Skip to content

G
go
Commit 68e86e6d authored by Keith Randall's avatar Keith Randall
Browse files
Options

cmd/compile: MOVBload and MOVBQZXload are the same op 

No need to have both ops when they do the same thing.
Just declare MOVBload to zero extend and we can get rid
of MOVBQZXload.  Same for W and L.

Kind of a followon cleanup for https://go-review.googlesource.com/c/19506/
Should enable an easier fix for #14920

Change-Id: I7cfac909a8ba387f433a6ae75c050740ebb34d42
Reviewed-on: https://go-review.googlesource.com/21004
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: default avatarBrad Fitzpatrick <bradfitz@golang.org>
parent 5cdb3d03
Expand all Hide whitespace changes
Inline Side-by-side
Showing with 133 additions and 196 deletions
src/cmd/compile/internal/amd64/ssa.go
View file @ 68e86e6d
...@@ -660,7 +660,7 @@ func ssaGenValue(s *gc.SSAGenState, v *ssa.Value) { ...@@ -660,7 +660,7 @@ func ssaGenValue(s *gc.SSAGenState, v *ssa.Value) {
p.From.Val = math.Float64frombits(uint64(v.AuxInt)) p.From.Val = math.Float64frombits(uint64(v.AuxInt))
p.To.Type = obj.TYPE_REG p.To.Type = obj.TYPE_REG
p.To.Reg = x p.To.Reg = x
case ssa.OpAMD64MOVQload, ssa.OpAMD64MOVSSload, ssa.OpAMD64MOVSDload, ssa.OpAMD64MOVLload, ssa.OpAMD64MOVWload, ssa.OpAMD64MOVBload, ssa.OpAMD64MOVBQSXload, ssa.OpAMD64MOVBQZXload, ssa.OpAMD64MOVWQSXload, ssa.OpAMD64MOVWQZXload, ssa.OpAMD64MOVLQSXload, ssa.OpAMD64MOVLQZXload, ssa.OpAMD64MOVOload: case ssa.OpAMD64MOVQload, ssa.OpAMD64MOVSSload, ssa.OpAMD64MOVSDload, ssa.OpAMD64MOVLload, ssa.OpAMD64MOVWload, ssa.OpAMD64MOVBload, ssa.OpAMD64MOVBQSXload, ssa.OpAMD64MOVWQSXload, ssa.OpAMD64MOVLQSXload, ssa.OpAMD64MOVOload:
p := gc.Prog(v.Op.Asm()) p := gc.Prog(v.Op.Asm())
p.From.Type = obj.TYPE_MEM p.From.Type = obj.TYPE_MEM
p.From.Reg = gc.SSARegNum(v.Args[0]) p.From.Reg = gc.SSARegNum(v.Args[0])
...@@ -1044,8 +1044,7 @@ func ssaGenValue(s *gc.SSAGenState, v *ssa.Value) { ...@@ -1044,8 +1044,7 @@ func ssaGenValue(s *gc.SSAGenState, v *ssa.Value) {
switch w.Op { switch w.Op {
case ssa.OpAMD64MOVQload, ssa.OpAMD64MOVLload, ssa.OpAMD64MOVWload, ssa.OpAMD64MOVBload, case ssa.OpAMD64MOVQload, ssa.OpAMD64MOVLload, ssa.OpAMD64MOVWload, ssa.OpAMD64MOVBload,
ssa.OpAMD64MOVQstore, ssa.OpAMD64MOVLstore, ssa.OpAMD64MOVWstore, ssa.OpAMD64MOVBstore, ssa.OpAMD64MOVQstore, ssa.OpAMD64MOVLstore, ssa.OpAMD64MOVWstore, ssa.OpAMD64MOVBstore,
ssa.OpAMD64MOVBQSXload, ssa.OpAMD64MOVBQZXload, ssa.OpAMD64MOVWQSXload, ssa.OpAMD64MOVBQSXload, ssa.OpAMD64MOVWQSXload, ssa.OpAMD64MOVLQSXload,
ssa.OpAMD64MOVWQZXload, ssa.OpAMD64MOVLQSXload, ssa.OpAMD64MOVLQZXload,
ssa.OpAMD64MOVSSload, ssa.OpAMD64MOVSDload, ssa.OpAMD64MOVOload, ssa.OpAMD64MOVSSload, ssa.OpAMD64MOVSDload, ssa.OpAMD64MOVOload,
ssa.OpAMD64MOVSSstore, ssa.OpAMD64MOVSDstore, ssa.OpAMD64MOVOstore: ssa.OpAMD64MOVSSstore, ssa.OpAMD64MOVSDstore, ssa.OpAMD64MOVOstore:
if w.Args[0] == v.Args[0] && w.Aux == nil && w.AuxInt >= 0 && w.AuxInt < minZeroPage { if w.Args[0] == v.Args[0] && w.Aux == nil && w.AuxInt >= 0 && w.AuxInt < minZeroPage {
......
