nkeynes@539: /**
nkeynes@586:  * $Id$
nkeynes@539:  * 
nkeynes@926:  * Provides the implementation for the ia32 ABI variant 
nkeynes@926:  * (eg prologue, epilogue, and calling conventions). Stack frame is
nkeynes@926:  * aligned on 16-byte boundaries for the benefit of OS X (which 
nkeynes@926:  * requires it).
nkeynes@995:  * 
nkeynes@995:  * Note: These should only be included from x86op.h
nkeynes@539:  *
nkeynes@539:  * Copyright (c) 2007 Nathan Keynes.
nkeynes@539:  *
nkeynes@539:  * This program is free software; you can redistribute it and/or modify
nkeynes@539:  * it under the terms of the GNU General Public License as published by
nkeynes@539:  * the Free Software Foundation; either version 2 of the License, or
nkeynes@539:  * (at your option) any later version.
nkeynes@539:  *
nkeynes@539:  * This program is distributed in the hope that it will be useful,
nkeynes@539:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
nkeynes@539:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
nkeynes@539:  * GNU General Public License for more details.
nkeynes@539:  */
nkeynes@539: 
nkeynes@995: #define REG_ARG1 REG_EAX
nkeynes@995: #define REG_ARG2 REG_EDX
nkeynes@995: #define REG_RESULT1 REG_EAX
nkeynes@995: #define MAX_REG_ARG 2
nkeynes@539: 
nkeynes@1004: static inline void decode_address( uintptr_t base, int addr_reg )
nkeynes@953: {
nkeynes@995:     MOVL_r32_r32( addr_reg, REG_ECX );
nkeynes@991:     SHRL_imm_r32( 12, REG_ECX ); 
nkeynes@991:     MOVP_sib_rptr( 2, REG_ECX, -1, base, REG_ECX );
nkeynes@953: }
nkeynes@953: 
nkeynes@539: /**
nkeynes@995:  * Note: clobbers ECX to make the indirect call - this isn't usually
nkeynes@995:  * a problem since the callee will generally clobber it anyway.
nkeynes@539:  */
nkeynes@995: static inline void CALL_ptr( void *ptr )
nkeynes@539: {
nkeynes@995:     MOVP_immptr_rptr( (uintptr_t)ptr, REG_ECX );
nkeynes@991:     CALL_r32(REG_ECX);
nkeynes@539: }
nkeynes@539: 
nkeynes@905: #ifdef HAVE_FASTCALL
nkeynes@1146: #define CALL1_PTR_MIN_SIZE 7
nkeynes@1146: 
nkeynes@995: static inline void CALL1_ptr_r32( void *ptr, int arg1 )
nkeynes@905: {
nkeynes@995:     if( arg1 != REG_ARG1 ) {
nkeynes@995:         MOVL_r32_r32( arg1, REG_ARG1 );
nkeynes@905:     }
nkeynes@995:     CALL_ptr(ptr);
nkeynes@953: }
nkeynes@953: 
nkeynes@995: static inline void CALL1_r32disp_r32( int preg, uint32_t disp, int arg1 )
nkeynes@953: {
nkeynes@995:     if( arg1 != REG_ARG1 ) {
nkeynes@995:         MOVL_r32_r32( arg1, REG_ARG1 );
nkeynes@953:     }
nkeynes@995:     CALL_r32disp(preg, disp);
nkeynes@953: }
nkeynes@953: 
nkeynes@995: static inline void CALL2_ptr_r32_r32( void *ptr, int arg1, int arg2 )
nkeynes@953: {
nkeynes@995:     if( arg2 != REG_ARG2 ) {
nkeynes@995:         MOVL_r32_r32( arg2, REG_ARG2 );
nkeynes@953:     }
nkeynes@995:     if( arg1 != REG_ARG1 ) {
nkeynes@995:         MOVL_r32_r32( arg1, REG_ARG1 );
nkeynes@995:     }
nkeynes@995:     CALL_ptr(ptr);
nkeynes@953: }
nkeynes@953: 
nkeynes@995: static inline void CALL2_r32disp_r32_r32( int preg, uint32_t disp, int arg1, int arg2 )
