piwalker/dolphin
1
0
Fork 0
mirror of https://github.com/dolphin-emu/dolphin.git synced 2025-07-24 14:49:42 -06:00
Code Issues Packages Projects Releases Wiki Activity

Core/DSPCore: Reorganize register layout for accessing accumulators

Browse Source 
(acc and ax) and product register with one read/write.

Gives a minuscule speedup of not more than 4%. In exchange, breaks all
your out-of-tree changes to dsp. Tests are not building again, yet.


git-svn-id: https://dolphin-emu.googlecode.com/svn/trunk@6680 8ced0084-cf51-0410-be5f-012b33b47a6e
This commit is contained in:
pierre
2010-12-29 02:12:06 +00:00
parent 0059b0f44f
commit fc1db5eaa0
23 changed files with 1224 additions and 728 deletions
Download Patch File Download Diff File Expand all files Collapse all files

84
Source/Core/DSPCore/Src/Jit/DSPJitArithmetic.cpp
View File

@ -21,6 +21,7 @@
#include "../DSPIntUtil.h"
#include "../DSPEmitter.h"
#include "../DSPAnalyzer.h"
#include "DSPJitUtil.h"
#include "x64Emitter.h"
#include "ABI.h"
using namespace Gen;
@ -92,16 +93,17 @@ void DSPEmitter::andcf(const UDSPInstruction opc)
get_acc_m(reg);
// Update_SR_LZ(((val & imm) == imm) ? true : false);
// if ((val & imm) == imm)
// g_dsp.r[DSP_REG_SR] |= SR_LOGIC_ZERO;
// g_dsp._r.sr |= SR_LOGIC_ZERO;
// else
// g_dsp.r[DSP_REG_SR] &= ~SR_LOGIC_ZERO;
// g_dsp._r.sr &= ~SR_LOGIC_ZERO;
AND(16, R(RAX), Imm16(imm));
CMP(16, R(RAX), Imm16(imm));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
FixupBranch notLogicZero = J_CC(CC_NE);
OR(16, MDisp(R11, DSP_REG_SR * 2), Imm16(SR_LOGIC_ZERO));
OR(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, sr)), Imm16(SR_LOGIC_ZERO));
FixupBranch exit = J();
SetJumpTarget(notLogicZero);
AND(16, MDisp(R11, DSP_REG_SR * 2), Imm16(~SR_LOGIC_ZERO));
AND(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, sr)), Imm16(~SR_LOGIC_ZERO));
SetJumpTarget(exit);
}
#else
@ -129,15 +131,16 @@ void DSPEmitter::andf(const UDSPInstruction opc)
get_acc_m(reg);
// Update_SR_LZ(((val & imm) == 0) ? true : false);
// if ((val & imm) == 0)
// g_dsp.r[DSP_REG_SR] |= SR_LOGIC_ZERO;
// g_dsp._r.sr |= SR_LOGIC_ZERO;
// else
// g_dsp.r[DSP_REG_SR] &= ~SR_LOGIC_ZERO;
// g_dsp._r.sr &= ~SR_LOGIC_ZERO;
TEST(16, R(RAX), Imm16(imm));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
FixupBranch notLogicZero = J_CC(CC_NE);
OR(16, MDisp(R11, DSP_REG_SR * 2), Imm16(SR_LOGIC_ZERO));
OR(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, sr)), Imm16(SR_LOGIC_ZERO));
FixupBranch exit = J();
SetJumpTarget(notLogicZero);
AND(16, MDisp(R11, DSP_REG_SR * 2), Imm16(~SR_LOGIC_ZERO));
AND(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, sr)), Imm16(~SR_LOGIC_ZERO));
SetJumpTarget(exit);
}
#else
@ -234,7 +237,7 @@ void DSPEmitter::cmpar(const UDSPInstruction opc)
// s64 sr = dsp_get_long_acc(sreg);
get_long_acc(sreg, RCX);
MOV(64, R(RAX), R(RCX));
// s64 rr = (s16)g_dsp.r[rreg];
// s64 rr = (s16)g_dsp._r.axh[rreg];
get_ax_h(rreg, RDX);
// rr <<= 16;
SHL(64, R(RDX), Imm8(16));
@ -323,11 +326,11 @@ void DSPEmitter::xorr(const UDSPInstruction opc)
#ifdef _M_X64
u8 dreg = (opc >> 8) & 0x1;
u8 sreg = (opc >> 9) & 0x1;
// u16 accm = g_dsp.r[DSP_REG_ACM0 + dreg] ^ g_dsp.r[DSP_REG_AXH0 + sreg];
// u16 accm = g_dsp._r.acm[dreg] ^ g_dsp._r.axh[sreg];
get_acc_m(dreg, RAX);
get_ax_h(sreg, RDX);
XOR(64, R(RAX), R(RDX));
// g_dsp.r[DSP_REG_ACM0 + dreg] = accm;
// g_dsp._r.acm[dreg] = accm;
set_acc_m(dreg);
// Update_SR_Register16((s16)accm, false, false, isOverS32(dsp_get_long_acc(dreg)));
if (!(DSPAnalyzer::code_flags[compilePC] & DSPAnalyzer::CODE_START_OF_INST) || (DSPAnalyzer::code_flags[compilePC] & DSPAnalyzer::CODE_UPDATE_SR))
@ -352,11 +355,11 @@ void DSPEmitter::andr(const UDSPInstruction opc)
#ifdef _M_X64
u8 dreg = (opc >> 8) & 0x1;
u8 sreg = (opc >> 9) & 0x1;
// u16 accm = g_dsp.r[DSP_REG_ACM0 + dreg] & g_dsp.r[DSP_REG_AXH0 + sreg];
// u16 accm = g_dsp._r.acm[dreg] & g_dsp._r.axh[sreg];
get_acc_m(dreg, RAX);
get_ax_h(sreg, RDX);
AND(64, R(RAX), R(RDX));
// g_dsp.r[DSP_REG_ACM0 + dreg] = accm;
// g_dsp._r.acm[dreg] = accm;
set_acc_m(dreg);
// Update_SR_Register16((s16)accm, false, false, isOverS32(dsp_get_long_acc(dreg)));
if (!(DSPAnalyzer::code_flags[compilePC] & DSPAnalyzer::CODE_START_OF_INST) || (DSPAnalyzer::code_flags[compilePC] & DSPAnalyzer::CODE_UPDATE_SR))
@ -381,11 +384,11 @@ void DSPEmitter::orr(const UDSPInstruction opc)
#ifdef _M_X64
u8 dreg = (opc >> 8) & 0x1;
u8 sreg = (opc >> 9) & 0x1;
// u16 accm = g_dsp.r[DSP_REG_ACM0 + dreg] | g_dsp.r[DSP_REG_AXH0 + sreg];
// u16 accm = g_dsp._r.acm[dreg] | g_dsp._r.axh[sreg];
get_acc_m(dreg, RAX);
get_ax_h(sreg, RDX);
OR(64, R(RAX), R(RDX));
// g_dsp.r[DSP_REG_ACM0 + dreg] = accm;
// g_dsp._r.acm[dreg] = accm;
set_acc_m(dreg);
// Update_SR_Register16((s16)accm, false, false, isOverS32(dsp_get_long_acc(dreg)));
if (!(DSPAnalyzer::code_flags[compilePC] & DSPAnalyzer::CODE_START_OF_INST) || (DSPAnalyzer::code_flags[compilePC] & DSPAnalyzer::CODE_UPDATE_SR))
@ -409,11 +412,11 @@ void DSPEmitter::andc(const UDSPInstruction opc)
{
#ifdef _M_X64
u8 dreg = (opc >> 8) & 0x1;
// u16 accm = g_dsp.r[DSP_REG_ACM0 + dreg] & g_dsp.r[DSP_REG_ACM0 + (1 - dreg)];
// u16 accm = g_dsp._r.acm[dreg] & g_dsp._r.acm[1 - dreg];
get_acc_m(dreg, RAX);
get_acc_m(1 - dreg, RDX);
AND(64, R(RAX), R(RDX));
// g_dsp.r[DSP_REG_ACM0 + dreg] = accm;
// g_dsp._r.acm[dreg] = accm;
set_acc_m(dreg);
// Update_SR_Register16((s16)accm, false, false, isOverS32(dsp_get_long_acc(dreg)));
if (!(DSPAnalyzer::code_flags[compilePC] & DSPAnalyzer::CODE_START_OF_INST) || (DSPAnalyzer::code_flags[compilePC] & DSPAnalyzer::CODE_UPDATE_SR))
@ -437,11 +440,11 @@ void DSPEmitter::orc(const UDSPInstruction opc)
{
#ifdef _M_X64
u8 dreg = (opc >> 8) & 0x1;
// u16 accm = g_dsp.r[DSP_REG_ACM0 + dreg] | g_dsp.r[DSP_REG_ACM0 + (1 - dreg)];
// u16 accm = g_dsp._r.acm[dreg] | g_dsp._r.acm[1 - dreg];
get_acc_m(dreg, RAX);
get_acc_m(1 - dreg, RDX);
OR(64, R(RAX), R(RDX));
// g_dsp.r[DSP_REG_ACM0 + dreg] = accm;
// g_dsp._r.acm[dreg] = accm;
set_acc_m(dreg);
// Update_SR_Register16((s16)accm, false, false, isOverS32(dsp_get_long_acc(dreg)));
if (!(DSPAnalyzer::code_flags[compilePC] & DSPAnalyzer::CODE_START_OF_INST) || (DSPAnalyzer::code_flags[compilePC] & DSPAnalyzer::CODE_UPDATE_SR))
@ -464,11 +467,11 @@ void DSPEmitter::xorc(const UDSPInstruction opc)
{
#ifdef _M_X64
u8 dreg = (opc >> 8) & 0x1;
// u16 accm = g_dsp.r[DSP_REG_ACM0 + dreg] ^ g_dsp.r[DSP_REG_ACM0 + (1 - dreg)];
// u16 accm = g_dsp._r.acm[dreg] ^ g_dsp._r.acm[1 - dreg];
get_acc_m(dreg, RAX);
get_acc_m(1 - dreg, RDX);
XOR(64, R(RAX), R(RDX));
// g_dsp.r[DSP_REG_ACM0 + dreg] = accm;
// g_dsp._r.acm[dreg] = accm;
set_acc_m(dreg);
// Update_SR_Register16((s16)accm, false, false, isOverS32(dsp_get_long_acc(dreg)));
if (!(DSPAnalyzer::code_flags[compilePC] & DSPAnalyzer::CODE_START_OF_INST) || (DSPAnalyzer::code_flags[compilePC] & DSPAnalyzer::CODE_UPDATE_SR))
@ -491,10 +494,10 @@ void DSPEmitter::notc(const UDSPInstruction opc)
{
#ifdef _M_X64
u8 dreg = (opc >> 8) & 0x1;
// u16 accm = g_dsp.r[DSP_REG_ACM0 + dreg] ^ 0xffff;
// u16 accm = g_dsp._r.acm[dreg] ^ 0xffff;
get_acc_m(dreg, RAX);
XOR(16, R(RAX), Imm16(0xffff));
// g_dsp.r[DSP_REG_ACM0 + dreg] = accm;
// g_dsp._r.acm[dreg] = accm;
set_acc_m(dreg);
// Update_SR_Register16((s16)accm, false, false, isOverS32(dsp_get_long_acc(dreg)));
if (!(DSPAnalyzer::code_flags[compilePC] & DSPAnalyzer::CODE_START_OF_INST) || (DSPAnalyzer::code_flags[compilePC] & DSPAnalyzer::CODE_UPDATE_SR))
@ -520,11 +523,11 @@ void DSPEmitter::xori(const UDSPInstruction opc)
u8 reg = (opc >> 8) & 0x1;
// u16 imm = dsp_fetch_code();
u16 imm = dsp_imem_read(compilePC+1);
// g_dsp.r[DSP_REG_ACM0 + reg] ^= imm;
// g_dsp._r.acm[reg] ^= imm;
get_acc_m(reg, RAX);
XOR(16, R(RAX), Imm16(imm));
set_acc_m(reg);
// Update_SR_Register16((s16)g_dsp.r[DSP_REG_ACM0 + reg], false, false, isOverS32(dsp_get_long_acc(reg)));
// Update_SR_Register16((s16)g_dsp._r.acm[reg], false, false, isOverS32(dsp_get_long_acc(reg)));
if (!(DSPAnalyzer::code_flags[compilePC] & DSPAnalyzer::CODE_START_OF_INST) || (DSPAnalyzer::code_flags[compilePC] & DSPAnalyzer::CODE_UPDATE_SR))
{
get_long_acc(reg, RSI);
@ -547,11 +550,11 @@ void DSPEmitter::andi(const UDSPInstruction opc)
u8 reg = (opc >> 8) & 0x1;
// u16 imm = dsp_fetch_code();
u16 imm = dsp_imem_read(compilePC+1);
// g_dsp.r[DSP_REG_ACM0 + reg] &= imm;
// g_dsp._r.acm[reg] &= imm;
get_acc_m(reg, RAX);
AND(16, R(RAX), Imm16(imm));
set_acc_m(reg);
// Update_SR_Register16((s16)g_dsp.r[DSP_REG_ACM0 + reg], false, false, isOverS32(dsp_get_long_acc(reg)));
// Update_SR_Register16((s16)g_dsp._r.acm[reg], false, false, isOverS32(dsp_get_long_acc(reg)));
if (!(DSPAnalyzer::code_flags[compilePC] & DSPAnalyzer::CODE_START_OF_INST) || (DSPAnalyzer::code_flags[compilePC] & DSPAnalyzer::CODE_UPDATE_SR))
{
get_long_acc(reg, RSI);
@ -574,11 +577,11 @@ void DSPEmitter::ori(const UDSPInstruction opc)
u8 reg = (opc >> 8) & 0x1;
// u16 imm = dsp_fetch_code();
u16 imm = dsp_imem_read(compilePC+1);
// g_dsp.r[DSP_REG_ACM0 + reg] |= imm;
// g_dsp._r.acm[reg] |= imm;
get_acc_m(reg, RAX);
OR(16, R(RAX), Imm16(imm));
set_acc_m(reg);
// Update_SR_Register16((s16)g_dsp.r[DSP_REG_ACM0 + reg], false, false, isOverS32(dsp_get_long_acc(reg)));
// Update_SR_Register16((s16)g_dsp._r.acm[reg], false, false, isOverS32(dsp_get_long_acc(reg)));
if (!(DSPAnalyzer::code_flags[compilePC] & DSPAnalyzer::CODE_START_OF_INST) || (DSPAnalyzer::code_flags[compilePC] & DSPAnalyzer::CODE_UPDATE_SR))
{
get_long_acc(reg, RSI);
@ -601,13 +604,14 @@ void DSPEmitter::addr(const UDSPInstruction opc)
#ifdef _M_X64
u8 dreg = (opc >> 8) & 0x1;
u8 sreg = ((opc >> 9) & 0x3) + DSP_REG_AXL0;
u16 *sregp = reg_ptr(sreg);
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
// s64 acc = dsp_get_long_acc(dreg);
get_long_acc(dreg, RCX);
MOV(64, R(RAX), R(RCX));
// s64 ax = (s16)g_dsp.r[sreg];
MOVSX(64, 16, RDX, MDisp(R11, sreg * 2));
MOVSX(64, 16, RDX, MDisp(R11, PtrOffset(sregp,&g_dsp._r)));
// ax <<= 16;
SHL(64, R(RDX), Imm8(16));
// s64 res = acc + ax;
@ -931,13 +935,14 @@ void DSPEmitter::subr(const UDSPInstruction opc)
#ifdef _M_X64
u8 dreg = (opc >> 8) & 0x1;
u8 sreg = ((opc >> 9) & 0x3) + DSP_REG_AXL0;
u16 *sregp = reg_ptr(sreg);
// s64 acc = dsp_get_long_acc(dreg);
get_long_acc(dreg, RCX);
MOV(64, R(RAX), R(RCX));
// s64 ax = (s16)g_dsp.r[sreg];
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVSX(64, 16, RDX, MDisp(R11, sreg * 2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVSX(64, 16, RDX, MDisp(R11, PtrOffset(sregp,&g_dsp._r)));
// ax <<= 16;
SHL(64, R(RDX), Imm8(16));
// s64 res = acc - ax;
@ -1205,10 +1210,11 @@ void DSPEmitter::movr(const UDSPInstruction opc)
#ifdef _M_X64
u8 areg = (opc >> 8) & 0x1;
u8 sreg = ((opc >> 9) & 0x3) + DSP_REG_AXL0;
// s64 acc = (s16)g_dsp.r[sreg];
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVSX(64, 16, RAX, MDisp(R11, sreg * 2));
u16 *sregp = reg_ptr(sreg);
// s64 acc = (s16)g_dsp._r[sreg];
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVSX(64, 16, RAX, MDisp(R11, PtrOffset(sregp,&g_dsp._r)));
// acc <<= 16;
SHL(64, R(RAX), Imm8(16));
// acc &= ~0xffff;
@ -1612,7 +1618,7 @@ void DSPEmitter::lsrnrx(const UDSPInstruction opc)
u8 sreg = (opc >> 9) & 0x1;
// s16 shift;
// u16 axh = g_dsp.r[DSP_REG_AXH0 + sreg];
// u16 axh = g_dsp._r.axh[sreg];
get_ax_h(sreg);
// u64 acc = dsp_get_long_acc(dreg);
get_long_acc(dreg, RDX);
@ -1676,7 +1682,7 @@ void DSPEmitter::asrnrx(const UDSPInstruction opc)
u8 sreg = (opc >> 9) & 0x1;
// s16 shift;
// u16 axh = g_dsp.r[DSP_REG_AXH0 + sreg];
// u16 axh = g_dsp._r.axh[sreg];
get_ax_h(sreg);
// s64 acc = dsp_get_long_acc(dreg);
get_long_acc(dreg, RDX);

