From 23f6479be645761c5ba7ba92cf062bbc6470065d Mon Sep 17 00:00:00 2001
From: Russ Cox <rsc@golang.org>
Date: Fri, 8 Apr 2011 13:53:59 -0400
Subject: [PATCH] 8g: optimize byte mov

Rewrite MOVB with less expensive
instruction when possible.

Suggested by atomic symbol.

benchmark                                        old ns/op    new ns/op    delta
crc32.BenchmarkCrc32KB                               13066         3942  -69.83%
crc64.BenchmarkCrc64KB                                8780         5949  -32.24%
lzw.BenchmarkDecoder1e4                             771224       636538  -17.46%
lzw.BenchmarkDecoder1e5                            7101218      6096634  -14.15%
lzw.BenchmarkDecoder1e6                           69762020     60789400  -12.86%
lzw.BenchmarkEncoder1e4                             707968       638812   -9.77%
lzw.BenchmarkEncoder1e5                            6567122      5965552   -9.16%
lzw.BenchmarkEncoder1e6                           65006000     58911680   -9.38%
utf8_test.BenchmarkRuneCountTenASCIIChars              166          165   -0.60%
utf8_test.BenchmarkRuneCountTenJapaneseChars           246          258   +4.88%
utf8_test.BenchmarkEncodeASCIIRune                      13           10  -23.08%
utf8_test.BenchmarkEncodeJapaneseRune                   37           16  -56.76%
utf8_test.BenchmarkDecodeASCIIRune                      23           21   -8.70%
utf8_test.BenchmarkDecodeJapaneseRune                   58           32  -44.83%

R=ken2
CC=golang-dev
https://golang.org/cl/4381045
---
 src/cmd/8g/peep.c | 29 +++++++++++++++++++++++++++--
 1 file changed, 27 insertions(+), 2 deletions(-)

diff --git a/src/cmd/8g/peep.c b/src/cmd/8g/peep.c
index 580b1a9223..5ad29e1b21 100644
--- a/src/cmd/8g/peep.c
+++ b/src/cmd/8g/peep.c
@@ -120,6 +120,25 @@ peep(void)
 			p = p->link;
 		}
 	}
+	
+	// movb elimination.
+	// movb is simulated by the linker
+	// when a register other than ax, bx, cx, dx
+	// is used, so rewrite to other instructions
+	// when possible.  a movb into a register
+	// can smash the entire 32-bit register without
+	// causing any trouble.
+	for(r=firstr; r!=R; r=r->link) {
+		p = r->prog;
+		if(p->as == AMOVB && regtyp(&p->to)) {
+			// movb into register.
+			// from another register or constant can be movl.
+			if(regtyp(&p->from) || p->from.type == D_CONST)
+				p->as = AMOVL;
+			else
+				p->as = AMOVBLZX;
+		}
+	}
 
 	// constant propagation
 	// find MOV $con,R followed by
@@ -152,6 +171,8 @@ loop1:
 	for(r=firstr; r!=R; r=r->link) {
 		p = r->prog;
 		switch(p->as) {
+		case AMOVB:
+		case AMOVW:
 		case AMOVL:
 			if(regtyp(&p->to))
 			if(regtyp(&p->from)) {
@@ -182,6 +203,7 @@ loop1:
 			}
 			break;
 
+		case AADDB:
 		case AADDL:
 		case AADDW:
 			if(p->from.type != D_CONST || needc(p->link))
@@ -204,6 +226,7 @@ loop1:
 			}
 			break;
 
+		case ASUBB:
 		case ASUBL:
 		case ASUBW:
 			if(p->from.type != D_CONST || needc(p->link))
@@ -380,6 +403,8 @@ subprop(Reg *r0)
 		case AMOVSL:
 			return 0;
 
+		case AMOVB:
+		case AMOVW:
 		case AMOVL:
 			if(p->to.type == v1->type)
 				goto gotit;
@@ -560,6 +585,8 @@ copyu(Prog *p, Adr *v, Adr *s)
 
 
 	case ANOP:	/* rhs store */
+	case AMOVB:
+	case AMOVW:
 	case AMOVL:
 	case AMOVBLSX:
 	case AMOVBLZX:
@@ -624,8 +651,6 @@ copyu(Prog *p, Adr *v, Adr *s)
 	case AXORB:
 	case AXORL:
 	case AXORW:
-	case AMOVB:
-	case AMOVW:
 		if(copyas(&p->to, v))
 			return 2;
 		goto caseread;