github/golang.go
2
0
Fork 0
mirror of https://github.com/golang/go.git synced 2026-04-02 17:30:01 +09:00
Code Issues Packages Projects Releases Wiki Activity

cmd/compile/internal/ssa: add codegen for Zicond extension on riscv64

Browse Source 
Zicond is a mandatory extension in rva23u64. This patch converts
certain branches to CondSelect and optimizes them to Zicond
instructions on RISC-V in appropriate cases, along with additional
optimization rules.

Zicond can provide performance benefits on unpredictable branches by
avoiding branch misprediction penalties. However, on simple predictable
branches, zicond uses 4 instructions vs 2 for traditional branches,
which can cause performance regressions.

To avoid regressions, we keep CondSelect globally disabled for riscv64
and only enable it for the ConstantTimeSelect intrinsic,
which has been shown to benefit from zicond:

goos: linux
goarch: riscv64
pkg: crypto/subtle
CPU: SG2044
                      │   nozicond.txt   │     zicond.txt     │
                      │     sec/op       │   sec/op   vs base │
ConstantTimeSelect-44   2.325n ± 4%   1.750n ± 2%  -24.69% (p=0.000 n=10)

Future work can explore enabling zicond for other cases that can benefit
from zicond.

Follow-up to CL 631595
Updates #75350

Co-authored-by: wangpengcheng.pp@bytedance.com
mengzhuo1203@gmail.com

Change-Id: If5d9555980e0d1e26fa924974f88943eb86b050b
GitHub-Last-Rev: 7a61508780
GitHub-Pull-Request: golang/go#75577
Reviewed-on: https://go-review.googlesource.com/c/go/+/705996
Reviewed-by: Mark Freeman <markfreeman@google.com>
Reviewed-by: David Chase <drchase@google.com>
Reviewed-by: Meng Zhuo <mengzhuo1203@gmail.com>
Reviewed-by: Joel Sing <joel@sing.id.au>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
This commit is contained in:
Xueqi Luo
2026-02-10 08:10:32 +00:00
committed by Meng Zhuo
parent cdf8951060
commit ca94cf1247
7 changed files with 839 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
src/cmd/compile/internal/riscv64/ssa.go
View File

@@ -294,7 +294,8 @@ func ssaGenValue(s *ssagen.State, v *ssa.Value) {
ssa.OpRISCV64FADDD, ssa.OpRISCV64FSUBD, ssa.OpRISCV64FMULD, ssa.OpRISCV64FDIVD,
ssa.OpRISCV64FEQD, ssa.OpRISCV64FNED, ssa.OpRISCV64FLTD, ssa.OpRISCV64FLED, ssa.OpRISCV64FSGNJD,
ssa.OpRISCV64MIN, ssa.OpRISCV64MAX, ssa.OpRISCV64MINU, ssa.OpRISCV64MAXU,
ssa.OpRISCV64SH1ADD, ssa.OpRISCV64SH2ADD, ssa.OpRISCV64SH3ADD:
ssa.OpRISCV64SH1ADD, ssa.OpRISCV64SH2ADD, ssa.OpRISCV64SH3ADD,
ssa.OpRISCV64CZEROEQZ, ssa.OpRISCV64CZERONEZ:
r := v.Reg()
r1 := v.Args[0].Reg()
r2 := v.Args[1].Reg()