62
Source/Core/DSPCore/Src/Jit/DSPJitBranch.cpp
View File

@ -14,9 +14,11 @@
// Official SVN repository and contact information can be found at
// http://code.google.com/p/dolphin-emu/
#include "../DSPMemoryMap.h"
#include "../DSPEmitter.h"
#include "../DSPStacks.h"
#include "DSPJitUtil.h"
#include "x64Emitter.h"
#include "ABI.h"
@ -41,9 +43,9 @@ const u8* CheckCondition(DSPEmitter& emitter, u8 cond, u8 skipCodeSize)
//emitter.INT3();
FixupBranch skipCode2;
#ifdef _M_IX86 // All32
emitter.MOV(16, R(EAX), M(&g_dsp.r[DSP_REG_SR]));
emitter.MOV(16, R(EAX), M(&g_dsp._r.sr));
#else
emitter.MOV(64, R(RAX), ImmPtr(&g_dsp.r[DSP_REG_SR]));
emitter.MOV(64, R(RAX), ImmPtr(&g_dsp._r.sr));
emitter.MOV(16, R(EAX), MatR(RAX));
#endif
switch(cond)
@ -66,9 +68,9 @@ const u8* CheckCondition(DSPEmitter& emitter, u8 cond, u8 skipCodeSize)
skipCode2 = emitter.J_CC(CC_NE);
//skipCode2 = emitter.J_CC((CCFlags)(CC_NE - (cond & 1)));
#ifdef _M_IX86 // All32
emitter.MOV(16, R(EAX), M(&g_dsp.r[DSP_REG_SR]));
emitter.MOV(16, R(EAX), M(&g_dsp._r.sr));
#else
emitter.MOV(64, R(RAX), ImmPtr(&g_dsp.r[DSP_REG_SR]));
emitter.MOV(64, R(RAX), ImmPtr(&g_dsp._r.sr));
emitter.MOV(16, R(EAX), MatR(RAX));
#endif
emitter.TEST(16, R(EAX), Imm16(SR_ARITH_ZERO));
@ -194,13 +196,14 @@ void DSPEmitter::jcc(const UDSPInstruction opc)
void r_jmprcc(const UDSPInstruction opc, DSPEmitter& emitter)
{
u8 reg = (opc >> 5) & 0x7;
u16 *regp = reg_ptr(reg);
//reg can only be DSP_REG_ARx and DSP_REG_IXx now,
//no need to handle DSP_REG_STx.
#ifdef _M_IX86 // All32
emitter.MOV(16, R(EAX), M(&g_dsp.r[reg]));
emitter.MOV(16, R(EAX), M(regp));
emitter.MOV(16, M(&g_dsp.pc), R(EAX));
#else
emitter.MOV(64, R(RSI), ImmPtr(&g_dsp.r[reg]));
emitter.MOV(64, R(RSI), ImmPtr(regp));
emitter.MOV(16, R(RSI), MatR(RSI));
emitter.MOV(64, R(RAX), ImmPtr(&(g_dsp.pc)));
emitter.MOV(16, MatR(RAX), R(RSI));
@ -264,13 +267,14 @@ void DSPEmitter::call(const UDSPInstruction opc)
void r_callr(const UDSPInstruction opc, DSPEmitter& emitter)
{
u8 reg = (opc >> 5) & 0x7;
u16 *regp = reg_ptr(reg);
emitter.MOV(16, R(DX), Imm16(emitter.compilePC + 1));
emitter.dsp_reg_store_stack(DSP_STACK_C);
#ifdef _M_IX86 // All32
emitter.MOV(16, R(EAX), M(&g_dsp.r[reg]));
emitter.MOV(16, R(EAX), M(regp));
emitter.MOV(16, M(&g_dsp.pc), R(EAX));
#else
emitter.MOV(64, R(RSI), ImmPtr(&g_dsp.r[reg]));
emitter.MOV(64, R(RSI), ImmPtr(regp));
emitter.MOV(16, R(RSI), MatR(RSI));
emitter.MOV(64, R(RAX), ImmPtr(&(g_dsp.pc)));
emitter.MOV(16, MatR(RAX), R(RSI));
@ -355,10 +359,10 @@ void DSPEmitter::rti(const UDSPInstruction opc)
// g_dsp.r[DSP_REG_SR] = dsp_reg_load_stack(DSP_STACK_D);
dsp_reg_load_stack(DSP_STACK_D);
#ifdef _M_IX86 // All32
MOV(16, M(&g_dsp.r[DSP_REG_SR]), R(DX));
MOV(16, M(&g_dsp._r.sr), R(DX));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, MDisp(R11,DSP_REG_SR*2), R(DX));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, MDisp(R11,STRUCT_OFFSET(g_dsp._r, sr)), R(DX));
#endif
// g_dsp.pc = dsp_reg_load_stack(DSP_STACK_C);
dsp_reg_load_stack(DSP_STACK_C);
@ -400,12 +404,12 @@ void DSPEmitter::halt(const UDSPInstruction opc)
void DSPEmitter::HandleLoop()
{
#ifdef _M_IX86 // All32
MOVZX(32, 16, EAX, M(&(g_dsp.r[DSP_REG_ST2])));
MOVZX(32, 16, ECX, M(&(g_dsp.r[DSP_REG_ST3])));
MOVZX(32, 16, EAX, M(&g_dsp._r.st[2]));
MOVZX(32, 16, ECX, M(&g_dsp._r.st[3]));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVZX(32, 16, EAX, MDisp(R11,DSP_REG_ST2*2));
MOVZX(32, 16, ECX, MDisp(R11,DSP_REG_ST3*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVZX(32, 16, EAX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, st[2])));
MOVZX(32, 16, ECX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, st[3])));
#endif
CMP(32, R(RCX), Imm32(0));
@ -414,19 +418,19 @@ void DSPEmitter::HandleLoop()
FixupBranch rLoopAddrG = J_CC(CC_NE, true);
#ifdef _M_IX86 // All32
SUB(16, M(&(g_dsp.r[DSP_REG_ST3])), Imm16(1));
CMP(16, M(&(g_dsp.r[DSP_REG_ST3])), Imm16(0));
SUB(16, M(&(g_dsp._r.st[3])), Imm16(1));
CMP(16, M(&(g_dsp._r.st[3])), Imm16(0));
#else
SUB(16, MDisp(R11,DSP_REG_ST3*2), Imm16(1));
CMP(16, MDisp(R11,DSP_REG_ST3*2), Imm16(0));
SUB(16, MDisp(R11,STRUCT_OFFSET(g_dsp._r, st[3])), Imm16(1));
CMP(16, MDisp(R11,STRUCT_OFFSET(g_dsp._r, st[3])), Imm16(0));
#endif
FixupBranch loadStack = J_CC(CC_LE, true);
#ifdef _M_IX86 // All32
MOVZX(32, 16, ECX, M(&(g_dsp.r[DSP_REG_ST0])));
MOVZX(32, 16, ECX, M(&(g_dsp._r.st[0])));
MOV(16, M(&g_dsp.pc), R(RCX));
#else
MOVZX(32, 16, RCX, MDisp(R11,DSP_REG_ST0*2));
MOVZX(32, 16, RCX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, st[0])));
MOV(64, R(RAX), ImmPtr(&(g_dsp.pc)));
MOV(16, MatR(RAX), R(RCX));
#endif
@ -454,12 +458,13 @@ void DSPEmitter::HandleLoop()
void DSPEmitter::loop(const UDSPInstruction opc)
{
u16 reg = opc & 0x1f;
u16 *regp = reg_ptr(reg);
// u16 cnt = g_dsp.r[reg];
#ifdef _M_IX86 // All32
MOVZX(32, 16, EDX, M(&(g_dsp.r[reg])));
MOVZX(32, 16, EDX, M(regp));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVZX(32, 16, EDX, MDisp(R11,reg*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVZX(32, 16, EDX, MDisp(R11,PtrOffset(regp, &g_dsp._r)));
#endif
u16 loop_pc = compilePC + 1;
@ -524,12 +529,13 @@ void DSPEmitter::loopi(const UDSPInstruction opc)
void DSPEmitter::bloop(const UDSPInstruction opc)
{
u16 reg = opc & 0x1f;
u16* regp = reg_ptr(reg);
// u16 cnt = g_dsp.r[reg];
#ifdef _M_IX86 // All32
MOVZX(32, 16, EDX, M(&(g_dsp.r[reg])));
MOVZX(32, 16, EDX, M(regp));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVZX(32, 16, EDX, MDisp(R11,reg*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVZX(32, 16, EDX, MDisp(R11,PtrOffset(regp, &g_dsp._r)));
#endif
u16 loop_pc = dsp_imem_read(compilePC + 1);

