From 294664b726db861678b66850db4274b55a782a42 Mon Sep 17 00:00:00 2001
From: gbeauche <>
Date: Mon, 6 Feb 2006 23:06:54 +0000
Subject: [PATCH] Fix SAHF_SETO_PROFITABLE code for x86-64 platforms.

This was only an experiment. Improvement was marginal: only +3% on AMD64
(an Athlon 64 3200+). However, it may be interesting to test it on EM64T
(e.g. newer P4s) since an older P3/800, hence in 32-bit mode, got a +15%
improvement in Speedometer 4 benchmarks.

Rationale: lahf/seto sequences avoid load/stores to the stack (push/pop)
and it was thus hoped to be faster.

Anyhow, SAHF_SETO_PROFITABLE can only be enabled manually at this time.
Edit your generated Makefile for testing, but first make sure your CPU
supports lahf in 64-bit mode (lahf_lm flag in /proc/cpuinfo).
---
 BasiliskII/src/uae_cpu/compiler/codegen_x86.cpp | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

diff --git a/BasiliskII/src/uae_cpu/compiler/codegen_x86.cpp b/BasiliskII/src/uae_cpu/compiler/codegen_x86.cpp
index a7a57986..c474e47e 100644
--- a/BasiliskII/src/uae_cpu/compiler/codegen_x86.cpp
+++ b/BasiliskII/src/uae_cpu/compiler/codegen_x86.cpp
@@ -52,6 +52,11 @@
 #define R14_INDEX 14
 #define R15_INDEX 15
 #endif
+/* XXX this has to match X86_Reg8H_Base + 4 */
+#define AH_INDEX (0x10+4+EAX_INDEX)
+#define CH_INDEX (0x10+4+ECX_INDEX)
+#define DH_INDEX (0x10+4+EDX_INDEX)
+#define BH_INDEX (0x10+4+EBX_INDEX)
 
 /* The register in which subroutines return an integer return value */
 #define REG_RESULT EAX_INDEX
@@ -2414,7 +2419,7 @@ LENDFUNC(NONE,READ,2,raw_mov_w_rm,(W2 d, IMM s))
 LOWFUNC(NONE,WRITE,2,raw_mov_b_mr,(IMM d, R1 s))
 {
     emit_byte(0x88);
-    emit_byte(0x05+8*s);
+    emit_byte(0x05+8*(s&0xf)); /* XXX this handles %ah case (defined as 0x10+4) and others */
     emit_long(d);
 }
 LENDFUNC(NONE,WRITE,2,raw_mov_b_mr,(IMM d, R1 s))
@@ -3221,7 +3226,6 @@ static __inline__ void raw_emit_nop_filler(int nbytes)
 #ifdef SAHF_SETO_PROFITABLE
 
 #define FLAG_NREG1 0  /* Set to -1 if any register will do */
-
 static __inline__ void raw_flags_to_reg(int r)
 {
   raw_lahf(0);  /* Most flags in AH */
@@ -3230,7 +3234,7 @@ static __inline__ void raw_flags_to_reg(int r)
   
 #if 1   /* Let's avoid those nasty partial register stalls */
   //raw_mov_b_mr((uintptr)live.state[FLAGTMP].mem,r);
-  raw_mov_b_mr(((uintptr)live.state[FLAGTMP].mem)+1,r+4);
+  raw_mov_b_mr(((uintptr)live.state[FLAGTMP].mem)+1,AH_INDEX);
   //live.state[FLAGTMP].status=CLEAN;
   live.state[FLAGTMP].status=INMEM;
   live.state[FLAGTMP].realreg=-1;