mleku/wazero
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

amd64: correctly select sign of zeros on f64/32 min/max (#731)

Browse Source 
Signed-off-by: Takeshi Yoneda <takeshi@tetrate.io>
This commit is contained in:
Takeshi Yoneda
2022-08-02 17:46:08 +09:00
committed by GitHub
parent 76c5d98786
commit b02b5513e8
2 changed files with 36 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
internal/engine/compiler/compiler_numeric_test.go
View File

@@ -843,7 +843,7 @@ func TestCompiler_compile_Min_Max_Copysign(t *testing.T) {
if math.IsNaN(float64(exp)) { // NaN cannot be compared with themselves, so we have to use IsNaN
require.True(t, math.IsNaN(float64(actual)))
} else {
require.Equal(t, exp, actual)
require.Equal(t, math.Float32bits(exp), math.Float32bits(actual))
}
},
},
@@ -860,7 +860,7 @@ func TestCompiler_compile_Min_Max_Copysign(t *testing.T) {
if math.IsNaN(exp) { // NaN cannot be compared with themselves, so we have to use IsNaN
require.True(t, math.IsNaN(actual))
} else {
require.Equal(t, exp, actual)
require.Equal(t, math.Float64bits(exp), math.Float64bits(actual))
}
},
},
@@ -877,7 +877,7 @@ func TestCompiler_compile_Min_Max_Copysign(t *testing.T) {
if math.IsNaN(float64(exp)) { // NaN cannot be compared with themselves, so we have to use IsNaN
require.True(t, math.IsNaN(float64(actual)))
} else {
require.Equal(t, exp, actual)
require.Equal(t, math.Float32bits(exp), math.Float32bits(actual))
}
},
},
@@ -894,7 +894,7 @@ func TestCompiler_compile_Min_Max_Copysign(t *testing.T) {
if math.IsNaN(exp) { // NaN cannot be compared with themselves, so we have to use IsNaN
require.True(t, math.IsNaN(actual))
} else {
require.Equal(t, exp, actual)
require.Equal(t, math.Float64bits(exp), math.Float64bits(actual))
}
},
},
@@ -938,6 +938,10 @@ func TestCompiler_compile_Min_Max_Copysign(t *testing.T) {
tc := tt
t.Run(tc.name, func(t *testing.T) {
for _, vs := range [][2]float64{
{math.Copysign(0, 1), math.Copysign(0, 1)},
{math.Copysign(0, -1), math.Copysign(0, 1)},
{math.Copysign(0, 1), math.Copysign(0, -1)},
{math.Copysign(0, -1), math.Copysign(0, -1)},
{100, -1.1}, {100, 0}, {0, 0}, {1, 1},
{-1, 100}, {100, 200}, {100.01234124, 100.01234124},
{100.01234124, -100.01234124}, {200.12315, 100},

38
internal/engine/compiler/impl_amd64.go
View File

@@ -1822,9 +1822,9 @@ func (c *amd64Compiler) compileRoundInstruction(is32Bit bool, mode int64) error
func (c *amd64Compiler) compileMin(o *wazeroir.OperationMin) error {
is32Bit := o.Type == wazeroir.Float32
if is32Bit {
return c.compileMinOrMax(is32Bit, amd64.MINSS)
return c.compileMinOrMax(is32Bit, true, amd64.MINSS)
} else {
return c.compileMinOrMax(is32Bit, amd64.MINSD)
return c.compileMinOrMax(is32Bit, true, amd64.MINSD)
}
}
@@ -1832,9 +1832,9 @@ func (c *amd64Compiler) compileMin(o *wazeroir.OperationMin) error {
func (c *amd64Compiler) compileMax(o *wazeroir.OperationMax) error {
is32Bit := o.Type == wazeroir.Float32
if is32Bit {
return c.compileMinOrMax(is32Bit, amd64.MAXSS)
return c.compileMinOrMax(is32Bit, false, amd64.MAXSS)
} else {
return c.compileMinOrMax(is32Bit, amd64.MAXSD)
return c.compileMinOrMax(is32Bit, false, amd64.MAXSD)
}
}
@@ -1850,7 +1850,7 @@ func (c *amd64Compiler) compileMax(o *wazeroir.OperationMax) error {
// the native min/max, which is why we cannot simply emit a native min/max instruction here.
//
// For the semantics, see wazeroir.Min and wazeroir.Max for detail.
func (c *amd64Compiler) compileMinOrMax(is32Bit bool, minOrMaxInstruction asm.Instruction) error {
func (c *amd64Compiler) compileMinOrMax(is32Bit, isMin bool, minOrMaxInstruction asm.Instruction) error {
x2 := c.locationStack.pop()
if err := c.compileEnsureOnRegister(x2); err != nil {
return err
@@ -1880,12 +1880,30 @@ func (c *amd64Compiler) compileMinOrMax(is32Bit bool, minOrMaxInstruction asm.In
// Start handling 2) and 3).
// Jump if two values are equal and NaN-free by checking the parity flag (PF).
// Here we use JPC to do the conditional jump when the parity flag is NOT set,
// and that is of 2).
equalExitJmp := c.assembler.CompileJump(amd64.JPC)
// Jump if one of two values is NaN by checking the parity flag (PF).
includeNaNJmp := c.assembler.CompileJump(amd64.JPS)
// Start handling 2).
// Before we exit this case, we have to ensure that positive zero (or negative zero for min instruction) is
// returned if two values are positive and negative zeros.
var inst asm.Instruction
switch {
case is32Bit && isMin:
inst = amd64.ORPS
case !is32Bit && isMin:
inst = amd64.ORPD
case is32Bit && !isMin:
inst = amd64.ANDPS
case !is32Bit && !isMin:
inst = amd64.ANDPD
}
c.assembler.CompileRegisterToRegister(inst, x2.register, x1.register)
sameExitJmp := c.assembler.CompileJump(amd64.JMP)
// Start handling 3).
c.assembler.SetJumpTargetOnNext(includeNaNJmp)
// We emit the ADD instruction to produce the NaN in x1.
if is32Bit {
@@ -1904,7 +1922,7 @@ func (c *amd64Compiler) compileMinOrMax(is32Bit bool, minOrMaxInstruction asm.In
c.assembler.CompileRegisterToRegister(minOrMaxInstruction, x2.register, x1.register)
// Set the jump target of 1) and 2) cases to the next instruction after 3) case.
c.assembler.SetJumpTargetOnNext(nanExitJmp, equalExitJmp)
c.assembler.SetJumpTargetOnNext(nanExitJmp, sameExitJmp)
// Record that we consumed the x2 and placed the minOrMax result in the x1's register.
c.locationStack.markRegisterUnused(x2.register)