36
Source/Core/DSPCore/Src/Jit/DSPJitCCUtil.cpp
View File

@ -36,14 +36,14 @@ void DSPEmitter::Update_SR_Register(Gen::X64Reg val)
// if (_Value == 0) g_dsp.r[DSP_REG_SR] |= SR_ARITH_ZERO;
CMP(64, R(val), Imm8(0));
FixupBranch notZero = J_CC(CC_NZ);
OR(16, MDisp(R11, DSP_REG_SR * 2), Imm16(SR_ARITH_ZERO));
OR(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, sr)), Imm16(SR_ARITH_ZERO));
SetJumpTarget(notZero);
// // 0x08
// if (_Value < 0) g_dsp.r[DSP_REG_SR] |= SR_SIGN;
CMP(64, R(val), Imm8(0));
FixupBranch greaterThanEqual = J_CC(CC_GE);
OR(16, MDisp(R11, DSP_REG_SR * 2), Imm16(SR_SIGN));
OR(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, sr)), Imm16(SR_SIGN));
SetJumpTarget(greaterThanEqual);
// // 0x10
@ -51,7 +51,7 @@ void DSPEmitter::Update_SR_Register(Gen::X64Reg val)
MOVSX(64, 32, RDX, R(val));
CMP(64, R(RDX), R(val));
FixupBranch noOverS32 = J_CC(CC_E);
OR(16, MDisp(R11, DSP_REG_SR * 2), Imm16(SR_OVER_S32));
OR(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, sr)), Imm16(SR_OVER_S32));
SetJumpTarget(noOverS32);
// // 0x20 - Checks if top bits of m are equal
@ -63,7 +63,7 @@ void DSPEmitter::Update_SR_Register(Gen::X64Reg val)
FixupBranch cC = J_CC(CC_NE);
SetJumpTarget(zeroC);
// g_dsp.r[DSP_REG_SR] |= SR_TOP2BITS;
OR(16, MDisp(R11, DSP_REG_SR * 2), Imm16(SR_TOP2BITS));
OR(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, sr)), Imm16(SR_TOP2BITS));
SetJumpTarget(cC);
#endif
}
@ -75,7 +75,7 @@ void DSPEmitter::Update_SR_Register64(Gen::X64Reg val)
{
#ifdef _M_X64
// g_dsp.r[DSP_REG_SR] &= ~SR_CMP_MASK;
AND(16, MDisp(R11, DSP_REG_SR * 2), Imm16(~SR_CMP_MASK));
AND(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, sr)), Imm16(~SR_CMP_MASK));
Update_SR_Register(val);
#endif
}
@ -87,7 +87,7 @@ void DSPEmitter::Update_SR_Register64_Carry(Gen::X64Reg val)
{
#ifdef _M_X64
// g_dsp.r[DSP_REG_SR] &= ~SR_CMP_MASK;
AND(16, MDisp(R11, DSP_REG_SR * 2), Imm16(~SR_CMP_MASK));
AND(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, sr)), Imm16(~SR_CMP_MASK));
CMP(64, R(RCX), R(val));
@ -95,7 +95,7 @@ void DSPEmitter::Update_SR_Register64_Carry(Gen::X64Reg val)
// g_dsp.r[DSP_REG_SR] |= SR_CARRY;
// Carry = (acc>res)
FixupBranch noCarry = J_CC(CC_BE);
OR(16, MDisp(R11, DSP_REG_SR * 2), Imm16(SR_CARRY));
OR(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, sr)), Imm16(SR_CARRY));
SetJumpTarget(noCarry);
// 0x02 and 0x80
@ -107,7 +107,7 @@ void DSPEmitter::Update_SR_Register64_Carry(Gen::X64Reg val)
AND(64, R(RCX), R(RDX));
CMP(64, R(RCX), Imm8(0));
FixupBranch noOverflow = J_CC(CC_GE);
OR(16, MDisp(R11, DSP_REG_SR * 2), Imm16(SR_OVERFLOW | SR_OVERFLOW_STICKY));
OR(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, sr)), Imm16(SR_OVERFLOW | SR_OVERFLOW_STICKY));
SetJumpTarget(noOverflow);
Update_SR_Register(val);
@ -121,7 +121,7 @@ void DSPEmitter::Update_SR_Register64_Carry2(Gen::X64Reg val)
{
#ifdef _M_X64
// g_dsp.r[DSP_REG_SR] &= ~SR_CMP_MASK;
AND(16, MDisp(R11, DSP_REG_SR * 2), Imm16(~SR_CMP_MASK));
AND(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, sr)), Imm16(~SR_CMP_MASK));
CMP(64, R(RCX), R(val));
@ -129,7 +129,7 @@ void DSPEmitter::Update_SR_Register64_Carry2(Gen::X64Reg val)
// g_dsp.r[DSP_REG_SR] |= SR_CARRY;
// Carry2 = (acc>=res)
FixupBranch noCarry2 = J_CC(CC_B);
OR(16, MDisp(R11, DSP_REG_SR * 2), Imm16(SR_CARRY));
OR(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, sr)), Imm16(SR_CARRY));
SetJumpTarget(noCarry2);
// 0x02 and 0x80
@ -141,7 +141,7 @@ void DSPEmitter::Update_SR_Register64_Carry2(Gen::X64Reg val)
AND(64, R(RCX), R(RDX));
CMP(64, R(RCX), Imm8(0));
FixupBranch noOverflow = J_CC(CC_GE);
OR(16, MDisp(R11, DSP_REG_SR * 2), Imm16(SR_OVERFLOW | SR_OVERFLOW_STICKY));
OR(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, sr)), Imm16(SR_OVERFLOW | SR_OVERFLOW_STICKY));
SetJumpTarget(noOverflow);
Update_SR_Register();
@ -164,20 +164,20 @@ void DSPEmitter::Update_SR_Register64_Carry2(Gen::X64Reg val)
void DSPEmitter::Update_SR_Register16(Gen::X64Reg val)
{
#ifdef _M_X64
AND(16, MDisp(R11, DSP_REG_SR * 2), Imm16(~SR_CMP_MASK));
AND(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, sr)), Imm16(~SR_CMP_MASK));
// // 0x04
// if (_Value == 0) g_dsp.r[DSP_REG_SR] |= SR_ARITH_ZERO;
CMP(64, R(val), Imm8(0));
FixupBranch notZero = J_CC(CC_NZ);
OR(16, MDisp(R11, DSP_REG_SR * 2), Imm16(SR_ARITH_ZERO));
OR(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, sr)), Imm16(SR_ARITH_ZERO));
SetJumpTarget(notZero);
// // 0x08
// if (_Value < 0) g_dsp.r[DSP_REG_SR] |= SR_SIGN;
CMP(64, R(val), Imm8(0));
FixupBranch greaterThanEqual = J_CC(CC_GE);
OR(16, MDisp(R11, DSP_REG_SR * 2), Imm16(SR_SIGN));
OR(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, sr)), Imm16(SR_SIGN));
SetJumpTarget(greaterThanEqual);
// // 0x20 - Checks if top bits of m are equal
@ -186,13 +186,13 @@ void DSPEmitter::Update_SR_Register16(Gen::X64Reg val)
SHR(16, R(val), Imm8(14));
CMP(16, R(val), Imm16(0));
FixupBranch nZero = J_CC(CC_NE);
OR(16, MDisp(R11, DSP_REG_SR * 2), Imm16(SR_TOP2BITS));
OR(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, sr)), Imm16(SR_TOP2BITS));
FixupBranch cC = J();
SetJumpTarget(nZero);
CMP(16, R(val), Imm16(3));
FixupBranch notThree = J_CC(CC_NE);
// g_dsp.r[DSP_REG_SR] |= SR_TOP2BITS;
OR(16, MDisp(R11, DSP_REG_SR * 2), Imm16(SR_TOP2BITS));
OR(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, sr)), Imm16(SR_TOP2BITS));
SetJumpTarget(notThree);
SetJumpTarget(cC);
#endif
@ -204,14 +204,14 @@ void DSPEmitter::Update_SR_Register16(Gen::X64Reg val)
void DSPEmitter::Update_SR_Register16_OverS32(Gen::X64Reg val)
{
#ifdef _M_X64
AND(16, MDisp(R11, DSP_REG_SR * 2), Imm16(~SR_CMP_MASK));
AND(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, sr)), Imm16(~SR_CMP_MASK));
// // 0x10
// if (_Value != (s32)_Value) g_dsp.r[DSP_REG_SR] |= SR_OVER_S32;
MOVSX(64, 32, RSI, R(val));
CMP(64, R(RSI), R(val));
FixupBranch noOverS32 = J_CC(CC_E);
OR(16, MDisp(R11, DSP_REG_SR * 2), Imm16(SR_OVER_S32));
OR(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, sr)), Imm16(SR_OVER_S32));
SetJumpTarget(noOverS32);
// // 0x20 - Checks if top bits of m are equal