19
src/cmd/compile/internal/ssa/_gen/RISCV64.rules
View File

@@ -859,3 +859,22 @@
(Max64 x y) && buildcfg.GORISCV64 >= 22 => (MAX x y)
(Min64u x y) && buildcfg.GORISCV64 >= 22 => (MINU x y)
(Max64u x y) && buildcfg.GORISCV64 >= 22 => (MAXU x y)
// Conditional selection (note that these will only be emitted for rva23u64 and above)
(CondSelect <t> x y cond) =>
(OR (CZEROEQZ <t> x (MOVBUreg <typ.UInt64> cond)) (CZERONEZ <t> y (MOVBUreg <typ.UInt64> cond)))
(CZERO(EQ|NE)Z x (SNEZ y)) => (CZERO(EQ|NE)Z x y)
(CZERO(EQ|NE)Z x (SEQZ y)) => (CZERO(NE|EQ)Z x y)
(CZEROEQZ x x) => x
(CZERONEZ x x) => (MOVDconst [0])
(CZERO(EQ|NE)Z (MOVDconst [0]) _) => (MOVDconst [0])
// Optimize conditional arithmetic operations
(OR (CZEROEQZ <t> x cond) (CZERONEZ <t> ((ADD|SUB|OR|XOR|SUBW) x y) cond)) => ((ADD|SUB|OR|XOR|SUBW) x (CZERONEZ <t> y cond))
(OR (CZEROEQZ <t> ((ADD|SUB|OR|XOR|SUBW) x y) cond) (CZERONEZ <t> x cond)) => ((ADD|SUB|OR|XOR|SUBW) x (CZEROEQZ <t> y cond))
(OR x:(CZEROEQZ z cond) (CZERONEZ y:(AND z _) cond)) => (OR y x)
(OR (CZEROEQZ x:(AND z _) cond) y:(CZERONEZ z cond)) => (OR x y)
(OR x:(CZEROEQZ z cond) (CZERONEZ y:(ANDI <t> [c] z) cond)) => (OR y x)
(OR (CZEROEQZ x:(ANDI <t> [c] z) cond) y:(CZERONEZ z cond)) => (OR x y)
(OR (CZEROEQZ <t> x cond) (CZERONEZ <t> ((ADDI|ORI|XORI) [c] x) cond)) => ((ADD|OR|XOR) x (CZERONEZ <t> (MOVDconst [c]) cond))
(OR (CZEROEQZ <t> ((ADDI|ORI|XORI) [c] x) cond) (CZERONEZ <t> x cond)) => ((ADD|OR|XOR) x (CZEROEQZ <t> (MOVDconst [c]) cond))

4
src/cmd/compile/internal/ssa/_gen/RISCV64Ops.go
View File

@@ -528,6 +528,10 @@ func init() {
// ====+=============================
{name: "FCLASSS", argLength: 1, reg: fpgp, asm: "FCLASSS", typ: "Int64"}, // classify float32
{name: "FCLASSD", argLength: 1, reg: fpgp, asm: "FCLASSD", typ: "Int64"}, // classify float64
// RISC-V Integer Conditional (Zicond) operations extension
{name: "CZEROEQZ", argLength: 2, reg: gp21, asm: "CZEROEQZ"}, // arg1 == 0 result is 0, else arg0
{name: "CZERONEZ", argLength: 2, reg: gp21, asm: "CZERONEZ"}, // arg1 != 0 result is 0, else arg0
}
RISCV64blocks := []blockData{

30
src/cmd/compile/internal/ssa/opGen.go
View File

@@ -5356,6 +5356,8 @@ const (
OpRISCV64LoweredFMAXD
OpRISCV64FCLASSS
OpRISCV64FCLASSD
OpRISCV64CZEROEQZ
OpRISCV64CZERONEZ
OpS390XFADDS
OpS390XFADD
@@ -81426,6 +81428,34 @@ var opcodeTable = [...]opInfo{
},
},
},
{
name: "CZEROEQZ",
argLen: 2,
asm: riscv.ACZEROEQZ,
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
{1, 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: "CZERONEZ",
argLen: 2,
asm: riscv.ACZERONEZ,
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
{1, 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: "FADDS",