src/cmd/compile/internal/ssa/gen/AMD64.rules
View file @ 68e86e6d
...@@ -619,11 +619,15 @@ ...@@ -619,11 +619,15 @@
// This prevents a single load from being split into multiple loads // This prevents a single load from being split into multiple loads
// which then might return different values. See test/atomicload.go. // which then might return different values. See test/atomicload.go.
(MOVBQSX x:(MOVBload [off] {sym} ptr mem)) && x.Uses == 1 -> @x.Block (MOVBQSXload <v.Type> [off] {sym} ptr mem) (MOVBQSX x:(MOVBload [off] {sym} ptr mem)) && x.Uses == 1 -> @x.Block (MOVBQSXload <v.Type> [off] {sym} ptr mem)
(MOVBQZX x:(MOVBload [off] {sym} ptr mem)) && x.Uses == 1 -> @x.Block (MOVBQZXload <v.Type> [off] {sym} ptr mem) (MOVBQZX x:(MOVBload [off] {sym} ptr mem)) && x.Uses == 1 -> @x.Block (MOVBload <v.Type> [off] {sym} ptr mem)
(MOVWQSX x:(MOVWload [off] {sym} ptr mem)) && x.Uses == 1 -> @x.Block (MOVWQSXload <v.Type> [off] {sym} ptr mem) (MOVWQSX x:(MOVWload [off] {sym} ptr mem)) && x.Uses == 1 -> @x.Block (MOVWQSXload <v.Type> [off] {sym} ptr mem)
(MOVWQZX x:(MOVWload [off] {sym} ptr mem)) && x.Uses == 1 -> @x.Block (MOVWQZXload <v.Type> [off] {sym} ptr mem) (MOVWQZX x:(MOVWload [off] {sym} ptr mem)) && x.Uses == 1 -> @x.Block (MOVWload <v.Type> [off] {sym} ptr mem)
(MOVLQSX x:(MOVLload [off] {sym} ptr mem)) && x.Uses == 1 -> @x.Block (MOVLQSXload <v.Type> [off] {sym} ptr mem) (MOVLQSX x:(MOVLload [off] {sym} ptr mem)) && x.Uses == 1 -> @x.Block (MOVLQSXload <v.Type> [off] {sym} ptr mem)
(MOVLQZX x:(MOVLload [off] {sym} ptr mem)) && x.Uses == 1 -> @x.Block (MOVLQZXload <v.Type> [off] {sym} ptr mem) (MOVLQZX x:(MOVLload [off] {sym} ptr mem)) && x.Uses == 1 -> @x.Block (MOVLload <v.Type> [off] {sym} ptr mem)
(MOVBQZX x:(MOVBloadidx1 [off] {sym} ptr idx mem)) && x.Uses == 1 -> @x.Block (MOVBloadidx1 <v.Type> [off] {sym} ptr idx mem)
(MOVWQZX x:(MOVWloadidx2 [off] {sym} ptr idx mem)) && x.Uses == 1 -> @x.Block (MOVWloadidx2 <v.Type> [off] {sym} ptr idx mem)
(MOVLQZX x:(MOVLloadidx4 [off] {sym} ptr idx mem)) && x.Uses == 1 -> @x.Block (MOVLloadidx4 <v.Type> [off] {sym} ptr idx mem)
// replace load from same location as preceding store with copy // replace load from same location as preceding store with copy
(MOVBload [off] {sym} ptr (MOVBstore [off2] {sym2} ptr2 x _)) && sym == sym2 && off == off2 && isSamePtr(ptr, ptr2) -> x (MOVBload [off] {sym} ptr (MOVBstore [off2] {sym2} ptr2 x _)) && sym == sym2 && off == off2 && isSamePtr(ptr, ptr2) -> x
...@@ -705,12 +709,6 @@ ...@@ -705,12 +709,6 @@
(MOVOload [off1] {sym1} (LEAQ [off2] {sym2} base) mem) && is32Bit(off1+off2) && canMergeSym(sym1, sym2) -> (MOVOload [off1] {sym1} (LEAQ [off2] {sym2} base) mem) && is32Bit(off1+off2) && canMergeSym(sym1, sym2) ->
(MOVOload [off1+off2] {mergeSym(sym1,sym2)} base mem) (MOVOload [off1+off2] {mergeSym(sym1,sym2)} base mem)
(MOVBQZXload [off1] {sym1} (LEAQ [off2] {sym2} base) mem) && is32Bit(off1+off2) && canMergeSym(sym1, sym2) ->
(MOVBQZXload [off1+off2] {mergeSym(sym1,sym2)} base mem)