273
Source/Core/DSPCore/Src/Jit/DSPJitExtOps.cpp
View File

@ -16,6 +16,7 @@
// http://code.google.com/p/dolphin-emu/
#include "../DSPMemoryMap.h"
#include "../DSPEmitter.h"
#include "DSPJitUtil.h"
#include "x64Emitter.h"
#include "ABI.h"
@ -68,7 +69,6 @@ void DSPEmitter::mv(const UDSPInstruction opc)
u8 sreg = (opc & 0x3) + DSP_REG_ACL0;
u8 dreg = ((opc >> 2) & 0x3);
pushExtValueFromReg(dreg + DSP_REG_AXL0, sreg);
// MOV(16, M(&g_dsp.r[dreg + DSP_REG_AXL0]), M(&g_dsp.r[sreg]));
}
// S @$arD, $acS.S
@ -79,14 +79,25 @@ void DSPEmitter::s(const UDSPInstruction opc)
{
u8 dreg = opc & 0x3;
u8 sreg = ((opc >> 3) & 0x3) + DSP_REG_ACL0;
u16 *sregp;
switch(sreg) {
case DSP_REG_ACL0:
case DSP_REG_ACL1:
sregp = &(g_dsp._r.ac[sreg-DSP_REG_ACL0].l);
break;
case DSP_REG_ACM0:
case DSP_REG_ACM1:
sregp = &(g_dsp._r.ac[sreg-DSP_REG_ACM0].m);
break;
}
// u16 addr = g_dsp.r[dest];
#ifdef _M_IX86 // All32
MOVZX(32, 16, EAX, M(&g_dsp.r[dreg]));
MOVZX(32, 16, ECX, M(&g_dsp.r[sreg]));
MOVZX(32, 16, EAX, M(&g_dsp._r.ar[dreg]));
MOVZX(32, 16, ECX, M(sregp));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVZX(64, 16, EAX, MDisp(R11,dreg*2));
MOVZX(64, 16, ECX, MDisp(R11,sreg*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVZX(64, 16, EAX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[dreg])));
MOVZX(64, 16, ECX, MDisp(R11,PtrOffset(sregp, &g_dsp._r)));
#endif
// u16 val = g_dsp.r[src];
dmem_write();
@ -101,13 +112,24 @@ void DSPEmitter::sn(const UDSPInstruction opc)
{
u8 dreg = opc & 0x3;
u8 sreg = ((opc >> 3) & 0x3) + DSP_REG_ACL0;
u16 *sregp;
switch(sreg) {
case DSP_REG_ACL0:
case DSP_REG_ACL1:
sregp = &(g_dsp._r.ac[sreg-DSP_REG_ACL0].l);
break;
case DSP_REG_ACM0:
case DSP_REG_ACM1:
sregp = &(g_dsp._r.ac[sreg-DSP_REG_ACM0].m);
break;
}
#ifdef _M_IX86 // All32
MOVZX(32, 16, EAX, M(&g_dsp.r[dreg]));
MOVZX(32, 16, ECX, M(&g_dsp.r[sreg]));
MOVZX(32, 16, EAX, M(&g_dsp._r.ar[dreg]));
MOVZX(32, 16, ECX, M(sregp));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVZX(64, 16, EAX, MDisp(R11,dreg*2));
MOVZX(64, 16, ECX, MDisp(R11,sreg*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVZX(64, 16, EAX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[dreg])));
MOVZX(64, 16, ECX, MDisp(R11,PtrOffset(sregp,&g_dsp._r)));
#endif
dmem_write();
increase_addr_reg(dreg);
@ -130,10 +152,10 @@ void DSPEmitter::l(const UDSPInstruction opc)
//store (up to) two registers in EBX,
//so store all of SR
#ifdef _M_IX86 // All32
MOV(16, R(EAX), M(&g_dsp.r[DSP_REG_SR]));
MOV(16, R(EAX), M(&g_dsp._r.sr));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, R(EAX), MDisp(R11,DSP_REG_SR*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r.sr));
MOV(16, R(EAX), MatR(R11));
#endif
SHL(32, R(EAX), Imm8(16));
OR(32, R(EBX), R(EAX));
@ -159,10 +181,10 @@ void DSPEmitter::ln(const UDSPInstruction opc)
//store (up to) two registers in EBX,
//so store all of SR
#ifdef _M_IX86 // All32
MOV(16, R(EAX), M(&g_dsp.r[DSP_REG_SR]));
MOV(16, R(EAX), M(&g_dsp._r.sr));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, R(EAX), MDisp(R11,DSP_REG_SR*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r.sr));
MOV(16, R(EAX), MatR(R11));
#endif
SHL(32, R(EAX), Imm8(16));
OR(32, R(EBX), R(EAX));
@ -178,15 +200,15 @@ void DSPEmitter::ln(const UDSPInstruction opc)
// register $ar3. Increment both $ar0 and $ar3.
void DSPEmitter::ls(const UDSPInstruction opc)
{
u8 sreg = (opc & 0x1) + DSP_REG_ACM0;
u8 sreg = opc & 0x1;
u8 dreg = ((opc >> 4) & 0x3) + DSP_REG_AXL0;
#ifdef _M_IX86 // All32
MOVZX(32, 16, EAX, M(&g_dsp.r[DSP_REG_AR3]));
MOVZX(32, 16, ECX, M(&g_dsp.r[sreg]));
MOVZX(32, 16, EAX, M(&g_dsp._r.ar[3]));
MOVZX(32, 16, ECX, M(&g_dsp._r.ac[sreg].m));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVZX(64, 16, EAX, MDisp(R11,DSP_REG_AR3*2));
MOVZX(64, 16, ECX, MDisp(R11,sreg*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVZX(64, 16, EAX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[3])));
MOVZX(64, 16, ECX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ac[sreg].m)));
#endif
dmem_write();
@ -205,15 +227,15 @@ void DSPEmitter::ls(const UDSPInstruction opc)
// register $ar0 and increment $ar3.
void DSPEmitter::lsn(const UDSPInstruction opc)
{
u8 sreg = (opc & 0x1) + DSP_REG_ACM0;
u8 sreg = opc & 0x1;
u8 dreg = ((opc >> 4) & 0x3) + DSP_REG_AXL0;
#ifdef _M_IX86 // All32
MOVZX(32, 16, EAX, M(&g_dsp.r[DSP_REG_AR3]));
MOVZX(32, 16, ECX, M(&g_dsp.r[sreg]));
MOVZX(32, 16, EAX, M(&g_dsp._r.ar[3]));
MOVZX(32, 16, ECX, M(&g_dsp._r.ac[sreg].m));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVZX(64, 16, EAX, MDisp(R11,DSP_REG_AR3*2));
MOVZX(64, 16, ECX, MDisp(R11,sreg*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVZX(64, 16, EAX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[3])));
MOVZX(64, 16, ECX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ac[sreg].m)));
#endif
dmem_write();
@ -231,15 +253,15 @@ void DSPEmitter::lsn(const UDSPInstruction opc)
// register $ar3 and increment $ar0.
void DSPEmitter::lsm(const UDSPInstruction opc)
{
u8 sreg = (opc & 0x1) + DSP_REG_ACM0;
u8 sreg = opc & 0x1;
u8 dreg = ((opc >> 4) & 0x3) + DSP_REG_AXL0;
#ifdef _M_IX86 // All32
MOVZX(32, 16, EAX, M(&g_dsp.r[DSP_REG_AR3]));
MOVZX(32, 16, ECX, M(&g_dsp.r[sreg]));
MOVZX(32, 16, EAX, M(&g_dsp._r.ar[3]));
MOVZX(32, 16, ECX, M(&g_dsp._r.ac[sreg].m));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVZX(64, 16, EAX, MDisp(R11,DSP_REG_AR3*2));
MOVZX(64, 16, ECX, MDisp(R11,sreg*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVZX(64, 16, EAX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[3])));
MOVZX(64, 16, ECX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ac[sreg].m)));
#endif
dmem_write();
@ -258,15 +280,15 @@ void DSPEmitter::lsm(const UDSPInstruction opc)
// register $ar3.
void DSPEmitter::lsnm(const UDSPInstruction opc)
{
u8 sreg = (opc & 0x1) + DSP_REG_ACM0;
u8 sreg = opc & 0x1;
u8 dreg = ((opc >> 4) & 0x3) + DSP_REG_AXL0;
#ifdef _M_IX86 // All32
MOVZX(32, 16, EAX, M(&g_dsp.r[DSP_REG_AR3]));
MOVZX(32, 16, ECX, M(&g_dsp.r[sreg]));
MOVZX(32, 16, EAX, M(&g_dsp._r.ar[3]));
MOVZX(32, 16, ECX, M(&g_dsp._r.ac[sreg].m));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVZX(64, 16, EAX, MDisp(R11,DSP_REG_AR3*2));
MOVZX(64, 16, ECX, MDisp(R11,sreg*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVZX(64, 16, EAX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[3])));
MOVZX(64, 16, ECX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ac[sreg].m)));
#endif
dmem_write();
@ -283,15 +305,15 @@ void DSPEmitter::lsnm(const UDSPInstruction opc)
// $ar3. Increment both $ar0 and $ar3.
void DSPEmitter::sl(const UDSPInstruction opc)
{
u8 sreg = (opc & 0x1) + DSP_REG_ACM0;
u8 sreg = opc & 0x1;
u8 dreg = ((opc >> 4) & 0x3) + DSP_REG_AXL0;
#ifdef _M_IX86 // All32
MOVZX(32, 16, EAX, M(&g_dsp.r[DSP_REG_AR0]));
MOVZX(32, 16, ECX, M(&g_dsp.r[sreg]));
MOVZX(32, 16, EAX, M(&g_dsp._r.ar[0]));
MOVZX(32, 16, ECX, M(&g_dsp._r.ac[sreg].m));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVZX(64, 16, EAX, MDisp(R11,DSP_REG_AR0*2));
MOVZX(64, 16, ECX, MDisp(R11,sreg*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVZX(64, 16, EAX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[0])));
MOVZX(64, 16, ECX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ac[sreg].m)));
#endif
dmem_write();
@ -309,15 +331,15 @@ void DSPEmitter::sl(const UDSPInstruction opc)
// and increment $ar3.
void DSPEmitter::sln(const UDSPInstruction opc)
{
u8 sreg = (opc & 0x1) + DSP_REG_ACM0;
u8 sreg = opc & 0x1;
u8 dreg = ((opc >> 4) & 0x3) + DSP_REG_AXL0;
#ifdef _M_IX86 // All32
MOVZX(32, 16, EAX, M(&g_dsp.r[DSP_REG_AR0]));
MOVZX(32, 16, ECX, M(&g_dsp.r[sreg]));
MOVZX(32, 16, EAX, M(&g_dsp._r.ar[0]));
MOVZX(32, 16, ECX, M(&g_dsp._r.ac[sreg].m));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVZX(64, 16, EAX, MDisp(R11,DSP_REG_AR0*2));
MOVZX(64, 16, ECX, MDisp(R11,sreg*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVZX(64, 16, EAX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[0])));
MOVZX(64, 16, ECX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ac[sreg].m)));
#endif
dmem_write();
@ -335,15 +357,15 @@ void DSPEmitter::sln(const UDSPInstruction opc)
// and increment $ar0.
void DSPEmitter::slm(const UDSPInstruction opc)
{
u8 sreg = (opc & 0x1) + DSP_REG_ACM0;
u8 sreg = opc & 0x1;
u8 dreg = ((opc >> 4) & 0x3) + DSP_REG_AXL0;
#ifdef _M_IX86 // All32
MOVZX(32, 16, EAX, M(&g_dsp.r[DSP_REG_AR0]));
MOVZX(32, 16, ECX, M(&g_dsp.r[sreg]));
MOVZX(32, 16, EAX, M(&g_dsp._r.ar[0]));
MOVZX(32, 16, ECX, M(&g_dsp._r.ac[sreg].m));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVZX(64, 16, EAX, MDisp(R11,DSP_REG_AR0*2));
MOVZX(64, 16, ECX, MDisp(R11,sreg*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVZX(64, 16, EAX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[0])));
MOVZX(64, 16, ECX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ac[sreg].m)));
#endif
dmem_write();
@ -361,15 +383,15 @@ void DSPEmitter::slm(const UDSPInstruction opc)
// and add corresponding indexing register $ix3 to addressing register $ar3.
void DSPEmitter::slnm(const UDSPInstruction opc)
{
u8 sreg = (opc & 0x1) + DSP_REG_ACM0;
u8 sreg = opc & 0x1;
u8 dreg = ((opc >> 4) & 0x3) + DSP_REG_AXL0;
#ifdef _M_IX86 // All32
MOVZX(32, 16, EAX, M(&g_dsp.r[DSP_REG_AR0]));
MOVZX(32, 16, ECX, M(&g_dsp.r[sreg]));
MOVZX(32, 16, EAX, M(&g_dsp._r.ar[0]));
MOVZX(32, 16, ECX, M(&g_dsp._r.ac[sreg].m));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVZX(64, 16, EAX, MDisp(R11,DSP_REG_AR0*2));
MOVZX(64, 16, ECX, MDisp(R11,sreg*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVZX(64, 16, EAX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[0])));
MOVZX(64, 16, ECX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ac[sreg].m)));
#endif
dmem_write();
@ -400,12 +422,12 @@ void DSPEmitter::ld(const UDSPInstruction opc)
// if (IsSameMemArea(g_dsp.r[sreg], g_dsp.r[DSP_REG_AR3])) {
#ifdef _M_IX86 // All32
MOV(16, R(ESI), M(&g_dsp.r[sreg]));
MOV(16, R(EDI), M(&g_dsp.r[DSP_REG_AR3]));
MOV(16, R(ESI), M(&g_dsp._r.ar[sreg]));
MOV(16, R(EDI), M(&g_dsp._r.ar[3]));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, R(ESI), MDisp(R11,sreg*2));
MOV(16, R(EDI), MDisp(R11,DSP_REG_AR3*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, R(ESI), MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[sreg])));
MOV(16, R(EDI), MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[3])));
#endif
SHR(16, R(ESI), Imm8(10));
SHR(16, R(EDI), Imm8(10));
@ -424,12 +446,12 @@ void DSPEmitter::ld(const UDSPInstruction opc)
//if (IsSameMemArea(g_dsp.r[dreg], g_dsp.r[DSP_REG_AR3])) {
#ifdef _M_IX86 // All32
MOV(16, R(ESI), M(&g_dsp.r[dreg]));
MOV(16, R(EDI), M(&g_dsp.r[DSP_REG_AR3]));
MOV(16, R(ESI), M(&g_dsp._r.ar[dreg]));
MOV(16, R(EDI), M(&g_dsp._r.ar[3]));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, R(ESI), MDisp(R11,dreg*2));
MOV(16, R(EDI), MDisp(R11,DSP_REG_AR3*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, R(ESI), MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[dreg])));
MOV(16, R(EDI), MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[3])));
#endif
SHR(16, R(ESI), Imm8(10));
SHR(16, R(EDI), Imm8(10));
@ -460,12 +482,12 @@ void DSPEmitter::ldn(const UDSPInstruction opc)
// if (IsSameMemArea(g_dsp.r[sreg], g_dsp.r[DSP_REG_AR3])) {
#ifdef _M_IX86 // All32
MOV(16, R(ESI), M(&g_dsp.r[sreg]));
MOV(16, R(EDI), M(&g_dsp.r[DSP_REG_AR3]));
MOV(16, R(ESI), M(&g_dsp._r.ar[sreg]));
MOV(16, R(EDI), M(&g_dsp._r.ar[3]));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, R(ESI), MDisp(R11,sreg*2));
MOV(16, R(EDI), MDisp(R11,DSP_REG_AR3*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, R(ESI), MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[sreg])));
MOV(16, R(EDI), MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[3])));
#endif
SHR(16, R(ESI), Imm8(10));
SHR(16, R(EDI), Imm8(10));
@ -483,12 +505,12 @@ void DSPEmitter::ldn(const UDSPInstruction opc)
//if (IsSameMemArea(g_dsp.r[dreg], g_dsp.r[DSP_REG_AR3])) {
#ifdef _M_IX86 // All32
MOV(16, R(ESI), M(&g_dsp.r[dreg]));
MOV(16, R(EDI), M(&g_dsp.r[DSP_REG_AR3]));
MOV(16, R(ESI), M(&g_dsp._r.ar[dreg]));
MOV(16, R(EDI), M(&g_dsp._r.ar[3]));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, R(ESI), MDisp(R11,dreg*2));
MOV(16, R(EDI), MDisp(R11,DSP_REG_AR3*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, R(ESI), MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[dreg])));
MOV(16, R(EDI), MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[3])));
#endif
SHR(16, R(ESI), Imm8(10));
SHR(16, R(EDI), Imm8(10));
@ -519,12 +541,12 @@ void DSPEmitter::ldm(const UDSPInstruction opc)
// if (IsSameMemArea(g_dsp.r[sreg], g_dsp.r[DSP_REG_AR3])) {
#ifdef _M_IX86 // All32
MOV(16, R(ESI), M(&g_dsp.r[sreg]));
MOV(16, R(EDI), M(&g_dsp.r[DSP_REG_AR3]));
MOV(16, R(ESI), M(&g_dsp._r.ar[sreg]));
MOV(16, R(EDI), M(&g_dsp._r.ar[3]));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, R(ESI), MDisp(R11,sreg*2));
MOV(16, R(EDI), MDisp(R11,DSP_REG_AR3*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, R(ESI), MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[sreg])));
MOV(16, R(EDI), MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[3])));
#endif
SHR(16, R(ESI), Imm8(10));
SHR(16, R(EDI), Imm8(10));
@ -542,12 +564,12 @@ void DSPEmitter::ldm(const UDSPInstruction opc)
//if (IsSameMemArea(g_dsp.r[dreg], g_dsp.r[DSP_REG_AR3])) {
#ifdef _M_IX86 // All32
MOV(16, R(ESI), M(&g_dsp.r[dreg]));
MOV(16, R(EDI), M(&g_dsp.r[DSP_REG_AR3]));
MOV(16, R(ESI), M(&g_dsp._r.ar[dreg]));
MOV(16, R(EDI), M(&g_dsp._r.ar[3]));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, R(ESI), MDisp(R11,dreg*2));
MOV(16, R(EDI), MDisp(R11,DSP_REG_AR3*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, R(ESI), MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[dreg])));
MOV(16, R(EDI), MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[3])));
#endif
SHR(16, R(ESI), Imm8(10));
SHR(16, R(EDI), Imm8(10));
@ -578,12 +600,12 @@ void DSPEmitter::ldnm(const UDSPInstruction opc)
// if (IsSameMemArea(g_dsp.r[sreg], g_dsp.r[DSP_REG_AR3])) {
#ifdef _M_IX86 // All32
MOV(16, R(ESI), M(&g_dsp.r[sreg]));
MOV(16, R(EDI), M(&g_dsp.r[DSP_REG_AR3]));
MOV(16, R(ESI), M(&g_dsp._r.ar[sreg]));
MOV(16, R(EDI), M(&g_dsp._r.ar[3]));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, R(ESI), MDisp(R11,sreg*2));
MOV(16, R(EDI), MDisp(R11,DSP_REG_AR3*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, R(ESI), MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[sreg])));
MOV(16, R(EDI), MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[3])));
#endif
SHR(16, R(ESI), Imm8(10));
SHR(16, R(EDI), Imm8(10));
@ -601,12 +623,12 @@ void DSPEmitter::ldnm(const UDSPInstruction opc)
//if (IsSameMemArea(g_dsp.r[dreg], g_dsp.r[DSP_REG_AR3])) {
#ifdef _M_IX86 // All32
MOV(16, R(ESI), M(&g_dsp.r[dreg]));
MOV(16, R(EDI), M(&g_dsp.r[DSP_REG_AR3]));
MOV(16, R(ESI), M(&g_dsp._r.ar[dreg]));
MOV(16, R(EDI), M(&g_dsp._r.ar[3]));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, R(ESI), MDisp(R11,dreg*2));
MOV(16, R(EDI), MDisp(R11,DSP_REG_AR3*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, R(ESI), MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[dreg])));
MOV(16, R(EDI), MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[3])));
#endif
SHR(16, R(ESI), Imm8(10));
SHR(16, R(EDI), Imm8(10));
@ -628,11 +650,12 @@ void DSPEmitter::ldnm(const UDSPInstruction opc)
// Push value from g_dsp.r[sreg] into EBX and stores the destinationindex in
// storeIndex
void DSPEmitter::pushExtValueFromReg(u16 dreg, u16 sreg) {
u16 *sregp = reg_ptr(sreg);
#ifdef _M_IX86 // All32
MOVZX(32, 16, EBX, M(&g_dsp.r[sreg]));
MOVZX(32, 16, EBX, M(sregp));
#else
MOV(64, R(RBX), ImmPtr(&g_dsp.r));
MOVZX(32, 16, EBX, MDisp(RBX,sreg*2));
MOV(64, R(RBX), ImmPtr(&g_dsp._r));
MOVZX(32, 16, EBX, MDisp(RBX,PtrOffset(sregp,&g_dsp._r)));
#endif
storeIndex = dreg;
}
@ -640,10 +663,10 @@ void DSPEmitter::pushExtValueFromReg(u16 dreg, u16 sreg) {
void DSPEmitter::pushExtValueFromMem(u16 dreg, u16 sreg) {
// u16 addr = g_dsp.r[addr];
#ifdef _M_IX86 // All32
MOVZX(32, 16, ECX, M(&g_dsp.r[sreg]));
MOVZX(32, 16, ECX, M(&g_dsp._r.ar[sreg]));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVZX(64, 16, ECX, MDisp(R11,sreg*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVZX(64, 16, ECX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[sreg])));
#endif
dmem_read();
MOVZX(32, 16, EBX, R(EAX));
@ -654,10 +677,10 @@ void DSPEmitter::pushExtValueFromMem(u16 dreg, u16 sreg) {
void DSPEmitter::pushExtValueFromMem2(u16 dreg, u16 sreg) {
// u16 addr = g_dsp.r[addr];
#ifdef _M_IX86 // All32
MOVZX(32, 16, ECX, M(&g_dsp.r[sreg]));
MOVZX(32, 16, ECX, M(&g_dsp._r.ar[sreg]));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVZX(64, 16, ECX, MDisp(R11,sreg*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVZX(64, 16, ECX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[sreg])));
#endif
dmem_read();
SHL(32,R(EAX),Imm8(16));
@ -675,11 +698,12 @@ void DSPEmitter::popExtValueToReg() {
// [nakeee] the or case never happens in real
// [nakeee] it's just how the hardware works so we added it
if (storeIndex != -1) {
u16 *dregp = reg_ptr(storeIndex);
#ifdef _M_IX86 // All32
MOV(16, M(&g_dsp.r[storeIndex]), R(EBX));
MOV(16, M(dregp), R(EBX));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, MDisp(R11,storeIndex*2), R(EBX));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, MDisp(R11,PtrOffset(dregp,&g_dsp._r)), R(EBX));
#endif
if (storeIndex >= DSP_REG_ACM0 && storeIndex2 == -1) {
TEST(32, R(EBX), Imm32(SR_40_MODE_BIT << 16));
@ -693,13 +717,13 @@ void DSPEmitter::popExtValueToReg() {
//g_dsp.r[reg - DSP_REG_ACM0 + DSP_REG_ACH0] = (val & 0x8000) ? 0xFFFF : 0x0000;
//g_dsp.r[reg - DSP_REG_ACM0 + DSP_REG_ACL0] = 0;
#ifdef _M_IX86 // All32
MOV(16,M(&g_dsp.r[storeIndex - DSP_REG_ACM0 + DSP_REG_ACH0]),
MOV(16,M(&g_dsp._r.ac[storeIndex - DSP_REG_ACM0].h),
R(EAX));
MOV(16,M(&g_dsp.r[storeIndex - DSP_REG_ACM0 + DSP_REG_ACL0]),
MOV(16,M(&g_dsp._r.ac[storeIndex - DSP_REG_ACM0].l),
Imm16(0));
#else
MOV(16, MDisp(R11,(storeIndex - DSP_REG_ACM0 + DSP_REG_ACH0)*2), R(EAX));
MOV(16, MDisp(R11,(storeIndex - DSP_REG_ACM0 + DSP_REG_ACL0)*2), Imm16(0));
MOV(16, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ac[storeIndex - DSP_REG_ACM0].h)), R(EAX));
MOV(16, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ac[storeIndex - DSP_REG_ACM0].l)), Imm16(0));
#endif
//}
SetJumpTarget(not_40bit);
@ -710,11 +734,12 @@ void DSPEmitter::popExtValueToReg() {
if (storeIndex2 != -1) {
SHR(32,R(EBX),Imm8(16));
u16 *dregp = reg_ptr(storeIndex2);
#ifdef _M_IX86 // All32
MOV(16, M(&g_dsp.r[storeIndex2]), R(EBX));
MOV(16, M(dregp), R(EBX));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, MDisp(R11,storeIndex2*2), R(EBX));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, MDisp(R11,PtrOffset(dregp, &g_dsp._r)), R(EBX));
#endif
}
storeIndex2 = -1;