778
src/cmd/compile/internal/ssa/rewriteRISCV64.go
View File

@@ -133,6 +133,8 @@ func rewriteValueRISCV64(v *Value) bool {
case OpCom8:
v.Op = OpRISCV64NOT
return true
case OpCondSelect:
return rewriteValueRISCV64_OpCondSelect(v)
case OpConst16:
return rewriteValueRISCV64_OpConst16(v)
case OpConst32:
@@ -514,6 +516,10 @@ func rewriteValueRISCV64(v *Value) bool {
return rewriteValueRISCV64_OpRISCV64AND(v)
case OpRISCV64ANDI:
return rewriteValueRISCV64_OpRISCV64ANDI(v)
case OpRISCV64CZEROEQZ:
return rewriteValueRISCV64_OpRISCV64CZEROEQZ(v)
case OpRISCV64CZERONEZ:
return rewriteValueRISCV64_OpRISCV64CZERONEZ(v)
case OpRISCV64FADDD:
return rewriteValueRISCV64_OpRISCV64FADDD(v)
case OpRISCV64FADDS:
@@ -1081,6 +1087,30 @@ func rewriteValueRISCV64_OpBswap32(v *Value) bool {
return true
}
}
func rewriteValueRISCV64_OpCondSelect(v *Value) bool {
v_2 := v.Args[2]
v_1 := v.Args[1]
v_0 := v.Args[0]
b := v.Block
typ := &b.Func.Config.Types
// match: (CondSelect <t> x y cond)
// result: (OR (CZEROEQZ <t> x (MOVBUreg <typ.UInt64> cond)) (CZERONEZ <t> y (MOVBUreg <typ.UInt64> cond)))
for {
t := v.Type
x := v_0
y := v_1
cond := v_2
v.reset(OpRISCV64OR)
v0 := b.NewValue0(v.Pos, OpRISCV64CZEROEQZ, t)
v1 := b.NewValue0(v.Pos, OpRISCV64MOVBUreg, typ.UInt64)
v1.AddArg(cond)
v0.AddArg2(x, v1)
v2 := b.NewValue0(v.Pos, OpRISCV64CZERONEZ, t)
v2.AddArg2(y, v1)
v.AddArg2(v0, v2)
return true
}
}
func rewriteValueRISCV64_OpConst16(v *Value) bool {
// match: (Const16 [val])
// result: (MOVDconst [int64(val)])
@@ -3528,6 +3558,105 @@ func rewriteValueRISCV64_OpRISCV64ANDI(v *Value) bool {
}
return false
}
func rewriteValueRISCV64_OpRISCV64CZEROEQZ(v *Value) bool {
v_1 := v.Args[1]
v_0 := v.Args[0]
// match: (CZEROEQZ x (SNEZ y))
// result: (CZEROEQZ x y)
for {
x := v_0
if v_1.Op != OpRISCV64SNEZ {
break
}
y := v_1.Args[0]
v.reset(OpRISCV64CZEROEQZ)
v.AddArg2(x, y)
return true
}
// match: (CZEROEQZ x (SEQZ y))
// result: (CZERONEZ x y)
for {
x := v_0
if v_1.Op != OpRISCV64SEQZ {
break
}
y := v_1.Args[0]
v.reset(OpRISCV64CZERONEZ)
v.AddArg2(x, y)
return true
}
// match: (CZEROEQZ x x)
// result: x
for {
x := v_0
if x != v_1 {
break
}
v.copyOf(x)
return true
}
// match: (CZEROEQZ (MOVDconst [0]) _)
// result: (MOVDconst [0])
for {
if v_0.Op != OpRISCV64MOVDconst || auxIntToInt64(v_0.AuxInt) != 0 {
break
}
v.reset(OpRISCV64MOVDconst)
v.AuxInt = int64ToAuxInt(0)
return true
}
return false
}
func rewriteValueRISCV64_OpRISCV64CZERONEZ(v *Value) bool {
v_1 := v.Args[1]
v_0 := v.Args[0]
// match: (CZERONEZ x (SNEZ y))
// result: (CZERONEZ x y)
for {
x := v_0
if v_1.Op != OpRISCV64SNEZ {
break
}
y := v_1.Args[0]
v.reset(OpRISCV64CZERONEZ)
v.AddArg2(x, y)
return true
}
// match: (CZERONEZ x (SEQZ y))
// result: (CZEROEQZ x y)
for {
x := v_0
if v_1.Op != OpRISCV64SEQZ {
break
}
y := v_1.Args[0]
v.reset(OpRISCV64CZEROEQZ)
v.AddArg2(x, y)
return true
}
// match: (CZERONEZ x x)
// result: (MOVDconst [0])
for {
x := v_0
if x != v_1 {
break
}
v.reset(OpRISCV64MOVDconst)
v.AuxInt = int64ToAuxInt(0)
return true
}
// match: (CZERONEZ (MOVDconst [0]) _)
// result: (MOVDconst [0])
for {
if v_0.Op != OpRISCV64MOVDconst || auxIntToInt64(v_0.AuxInt) != 0 {
break
}
v.reset(OpRISCV64MOVDconst)
v.AuxInt = int64ToAuxInt(0)
return true
}
return false
}
func rewriteValueRISCV64_OpRISCV64FADDD(v *Value) bool {
v_1 := v.Args[1]
v_0 := v.Args[0]
@@ -6925,6 +7054,8 @@ func rewriteValueRISCV64_OpRISCV64NEGW(v *Value) bool {
