qbit/go
1
0
Fork 0
mirror of https://github.com/golang/go synced 2024-11-17 11:04:53 -07:00
Code Releases Activity

cmd/compile: fold LEAQ with constant scale into LEA

Browse Source 
Discovered this after rebasing CL196679 (use poset bounds in prove).
Some tests fail with that CL applied:

codegen/smallintiface.go:11: linux/amd64/: opcode not found: "^LEAQ\\truntime.staticuint64s\\+8\\(SB\\)"
codegen/smallintiface.go:16: linux/amd64/: opcode not found: "^LEAQ\\truntime.staticuint64s\\+2024\\(SB\\)"
codegen/smallintiface.go:21: linux/amd64/: opcode not found: "^LEAQ\\truntime.staticuint64s\\+24\\(SB\\)"

The only difference in prove SSA dumps is that a single Lsh64x64
op with constant shift (<< 3) is marked as bounded. This triggers
a different rule matching sequence in lower, which manages to generate
worse code for the above testcases.

This CL fixes the above test after CL196679 is applied. Right now,
these rules never trigger (this CL passes toolstash -cmp), so I can't
write a test.

Change-Id: I353f1c79c1875cac1da82cd8afa1e05e42684f1c
Reviewed-on: https://go-review.googlesource.com/c/go/+/224877
Run-TryBot: Giovanni Bajo <rasky@develer.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
This commit is contained in:
Giovanni Bajo 2020-03-23 09:21:44 +01:00
parent 5d47f870a6
commit 24925c7ed9
2 changed files with 128 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
src/cmd/compile/internal/ssa/gen/AMD64.rules
View File

@ -1202,6 +1202,14 @@
(LEAQ8 [off1+4*off2] {sym1} x y)
// TODO: more?
// Lower LEAQ2/4/8 when the offset is a constant
(LEAQ2 [off] {sym} x (MOV(Q|L)const [scale])) && is32Bit(off+scale*2) ->
(LEAQ [off+scale*2] {sym} x)
(LEAQ4 [off] {sym} x (MOV(Q|L)const [scale])) && is32Bit(off+scale*4) ->
(LEAQ [off+scale*4] {sym} x)
(LEAQ8 [off] {sym} x (MOV(Q|L)const [scale])) && is32Bit(off+scale*8) ->
(LEAQ [off+scale*8] {sym} x)
// Absorb InvertFlags into branches.
(LT (InvertFlags cmp) yes no) -> (GT cmp yes no)
(GT (InvertFlags cmp) yes no) -> (LT cmp yes no)

120
src/cmd/compile/internal/ssa/rewriteAMD64.go
View File

@ -9486,6 +9486,46 @@ func rewriteValueAMD64_OpAMD64LEAQ2(v *Value) bool {
v.AddArg2(x, y)
return true
}
// match: (LEAQ2 [off] {sym} x (MOVQconst [scale]))
// cond: is32Bit(off+scale*2)
// result: (LEAQ [off+scale*2] {sym} x)
for {
off := v.AuxInt
sym := v.Aux
x := v_0
if v_1.Op != OpAMD64MOVQconst {
break
}
scale := v_1.AuxInt
if !(is32Bit(off + scale*2)) {
break
}
v.reset(OpAMD64LEAQ)
v.AuxInt = off + scale*2
v.Aux = sym
v.AddArg(x)
return true
}
// match: (LEAQ2 [off] {sym} x (MOVLconst [scale]))
// cond: is32Bit(off+scale*2)
// result: (LEAQ [off+scale*2] {sym} x)
for {
off := v.AuxInt
sym := v.Aux
x := v_0
if v_1.Op != OpAMD64MOVLconst {
break
}
scale := v_1.AuxInt
if !(is32Bit(off + scale*2)) {
break
}
v.reset(OpAMD64LEAQ)
v.AuxInt = off + scale*2
v.Aux = sym
v.AddArg(x)
return true
}
return false
}
func rewriteValueAMD64_OpAMD64LEAQ4(v *Value) bool {
@ -9593,6 +9633,46 @@ func rewriteValueAMD64_OpAMD64LEAQ4(v *Value) bool {
v.AddArg2(x, y)
return true
}
// match: (LEAQ4 [off] {sym} x (MOVQconst [scale]))
// cond: is32Bit(off+scale*4)
// result: (LEAQ [off+scale*4] {sym} x)
for {
off := v.AuxInt
sym := v.Aux
x := v_0
if v_1.Op != OpAMD64MOVQconst {
break
}
scale := v_1.AuxInt
if !(is32Bit(off + scale*4)) {
break
}
v.reset(OpAMD64LEAQ)
v.AuxInt = off + scale*4
v.Aux = sym
v.AddArg(x)
return true
}
// match: (LEAQ4 [off] {sym} x (MOVLconst [scale]))
// cond: is32Bit(off+scale*4)
// result: (LEAQ [off+scale*4] {sym} x)
for {
off := v.AuxInt
sym := v.Aux
x := v_0
if v_1.Op != OpAMD64MOVLconst {
break
}
scale := v_1.AuxInt
if !(is32Bit(off + scale*4)) {
break
}
v.reset(OpAMD64LEAQ)
v.AuxInt = off + scale*4
v.Aux = sym
v.AddArg(x)
return true
}
return false
}
func rewriteValueAMD64_OpAMD64LEAQ8(v *Value) bool {
@ -9662,6 +9742,46 @@ func rewriteValueAMD64_OpAMD64LEAQ8(v *Value) bool {
v.AddArg2(x, y)
return true
}
// match: (LEAQ8 [off] {sym} x (MOVQconst [scale]))
// cond: is32Bit(off+scale*8)
// result: (LEAQ [off+scale*8] {sym} x)
for {
off := v.AuxInt
sym := v.Aux
x := v_0
if v_1.Op != OpAMD64MOVQconst {
break
}
scale := v_1.AuxInt
if !(is32Bit(off + scale*8)) {
break
}
v.reset(OpAMD64LEAQ)
v.AuxInt = off + scale*8
v.Aux = sym
v.AddArg(x)
return true
}
// match: (LEAQ8 [off] {sym} x (MOVLconst [scale]))
// cond: is32Bit(off+scale*8)
// result: (LEAQ [off+scale*8] {sym} x)
for {
off := v.AuxInt
sym := v.Aux
x := v_0
if v_1.Op != OpAMD64MOVLconst {
break
}
scale := v_1.AuxInt
if !(is32Bit(off + scale*8)) {
break
}
v.reset(OpAMD64LEAQ)
v.AuxInt = off + scale*8
v.Aux = sym
v.AddArg(x)
return true
}
return false
}
func rewriteValueAMD64_OpAMD64MOVBQSX(v *Value) bool {