107
Source/Core/DSPCore/Src/Jit/DSPJitLoadStore.cpp
View File

@ -20,6 +20,7 @@
#include "../DSPIntCCUtil.h"
#include "../DSPIntUtil.h"
#include "../DSPEmitter.h"
#include "DSPJitUtil.h"
#include "x64Emitter.h"
#include "ABI.h"
using namespace Gen;
@ -33,14 +34,15 @@ using namespace Gen;
void DSPEmitter::srs(const UDSPInstruction opc)
{
u8 reg = ((opc >> 8) & 0x7) + 0x18;
//u16 addr = (g_dsp.r[DSP_REG_CR] << 8) | (opc & 0xFF);
u16 *regp = reg_ptr(reg);
//u16 addr = (g_dsp._r.cr << 8) | (opc & 0xFF);
#ifdef _M_IX86 // All32
MOVZX(32, 16, ECX, M(&g_dsp.r[reg]));
MOVZX(32, 8, EAX, M(&g_dsp.r[DSP_REG_CR]));
MOVZX(32, 16, ECX, M(regp));
MOVZX(32, 8, EAX, M(&g_dsp._r.cr));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVZX(64, 16, RCX, MDisp(R11,reg*2));
MOVZX(64, 8, RAX, MDisp(R11,DSP_REG_CR*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVZX(64, 16, RCX, MDisp(R11,PtrOffset(regp,&g_dsp._r)));
MOVZX(64, 8, RAX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, cr)));
#endif
SHL(16, R(EAX), Imm8(8));
OR(8, R(EAX), Imm8(opc & 0xFF));
@ -55,21 +57,22 @@ void DSPEmitter::srs(const UDSPInstruction opc)
void DSPEmitter::lrs(const UDSPInstruction opc)
{
u8 reg = ((opc >> 8) & 0x7) + 0x18;
u16 *regp = reg_ptr(reg);
//u16 addr = (g_dsp.r[DSP_REG_CR] << 8) | (opc & 0xFF);
#ifdef _M_IX86 // All32
MOVZX(32, 8, ECX, M(&g_dsp.r[DSP_REG_CR]));
MOVZX(32, 8, ECX, M(&g_dsp._r.cr));
SHL(16, R(ECX), Imm8(8));
OR(8, R(ECX), Imm8(opc & 0xFF));
dmem_read();
MOV(16, M(&g_dsp.r[reg]), R(EAX));
MOV(16, M(regp), R(EAX));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVZX(64, 8, RCX, MDisp(R11,DSP_REG_CR*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVZX(64, 8, RCX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, cr)));
SHL(16, R(ECX), Imm8(8));
OR(8, R(ECX), Imm8(opc & 0xFF));
dmem_read();
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, MDisp(R11,reg*2), R(RAX));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, MDisp(R11,PtrOffset(regp, &g_dsp._r)), R(RAX));
#endif
dsp_conditional_extend_accum(reg);
}
@ -192,10 +195,10 @@ void DSPEmitter::srr(const UDSPInstruction opc)
dsp_op_read_reg(sreg, ECX);
#ifdef _M_IX86 // All32
MOVZX(32, 16, EAX, M(&g_dsp.r[dreg]));
MOVZX(32, 16, EAX, M(&g_dsp._r.ar[dreg]));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVZX(64, 16, RAX, MDisp(R11,dreg*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVZX(64, 16, RAX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[dreg])));
#endif
dmem_write();
}
@ -212,10 +215,10 @@ void DSPEmitter::srrd(const UDSPInstruction opc)
dsp_op_read_reg(sreg, ECX);
#ifdef _M_IX86 // All32
MOVZX(32, 16, EAX, M(&g_dsp.r[dreg]));
MOVZX(32, 16, EAX, M(&g_dsp._r.ar[dreg]));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVZX(64, 16, RAX, MDisp(R11,dreg*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVZX(64, 16, RAX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[dreg])));
#endif
dmem_write();
decrement_addr_reg(dreg);
@ -233,10 +236,10 @@ void DSPEmitter::srri(const UDSPInstruction opc)
dsp_op_read_reg(sreg, ECX);
#ifdef _M_IX86 // All32
MOVZX(32, 16, EAX, M(&g_dsp.r[dreg]));
MOVZX(32, 16, EAX, M(&g_dsp._r.ar[dreg]));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVZX(64, 16, RAX, MDisp(R11,dreg*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVZX(64, 16, RAX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[dreg])));
#endif
dmem_write();
increment_addr_reg(dreg);
@ -254,10 +257,10 @@ void DSPEmitter::srrn(const UDSPInstruction opc)
dsp_op_read_reg(sreg, ECX);
#ifdef _M_IX86 // All32
MOVZX(32, 16, EAX, M(&g_dsp.r[dreg]));
MOVZX(32, 16, EAX, M(&g_dsp._r.ar[dreg]));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVZX(64, 16, RAX, MDisp(R11,dreg*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVZX(64, 16, RAX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[dreg])));
#endif
dmem_write();
increase_addr_reg(dreg);
@ -270,20 +273,20 @@ void DSPEmitter::srrn(const UDSPInstruction opc)
void DSPEmitter::ilrr(const UDSPInstruction opc)
{
u16 reg = opc & 0x3;
u16 dreg = DSP_REG_ACM0 + ((opc >> 8) & 1);
u16 dreg = (opc >> 8) & 1;
#ifdef _M_IX86 // All32
MOVZX(32, 16, ECX, M(&g_dsp.r[reg]));
MOVZX(32, 16, ECX, M(&g_dsp._r.ar[reg]));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVZX(64, 16, RCX, MDisp(R11,reg*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVZX(64, 16, RCX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[reg])));
#endif
imem_read();
#ifdef _M_IX86 // All32
MOV(16, M(&g_dsp.r[dreg]), R(EAX));
MOV(16, M(&g_dsp._r.ac[dreg].m), R(EAX));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, MDisp(R11,dreg*2), R(RAX));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ac[dreg].m)), R(RAX));
#endif
dsp_conditional_extend_accum(dreg);
}
@ -295,20 +298,20 @@ void DSPEmitter::ilrr(const UDSPInstruction opc)
void DSPEmitter::ilrrd(const UDSPInstruction opc)
{
u16 reg = opc & 0x3;
u16 dreg = DSP_REG_ACM0 + ((opc >> 8) & 1);
u16 dreg = (opc >> 8) & 1;
#ifdef _M_IX86 // All32
MOVZX(32, 16, ECX, M(&g_dsp.r[reg]));
MOVZX(32, 16, ECX, M(&g_dsp._r.ar[reg]));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVZX(64, 16, RCX, MDisp(R11,reg*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVZX(64, 16, RCX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[reg])));
#endif
imem_read();
#ifdef _M_IX86 // All32
MOV(16, M(&g_dsp.r[dreg]), R(EAX));
MOV(16, M(&g_dsp._r.ac[dreg].m), R(EAX));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, MDisp(R11,dreg*2), R(RAX));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ac[dreg].m)), R(RAX));
#endif
dsp_conditional_extend_accum(dreg);
decrement_addr_reg(reg);
@ -321,20 +324,20 @@ void DSPEmitter::ilrrd(const UDSPInstruction opc)
void DSPEmitter::ilrri(const UDSPInstruction opc)
{
u16 reg = opc & 0x3;
u16 dreg = DSP_REG_ACM0 + ((opc >> 8) & 1);
u16 dreg = (opc >> 8) & 1;
#ifdef _M_IX86 // All32
MOVZX(32, 16, ECX, M(&g_dsp.r[reg]));
MOVZX(32, 16, ECX, M(&g_dsp._r.ar[reg]));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVZX(64, 16, RCX, MDisp(R11,reg*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVZX(64, 16, RCX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[reg])));
#endif
imem_read();
#ifdef _M_IX86 // All32
MOV(16, M(&g_dsp.r[dreg]), R(EAX));
MOV(16, M(&g_dsp._r.ac[dreg].m), R(EAX));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, MDisp(R11,dreg*2), R(RAX));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ac[dreg].m)), R(RAX));
#endif
dsp_conditional_extend_accum(dreg);
increment_addr_reg(reg);
@ -348,20 +351,20 @@ void DSPEmitter::ilrri(const UDSPInstruction opc)
void DSPEmitter::ilrrn(const UDSPInstruction opc)
{
u16 reg = opc & 0x3;
u16 dreg = DSP_REG_ACM0 + ((opc >> 8) & 1);
u16 dreg = (opc >> 8) & 1;
#ifdef _M_IX86 // All32
MOVZX(32, 16, ECX, M(&g_dsp.r[reg]));
MOVZX(32, 16, ECX, M(&g_dsp._r.ar[reg]));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVZX(64, 16, RCX, MDisp(R11,reg*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVZX(64, 16, RCX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[reg])));
#endif
imem_read();
#ifdef _M_IX86 // All32
MOV(16, M(&g_dsp.r[dreg]), R(EAX));
MOV(16, M(&g_dsp._r.ac[dreg].m), R(EAX));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, MDisp(R11,dreg*2), R(RAX));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ac[dreg].m)), R(RAX));
#endif
dsp_conditional_extend_accum(dreg);
increase_addr_reg(reg);

