[LA64_DYNAREC] Micro-optimization for emit_shr32c (#3317)

ksco · web-flow · commit 816f807a2dc8 · 2025-12-29T11:09:04.000+01:00
diff --git a/src/dynarec/la64/dynarec_la64_emit_shift.c b/src/dynarec/la64/dynarec_la64_emit_shift.c
@@ -760,8 +760,7 @@ void emit_shr32c(dynarec_la64_t* dyn, int ninst, rex_t rex, int s1, uint32_t c,
             BSTRINS_D(s3, xZR, F_AF, F_AF);
             X64_SET_EFLAGS(s3, X_OF | X_AF);
         }
-        SRLIxw(s1, s1, c);
-        if (!rex.w) ZEROUP(s1);
+        SRLIxw(s1, s1, c); // no need to zeroup on !rex.w
         IFX (X_PEND) {
             SDxw(s1, xEmu, offsetof(x64emu_t, res));
         }
@@ -789,16 +788,13 @@ void emit_shr32c(dynarec_la64_t* dyn, int ninst, rex_t rex, int s1, uint32_t c,
     if (rex.w) {
         SRLI_D(s1, s1, c);
     } else {
-        SRLI_W(s1, s1, c);
+        SRLI_W(s1, s1, c); // no need to zeroup
     }
 
     IFX (X_SF) {
         BGE(s1, xZR, 8);
         ORI(xFlags, xFlags, 1 << F_SF);
     }
-    if (!rex.w) {
-        ZEROUP(s1);
-    }
     IFX (X_PEND) {
         SDxw(s1, xEmu, offsetof(x64emu_t, res));
     }

Original file line number	Diff line number	Diff line change
`@@ -760,8 +760,7 @@ void emit_shr32c(dynarec_la64_t* dyn, int ninst, rex_t rex, int s1, uint32_t c,`
`760`	`760`	`BSTRINS_D(s3, xZR, F_AF, F_AF);`
`761`	`761`	`X64_SET_EFLAGS(s3, X_OF \| X_AF);`
`762`	`762`	`}`
`763`		`- SRLIxw(s1, s1, c);`
`764`		`- if (!rex.w) ZEROUP(s1);`
	`763`	`+ SRLIxw(s1, s1, c); // no need to zeroup on !rex.w`
`765`	`764`	`IFX (X_PEND) {`
`766`	`765`	`SDxw(s1, xEmu, offsetof(x64emu_t, res));`
`767`	`766`	`}`
`@@ -789,16 +788,13 @@ void emit_shr32c(dynarec_la64_t* dyn, int ninst, rex_t rex, int s1, uint32_t c,`
`789`	`788`	`if (rex.w) {`
`790`	`789`	`SRLI_D(s1, s1, c);`
`791`	`790`	`} else {`
`792`		`- SRLI_W(s1, s1, c);`
	`791`	`+ SRLI_W(s1, s1, c); // no need to zeroup`
`793`	`792`	`}`
`794`	`793`
`795`	`794`	`IFX (X_SF) {`
`796`	`795`	`BGE(s1, xZR, 8);`
`797`	`796`	`ORI(xFlags, xFlags, 1 << F_SF);`
`798`	`797`	`}`
`799`		`- if (!rex.w) {`
`800`		`- ZEROUP(s1);`
`801`		`- }`
`802`	`798`	`IFX (X_PEND) {`
`803`	`799`	`SDxw(s1, xEmu, offsetof(x64emu_t, res));`
`804`	`800`	`}`