(MOVWQZXload [off1] {sym1} (LEAQ [off2] {sym2} base) mem) && is32Bit(off1+off2) && canMergeSym(sym1, sym2) ->
(MOVWQZXload [off1+off2] {mergeSym(sym1,sym2)} base mem)
(MOVLQZXload [off1] {sym1} (LEAQ [off2] {sym2} base) mem) && is32Bit(off1+off2) && canMergeSym(sym1, sym2) ->
(MOVLQZXload [off1+off2] {mergeSym(sym1,sym2)} base mem)
(MOVBQSXload [off1] {sym1} (LEAQ [off2] {sym2} base) mem) && is32Bit(off1+off2) && canMergeSym(sym1, sym2) -> (MOVBQSXload [off1] {sym1} (LEAQ [off2] {sym2} base) mem) && is32Bit(off1+off2) && canMergeSym(sym1, sym2) ->
(MOVBQSXload [off1+off2] {mergeSym(sym1,sym2)} base mem) (MOVBQSXload [off1+off2] {mergeSym(sym1,sym2)} base mem)
(MOVWQSXload [off1] {sym1} (LEAQ [off2] {sym2} base) mem) && is32Bit(off1+off2) && canMergeSym(sym1, sym2) -> (MOVWQSXload [off1] {sym1} (LEAQ [off2] {sym2} base) mem) && is32Bit(off1+off2) && canMergeSym(sym1, sym2) ->
...@@ -1197,21 +1195,21 @@ ...@@ -1197,21 +1195,21 @@
// Combining byte loads into larger (unaligned) loads. // Combining byte loads into larger (unaligned) loads.
// There are many ways these combinations could occur. This is // There are many ways these combinations could occur. This is
// designed to match the way encoding/binary.LittleEndian does it. // designed to match the way encoding/binary.LittleEndian does it.
(ORW x:(MOVBQZXload [i] {s} p mem) (ORW x:(MOVBload [i] {s} p mem)
(SHLWconst [8] (MOVBQZXload [i+1] {s} p mem))) -> @x.Block (MOVWload [i] {s} p mem) (SHLWconst [8] (MOVBload [i+1] {s} p mem))) -> @x.Block (MOVWload [i] {s} p mem)
(ORL (ORL (ORL (ORL (ORL (ORL
x:(MOVBQZXload [i] {s} p mem) x:(MOVBload [i] {s} p mem)
(SHLLconst [8] (MOVBQZXload [i+1] {s} p mem))) (SHLLconst [8] (MOVBload [i+1] {s} p mem)))
(SHLLconst [16] (MOVBQZXload [i+2] {s} p mem))) (SHLLconst [16] (MOVBload [i+2] {s} p mem)))
(SHLLconst [24] (MOVBQZXload [i+3] {s} p mem))) -> @x.Block (MOVLload [i] {s} p mem) (SHLLconst [24] (MOVBload [i+3] {s} p mem))) -> @x.Block (MOVLload [i] {s} p mem)
(ORQ (ORQ (ORQ (ORQ (ORQ (ORQ (ORQ (ORQ (ORQ (ORQ (ORQ (ORQ (ORQ (ORQ
x:(MOVBQZXload [i] {s} p mem) x:(MOVBload [i] {s} p mem)
(SHLQconst [8] (MOVBQZXload [i+1] {s} p mem))) (SHLQconst [8] (MOVBload [i+1] {s} p mem)))
(SHLQconst [16] (MOVBQZXload [i+2] {s} p mem))) (SHLQconst [16] (MOVBload [i+2] {s} p mem)))
(SHLQconst [24] (MOVBQZXload [i+3] {s} p mem))) (SHLQconst [24] (MOVBload [i+3] {s} p mem)))
(SHLQconst [32] (MOVBQZXload [i+4] {s} p mem))) (SHLQconst [32] (MOVBload [i+4] {s} p mem)))
(SHLQconst [40] (MOVBQZXload [i+5] {s} p mem))) (SHLQconst [40] (MOVBload [i+5] {s} p mem)))
(SHLQconst [48] (MOVBQZXload [i+6] {s} p mem))) (SHLQconst [48] (MOVBload [i+6] {s} p mem)))
(SHLQconst [56] (MOVBQZXload [i+7] {s} p mem))) -> @x.Block (MOVQload [i] {s} p mem) (SHLQconst [56] (MOVBload [i+7] {s} p mem))) -> @x.Block (MOVQload [i] {s} p mem)
src/cmd/compile/internal/ssa/gen/AMD64Ops.go
View file @ 68e86e6d
...@@ -367,15 +367,12 @@ func init() { ...@@ -367,15 +367,12 @@ func init() {
// Note: LEAQ{1,2,4,8} must not have OpSB as either argument. // Note: LEAQ{1,2,4,8} must not have OpSB as either argument.