func rewriteValueRISCV64_OpRISCV64OR(v *Value) bool {
v_1 := v.Args[1]
v_0 := v.Args[0]
b := v.Block
typ := &b.Func.Config.Types
// match: (OR (MOVDconst [val]) x)
// cond: is32Bit(val)
// result: (ORI [val] x)
@@ -6955,6 +7086,653 @@ func rewriteValueRISCV64_OpRISCV64OR(v *Value) bool {
v.copyOf(x)
return true
}
// match: (OR (CZEROEQZ <t> x cond) (CZERONEZ <t> (ADD x y) cond))
// result: (ADD x (CZERONEZ <t> y cond))
for {
for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
if v_0.Op != OpRISCV64CZEROEQZ {
continue
}
t := v_0.Type
cond := v_0.Args[1]
x := v_0.Args[0]
if v_1.Op != OpRISCV64CZERONEZ || v_1.Type != t {
continue
}
_ = v_1.Args[1]
v_1_0 := v_1.Args[0]
if v_1_0.Op != OpRISCV64ADD {
continue
}
_ = v_1_0.Args[1]
v_1_0_0 := v_1_0.Args[0]
v_1_0_1 := v_1_0.Args[1]
for _i1 := 0; _i1 <= 1; _i1, v_1_0_0, v_1_0_1 = _i1+1, v_1_0_1, v_1_0_0 {
if x != v_1_0_0 {
continue
}
y := v_1_0_1
if cond != v_1.Args[1] {
continue
}
v.reset(OpRISCV64ADD)
v0 := b.NewValue0(v.Pos, OpRISCV64CZERONEZ, t)
v0.AddArg2(y, cond)
v.AddArg2(x, v0)
return true
}
}
break
}
// match: (OR (CZEROEQZ <t> x cond) (CZERONEZ <t> (SUB x y) cond))
// result: (SUB x (CZERONEZ <t> y cond))
for {
for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
if v_0.Op != OpRISCV64CZEROEQZ {
continue
}
t := v_0.Type
cond := v_0.Args[1]
x := v_0.Args[0]
if v_1.Op != OpRISCV64CZERONEZ || v_1.Type != t {
continue
}
_ = v_1.Args[1]
v_1_0 := v_1.Args[0]
if v_1_0.Op != OpRISCV64SUB {
continue
}
y := v_1_0.Args[1]
if x != v_1_0.Args[0] || cond != v_1.Args[1] {
continue
}
v.reset(OpRISCV64SUB)
v0 := b.NewValue0(v.Pos, OpRISCV64CZERONEZ, t)
v0.AddArg2(y, cond)
v.AddArg2(x, v0)
return true
}
break
}
// match: (OR (CZEROEQZ <t> x cond) (CZERONEZ <t> (OR x y) cond))
// result: (OR x (CZERONEZ <t> y cond))
for {
for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
if v_0.Op != OpRISCV64CZEROEQZ {
continue
}
t := v_0.Type
cond := v_0.Args[1]
x := v_0.Args[0]
if v_1.Op != OpRISCV64CZERONEZ || v_1.Type != t {
continue
}
_ = v_1.Args[1]
v_1_0 := v_1.Args[0]
if v_1_0.Op != OpRISCV64OR {
continue
}
_ = v_1_0.Args[1]
v_1_0_0 := v_1_0.Args[0]
v_1_0_1 := v_1_0.Args[1]
for _i1 := 0; _i1 <= 1; _i1, v_1_0_0, v_1_0_1 = _i1+1, v_1_0_1, v_1_0_0 {
if x != v_1_0_0 {
continue
}
y := v_1_0_1
if cond != v_1.Args[1] {
continue
}
v.reset(OpRISCV64OR)
v0 := b.NewValue0(v.Pos, OpRISCV64CZERONEZ, t)
v0.AddArg2(y, cond)
v.AddArg2(x, v0)
return true
}
}
break
}
// match: (OR (CZEROEQZ <t> x cond) (CZERONEZ <t> (XOR x y) cond))
// result: (XOR x (CZERONEZ <t> y cond))
for {
for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
if v_0.Op != OpRISCV64CZEROEQZ {
continue
}
t := v_0.Type
cond := v_0.Args[1]
x := v_0.Args[0]
if v_1.Op != OpRISCV64CZERONEZ || v_1.Type != t {
continue
}
_ = v_1.Args[1]
v_1_0 := v_1.Args[0]
if v_1_0.Op != OpRISCV64XOR {
continue
}
_ = v_1_0.Args[1]
v_1_0_0 := v_1_0.Args[0]
v_1_0_1 := v_1_0.Args[1]
for _i1 := 0; _i1 <= 1; _i1, v_1_0_0, v_1_0_1 = _i1+1, v_1_0_1, v_1_0_0 {
if x != v_1_0_0 {
continue
}
y := v_1_0_1
if cond != v_1.Args[1] {
continue
}
v.reset(OpRISCV64XOR)
v0 := b.NewValue0(v.Pos, OpRISCV64CZERONEZ, t)
v0.AddArg2(y, cond)
v.AddArg2(x, v0)
return true
}
}
break
}
// match: (OR (CZEROEQZ <t> x cond) (CZERONEZ <t> (SUBW x y) cond))
// result: (SUBW x (CZERONEZ <t> y cond))
for {
for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
if v_0.Op != OpRISCV64CZEROEQZ {
continue
}
t := v_0.Type