114
Source/Core/DSPCore/Src/Jit/DSPJitMisc.cpp
View File

@ -18,6 +18,7 @@
#include "../DSPIntUtil.h"
#include "../DSPEmitter.h"
#include "DSPJitUtil.h"
#include "x64Emitter.h"
#include "ABI.h"
using namespace Gen;
@ -47,12 +48,12 @@ void DSPEmitter::dsp_reg_stack_push(int stack_reg)
//g_dsp.reg_stack[stack_reg][g_dsp.reg_stack_ptr[stack_reg]] = g_dsp.r[DSP_REG_ST0 + stack_reg];
#ifdef _M_IX86 // All32
MOV(16, R(CX), M(&g_dsp.r[DSP_REG_ST0 + stack_reg]));
MOV(16, R(CX), M(&g_dsp._r.st[stack_reg]));
MOVZX(32, 8, EAX, R(AL));
MOV(16, MComplex(EAX,EAX,1,(u32)&g_dsp.reg_stack[stack_reg][0]), R(CX));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, R(CX), MDisp(R11,(DSP_REG_ST0 + stack_reg)*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, R(CX), MDisp(R11,STRUCT_OFFSET(g_dsp._r, st[stack_reg])));
MOVZX(64, 8, RAX, R(AL));
MOV(64, R(R10), ImmPtr(&g_dsp.reg_stack[stack_reg][0]));
MOV(16, MComplex(R10,RAX,2,0), R(CX));
@ -76,13 +77,13 @@ void DSPEmitter::dsp_reg_stack_pop(int stack_reg)
#ifdef _M_IX86 // All32
MOVZX(32, 8, EAX, R(AL));
MOV(16, R(CX), MComplex(EAX,EAX,1,(u32)&g_dsp.reg_stack[stack_reg][0]));
MOV(16, M(&g_dsp.r[DSP_REG_ST0 + stack_reg]), R(CX));
MOV(16, M(&g_dsp._r.st[stack_reg]), R(CX));
#else
MOVZX(64, 8, RAX, R(AL));
MOV(64, R(R10), ImmPtr(&g_dsp.reg_stack[stack_reg][0]));
MOV(16, R(CX), MComplex(R10,RAX,2,0));
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, MDisp(R11,(DSP_REG_ST0 + stack_reg)*2), R(CX));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, MDisp(R11,STRUCT_OFFSET(g_dsp._r, st[stack_reg])), R(CX));
#endif
//g_dsp.reg_stack_ptr[stack_reg]--;
@ -106,10 +107,10 @@ void DSPEmitter::dsp_reg_store_stack(int stack_reg, Gen::X64Reg host_sreg)
dsp_reg_stack_push(stack_reg);
//g_dsp.r[DSP_REG_ST0 + stack_reg] = val;
#ifdef _M_IX86 // All32
MOV(16, M(&g_dsp.r[DSP_REG_ST0+stack_reg]), R(EDX));
MOV(16, M(&g_dsp._r.st[stack_reg]), R(EDX));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, MDisp(R11,(DSP_REG_ST0+stack_reg)*2), R(EDX));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, MDisp(R11,STRUCT_OFFSET(g_dsp._r, st[stack_reg])), R(EDX));
#endif
}
@ -117,10 +118,10 @@ void DSPEmitter::dsp_reg_load_stack(int stack_reg, Gen::X64Reg host_dreg)
{
//u16 val = g_dsp.r[DSP_REG_ST0 + stack_reg];
#ifdef _M_IX86 // All32
MOV(16, R(EDX), M(&g_dsp.r[DSP_REG_ST0+stack_reg]));
MOV(16, R(EDX), M(&g_dsp._r.st[stack_reg]));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, R(EDX), MDisp(R11,(DSP_REG_ST0+stack_reg)*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, R(EDX), MDisp(R11,STRUCT_OFFSET(g_dsp._r, st[stack_reg])));
#endif
dsp_reg_stack_pop(stack_reg);
if (host_dreg != EDX) {
@ -133,10 +134,10 @@ void DSPEmitter::dsp_reg_store_stack_imm(int stack_reg, u16 val)
dsp_reg_stack_push(stack_reg);
//g_dsp.r[DSP_REG_ST0 + stack_reg] = val;
#ifdef _M_IX86 // All32
MOV(16, M(&g_dsp.r[DSP_REG_ST0+stack_reg]), Imm16(val));
MOV(16, M(&g_dsp._r.st[stack_reg]), Imm16(val));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, MDisp(R11,(DSP_REG_ST0+stack_reg)*2), Imm16(val));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, MDisp(R11,STRUCT_OFFSET(g_dsp._r, st[stack_reg])), Imm16(val));
#endif
}
@ -149,10 +150,10 @@ void DSPEmitter::dsp_op_write_reg(int reg, Gen::X64Reg host_sreg)
// sign extend from the bottom 8 bits.
MOVSX(16,8,host_sreg,R(host_sreg));
#ifdef _M_IX86 // All32
MOV(16, M(&g_dsp.r[reg]), R(host_sreg));
MOV(16, M(&g_dsp._r.ac[reg-DSP_REG_ACH0].h), R(host_sreg));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, MDisp(R11,reg*2), R(host_sreg));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ac[reg-DSP_REG_ACH0].h)), R(host_sreg));
#endif
break;
@ -165,14 +166,17 @@ void DSPEmitter::dsp_op_write_reg(int reg, Gen::X64Reg host_sreg)
break;
default:
{
u16 *regp = reg_ptr(reg);
#ifdef _M_IX86 // All32
MOV(16, M(&g_dsp.r[reg]), R(host_sreg));
MOV(16, M(regp), R(host_sreg));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, MDisp(R11,reg*2), R(host_sreg));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, MDisp(R11,PtrOffset(regp,&g_dsp._r)), R(host_sreg));
#endif
break;
}
}
}
void DSPEmitter::dsp_op_write_reg_imm(int reg, u16 val)
@ -183,10 +187,10 @@ void DSPEmitter::dsp_op_write_reg_imm(int reg, u16 val)
case DSP_REG_ACH1:
// sign extend from the bottom 8 bits.
#ifdef _M_IX86 // All32
MOV(16, M(&g_dsp.r[reg]), Imm16((u16)(s16)(s8)(u8)val));
MOV(16, M(&g_dsp._r.ac[reg-DSP_REG_ACH0].h), Imm16((u16)(s16)(s8)(u8)val));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, MDisp(R11,reg*2), Imm16((u16)(s16)(s8)(u8)val));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ac[reg-DSP_REG_ACH0].h)), Imm16((u16)(s16)(s8)(u8)val));
#endif
break;
@ -199,14 +203,17 @@ void DSPEmitter::dsp_op_write_reg_imm(int reg, u16 val)
break;
default:
{
u16 *regp = reg_ptr(reg);
#ifdef _M_IX86 // All32
MOV(16, M(&g_dsp.r[reg]), Imm16(val));
MOV(16, M(regp), Imm16(val));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, MDisp(R11,reg*2), Imm16(val));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, MDisp(R11,PtrOffset(regp,&g_dsp._r)), Imm16(val));
#endif
break;
}
}
}
void DSPEmitter::dsp_conditional_extend_accum(int reg)
@ -217,10 +224,10 @@ void DSPEmitter::dsp_conditional_extend_accum(int reg)
case DSP_REG_ACM1:
{
#ifdef _M_IX86 // All32
MOV(16, R(EAX), M(&g_dsp.r[DSP_REG_SR]));
MOV(16, R(EAX), M(&g_dsp._r.sr));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, R(EAX), MDisp(R11,DSP_REG_SR*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, R(EAX), MDisp(R11,STRUCT_OFFSET(g_dsp._r, sr)));
#endif
TEST(16, R(EAX), Imm16(SR_40_MODE_BIT));
FixupBranch not_40bit = J_CC(CC_Z);
@ -229,21 +236,21 @@ void DSPEmitter::dsp_conditional_extend_accum(int reg)
// Sign extend into whole accum.
//u16 val = g_dsp.r[reg];
#ifdef _M_IX86 // All32
MOVSX(32, 16, EAX, M(&g_dsp.r[reg]));
MOVSX(32, 16, EAX, M(&g_dsp._r.ac[reg-DSP_REG_ACM0].m));
#else
MOVSX(64, 16, EAX, MDisp(R11,reg*2));
MOVSX(64, 16, EAX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ac[reg-DSP_REG_ACM0].m)));
#endif
SHR(32,R(EAX),Imm8(16));
//g_dsp.r[reg - DSP_REG_ACM0 + DSP_REG_ACH0] = (val & 0x8000) ? 0xFFFF : 0x0000;
//g_dsp.r[reg - DSP_REG_ACM0 + DSP_REG_ACL0] = 0;
#ifdef _M_IX86 // All32
MOV(16,M(&g_dsp.r[reg - DSP_REG_ACM0 + DSP_REG_ACH0]),
MOV(16,M(&g_dsp._r.ac[reg - DSP_REG_ACM0].h),
R(EAX));
MOV(16,M(&g_dsp.r[reg - DSP_REG_ACM0 + DSP_REG_ACL0]),
MOV(16,M(&g_dsp._r.ac[reg - DSP_REG_ACM0].l),
Imm16(0));
#else
MOV(16, MDisp(R11,(reg - DSP_REG_ACM0 + DSP_REG_ACH0)*2), R(EAX));
MOV(16, MDisp(R11,(reg - DSP_REG_ACM0 + DSP_REG_ACL0)*2), Imm16(0));
MOV(16, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ac[reg-DSP_REG_ACM0].h)), R(EAX));
MOV(16, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ac[reg-DSP_REG_ACM0].l)), Imm16(0));
#endif
//}
SetJumpTarget(not_40bit);
@ -259,10 +266,10 @@ void DSPEmitter::dsp_conditional_extend_accum_imm(int reg, u16 val)
case DSP_REG_ACM1:
{
#ifdef _M_IX86 // All32
MOV(16, R(EAX), M(&g_dsp.r[DSP_REG_SR]));
MOV(16, R(EAX), M(&g_dsp._r.sr));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, R(EAX), MDisp(R11,DSP_REG_SR*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, R(EAX), MDisp(R11,STRUCT_OFFSET(g_dsp._r, sr)));
#endif
TEST(16, R(EAX), Imm16(SR_40_MODE_BIT));
FixupBranch not_40bit = J_CC(CC_Z);
@ -272,14 +279,14 @@ void DSPEmitter::dsp_conditional_extend_accum_imm(int reg, u16 val)
//g_dsp.r[reg - DSP_REG_ACM0 + DSP_REG_ACH0] = (val & 0x8000) ? 0xFFFF : 0x0000;
//g_dsp.r[reg - DSP_REG_ACM0 + DSP_REG_ACL0] = 0;
#ifdef _M_IX86 // All32
MOV(16,M(&g_dsp.r[reg - DSP_REG_ACM0 + DSP_REG_ACH0]),
MOV(16,M(&g_dsp._r.ac[reg - DSP_REG_ACM0].h),
Imm16((val & 0x8000)?0xffff:0x0000));
MOV(16,M(&g_dsp.r[reg - DSP_REG_ACM0 + DSP_REG_ACL0]),
MOV(16,M(&g_dsp._r.ac[reg - DSP_REG_ACM0].l),
Imm16(0));
#else
MOV(16, MDisp(R11,(reg - DSP_REG_ACM0 + DSP_REG_ACH0)*2),
MOV(16, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ac[reg-DSP_REG_ACM0].h)),
Imm16((val & 0x8000)?0xffff:0x0000));
MOV(16, MDisp(R11,(reg - DSP_REG_ACM0 + DSP_REG_ACL0)*2),
MOV(16, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ac[reg-DSP_REG_ACM0].l)),
Imm16(0));
#endif
//}
@ -297,14 +304,17 @@ void DSPEmitter::dsp_op_read_reg(int reg, Gen::X64Reg host_dreg)
case DSP_REG_ST3:
return dsp_reg_load_stack(reg - DSP_REG_ST0, host_dreg);
default:
{
u16 *regp = reg_ptr(reg);
//return g_dsp.r[reg];
#ifdef _M_IX86 // All32
MOV(16, R(host_dreg), M(&g_dsp.r[reg]));
MOV(16, R(host_dreg), M(regp));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, R(host_dreg), MDisp(R11,reg*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, R(host_dreg), MDisp(R11,PtrOffset(regp,&g_dsp._r)));
#endif
}
}
}
// MRR $D, $S
@ -415,10 +425,10 @@ void DSPEmitter::setCompileSR(u16 bit) {
// g_dsp.r[DSP_REG_SR] |= bit
#ifdef _M_IX86 // All32
OR(16, M(&g_dsp.r[DSP_REG_SR]), Imm16(bit));
OR(16, M(&g_dsp._r.sr), Imm16(bit));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
OR(16, MDisp(R11,DSP_REG_SR*2), Imm16(bit));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
OR(16, MDisp(R11,STRUCT_OFFSET(g_dsp._r, sr)), Imm16(bit));
#endif
compileSR |= bit;
@ -428,10 +438,10 @@ void DSPEmitter::clrCompileSR(u16 bit) {
// g_dsp.r[DSP_REG_SR] &= bit
#ifdef _M_IX86 // All32
AND(16, M(&g_dsp.r[DSP_REG_SR]), Imm16(~bit));
AND(16, M(&g_dsp._r.sr), Imm16(~bit));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
AND(16, MDisp(R11,DSP_REG_SR*2), Imm16(~bit));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
AND(16, MDisp(R11,STRUCT_OFFSET(g_dsp._r, sr)), Imm16(~bit));
#endif
compileSR &= ~bit;

127
Source/Core/DSPCore/Src/Jit/DSPJitMultiplier.cpp
View File

@ -23,6 +23,7 @@
#include "../DSPIntUtil.h"
#include "../DSPEmitter.h"
#include "../DSPAnalyzer.h"
#include "DSPJitUtil.h"
#include "x64Emitter.h"
#include "ABI.h"
using namespace Gen;
@ -37,8 +38,8 @@ void DSPEmitter::multiply()
IMUL(64, R(ESI));
// Conditionally multiply by 2.
// if ((g_dsp.r[DSP_REG_SR] & SR_MUL_MODIFY) == 0)
TEST(16, MDisp(R11, DSP_REG_SR * 2), Imm16(SR_MUL_MODIFY));
// if ((g_dsp._r.sr & SR_MUL_MODIFY) == 0)
TEST(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, sr)), Imm16(SR_MUL_MODIFY));
FixupBranch noMult2 = J_CC(CC_NZ);
// prod <<= 1;
SHL(64, R(EAX), Imm8(1));
@ -89,8 +90,8 @@ void DSPEmitter::multiply_mulx(u8 axh0, u8 axh1)
// result = dsp_multiply(val1, val2, 0); // unsigned support OFF if both ax?.h regs are used
// if ((sign == 1) && (g_dsp.r[DSP_REG_SR] & SR_MUL_UNSIGNED)) //unsigned
TEST(16, MDisp(R11, DSP_REG_SR * 2), Imm16(SR_MUL_UNSIGNED));
// if ((sign == 1) && (g_dsp._r.sr & SR_MUL_UNSIGNED)) //unsigned
TEST(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, sr)), Imm16(SR_MUL_UNSIGNED));
FixupBranch unsignedMul = J_CC(CC_NZ);
// prod = (s16)a * (s16)b; //signed
MOVSX(64, 16, RAX, R(RDI));
@ -131,8 +132,8 @@ void DSPEmitter::multiply_mulx(u8 axh0, u8 axh1)
SetJumpTarget(signedMul);
// Conditionally multiply by 2.
// if ((g_dsp.r[DSP_REG_SR] & SR_MUL_MODIFY) == 0)
TEST(16, MDisp(R11, DSP_REG_SR * 2), Imm16(SR_MUL_MODIFY));
// if ((g_dsp._r.sr & SR_MUL_MODIFY) == 0)
TEST(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, sr)), Imm16(SR_MUL_MODIFY));
FixupBranch noMult2 = J_CC(CC_NZ);
// prod <<= 1;
SHL(64, R(RAX), Imm8(1));
@ -155,14 +156,14 @@ void DSPEmitter::clrp(const UDSPInstruction opc)
{
#ifdef _M_X64
// g_dsp.r[DSP_REG_PRODL] = 0x0000;
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, MDisp(R11, DSP_REG_PRODL * 2), Imm16(0x0000));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, prod.l)), Imm16(0x0000));
// g_dsp.r[DSP_REG_PRODM] = 0xfff0;
MOV(16, MDisp(R11, DSP_REG_PRODM * 2), Imm16(0xfff0));
MOV(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, prod.m)), Imm16(0xfff0));
// g_dsp.r[DSP_REG_PRODH] = 0x00ff;
MOV(16, MDisp(R11, DSP_REG_PRODH * 2), Imm16(0x00ff));
MOV(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, prod.h)), Imm16(0x00ff));
// g_dsp.r[DSP_REG_PRODM2] = 0x0010;
MOV(16, MDisp(R11, DSP_REG_PRODM2 * 2), Imm16(0x0010));
MOV(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, prod.m2)), Imm16(0x0010));
#else
Default(opc);
#endif
@ -316,8 +317,8 @@ void DSPEmitter::mulaxh(const UDSPInstruction opc)
{
#ifdef _M_X64
// s64 prod = dsp_multiply(dsp_get_ax_h(0), dsp_get_ax_h(0));
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVSX(64, 16, RSI, MDisp(R11, DSP_REG_AXH0 * 2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVSX(64, 16, RSI, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ax[0].h)));
MOV(64, R(RDI), R(RSI));
multiply();
// dsp_set_long_prod(prod);
@ -339,10 +340,10 @@ void DSPEmitter::mul(const UDSPInstruction opc)
u8 sreg = (opc >> 11) & 0x1;
// u16 axl = dsp_get_ax_l(sreg);
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVSX(64, 16, RSI, MDisp(R11, (DSP_REG_AXL0 + sreg) * 2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVSX(64, 16, RSI, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ax[sreg].l)));
// u16 axh = dsp_get_ax_h(sreg);
MOVSX(64, 16, RDI, MDisp(R11, (DSP_REG_AXH0 + sreg) * 2));
MOVSX(64, 16, RDI, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ax[sreg].h)));
// s64 prod = dsp_multiply(axh, axl);
multiply();
// dsp_set_long_prod(prod);
@ -372,9 +373,9 @@ void DSPEmitter::mulac(const UDSPInstruction opc)
ADD(64, R(RAX), R(RDX));
PUSH(64, R(RAX));
// u16 axl = dsp_get_ax_l(sreg);
MOVSX(64, 16, RSI, MDisp(R11, (DSP_REG_AXL0 + sreg) * 2));
MOVSX(64, 16, RSI, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ax[sreg].l)));
// u16 axh = dsp_get_ax_h(sreg);
MOVSX(64, 16, RDI, MDisp(R11, (DSP_REG_AXH0 + sreg) * 2));
MOVSX(64, 16, RDI, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ax[sreg].h)));
// s64 prod = dsp_multiply(axl, axh);
multiply();
// dsp_set_long_prod(prod);
@ -460,12 +461,14 @@ void DSPEmitter::mulx(const UDSPInstruction opc)
#ifdef _M_X64
u8 treg = ((opc >> 11) & 0x1);
u8 sreg = ((opc >> 12) & 0x1);
u16 *sregp = reg_ptr(DSP_REG_AXL0 + sreg*2);
u16 *tregp = reg_ptr(DSP_REG_AXL1 + treg*2);
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
// u16 val1 = (sreg == 0) ? dsp_get_ax_l(0) : dsp_get_ax_h(0);
MOVSX(64, 16, RSI, MDisp(R11, (DSP_REG_AXL0 + sreg*2) * 2));
MOVSX(64, 16, RSI, MDisp(R11, PtrOffset(sregp,&g_dsp._r)));
// u16 val2 = (treg == 0) ? dsp_get_ax_l(1) : dsp_get_ax_h(1);
MOVSX(64, 16, RDI, MDisp(R11, (DSP_REG_AXL1 + treg*2) * 2));
MOVSX(64, 16, RDI, MDisp(R11, PtrOffset(tregp,&g_dsp._r)));
// s64 prod = dsp_multiply_mulx(sreg, treg, val1, val2);
multiply_mulx(sreg, treg);
// dsp_set_long_prod(prod);
@ -488,15 +491,17 @@ void DSPEmitter::mulxac(const UDSPInstruction opc)
u8 rreg = (opc >> 8) & 0x1;
u8 treg = (opc >> 11) & 0x1;
u8 sreg = (opc >> 12) & 0x1;
u16 *sregp = reg_ptr(DSP_REG_AXL0 + sreg*2);
u16 *tregp = reg_ptr(DSP_REG_AXL1 + treg*2);
// s64 acc = dsp_get_long_acc(rreg) + dsp_get_long_prod();
get_long_acc(rreg, RCX);
get_long_prod();
ADD(64, R(RCX), R(RAX));
// u16 val1 = (sreg == 0) ? dsp_get_ax_l(0) : dsp_get_ax_h(0);
MOVSX(64, 16, RSI, MDisp(R11, (DSP_REG_AXL0 + sreg*2) * 2));
MOVSX(64, 16, RSI, MDisp(R11, PtrOffset(sregp, &g_dsp._r)));
// u16 val2 = (treg == 0) ? dsp_get_ax_l(1) : dsp_get_ax_h(1);
MOVSX(64, 16, RDI, MDisp(R11, (DSP_REG_AXL1 + treg*2) * 2));
MOVSX(64, 16, RDI, MDisp(R11, PtrOffset(tregp, &g_dsp._r)));
// s64 prod = dsp_multiply_mulx(sreg, treg, val1, val2);
multiply_mulx(sreg, treg);
@ -527,13 +532,15 @@ void DSPEmitter::mulxmv(const UDSPInstruction opc)
u8 rreg = ((opc >> 8) & 0x1);
u8 treg = (opc >> 11) & 0x1;
u8 sreg = (opc >> 12) & 0x1;
u16 *sregp = reg_ptr(DSP_REG_AXL0 + sreg*2);
u16 *tregp = reg_ptr(DSP_REG_AXL1 + treg*2);
// s64 acc = dsp_get_long_prod();
get_long_prod(RCX);
// u16 val1 = (sreg == 0) ? dsp_get_ax_l(0) : dsp_get_ax_h(0);
MOVSX(64, 16, RSI, MDisp(R11, (DSP_REG_AXL0 + sreg*2) * 2));
MOVSX(64, 16, RSI, MDisp(R11, PtrOffset(sregp, &g_dsp._r)));
// u16 val2 = (treg == 0) ? dsp_get_ax_l(1) : dsp_get_ax_h(1);
MOVSX(64, 16, RDI, MDisp(R11, (DSP_REG_AXL1 + treg*2) * 2));
MOVSX(64, 16, RDI, MDisp(R11, PtrOffset(tregp, &g_dsp._r)));
// s64 prod = dsp_multiply_mulx(sreg, treg, val1, val2);
multiply_mulx(sreg, treg);
@ -565,13 +572,15 @@ void DSPEmitter::mulxmvz(const UDSPInstruction opc)
u8 rreg = (opc >> 8) & 0x1;
u8 treg = (opc >> 11) & 0x1;
u8 sreg = (opc >> 12) & 0x1;
u16 *sregp = reg_ptr(DSP_REG_AXL0 + sreg*2);
u16 *tregp = reg_ptr(DSP_REG_AXL1 + treg*2);
// s64 acc = dsp_get_long_prod_round_prodl();
get_long_prod_round_prodl(RCX);
// u16 val1 = (sreg == 0) ? dsp_get_ax_l(0) : dsp_get_ax_h(0);
MOVSX(64, 16, RSI, MDisp(R11, (DSP_REG_AXL0 + sreg*2) * 2));
MOVSX(64, 16, RSI, MDisp(R11, PtrOffset(sregp, &g_dsp._r)));
// u16 val2 = (treg == 0) ? dsp_get_ax_l(1) : dsp_get_ax_h(1);
MOVSX(64, 16, RDI, MDisp(R11, (DSP_REG_AXL1 + treg*2) * 2));
MOVSX(64, 16, RDI, MDisp(R11, PtrOffset(tregp, &g_dsp._r)));
// s64 prod = dsp_multiply_mulx(sreg, treg, val1, val2);
multiply_mulx(sreg, treg);
@ -602,10 +611,10 @@ void DSPEmitter::mulc(const UDSPInstruction opc)
u8 sreg = (opc >> 12) & 0x1;
// u16 accm = dsp_get_acc_m(sreg);
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVSX(64, 16, ESI, MDisp(R11, (DSP_REG_ACM0 + sreg) * 2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVSX(64, 16, ESI, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ac[sreg].m)));
// u16 axh = dsp_get_ax_h(treg);
MOVSX(64, 16, EDI, MDisp(R11, (DSP_REG_AXH0 + treg) * 2));
MOVSX(64, 16, EDI, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ax[treg].h)));
// s64 prod = dsp_multiply(accm, axh);
multiply();
// dsp_set_long_prod(prod);
@ -636,9 +645,9 @@ void DSPEmitter::mulcac(const UDSPInstruction opc)
ADD(64, R(RAX), R(RDX));
PUSH(64, R(RAX));
// u16 accm = dsp_get_acc_m(sreg);
MOVSX(64, 16, RSI, MDisp(R11, (DSP_REG_ACM0 + sreg) * 2));
MOVSX(64, 16, RSI, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ac[sreg].m)));
// u16 axh = dsp_get_ax_h(treg);
MOVSX(64, 16, RDI, MDisp(R11, (DSP_REG_AXH0 + treg) * 2));
MOVSX(64, 16, RDI, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ax[treg].h)));
// s64 prod = dsp_multiply(accm, axh);
multiply();
// dsp_set_long_prod(prod);
@ -675,9 +684,9 @@ void DSPEmitter::mulcmv(const UDSPInstruction opc)
get_long_prod();
PUSH(64, R(RAX));
// u16 accm = dsp_get_acc_m(sreg);
MOVSX(64, 16, RSI, MDisp(R11, (DSP_REG_ACM0 + sreg) * 2));
MOVSX(64, 16, RSI, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ac[sreg].m)));
// u16 axh = dsp_get_ax_h(treg);
MOVSX(64, 16, RDI, MDisp(R11, (DSP_REG_AXH0 + treg) * 2));
MOVSX(64, 16, RDI, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ax[treg].h)));
// s64 prod = dsp_multiply(accm, axh);
multiply();
// dsp_set_long_prod(prod);
@ -711,14 +720,14 @@ void DSPEmitter::mulcmvz(const UDSPInstruction opc)
u8 treg = (opc >> 11) & 0x1;
u8 sreg = (opc >> 12) & 0x1;
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
// s64 acc = dsp_get_long_prod_round_prodl();
get_long_prod_round_prodl();
PUSH(64, R(RAX));
// u16 accm = dsp_get_acc_m(sreg);
MOVSX(64, 16, RSI, MDisp(R11, (DSP_REG_ACM0 + sreg) * 2));
MOVSX(64, 16, RSI, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ac[sreg].m)));
// u16 axh = dsp_get_ax_h(treg);
MOVSX(64, 16, RDI, MDisp(R11, (DSP_REG_AXH0 + treg) * 2));
MOVSX(64, 16, RDI, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ax[treg].h)));
// s64 prod = dsp_multiply(accm, axh);
multiply();
// dsp_set_long_prod(prod);
@ -748,12 +757,14 @@ void DSPEmitter::maddx(const UDSPInstruction opc)
#ifdef _M_X64
u8 treg = (opc >> 8) & 0x1;
u8 sreg = (opc >> 9) & 0x1;
u16 *sregp = reg_ptr(DSP_REG_AXL0 + sreg*2);
u16 *tregp = reg_ptr(DSP_REG_AXL1 + treg*2);
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
// u16 val1 = (sreg == 0) ? dsp_get_ax_l(0) : dsp_get_ax_h(0);
MOVSX(64, 16, RSI, MDisp(R11, (DSP_REG_AXL0 + sreg*2) * 2));
MOVSX(64, 16, RSI, MDisp(R11, PtrOffset(sregp, &g_dsp._r)));
// u16 val2 = (treg == 0) ? dsp_get_ax_l(1) : dsp_get_ax_h(1);
MOVSX(64, 16, RDI, MDisp(R11, (DSP_REG_AXL1 + treg*2) * 2));
MOVSX(64, 16, RDI, MDisp(R11, PtrOffset(tregp, &g_dsp._r)));
// s64 prod = dsp_multiply_add(val1, val2);
multiply_add();
// dsp_set_long_prod(prod);
@ -773,12 +784,14 @@ void DSPEmitter::msubx(const UDSPInstruction opc)
#ifdef _M_X64
u8 treg = (opc >> 8) & 0x1;
u8 sreg = (opc >> 9) & 0x1;
u16 *sregp = reg_ptr(DSP_REG_AXL0 + sreg*2);
u16 *tregp = reg_ptr(DSP_REG_AXL1 + treg*2);
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
// u16 val1 = (sreg == 0) ? dsp_get_ax_l(0) : dsp_get_ax_h(0);
MOVSX(64, 16, RSI, MDisp(R11, (DSP_REG_AXL0 + sreg*2) * 2));
MOVSX(64, 16, RSI, MDisp(R11, PtrOffset(sregp, &g_dsp._r)));
// u16 val2 = (treg == 0) ? dsp_get_ax_l(1) : dsp_get_ax_h(1);
MOVSX(64, 16, RDI, MDisp(R11, (DSP_REG_AXL1 + treg*2) * 2));
MOVSX(64, 16, RDI, MDisp(R11, PtrOffset(tregp, &g_dsp._r)));
// s64 prod = dsp_multiply_sub(val1, val2);
multiply_sub();
// dsp_set_long_prod(prod);
@ -799,11 +812,11 @@ void DSPEmitter::maddc(const UDSPInstruction opc)
u8 treg = (opc >> 8) & 0x1;
u8 sreg = (opc >> 9) & 0x1;
MOV(64, R(R11), ImmPtr(&g_dsp.r));
// u16 accm = dsp_get_acc_m(sreg);
MOVSX(64, 16, RSI, MDisp(R11, (DSP_REG_ACM0 + sreg) * 2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
// u16 accm = dsp_get_acc_m(sreg);
MOVSX(64, 16, RSI, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ac[sreg].m)));
// u16 axh = dsp_get_ax_h(treg);
MOVSX(64, 16, RDI, MDisp(R11, (DSP_REG_AXH0 + treg) * 2));
MOVSX(64, 16, RDI, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ax[treg].h)));
// s64 prod = dsp_multiply_add(accm, axh);
multiply_add();
// dsp_set_long_prod(prod);
@ -823,12 +836,12 @@ void DSPEmitter::msubc(const UDSPInstruction opc)
#ifdef _M_X64
u8 treg = (opc >> 8) & 0x1;
u8 sreg = (opc >> 9) & 0x1;
//
// u16 accm = dsp_get_acc_m(sreg);
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVSX(64, 16, RSI, MDisp(R11, (DSP_REG_ACM0 + sreg) * 2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVSX(64, 16, RSI, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ac[sreg].m)));
// u16 axh = dsp_get_ax_h(treg);
MOVSX(64, 16, RDI, MDisp(R11, (DSP_REG_AXH0 + treg) * 2));
MOVSX(64, 16, RDI, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ax[treg].h)));
// s64 prod = dsp_multiply_sub(accm, axh);
multiply_sub();
// dsp_set_long_prod(prod);
@ -848,11 +861,11 @@ void DSPEmitter::madd(const UDSPInstruction opc)
#ifdef _M_X64
u8 sreg = (opc >> 8) & 0x1;
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
// u16 axl = dsp_get_ax_l(sreg);
MOVSX(64, 16, RSI, MDisp(R11, (DSP_REG_AXL0 + sreg) * 2));
MOVSX(64, 16, RSI, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ax[sreg].l)));
// u16 axh = dsp_get_ax_h(sreg);
MOVSX(64, 16, RDI, MDisp(R11, (DSP_REG_AXH0 + sreg) * 2));
MOVSX(64, 16, RDI, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ax[sreg].h)));
// s64 prod = dsp_multiply_add(axl, axh);
multiply_add();
// dsp_set_long_prod(prod);
@ -873,10 +886,10 @@ void DSPEmitter::msub(const UDSPInstruction opc)
u8 sreg = (opc >> 8) & 0x1;
//
// u16 axl = dsp_get_ax_l(sreg);
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVSX(64, 16, RSI, MDisp(R11, (DSP_REG_AXL0 + sreg) * 2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVSX(64, 16, RSI, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ax[sreg].l)));
// u16 axh = dsp_get_ax_h(sreg);
MOVSX(64, 16, RDI, MDisp(R11, (DSP_REG_AXH0 + sreg) * 2));
MOVSX(64, 16, RDI, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ax[sreg].h)));
// s64 prod = dsp_multiply_sub(axl, axh);
multiply_sub();
// dsp_set_long_prod(prod);