// auxint+aux == add auxint and the offset of the symbol in aux (if any) to the effective address // auxint+aux == add auxint and the offset of the symbol in aux (if any) to the effective address
{name: "MOVBload", argLength: 2, reg: gpload, asm: "MOVBLZX", aux: "SymOff", typ: "UInt8"}, // load byte from arg0+auxint+aux. arg1=mem {name: "MOVBload", argLength: 2, reg: gpload, asm: "MOVBLZX", aux: "SymOff", typ: "UInt8"}, // load byte from arg0+auxint+aux. arg1=mem. Zero extend.
{name: "MOVBQSXload", argLength: 2, reg: gpload, asm: "MOVBQSX", aux: "SymOff"}, // ditto, extend to int64 {name: "MOVBQSXload", argLength: 2, reg: gpload, asm: "MOVBQSX", aux: "SymOff"}, // ditto, sign extend to int64
{name: "MOVBQZXload", argLength: 2, reg: gpload, asm: "MOVBLZX", aux: "SymOff"}, // ditto, extend to uint64 {name: "MOVWload", argLength: 2, reg: gpload, asm: "MOVWLZX", aux: "SymOff", typ: "UInt16"}, // load 2 bytes from arg0+auxint+aux. arg1=mem. Zero extend.
{name: "MOVWload", argLength: 2, reg: gpload, asm: "MOVWLZX", aux: "SymOff", typ: "UInt16"}, // load 2 bytes from arg0+auxint+aux. arg1=mem {name: "MOVWQSXload", argLength: 2, reg: gpload, asm: "MOVWQSX", aux: "SymOff"}, // ditto, sign extend to int64
{name: "MOVWQSXload", argLength: 2, reg: gpload, asm: "MOVWQSX", aux: "SymOff"}, // ditto, extend to int64 {name: "MOVLload", argLength: 2, reg: gpload, asm: "MOVL", aux: "SymOff", typ: "UInt32"}, // load 4 bytes from arg0+auxint+aux. arg1=mem. Zero extend.
{name: "MOVWQZXload", argLength: 2, reg: gpload, asm: "MOVWLZX", aux: "SymOff"}, // ditto, extend to uint64 {name: "MOVLQSXload", argLength: 2, reg: gpload, asm: "MOVLQSX", aux: "SymOff"}, // ditto, sign extend to int64
{name: "MOVLload", argLength: 2, reg: gpload, asm: "MOVL", aux: "SymOff", typ: "UInt32"}, // load 4 bytes from arg0+auxint+aux. arg1=mem
{name: "MOVLQSXload", argLength: 2, reg: gpload, asm: "MOVLQSX", aux: "SymOff"}, // ditto, extend to int64
{name: "MOVLQZXload", argLength: 2, reg: gpload, asm: "MOVL", aux: "SymOff"}, // ditto, extend to uint64
{name: "MOVQload", argLength: 2, reg: gpload, asm: "MOVQ", aux: "SymOff", typ: "UInt64"}, // load 8 bytes from arg0+auxint+aux. arg1=mem {name: "MOVQload", argLength: 2, reg: gpload, asm: "MOVQ", aux: "SymOff", typ: "UInt64"}, // load 8 bytes from arg0+auxint+aux. arg1=mem
{name: "MOVBstore", argLength: 3, reg: gpstore, asm: "MOVB", aux: "SymOff", typ: "Mem"}, // store byte in arg1 to arg0+auxint+aux. arg2=mem {name: "MOVBstore", argLength: 3, reg: gpstore, asm: "MOVB", aux: "SymOff", typ: "Mem"}, // store byte in arg1 to arg0+auxint+aux. arg2=mem
{name: "MOVWstore", argLength: 3, reg: gpstore, asm: "MOVW", aux: "SymOff", typ: "Mem"}, // store 2 bytes in arg1 to arg0+auxint+aux. arg2=mem {name: "MOVWstore", argLength: 3, reg: gpstore, asm: "MOVW", aux: "SymOff", typ: "Mem"}, // store 2 bytes in arg1 to arg0+auxint+aux. arg2=mem
......