cond := v_0.Args[1]
x := v_0.Args[0]
if v_1.Op != OpRISCV64CZERONEZ || v_1.Type != t {
continue
}
_ = v_1.Args[1]
v_1_0 := v_1.Args[0]
if v_1_0.Op != OpRISCV64SUBW {
continue
}
y := v_1_0.Args[1]
if x != v_1_0.Args[0] || cond != v_1.Args[1] {
continue
}
v.reset(OpRISCV64SUBW)
v0 := b.NewValue0(v.Pos, OpRISCV64CZERONEZ, t)
v0.AddArg2(y, cond)
v.AddArg2(x, v0)
return true
}
break
}
// match: (OR (CZEROEQZ <t> (ADD x y) cond) (CZERONEZ <t> x cond))
// result: (ADD x (CZEROEQZ <t> y cond))
for {
for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
if v_0.Op != OpRISCV64CZEROEQZ {
continue
}
t := v_0.Type
cond := v_0.Args[1]
v_0_0 := v_0.Args[0]
if v_0_0.Op != OpRISCV64ADD {
continue
}
_ = v_0_0.Args[1]
v_0_0_0 := v_0_0.Args[0]
v_0_0_1 := v_0_0.Args[1]
for _i1 := 0; _i1 <= 1; _i1, v_0_0_0, v_0_0_1 = _i1+1, v_0_0_1, v_0_0_0 {
x := v_0_0_0
y := v_0_0_1
if v_1.Op != OpRISCV64CZERONEZ || v_1.Type != t {
continue
}
_ = v_1.Args[1]
if x != v_1.Args[0] || cond != v_1.Args[1] {
continue
}
v.reset(OpRISCV64ADD)
v0 := b.NewValue0(v.Pos, OpRISCV64CZEROEQZ, t)
v0.AddArg2(y, cond)
v.AddArg2(x, v0)
return true
}
}
break
}
// match: (OR (CZEROEQZ <t> (SUB x y) cond) (CZERONEZ <t> x cond))
// result: (SUB x (CZEROEQZ <t> y cond))
for {
for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
if v_0.Op != OpRISCV64CZEROEQZ {
continue
}
t := v_0.Type
cond := v_0.Args[1]
v_0_0 := v_0.Args[0]
if v_0_0.Op != OpRISCV64SUB {
continue
}
y := v_0_0.Args[1]
x := v_0_0.Args[0]
if v_1.Op != OpRISCV64CZERONEZ || v_1.Type != t {
continue
}
_ = v_1.Args[1]
if x != v_1.Args[0] || cond != v_1.Args[1] {
continue
}
v.reset(OpRISCV64SUB)
v0 := b.NewValue0(v.Pos, OpRISCV64CZEROEQZ, t)
v0.AddArg2(y, cond)
v.AddArg2(x, v0)
return true
}
break
}
// match: (OR (CZEROEQZ <t> (OR x y) cond) (CZERONEZ <t> x cond))
// result: (OR x (CZEROEQZ <t> y cond))
for {
for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
if v_0.Op != OpRISCV64CZEROEQZ {
continue
}
t := v_0.Type
cond := v_0.Args[1]
v_0_0 := v_0.Args[0]
if v_0_0.Op != OpRISCV64OR {
continue
}
_ = v_0_0.Args[1]
v_0_0_0 := v_0_0.Args[0]
v_0_0_1 := v_0_0.Args[1]
for _i1 := 0; _i1 <= 1; _i1, v_0_0_0, v_0_0_1 = _i1+1, v_0_0_1, v_0_0_0 {
x := v_0_0_0
y := v_0_0_1
if v_1.Op != OpRISCV64CZERONEZ || v_1.Type != t {
continue
}
_ = v_1.Args[1]
if x != v_1.Args[0] || cond != v_1.Args[1] {
continue
}
v.reset(OpRISCV64OR)
v0 := b.NewValue0(v.Pos, OpRISCV64CZEROEQZ, t)
v0.AddArg2(y, cond)
v.AddArg2(x, v0)
return true
}
}
break
}
// match: (OR (CZEROEQZ <t> (XOR x y) cond) (CZERONEZ <t> x cond))
// result: (XOR x (CZEROEQZ <t> y cond))
for {
for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
if v_0.Op != OpRISCV64CZEROEQZ {
continue
}
t := v_0.Type
cond := v_0.Args[1]
v_0_0 := v_0.Args[0]
if v_0_0.Op != OpRISCV64XOR {
continue
}
_ = v_0_0.Args[1]
v_0_0_0 := v_0_0.Args[0]
v_0_0_1 := v_0_0.Args[1]
for _i1 := 0; _i1 <= 1; _i1, v_0_0_0, v_0_0_1 = _i1+1, v_0_0_1, v_0_0_0 {
x := v_0_0_0
y := v_0_0_1
if v_1.Op != OpRISCV64CZERONEZ || v_1.Type != t {
continue
}
_ = v_1.Args[1]
if x != v_1.Args[0] || cond != v_1.Args[1] {
continue
}
v.reset(OpRISCV64XOR)
v0 := b.NewValue0(v.Pos, OpRISCV64CZEROEQZ, t)
v0.AddArg2(y, cond)
v.AddArg2(x, v0)
return true
}
}
break
}
// match: (OR (CZEROEQZ <t> (SUBW x y) cond) (CZERONEZ <t> x cond))
// result: (SUBW x (CZEROEQZ <t> y cond))
for {
for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
if v_0.Op != OpRISCV64CZEROEQZ {
continue
}