nkeynes@953: {
nkeynes@995:     if( arg2 != REG_ARG2 ) {
nkeynes@995:         MOVL_r32_r32( arg2, REG_ARG2 );
nkeynes@953:     }
nkeynes@995:     if( arg1 != REG_ARG1 ) {
nkeynes@995:         MOVL_r32_r32( arg1, REG_ARG1 );
nkeynes@995:     }
nkeynes@995:     CALL_r32disp(preg, disp);
nkeynes@905: }
nkeynes@905: 
nkeynes@995: #define CALL3_r32disp_r32_r32_r32(preg,disp,arg1,arg2,arg3) CALL2_r32disp_r32_r32(preg,disp,arg1,arg2)
nkeynes@927: 
nkeynes@905: #else
nkeynes@1146: 
nkeynes@1146: #define CALL1_PTR_MIN_SIZE (3+1+7+3)
nkeynes@1146: 
nkeynes@1146: static inline void CALL1_ptr_r32( void *ptr, int arg1 )
nkeynes@539: {
nkeynes@991:     SUBL_imms_r32( 12, REG_ESP );
nkeynes@539:     PUSH_r32(arg1);
nkeynes@995:     CALL_ptr(ptr);
nkeynes@991:     ADDL_imms_r32( 16, REG_ESP );
nkeynes@539: }
nkeynes@539: 
nkeynes@995: static inline void CALL1_r32disp_r32( int preg, uint32_t disp, int arg1 )
nkeynes@995: {
nkeynes@995:     SUBL_imms_r32( 12, REG_ESP );
nkeynes@995:     PUSH_r32(arg1);
nkeynes@995:     CALL_r32disp(preg, disp);
nkeynes@995:     ADDL_imms_r32( 16, REG_ESP );
nkeynes@995: }
nkeynes@995: 
nkeynes@995: static inline void CALL2_ptr_r32_r32( void *ptr, int arg1, int arg2 )
nkeynes@539: {
nkeynes@991:     SUBL_imms_r32( 8, REG_ESP );
nkeynes@539:     PUSH_r32(arg2);
nkeynes@539:     PUSH_r32(arg1);
nkeynes@995:     CALL_ptr(ptr);
nkeynes@995:     ADDL_imms_r32( 16, REG_ESP );
nkeynes@995: }
nkeynes@995: 
nkeynes@995: static inline void CALL2_r32disp_r32_r32( int preg, uint32_t disp, int arg1, int arg2 )
nkeynes@995: {
nkeynes@995:     SUBL_imms_r32( 8, REG_ESP );
nkeynes@995:     PUSH_r32(arg2);
nkeynes@995:     PUSH_r32(arg1);
nkeynes@995:     CALL_r32disp(preg, disp);
nkeynes@995:     ADDL_imms_r32( 16, REG_ESP );
nkeynes@995: }
nkeynes@995: 
nkeynes@995: static inline void CALL3_r32disp_r32_r32_r32( int preg, uint32_t disp, int arg1, int arg2, int arg3 )
nkeynes@995: {
nkeynes@995:     SUBL_imms_r32( 8, REG_ESP );
nkeynes@995:     PUSH_r32(arg2);
nkeynes@995:     PUSH_r32(arg1);
nkeynes@995:     MOVL_rspdisp_r32( 16, REG_EAX );
nkeynes@1146:     MOVL_r32_rspdisp( REG_EAX, 8 );
nkeynes@995:     CALL_r32disp(preg,disp);
nkeynes@991:     ADDL_imms_r32( 16, REG_ESP );
nkeynes@539: }
nkeynes@926: 
nkeynes@905: #endif
nkeynes@539: 
nkeynes@1112: #define PROLOGUE_SIZE 9
nkeynes@1112: 
nkeynes@539: /**
nkeynes@539:  * Emit the 'start of block' assembly. Sets up the stack frame and save
nkeynes@539:  * SI/DI as required
nkeynes@926:  * Allocates 8 bytes for local variables, which also has the convenient
nkeynes@926:  * side-effect of aligning the stack.
nkeynes@539:  */
nkeynes@1125: static inline void emit_prologue( )
nkeynes@539: {
nkeynes@991:     PUSH_r32(REG_EBP);
nkeynes@991:     SUBL_imms_r32( 8, REG_ESP ); 
nkeynes@1112:     MOVP_immptr_rptr( ((uint8_t *)&sh4r) + 128, REG_EBP );
nkeynes@926: }
nkeynes@736: 
nkeynes@1125: static inline void emit_epilogue( )
nkeynes@926: {
nkeynes@991:     ADDL_imms_r32( 8, REG_ESP );
nkeynes@991:     POP_r32(REG_EBP);
nkeynes@539: }
nkeynes@1125: