From da34ddf24b4ed9b495143f58c6c5ded918ce4d16 Mon Sep 17 00:00:00 2001
From: Ilya Tocar <ilya.tocar@intel.com>
Date: Wed, 16 Aug 2017 14:01:48 -0500
Subject: [PATCH] cmd/compile/internal/ssa: combine more const stores

We already combine const stores up-to MOVQstoreconst.
Combine 2 64-bit stores of const zero into 1 sse store of 128-bit zero.

Shaves significant (>1%) amount of code from go tool:
/localdisk/itocar/golang/bin/go 10334877
go_old 10388125 [53248 bytes]

global text (code) = 51041 bytes (1.343944%)
read-only data = 663 bytes (0.039617%)
Total difference 51704 bytes (0.873981%)

Change-Id: I7bc40968023c3a69f379b10fbb433cdb11364f1b
Reviewed-on: https://go-review.googlesource.com/56250
Run-TryBot: Ilya Tocar <ilya.tocar@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Giovanni Bajo <rasky@develer.com>
Reviewed-by: Keith Randall <khr@golang.org>
---
 src/cmd/compile/internal/gc/asm_test.go      |  4 +--
 src/cmd/compile/internal/ssa/gen/AMD64.rules |  7 ++++
 src/cmd/compile/internal/ssa/rewriteAMD64.go | 36 ++++++++++++++++++++
 3 files changed, 45 insertions(+), 2 deletions(-)

diff --git a/src/cmd/compile/internal/gc/asm_test.go b/src/cmd/compile/internal/gc/asm_test.go
index 23b70ae41d..d3e56230bb 100644
--- a/src/cmd/compile/internal/gc/asm_test.go
+++ b/src/cmd/compile/internal/gc/asm_test.go
@@ -362,7 +362,7 @@ var linuxAMD64Tests = []*asmTest{
 			*t = T1{}
 		}
 		`,
-		[]string{"\tMOVQ\t\\$0, \\(.*\\)", "\tMOVQ\t\\$0, 8\\(.*\\)", "\tMOVQ\t\\$0, 16\\(.*\\)"},
+		[]string{"\tXORPS\tX., X", "\tMOVUPS\tX., \\(.*\\)", "\tMOVQ\t\\$0, 16\\(.*\\)"},
 	},
 	// SSA-able composite literal initialization. Issue 18872.
 	{
@@ -387,7 +387,7 @@ var linuxAMD64Tests = []*asmTest{
 			*t = T2{}
 		}
 		`,
-		[]string{"\tMOVQ\t\\$0, \\(.*\\)", "\tMOVQ\t\\$0, 8\\(.*\\)", "\tMOVQ\t\\$0, 16\\(.*\\)", "\tCALL\truntime\\.writebarrierptr\\(SB\\)"},
+		[]string{"\tXORPS\tX., X", "\tMOVUPS\tX., \\(.*\\)", "\tMOVQ\t\\$0, 16\\(.*\\)", "\tCALL\truntime\\.writebarrierptr\\(SB\\)"},
 	},
 	// Rotate tests
 	{
diff --git a/src/cmd/compile/internal/ssa/gen/AMD64.rules b/src/cmd/compile/internal/ssa/gen/AMD64.rules
index 2f38a7d5cc..c31c7ced02 100644
--- a/src/cmd/compile/internal/ssa/gen/AMD64.rules
+++ b/src/cmd/compile/internal/ssa/gen/AMD64.rules
@@ -2204,6 +2204,13 @@
   && ValAndOff(a).Off() + 4 == ValAndOff(c).Off()
   && clobber(x)
   -> (MOVQstore [ValAndOff(a).Off()] {s} p (MOVQconst [ValAndOff(a).Val()&0xffffffff | ValAndOff(c).Val()<<32]) mem)
+(MOVQstoreconst [c] {s} p x:(MOVQstoreconst [c2] {s} p mem))
+  && x.Uses == 1
+  && ValAndOff(c2).Off() + 8 == ValAndOff(c).Off()
+  && ValAndOff(c).Val() == 0
+  && ValAndOff(c2).Val() == 0
+  && clobber(x)
+  -> (MOVOstore [ValAndOff(c2).Off()] {s} p (MOVOconst [0]) mem)
 
 (MOVBstoreconstidx1 [c] {s} p i x:(MOVBstoreconstidx1 [a] {s} p i mem))
   && x.Uses == 1
diff --git a/src/cmd/compile/internal/ssa/rewriteAMD64.go b/src/cmd/compile/internal/ssa/rewriteAMD64.go
index 437ca36064..4ce6415ba5 100644
--- a/src/cmd/compile/internal/ssa/rewriteAMD64.go
+++ b/src/cmd/compile/internal/ssa/rewriteAMD64.go
@@ -9009,6 +9009,8 @@ func rewriteValueAMD64_OpAMD64MOVQstore_0(v *Value) bool {
 	return false
 }
 func rewriteValueAMD64_OpAMD64MOVQstoreconst_0(v *Value) bool {
+	b := v.Block
+	_ = b
 	// match: (MOVQstoreconst [sc] {s} (ADDQconst [off] ptr) mem)
 	// cond: ValAndOff(sc).canAdd(off)
 	// result: (MOVQstoreconst [ValAndOff(sc).add(off)] {s} ptr mem)
@@ -9137,6 +9139,40 @@ func rewriteValueAMD64_OpAMD64MOVQstoreconst_0(v *Value) bool {
 		v.AddArg(mem)
 		return true
 	}
+	// match: (MOVQstoreconst [c] {s} p x:(MOVQstoreconst [c2] {s} p mem))
+	// cond: x.Uses == 1   && ValAndOff(c2).Off() + 8 == ValAndOff(c).Off()   && ValAndOff(c).Val() == 0   && ValAndOff(c2).Val() == 0   && clobber(x)
+	// result: (MOVOstore [ValAndOff(c2).Off()] {s} p (MOVOconst [0]) mem)
+	for {
+		c := v.AuxInt
+		s := v.Aux
+		_ = v.Args[1]
+		p := v.Args[0]
+		x := v.Args[1]
+		if x.Op != OpAMD64MOVQstoreconst {
+			break
+		}
+		c2 := x.AuxInt
+		if x.Aux != s {
+			break
+		}
+		_ = x.Args[1]
+		if p != x.Args[0] {
+			break
+		}
+		mem := x.Args[1]
+		if !(x.Uses == 1 && ValAndOff(c2).Off()+8 == ValAndOff(c).Off() && ValAndOff(c).Val() == 0 && ValAndOff(c2).Val() == 0 && clobber(x)) {
+			break
+		}
+		v.reset(OpAMD64MOVOstore)
+		v.AuxInt = ValAndOff(c2).Off()
+		v.Aux = s
+		v.AddArg(p)
+		v0 := b.NewValue0(v.Pos, OpAMD64MOVOconst, types.TypeInt128)
+		v0.AuxInt = 0
+		v.AddArg(v0)
+		v.AddArg(mem)
+		return true
+	}
 	// match: (MOVQstoreconst [sc] {sym1} (LEAL [off] {sym2} ptr) mem)
 	// cond: canMergeSym(sym1, sym2) && ValAndOff(sc).canAdd(off)
 	// result: (MOVQstoreconst [ValAndOff(sc).add(off)] {mergeSym(sym1, sym2)} ptr mem)
-- 
2.50.0