t := v_0.Type
cond := v_0.Args[1]
v_0_0 := v_0.Args[0]
if v_0_0.Op != OpRISCV64SUBW {
continue
}
y := v_0_0.Args[1]
x := v_0_0.Args[0]
if v_1.Op != OpRISCV64CZERONEZ || v_1.Type != t {
continue
}
_ = v_1.Args[1]
if x != v_1.Args[0] || cond != v_1.Args[1] {
continue
}
v.reset(OpRISCV64SUBW)
v0 := b.NewValue0(v.Pos, OpRISCV64CZEROEQZ, t)
v0.AddArg2(y, cond)
v.AddArg2(x, v0)
return true
}
break
}
// match: (OR x:(CZEROEQZ z cond) (CZERONEZ y:(AND z _) cond))
// result: (OR y x)
for {
for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
x := v_0
if x.Op != OpRISCV64CZEROEQZ {
continue
}
cond := x.Args[1]
z := x.Args[0]
if v_1.Op != OpRISCV64CZERONEZ {
continue
}
_ = v_1.Args[1]
y := v_1.Args[0]
if y.Op != OpRISCV64AND {
continue
}
y_0 := y.Args[0]
y_1 := y.Args[1]
for _i1 := 0; _i1 <= 1; _i1, y_0, y_1 = _i1+1, y_1, y_0 {
if z != y_0 || cond != v_1.Args[1] {
continue
}
v.reset(OpRISCV64OR)
v.AddArg2(y, x)
return true
}
}
break
}
// match: (OR (CZEROEQZ x:(AND z _) cond) y:(CZERONEZ z cond))
// result: (OR x y)
for {
for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
if v_0.Op != OpRISCV64CZEROEQZ {
continue
}
cond := v_0.Args[1]
x := v_0.Args[0]
if x.Op != OpRISCV64AND {
continue
}
x_0 := x.Args[0]
x_1 := x.Args[1]
for _i1 := 0; _i1 <= 1; _i1, x_0, x_1 = _i1+1, x_1, x_0 {
z := x_0
y := v_1
if y.Op != OpRISCV64CZERONEZ {
continue
}
_ = y.Args[1]
if z != y.Args[0] || cond != y.Args[1] {
continue
}
v.reset(OpRISCV64OR)
v.AddArg2(x, y)
return true
}
}
break
}
// match: (OR x:(CZEROEQZ z cond) (CZERONEZ y:(ANDI <t> [c] z) cond))
// result: (OR y x)
for {
for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
x := v_0
if x.Op != OpRISCV64CZEROEQZ {
continue
}
cond := x.Args[1]
z := x.Args[0]
if v_1.Op != OpRISCV64CZERONEZ {
continue
}
_ = v_1.Args[1]
y := v_1.Args[0]
if y.Op != OpRISCV64ANDI {
continue
}
if z != y.Args[0] || cond != v_1.Args[1] {
continue
}
v.reset(OpRISCV64OR)
v.AddArg2(y, x)
return true
}
break
}
// match: (OR (CZEROEQZ x:(ANDI <t> [c] z) cond) y:(CZERONEZ z cond))
// result: (OR x y)
for {
for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
if v_0.Op != OpRISCV64CZEROEQZ {
continue
}
cond := v_0.Args[1]
x := v_0.Args[0]
if x.Op != OpRISCV64ANDI {
continue
}
z := x.Args[0]
y := v_1
if y.Op != OpRISCV64CZERONEZ {
continue
}
_ = y.Args[1]
if z != y.Args[0] || cond != y.Args[1] {
continue
}
v.reset(OpRISCV64OR)
v.AddArg2(x, y)
return true
}
break
}
// match: (OR (CZEROEQZ <t> x cond) (CZERONEZ <t> (ADDI [c] x) cond))
// result: (ADD x (CZERONEZ <t> (MOVDconst [c]) cond))
for {
for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
if v_0.Op != OpRISCV64CZEROEQZ {
continue
}
t := v_0.Type
cond := v_0.Args[1]
x := v_0.Args[0]
if v_1.Op != OpRISCV64CZERONEZ || v_1.Type != t {
continue
}
_ = v_1.Args[1]
v_1_0 := v_1.Args[0]
if v_1_0.Op != OpRISCV64ADDI {
continue
}
c := auxIntToInt64(v_1_0.AuxInt)
if x != v_1_0.Args[0] || cond != v_1.Args[1] {
continue
}
v.reset(OpRISCV64ADD)
v0 := b.NewValue0(v.Pos, OpRISCV64CZERONEZ, t)
v1 := b.NewValue0(v.Pos, OpRISCV64MOVDconst, typ.UInt64)
v1.AuxInt = int64ToAuxInt(c)
v0.AddArg2(v1, cond)
v.AddArg2(x, v0)
return true
}
break
}
// match: (OR (CZEROEQZ <t> x cond) (CZERONEZ <t> (ORI [c] x) cond))
// result: (OR x (CZERONEZ <t> (MOVDconst [c]) cond))
for {
for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
if v_0.Op != OpRISCV64CZEROEQZ {
continue
}
t := v_0.Type
cond := v_0.Args[1]
x := v_0.Args[0]
