Emulator-Archive/box86
1
0
Fork 0
mirror of https://github.com/ptitSeb/box86.git synced 2024-11-27 08:50:28 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Added USE_FLOAT build option (hopefully a bit faster on the Pandora)

Browse Source
This commit is contained in:
ptitSeb 2019-02-24 13:58:31 +01:00
parent b98412de11
commit fbd494d348
17 changed files with 724 additions and 61 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

36
CMakeLists.txt
View File

@ -9,7 +9,7 @@ option(USE_CCACHE "Set to ON to use ccache if present in the system" ${USE_CCACH
option(LD80BITS "Set to ON if host device have 80bits long double (i.e. i386)" ${LD80BITS})
option(NOALIGN "Set to ON if host device doesn't need re-align (i.e. i386)" ${NOALIGN})
option(HAVE_TRACE "Set to ON to have Trace ability (needs ZydisInfo library)" ${HAVE_TRACE})
option(USE_FLOAT "Set to ON to use only float, no double, in all x87 Emulation" ${USE_FLOAT})
# Pandora
if(PANDORA)
@ -29,6 +29,10 @@ if(HAVE_TRACE)
add_definitions(-DHAVE_TRACE)
endif()
if(USE_FLOAT)
add_definitions(-DUSE_FLOAT)
endif()
add_definitions(-g -std=gnu99 -funwind-tables -O3 -fvisibility=hidden -no-pie -fjump-tables)
if(USE_CCACHE)
@ -115,52 +119,58 @@ set(WRAPPER "${CMAKE_HOME_DIRECTORY}/src/wrapper.c" "${CMAKE_HOME_DIRECTORY}/src
add_custom_command(OUTPUT ${WRAPPER} COMMAND "${CMAKE_HOME_DIRECTORY}/rebuild_wrappers.py" "${CMAKE_HOME_DIRECTORY}" MAIN_DEPENDENCY "${CMAKE_HOME_DIRECTORY}/rebuild_wrappers.py" DEPENDS ${WRAPPEDS})
set_source_files_properties(${WRAPPER} PROPERTIES GENERATED TRUE)
add_executable(box86 ${ELFLOADER_SRC} ${WRAPPEDS})
target_link_libraries(box86 m dl rt pthread)
if(USE_FLOAT)
set(BOX86 box86f)
else()
set(BOX86 box86)
endif()
target_link_libraries(box86 -no-pie -Wl,-Ttext-segment,0xA8000000)
add_executable(${BOX86} ${ELFLOADER_SRC} ${WRAPPEDS})
target_link_libraries(${BOX86} m dl rt pthread)
add_test(test01 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/box86
target_link_libraries(${BOX86} -no-pie -Wl,-Ttext-segment,0xA8000000)
add_test(test01 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86}
-D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test01 -D TEST_OUTPUT=tmpfile.txt
-D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref01.txt
-P ${CMAKE_SOURCE_DIR}/runTest.cmake )
add_test(test02 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/box86
add_test(test02 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86}
-D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test02 -D TEST_OUTPUT=tmpfile.txt
-D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref02.txt
-P ${CMAKE_SOURCE_DIR}/runTest.cmake )
add_test(test03 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/box86
add_test(test03 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86}
-D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test03 -D TEST_OUTPUT=tmpfile.txt
-D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref03.txt
-P ${CMAKE_SOURCE_DIR}/runTest.cmake )
add_test(test04 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/box86
add_test(test04 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86}
-D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test04 -D TEST_ARGS2=yeah -D TEST_OUTPUT=tmpfile.txt
-D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref04.txt
-P ${CMAKE_SOURCE_DIR}/runTest.cmake )
add_test(test05 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/box86
add_test(test05 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86}
-D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test05 -D TEST_ARGS2=7 -D TEST_OUTPUT=tmpfile.txt
-D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref05.txt
-P ${CMAKE_SOURCE_DIR}/runTest.cmake )
add_test(test06 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/box86
add_test(test06 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86}
-D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test06 -D TEST_OUTPUT=tmpfile.txt
-D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref06.txt
-P ${CMAKE_SOURCE_DIR}/runTest.cmake )
add_test(test07 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/box86
add_test(test07 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86}
-D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test07 -D TEST_OUTPUT=tmpfile.txt
-D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref07.txt
-P ${CMAKE_SOURCE_DIR}/runTest.cmake )
add_test(test08 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/box86
add_test(test08 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86}
-D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test08 -D TEST_OUTPUT=tmpfile.txt
-D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref08.txt
-P ${CMAKE_SOURCE_DIR}/runTest.cmake )
add_test(test09 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/box86
add_test(test09 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86}
-D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test09 -D TEST_OUTPUT=tmpfile.txt
-D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref09.txt
-P ${CMAKE_SOURCE_DIR}/runTest.cmake )

