valter/dosbox-staging
1
0
Fork 0
Code Releases Activity

A little cleanup and small speed up (jmarsh)

Browse source 
Imported-from: https://svn.code.sf.net/p/dosbox/code-0/dosbox/trunk@4271
This commit is contained in:
Peter Veenstra 2019-10-10 15:54:19 +00:00
parent 195484d5e9
commit 26a9f75b50
2 changed files with 17 additions and 19 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
src/cpu/core_dyn_x86.cpp
View file

@ -170,7 +170,7 @@ static struct {
} core_dyn;
#if defined(X86_DYNFPU_DH_ENABLED)
static struct {
static struct dyn_dh_fpu {
Bit16u cw,host_cw;
bool state_used;
// some fields expanded here for alignment purposes

34
src/cpu/core_dyn_x86/risc_x64.h
View file

@ -18,10 +18,10 @@
#if defined(_WIN64)
enum {
X64_REG_RAX,
X64_REG_RBX,
X64_REG_RCX,
X64_REG_RDX,
X64_REG_RCX,
X64_REG_RAX,
// volatiles
X64_REG_R8,
X64_REG_R9,
@ -305,11 +305,9 @@ static BlockReturn gen_runcodeInit(Bit8u *code) {
*(Bit32u*)diff = (Bit32u)(cache.pos - diff - 4);
// eax = return value, ecx = flags
opcode(2).setea(5,-1,0,offsetof(CPU_Regs,flags)).Emit8(0x8B); // mov edx, [reg_flags(rbp)]
opcode(4).setrm(1).setimm(FMASK_TEST,4).Emit8(0x81); // and ecx,FMASK_TEST
opcode(4).setrm(2).setimm(~FMASK_TEST,4).Emit8(0x81); // and edx,~FMASK_TEST
opcode(1).setrm(2).Emit8(0x0B); // or ecx,edx
opcode(1).setea(5,-1,0,offsetof(CPU_Regs,flags)).Emit8(0x89); // mov [reg_flags(rbp)],ecx
opcode(1).setea(5,-1,0,offsetof(CPU_Regs,flags)).Emit8(0x33); // xor ecx, reg_flags
opcode(4).setrm(1).setimm(FMASK_TEST,4).Emit8(0x81); // and ecx,FMASK_TEST
opcode(1).setea(5,-1,0,offsetof(CPU_Regs,flags)).Emit8(0x31); // xor reg_flags, ecx
opcode(4).set64().setea(4,-1,0,CALLSTACK+8).Emit8(0x8B); // mov rsp, [rsp+8/40]
#if defined(_WIN64)
@ -1201,7 +1199,7 @@ static void gen_return(BlockReturn retcode) {
gen_protectflags();
opcode(1).setea(4,-1,0,CALLSTACK+8).Emit8(0x8B); // mov ecx, [rsp+8/40]
opcode(0).set64().setrm(4).setimm(CALLSTACK+16,1).Emit8(0x83); // add rsp,16/48
if (retcode==0) cache_addw(0xc033); //MOV EAX, 0
if (retcode==0) cache_addw(0xc033); // xor eax,eax
else {
cache_addb(0xb8); //MOV EAX, retcode
cache_addd(retcode);
@ -1214,7 +1212,7 @@ static void gen_return_fast(BlockReturn retcode,bool ret_exception=false) {
opcode(1).setabsaddr(&reg_flags).Emit8(0x8B); // mov ECX, [cpu_regs.flags]
if (!ret_exception) {
opcode(0).set64().setrm(4).setimm(CALLSTACK+16,1).Emit8(0x83); // add rsp,16/48
if (retcode==0) cache_addw(0xc033); //MOV EAX, 0
if (retcode==0) cache_addw(0xc033); // xor eax,eax
else {
cache_addb(0xb8); //MOV EAX, retcode
cache_addd(retcode);
@ -1250,19 +1248,19 @@ static void gen_dh_fpu_saveInit(void) {
cache.pos = &cache_code_link_blocks[64];
gen_dh_fpu_save = (void(*)(void))cache.pos;
Bits addr = (Bits)&dyn_dh_fpu;
Bitu addr = (Bitu)&dyn_dh_fpu;
// mov RAX, &dyn_dh_fpu
if ((Bit32u)addr == addr) opcode(0).setimm(addr,4).Emit8Reg(0xB8);
else opcode(0).set64().setimm(addr,8).Emit8Reg(0xB8);
// fnsave [RAX+offs8]
cache_addw(0x70DD);cache_addb((Bits)&dyn_dh_fpu.state-addr);
// fldcw [RAX+offs8]
cache_addw(0x68D9);cache_addb((Bits)&dyn_dh_fpu.host_cw-addr);
// mov byte [RAX+offs8], 0
cache_addw(0x40C6);cache_addw((Bit8u)((Bits)&dyn_dh_fpu.state_used-addr));
// or byte [RAX+offs8], 0x3F
cache_addw(0x4880);cache_addb((Bits)&dyn_dh_fpu.state.cw-addr);cache_addb(0x3F);
// fnsave [dyn_dh_fpu.state]
opcode(6).setea(0,-1,0,offsetof(struct dyn_dh_fpu,state)).Emit8(0xdd);
// fldcw [dyn_dh_fpu.host_cw]
opcode(5).setea(0,-1,0,offsetof(struct dyn_dh_fpu,host_cw)).Emit8(0xd9);
// mov byte [dyn_dh_fpu.state_used], 0
opcode(0).setimm(0,1).setea(0,-1,0,offsetof(struct dyn_dh_fpu,state_used)).Emit8(0xc6);
// or byte [dyn_dh_fpu.state.cw], 0x3F
opcode(1).setimm(0x3F,1).setea(0,-1,0,offsetof(struct dyn_dh_fpu,state.cw)).Emit8(0x80);
cache_addb(0xC3); // RET
cache.pos = oldpos;