if v_1.Op != OpRISCV64CZERONEZ || v_1.Type != t {
continue
}
_ = v_1.Args[1]
v_1_0 := v_1.Args[0]
if v_1_0.Op != OpRISCV64ORI {
continue
}
c := auxIntToInt64(v_1_0.AuxInt)
if x != v_1_0.Args[0] || cond != v_1.Args[1] {
continue
}
v.reset(OpRISCV64OR)
v0 := b.NewValue0(v.Pos, OpRISCV64CZERONEZ, t)
v1 := b.NewValue0(v.Pos, OpRISCV64MOVDconst, typ.UInt64)
v1.AuxInt = int64ToAuxInt(c)
v0.AddArg2(v1, cond)
v.AddArg2(x, v0)
return true
}
break
}
// match: (OR (CZEROEQZ <t> x cond) (CZERONEZ <t> (XORI [c] x) cond))
// result: (XOR x (CZERONEZ <t> (MOVDconst [c]) cond))
for {
for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
if v_0.Op != OpRISCV64CZEROEQZ {
continue
}
t := v_0.Type
cond := v_0.Args[1]
x := v_0.Args[0]
if v_1.Op != OpRISCV64CZERONEZ || v_1.Type != t {
continue
}
_ = v_1.Args[1]
v_1_0 := v_1.Args[0]
if v_1_0.Op != OpRISCV64XORI {
continue
}
c := auxIntToInt64(v_1_0.AuxInt)
if x != v_1_0.Args[0] || cond != v_1.Args[1] {
continue
}
v.reset(OpRISCV64XOR)
v0 := b.NewValue0(v.Pos, OpRISCV64CZERONEZ, t)
v1 := b.NewValue0(v.Pos, OpRISCV64MOVDconst, typ.UInt64)
v1.AuxInt = int64ToAuxInt(c)
v0.AddArg2(v1, cond)
v.AddArg2(x, v0)
return true
}
break
}
// match: (OR (CZEROEQZ <t> (ADDI [c] x) cond) (CZERONEZ <t> x cond))
// result: (ADD x (CZEROEQZ <t> (MOVDconst [c]) cond))
for {
for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
if v_0.Op != OpRISCV64CZEROEQZ {
continue
}
t := v_0.Type
cond := v_0.Args[1]
v_0_0 := v_0.Args[0]
if v_0_0.Op != OpRISCV64ADDI {
continue
}
c := auxIntToInt64(v_0_0.AuxInt)
x := v_0_0.Args[0]
if v_1.Op != OpRISCV64CZERONEZ || v_1.Type != t {
continue
}
_ = v_1.Args[1]
if x != v_1.Args[0] || cond != v_1.Args[1] {
continue
}
v.reset(OpRISCV64ADD)
v0 := b.NewValue0(v.Pos, OpRISCV64CZEROEQZ, t)
v1 := b.NewValue0(v.Pos, OpRISCV64MOVDconst, typ.UInt64)
v1.AuxInt = int64ToAuxInt(c)
v0.AddArg2(v1, cond)
v.AddArg2(x, v0)
return true
}
break
}
// match: (OR (CZEROEQZ <t> (ORI [c] x) cond) (CZERONEZ <t> x cond))
// result: (OR x (CZEROEQZ <t> (MOVDconst [c]) cond))
for {
for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
if v_0.Op != OpRISCV64CZEROEQZ {
continue
}
t := v_0.Type
cond := v_0.Args[1]
v_0_0 := v_0.Args[0]
if v_0_0.Op != OpRISCV64ORI {
continue
}
c := auxIntToInt64(v_0_0.AuxInt)
x := v_0_0.Args[0]
if v_1.Op != OpRISCV64CZERONEZ || v_1.Type != t {
continue
}
_ = v_1.Args[1]
if x != v_1.Args[0] || cond != v_1.Args[1] {
continue
}
v.reset(OpRISCV64OR)
v0 := b.NewValue0(v.Pos, OpRISCV64CZEROEQZ, t)
v1 := b.NewValue0(v.Pos, OpRISCV64MOVDconst, typ.UInt64)
v1.AuxInt = int64ToAuxInt(c)
v0.AddArg2(v1, cond)
v.AddArg2(x, v0)
return true
}
break
}
// match: (OR (CZEROEQZ <t> (XORI [c] x) cond) (CZERONEZ <t> x cond))
// result: (XOR x (CZEROEQZ <t> (MOVDconst [c]) cond))
for {
for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
if v_0.Op != OpRISCV64CZEROEQZ {
continue
}
t := v_0.Type
cond := v_0.Args[1]
v_0_0 := v_0.Args[0]
if v_0_0.Op != OpRISCV64XORI {
continue
}
c := auxIntToInt64(v_0_0.AuxInt)
x := v_0_0.Args[0]
if v_1.Op != OpRISCV64CZERONEZ || v_1.Type != t {
continue
}
_ = v_1.Args[1]
if x != v_1.Args[0] || cond != v_1.Args[1] {
continue
}
v.reset(OpRISCV64XOR)
v0 := b.NewValue0(v.Pos, OpRISCV64CZEROEQZ, t)
v1 := b.NewValue0(v.Pos, OpRISCV64MOVDconst, typ.UInt64)
v1.AuxInt = int64ToAuxInt(c)
v0.AddArg2(v1, cond)
v.AddArg2(x, v0)
return true
}
break
}
return false
}
func rewriteValueRISCV64_OpRISCV64ORI(v *Value) bool {