184
Source/Core/DSPCore/Src/Jit/DSPJitUtil.cpp
View File

@ -15,9 +15,6 @@
// Official SVN repository and contact information can be found at
// http://code.google.com/p/dolphin-emu/
#ifndef _DSP_JIT_UTIL_H
#define _DSP_JIT_UTIL_H
#include "../DSPMemoryMap.h"
#include "../DSPHWInterface.h"
#include "../DSPEmitter.h"
@ -30,6 +27,10 @@ using namespace Gen;
// clobbers RCX
void DSPEmitter::ToMask(X64Reg value_reg)
{
//microbenchmarking results(1 000 000 000 iterations):
// cpu\variant| 1 | 2
//intel mobile C2D@2.5GHz | 5.5s | 4.0s
//amd athlon64x2@3GHz | 6.1s | 6.4s
#if 0
MOV(16, R(CX), R(value_reg));
SHR(16, R(CX), Imm8(8));
@ -77,12 +78,12 @@ void DSPEmitter::increment_addr_reg(int reg)
// s16 tmp = g_dsp.r[reg];
#ifdef _M_IX86 // All32
MOV(16, R(AX), M(&g_dsp.r[reg]));
MOV(16, R(DX), M(&g_dsp.r[DSP_REG_WR0 + reg]));
MOV(16, R(AX), M(&g_dsp._r.ar[reg]));
MOV(16, R(DX), M(&g_dsp._r.wr[reg]));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, R(AX), MDisp(R11,reg*2));
MOV(16, R(DX), MDisp(R11,(DSP_REG_WR0 + reg)*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, R(AX), MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[reg])));
MOV(16, R(DX), MDisp(R11,STRUCT_OFFSET(g_dsp._r, wr[reg])));
#endif
MOV(16,R(DI), R(AX));
@ -102,10 +103,10 @@ void DSPEmitter::increment_addr_reg(int reg)
// g_dsp.r[reg] = tmp;
#ifdef _M_IX86 // All32
MOV(16, M(&g_dsp.r[reg]), R(AX));
MOV(16, M(&g_dsp._r.ar[reg]), R(AX));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, MDisp(R11,reg*2), R(AX));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[reg])), R(AX));
#endif
}
@ -126,12 +127,12 @@ void DSPEmitter::decrement_addr_reg(int reg)
// s16 ar = g_dsp.r[reg];
#ifdef _M_IX86 // All32
MOV(16, R(AX), M(&g_dsp.r[reg]));
MOVZX(32, 16, EDX, M(&g_dsp.r[DSP_REG_WR0 + reg]));
MOV(16, R(AX), M(&g_dsp._r.ar[reg]));
MOVZX(32, 16, EDX, M(&g_dsp._r.wr[reg]));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, R(AX), MDisp(R11,reg*2));
MOVZX(32, 16, EDX, MDisp(R11,(DSP_REG_WR0 + reg)*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, R(AX), MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[reg])));
MOVZX(32, 16, EDX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, wr[reg])));
#endif
// ToMask(WR0), calculating it into EDI
@ -158,10 +159,10 @@ void DSPEmitter::decrement_addr_reg(int reg)
// g_dsp.r[reg] = tmp;
#ifdef _M_IX86 // All32
MOV(16, M(&g_dsp.r[reg]), R(AX));
MOV(16, M(&g_dsp._r.ar[reg]), R(AX));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, MDisp(R11,reg*2), R(AX));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[reg])), R(AX));
#endif
}
@ -189,14 +190,14 @@ void DSPEmitter::increase_addr_reg(int reg)
*/
#ifdef _M_IX86 // All32
MOV(16, R(SI), M(&g_dsp.r[DSP_REG_IX0 + reg]));
MOV(16, R(AX), M(&g_dsp.r[reg]));
MOVZX(32, 16, EDX, M(&g_dsp.r[DSP_REG_WR0 + reg]));
MOV(16, R(SI), M(&g_dsp._r.ix[reg]));
MOV(16, R(AX), M(&g_dsp._r.ar[reg]));
MOVZX(32, 16, EDX, M(&g_dsp._r.wr[reg]));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, R(SI), MDisp(R11,(DSP_REG_IX0 + reg)*2));
MOV(16, R(AX), MDisp(R11,reg*2));
MOVZX(32, 16, EDX, MDisp(R11,(DSP_REG_WR0 + reg)*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, R(SI), MDisp(R11,STRUCT_OFFSET(g_dsp._r, ix[reg])));
MOV(16, R(AX), MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[reg])));
MOVZX(32, 16, EDX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, wr[reg])));
#endif
// ToMask(WR0), calculating it into EDI
@ -239,10 +240,10 @@ void DSPEmitter::increase_addr_reg(int reg)
// g_dsp.r[reg] = tmp;
#ifdef _M_IX86 // All32
MOV(16, M(&g_dsp.r[reg]), R(EAX));
MOV(16, M(&g_dsp._r.ar[reg]), R(EAX));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, MDisp(R11,reg*2), R(EAX));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[reg])), R(EAX));
#endif
}
@ -270,14 +271,14 @@ void DSPEmitter::decrease_addr_reg(int reg)
*/
#ifdef _M_IX86 // All32
MOV(16, R(SI), M(&g_dsp.r[DSP_REG_IX0 + reg]));
MOV(16, R(AX), M(&g_dsp.r[reg]));
MOVZX(32, 16, EDX, M(&g_dsp.r[DSP_REG_WR0 + reg]));
MOV(16, R(SI), M(&g_dsp._r.ix[reg]));
MOV(16, R(AX), M(&g_dsp._r.ar[reg]));
MOVZX(32, 16, EDX, M(&g_dsp._r.wr[reg]));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, R(SI), MDisp(R11,(DSP_REG_IX0 + reg)*2));
MOV(16, R(AX), MDisp(R11,reg*2));
MOVZX(32, 16, EDX, MDisp(R11,(DSP_REG_WR0 + reg)*2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, R(SI), MDisp(R11,STRUCT_OFFSET(g_dsp._r, ix[reg])));
MOV(16, R(AX), MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[reg])));
MOVZX(32, 16, EDX, MDisp(R11,STRUCT_OFFSET(g_dsp._r, wr[reg])));
#endif
// ToMask(WR0), calculating it into EDI
@ -318,10 +319,10 @@ void DSPEmitter::decrease_addr_reg(int reg)
// g_dsp.r[reg] = tmp;
#ifdef _M_IX86 // All32
MOV(16, M(&g_dsp.r[reg]), R(EAX));
MOV(16, M(&g_dsp._r.ar[reg]), R(EAX));
#else
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, MDisp(R11,reg*2), R(EAX));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, MDisp(R11,STRUCT_OFFSET(g_dsp._r, ar[reg])), R(EAX));
#endif
}
@ -492,20 +493,34 @@ void DSPEmitter::dmem_read_imm(u16 address)
void DSPEmitter::get_long_prod(X64Reg long_prod)
{
#ifdef _M_X64
MOV(64, R(R11), ImmPtr(&g_dsp.r));
//s64 val = (s8)(u8)g_dsp.r[DSP_REG_PRODH];
MOVSX(64, 8, long_prod, MDisp(R11,DSP_REG_PRODH*2));
#if 0
MOV(64, R(R11), ImmPtr(&g_dsp._r));
//s64 val = (s8)(u8)g_dsp.r[DSP_REG_PRODH];
MOVSX(64, 8, long_prod, MDisp(R11,STRUCT_OFFSET(g_dsp._r, prod.h)));
//val <<= 32;
SHL(64, R(long_prod), Imm8(16));
//s64 low_prod = g_dsp.r[DSP_REG_PRODM];
OR(16, R(long_prod), MDisp(R11,DSP_REG_PRODM*2));
OR(16, R(long_prod), MDisp(R11,STRUCT_OFFSET(g_dsp._r, prod.m)));
//low_prod += g_dsp.r[DSP_REG_PRODM2];
ADD(16, R(long_prod), MDisp(R11,DSP_REG_PRODM2*2));
ADD(16, R(long_prod), MDisp(R11,STRUCT_OFFSET(g_dsp._r, prod.m2)));
//low_prod <<= 16;
SHL(64, R(long_prod), Imm8(16));
//low_prod |= g_dsp.r[DSP_REG_PRODL];
OR(16, R(long_prod), MDisp(R11,DSP_REG_PRODL*2));
OR(16, R(long_prod), MDisp(R11,STRUCT_OFFSET(g_dsp._r, prod.l)));
//return val;
#else
MOV(64, R(R11), ImmPtr(&g_dsp._r));
//s64 val = (s8)(u8)g_dsp.r[DSP_REG_PRODH];
MOV(64, R(long_prod), MDisp(R11,STRUCT_OFFSET(g_dsp._r, prod.val)));
MOV(64, R(R11), R(long_prod));
SHL(64, R(long_prod), Imm8(64-40));//sign extend
SAR(64, R(long_prod), Imm8(64-40));
SHR(64, R(R11), Imm8(48));
SHL(64, R(R11), Imm8(16));
ADD(64, R(long_prod), R(R11));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
#endif
#endif
}
@ -540,20 +555,28 @@ void DSPEmitter::get_long_prod_round_prodl(X64Reg long_prod)
void DSPEmitter::set_long_prod()
{
#ifdef _M_X64
#if 0
// g_dsp.r[DSP_REG_PRODL] = (u16)val;
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, MDisp(R11, DSP_REG_PRODL * 2), R(AX));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, prod.l)), R(AX));
// val >>= 16;
SAR(64, R(RAX), Imm8(16));
// g_dsp.r[DSP_REG_PRODM] = (u16)val;
MOV(16, MDisp(R11, DSP_REG_PRODM * 2), R(AX));
MOV(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, prod.m)), R(AX));
// val >>= 16;
SAR(64, R(RAX), Imm8(16));
// g_dsp.r[DSP_REG_PRODH] = (u8)val;
MOVSX(64, 8, RAX, R(AL));
MOV(8, MDisp(R11, DSP_REG_PRODH * 2), R(AL));
MOV(8, MDisp(R11, STRUCT_OFFSET(g_dsp._r, prod.h)), R(AL));
// g_dsp.r[DSP_REG_PRODM2] = 0;
MOV(16, MDisp(R11, DSP_REG_PRODM2 * 2), Imm16(0));
MOV(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, prod.m2)), Imm16(0));
#else
MOV(64, R(R11), Imm64(0x000000ffffffffffULL));
AND(64, R(RAX), R(R11));
// g_dsp.r[DSP_REG_PRODL] = (u16)val;
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(64, MDisp(R11, STRUCT_OFFSET(g_dsp._r, prod.val)), R(RAX));
#endif
#endif
}
@ -583,15 +606,23 @@ void DSPEmitter::round_long_acc(X64Reg long_acc)
void DSPEmitter::get_long_acc(int _reg, X64Reg acc)
{
#ifdef _M_X64
#if 0
// s64 high = (s64)(s8)g_dsp.r[DSP_REG_ACH0 + reg] << 32;
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVSX(64, 8, acc, MDisp(R11, (DSP_REG_ACH0 + _reg) * 2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVSX(64, 8, acc, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ac[_reg].h)));
SHL(64, R(acc), Imm8(16));
// u32 mid_low = ((u32)g_dsp.r[DSP_REG_ACM0 + reg] << 16) | g_dsp.r[DSP_REG_ACL0 + reg];
OR(16, R(acc), MDisp(R11, (DSP_REG_ACM0 + _reg) * 2));
OR(16, R(acc), MDisp(R11, STRUCT_OFFSET(g_dsp._r, ac[_reg].m)));
SHL(64, R(acc), Imm8(16));
OR(16, R(acc), MDisp(R11, (DSP_REG_ACL0 + _reg) * 2));
OR(16, R(acc), MDisp(R11, STRUCT_OFFSET(g_dsp._r, ac[_reg].l)));
// return high | mid_low;
#else
// s64 high = (s64)(s8)g_dsp.r[DSP_REG_ACH0 + reg] << 32;
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(64, R(acc), MDisp(R11, STRUCT_OFFSET(g_dsp._r, ac[_reg].val)));
SHL(64, R(acc), Imm8(64-40));//sign extend
SAR(64, R(acc), Imm8(64-40));
#endif
#endif
}
@ -600,18 +631,26 @@ void DSPEmitter::get_long_acc(int _reg, X64Reg acc)
void DSPEmitter::set_long_acc(int _reg, X64Reg acc)
{
#ifdef _M_X64
#if 0
// g_dsp.r[DSP_REG_ACL0 + _reg] = (u16)val;
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, MDisp(R11, (DSP_REG_ACL0 + _reg) * 2), R(acc));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ac[_reg].l)), R(acc));
// val >>= 16;
SHR(64, R(acc), Imm8(16));
// g_dsp.r[DSP_REG_ACM0 + _reg] = (u16)val;
MOV(16, MDisp(R11, (DSP_REG_ACM0 + _reg) * 2), R(acc));
MOV(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ac[_reg].m)), R(acc));
// val >>= 16;
SHR(64, R(acc), Imm8(16));
// g_dsp.r[DSP_REG_ACH0 + _reg] = (u16)(s16)(s8)(u8)val;
MOVSX(64, 8, acc, R(acc));
MOV(16, MDisp(R11, (DSP_REG_ACH0 + _reg) * 2), R(acc));
MOV(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ac[_reg].h)), R(acc));
#else
SHL(64, R(acc), Imm8(64-40));//sign extend
SAR(64, R(acc), Imm8(64-40));
// g_dsp.r[DSP_REG_ACL0 + _reg] = (u16)val;
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(64, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ac[_reg].val)), R(acc));
#endif
#endif
}
@ -620,8 +659,8 @@ void DSPEmitter::get_acc_m(int _reg, X64Reg acm)
{
// return g_dsp.r[DSP_REG_ACM0 + _reg];
#ifdef _M_X64
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVSX(64, 16, acm, MDisp(R11, (DSP_REG_ACM0 + _reg) * 2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVSX(64, 16, acm, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ac[_reg].m)));
#endif
}
@ -630,8 +669,8 @@ void DSPEmitter::set_acc_m(int _reg)
{
// return g_dsp.r[DSP_REG_ACM0 + _reg];
#ifdef _M_X64
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOV(16, MDisp(R11, (DSP_REG_ACM0 + _reg) * 2), R(RAX));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOV(16, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ac[_reg].m)), R(RAX));
#endif
}
@ -640,10 +679,15 @@ void DSPEmitter::get_long_acx(int _reg, X64Reg acx)
{
// return ((u32)g_dsp.r[DSP_REG_AXH0 + _reg] << 16) | g_dsp.r[DSP_REG_AXL0 + _reg];
#ifdef _M_X64
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVSX(64, 16, acx, MDisp(R11, (DSP_REG_AXH0 + _reg) * 2));
#if 0
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVSX(64, 16, acx, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ax[_reg].h)));
SHL(64, R(acx), Imm8(16));
OR(16, R(acx), MDisp(R11, (DSP_REG_AXL0 + _reg) * 2));
OR(16, R(acx), MDisp(R11, STRUCT_OFFSET(g_dsp._r, ax[_reg].l)));
#else
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVSX(64, 32, acx, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ax[_reg].val)));
#endif
#endif
}
@ -652,8 +696,8 @@ void DSPEmitter::get_ax_l(int _reg, X64Reg axl)
{
// return (s16)g_dsp.r[DSP_REG_AXL0 + _reg];
#ifdef _M_X64
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVSX(64, 16, axl, MDisp(R11, (DSP_REG_AXL0 + _reg) * 2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVSX(64, 16, axl, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ax[_reg].l)));
#endif
}
@ -662,9 +706,7 @@ void DSPEmitter::get_ax_h(int _reg, X64Reg axh)
{
// return (s16)g_dsp.r[DSP_REG_AXH0 + _reg];
#ifdef _M_X64
MOV(64, R(R11), ImmPtr(&g_dsp.r));
MOVSX(64, 16, axh, MDisp(R11, (DSP_REG_AXH0 + _reg) * 2));
MOV(64, R(R11), ImmPtr(&g_dsp._r));
MOVSX(64, 16, axh, MDisp(R11, STRUCT_OFFSET(g_dsp._r, ax[_reg].h)));
#endif
}
#endif

74
Source/Core/DSPCore/Src/Jit/DSPJitUtil.h Normal file
View File

@ -0,0 +1,74 @@
// Copyright (C) 2010 Dolphin Project.
// This program is free software: you can redistribute it and/or modify
// it under the terms of the GNU General Public License as published by
// the Free Software Foundation, version 2.0.
// This program is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU General Public License 2.0 for more details.
// A copy of the GPL 2.0 should have been included with the program.
// If not, see http://www.gnu.org/licenses/
// Official SVN repository and contact information can be found at
// http://code.google.com/p/dolphin-emu/
#ifndef __DSPJITUTIL_H__
#define __DSPJITUTIL_H__
#include "../DSPMemoryMap.h"
#include "../DSPHWInterface.h"
#include "../DSPEmitter.h"
static u16 *reg_ptr(int reg) {
switch(reg) {
case DSP_REG_AR0:
case DSP_REG_AR1:
case DSP_REG_AR2:
case DSP_REG_AR3:
return &g_dsp._r.ar[reg - DSP_REG_AR0];
case DSP_REG_IX0:
case DSP_REG_IX1:
case DSP_REG_IX2:
case DSP_REG_IX3:
return &g_dsp._r.ix[reg - DSP_REG_IX0];
case DSP_REG_WR0:
case DSP_REG_WR1:
case DSP_REG_WR2:
case DSP_REG_WR3:
return &g_dsp._r.wr[reg - DSP_REG_WR0];
case DSP_REG_ST0:
case DSP_REG_ST1:
case DSP_REG_ST2:
case DSP_REG_ST3:
return &g_dsp._r.st[reg - DSP_REG_ST0];
case DSP_REG_ACH0:
case DSP_REG_ACH1:
return &g_dsp._r.ac[reg - DSP_REG_ACH0].h;
case DSP_REG_CR: return &g_dsp._r.cr;
case DSP_REG_SR: return &g_dsp._r.sr;
case DSP_REG_PRODL: return &g_dsp._r.prod.l;
case DSP_REG_PRODM: return &g_dsp._r.prod.m;
case DSP_REG_PRODH: return &g_dsp._r.prod.h;
case DSP_REG_PRODM2: return &g_dsp._r.prod.m2;
case DSP_REG_AXL0:
case DSP_REG_AXL1:
return &g_dsp._r.ax[reg - DSP_REG_AXL0].l;
case DSP_REG_AXH0:
case DSP_REG_AXH1:
return &g_dsp._r.ax[reg - DSP_REG_AXH0].h;
case DSP_REG_ACL0:
case DSP_REG_ACL1:
return &g_dsp._r.ac[reg - DSP_REG_ACL0].l;
case DSP_REG_ACM0:
case DSP_REG_ACM1:
return &g_dsp._r.ac[reg - DSP_REG_ACM0].m;
default:
_assert_msg_(DSP_JIT, 0, "cannot happen");
return NULL;
}
}
#endif /*__DSPJITUTIL_H__*/