src/cmd/compile/internal/ssa/opGen.go
View file @ 68e86e6d
...@@ -261,13 +261,10 @@ const ( ...@@ -261,13 +261,10 @@ const (
OpAMD64LEAQ8 OpAMD64LEAQ8
OpAMD64MOVBload OpAMD64MOVBload
OpAMD64MOVBQSXload OpAMD64MOVBQSXload
OpAMD64MOVBQZXload
OpAMD64MOVWload OpAMD64MOVWload
OpAMD64MOVWQSXload OpAMD64MOVWQSXload
OpAMD64MOVWQZXload
OpAMD64MOVLload OpAMD64MOVLload
OpAMD64MOVLQSXload OpAMD64MOVLQSXload
OpAMD64MOVLQZXload
OpAMD64MOVQload OpAMD64MOVQload
OpAMD64MOVBstore OpAMD64MOVBstore
OpAMD64MOVWstore OpAMD64MOVWstore
...@@ -3393,20 +3390,6 @@ var opcodeTable = [...]opInfo{ ...@@ -3393,20 +3390,6 @@ var opcodeTable = [...]opInfo{
}, },
}, },
}, },
{
name: "MOVBQZXload",
auxType: auxSymOff,
argLen: 2,
asm: x86.AMOVBLZX,
reg: regInfo{
inputs: []inputInfo{
{0, 4295032831}, // AX CX DX BX SP BP SI DI R8 R9 R10 R11 R12 R13 R14 R15 SB
},
outputs: []regMask{
65519, // AX CX DX BX BP SI DI R8 R9 R10 R11 R12 R13 R14 R15
},
},
},
{ {
name: "MOVWload", name: "MOVWload",
auxType: auxSymOff, auxType: auxSymOff,
...@@ -3435,20 +3418,6 @@ var opcodeTable = [...]opInfo{ ...@@ -3435,20 +3418,6 @@ var opcodeTable = [...]opInfo{
}, },
}, },
}, },
{
name: "MOVWQZXload",
auxType: auxSymOff,
argLen: 2,
asm: x86.AMOVWLZX,
reg: regInfo{
inputs: []inputInfo{
{0, 4295032831}, // AX CX DX BX SP BP SI DI R8 R9 R10 R11 R12 R13 R14 R15 SB
},
outputs: []regMask{
65519, // AX CX DX BX BP SI DI R8 R9 R10 R11 R12 R13 R14 R15
},
},
},
{ {
name: "MOVLload", name: "MOVLload",
auxType: auxSymOff, auxType: auxSymOff,
...@@ -3477,20 +3446,6 @@ var opcodeTable = [...]opInfo{ ...@@ -3477,20 +3446,6 @@ var opcodeTable = [...]opInfo{
}, },
}, },
}, },
{
name: "MOVLQZXload",
auxType: auxSymOff,
argLen: 2,
asm: x86.AMOVL,
reg: regInfo{
inputs: []inputInfo{
{0, 4295032831}, // AX CX DX BX SP BP SI DI R8 R9 R10 R11 R12 R13 R14 R15 SB
},
outputs: []regMask{
65519, // AX CX DX BX BP SI DI R8 R9 R10 R11 R12 R13 R14 R15
},
},
},
{ {
name: "MOVQload", name: "MOVQload",
auxType: auxSymOff, auxType: auxSymOff,
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7