7
src/cmd/compile/internal/ssagen/intrinsics.go
View File

@@ -1612,6 +1612,10 @@ func initIntrinsics(cfg *intrinsicBuildConfig) {
/******** crypto/internal/constanttime ********/
// We implement a superset of the Select promise:
// Select returns x if v != 0 and y if v == 0.
hasCMOV := []*sys.Arch{sys.ArchAMD64, sys.ArchARM64, sys.ArchLoong64, sys.ArchPPC64, sys.ArchPPC64LE, sys.ArchWasm}
if cfg.goriscv64 >= 23 {
hasCMOV = append(hasCMOV, sys.ArchRISCV64)
}
add("crypto/internal/constanttime", "Select",
func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value {
v, x, y := args[0], args[1], args[2]
@@ -1631,8 +1635,7 @@ func initIntrinsics(cfg *intrinsicBuildConfig) {
check := s.newValue2(checkOp, types.Types[types.TBOOL], zero, v)
return s.newValue3(ssa.OpCondSelect, types.Types[types.TINT], x, y, check)
},
sys.ArchAMD64, sys.ArchARM64, sys.ArchLoong64, sys.ArchPPC64, sys.ArchPPC64LE, sys.ArchWasm) // all with CMOV support.
}, hasCMOV...) // all with CMOV support.
add("crypto/internal/constanttime", "boolToUint8",
func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value {
return s.newValue1(ssa.OpCvtBoolToUint8, types.Types[types.TUINT8], args[0])

1
src/cmd/compile/internal/ssagen/intrinsics_test.go
View File

@@ -1204,6 +1204,7 @@ var wantIntrinsics = map[testIntrinsicKey]struct{}{
{"riscv64", "sync/atomic", "SwapUint32"}: struct{}{},
{"riscv64", "sync/atomic", "SwapUint64"}: struct{}{},
{"riscv64", "sync/atomic", "SwapUintptr"}: struct{}{},
{"riscv64", "crypto/internal/constanttime", "Select"}: struct{}{},
{"riscv64", "crypto/internal/constanttime", "boolToUint8"}: struct{}{},
{"s390x", "internal/runtime/atomic", "And"}: struct{}{},
{"s390x", "internal/runtime/atomic", "And8"}: struct{}{},