC64, sys.S390X, sys.Wasm)
+
+	if cfg.goriscv64 >= 22 {
+		addF("math/bits", "Len64",
+			func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value {
+				return s.newValue1(ssa.OpBitLen64, types.Types[types.TINT], args[0])
+			},
+			sys.RISCV64)
+		addF("math/bits", "Len32",
+			func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value {
+				return s.newValue1(ssa.OpBitLen32, types.Types[types.TINT], args[0])
+			},
+			sys.RISCV64)
+		addF("math/bits", "Len16",
+			func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value {
+				return s.newValue1(ssa.OpBitLen16, types.Types[types.TINT], args[0])
+			},
+			sys.RISCV64)
+		addF("math/bits", "Len8",
+			func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value {
+				return s.newValue1(ssa.OpBitLen8, types.Types[types.TINT], args[0])
+			},
+			sys.RISCV64)
+	}
+
 	alias("math/bits", "Len", "math/bits", "Len64", p8...)
 	alias("math/bits", "Len", "math/bits", "Len32", p4...)
 
diff --git a/src/cmd/compile/internal/ssagen/intrinsics_test.go b/src/cmd/compile/internal/ssagen/intrinsics_test.go
index a06fdeedb2..230a7bdf67 100644
--- a/src/cmd/compile/internal/ssagen/intrinsics_test.go
+++ b/src/cmd/compile/internal/ssagen/intrinsics_test.go
@@ -1110,6 +1110,8 @@ var wantIntrinsics = map[testIntrinsicKey]struct{}{
 	{"riscv64", "internal/runtime/sys", "GetCallerPC"}:                 struct{}{},
 	{"riscv64", "internal/runtime/sys", "GetCallerSP"}:                 struct{}{},
 	{"riscv64", "internal/runtime/sys", "GetClosurePtr"}:               struct{}{},
+	{"riscv64", "internal/runtime/sys", "Len64"}:                       struct{}{},
+	{"riscv64", "internal/runtime/sys", "Len8"}:                        struct{}{},
 	{"riscv64", "internal/runtime/sys", "TrailingZeros32"}:             struct{}{},
 	{"riscv64", "internal/runtime/sys", "TrailingZeros64"}:             struct{}{},
 	{"riscv64", "internal/runtime/sys", "TrailingZeros8"}:              struct{}{},
@@ -1120,6 +1122,11 @@ var wantIntrinsics = map[testIntrinsicKey]struct{}{
 	{"riscv64", "math/big", "mulWW"}:                                   struct{}{},
 	{"riscv64", "math/bits", "Add"}:                                    struct{}{},
 	{"riscv64", "math/bits", "Add64"}:                                  struct{}{},
+	{"riscv64", "math/bits", "Len"}:                                    struct{}{},
+	{"riscv64", "math/bits", "Len16"}:                                  struct{}{},
+	{"riscv64", "math/bits", "Len32"}:                                  struct{}{},
+	{"riscv64", "math/bits", "Len64"}:                                  struct{}{},
+	{"riscv64", "math/bits", "Len8"}:                                   struct{}{},
 	{"riscv64", "math/bits", "Mul"}:                                    struct{}{},
 	{"riscv64", "math/bits", "Mul64"}:                                  struct{}{},
 	{"riscv64", "math/bits", "RotateLeft"}:                             struct{}{},