14
rebuild_wrappers.py
View File

@ -88,6 +88,12 @@ typedef union ui64_s {
uint32_t d[2];
} ui64_t;
#ifdef USE_FLOAT
#define ST0val ST0.f
#else
#define ST0val ST0.d
#endif
""",
"wrapper.h": """/*****************************************************************
* File automatically generated by rebuild_wrappers.py (v1.0.0.02)
@ -202,10 +208,10 @@ typedef void (*wrapper_t)(x86emu_t* emu, uintptr_t fnc);
"R_EAX=(unsigned short)fn({0});", # W
"R_EAX=(uint32_t)fn({0});", # u
"ui64_t r; r.u=(uint64_t)fn({0}); R_EAX=r.d[0]; R_EDX=r.d[1];", # U
"float fl=fn({0}); fpu_do_push(emu); ST0.d = fl;", # f
"double db=fn({0}); fpu_do_push(emu); ST0.d = db;", # d
"long double ld=fn({0}); fpu_do_push(emu); ST0.d = ld;", # D
"double db=fn({0}); fpu_do_push(emu); ST0.d = db;", # L
"float fl=fn({0}); fpu_do_push(emu); ST0val = fl;", # f
"double db=fn({0}); fpu_do_push(emu); ST0val = db;", # d
"long double ld=fn({0}); fpu_do_push(emu); ST0val = ld;", # D
"double db=fn({0}); fpu_do_push(emu); ST0val = db;", # L
"R_EAX=(uintptr_t)fn({0});", # p
"\n#error Invalid return type: va_list\n", # V
]

8
src/main.c
View File

@ -172,7 +172,13 @@ int main(int argc, const char **argv, const char **env) {
// trying to open and load 1st arg
if(argc==1) {
printf("Box86 v%d.%d.%d\n", BOX86_MAJOR, BOX86_MINOR, BOX86_REVISION);
printf("Box86%s v%d.%d.%d\n",
#ifdef USE_FLOAT
" (float only)",
#else
"",
#endif
BOX86_MAJOR, BOX86_MINOR, BOX86_REVISION);
PrintHelp();
return 1;
}

5
src/regs.h
View File

@ -76,6 +76,10 @@ typedef enum {
#pragma pack(push, 1)
typedef union {
#ifdef USE_FLOAT
float f;
uint32_t ll; // same, to avoid more #ifdef...
#else
double d;
struct {
uint32_t lower;
@ -86,6 +90,7 @@ typedef union {
float upper;
} f;
int64_t ll;
#endif
} fpu_reg_t;
typedef union {

80
src/rund8.h
View File

@ -8,7 +8,11 @@
case 0xC5:
case 0xC6:
case 0xC7: /* FADD */
#ifdef USE_FLOAT
ST0.f += ST(nextop&7).f;
#else
ST0.d += ST(nextop&7).d;
#endif
break;
case 0xC8:
case 0xC9:
@ -18,7 +22,11 @@
case 0xCD:
case 0xCE:
case 0xCF: /* FMUL */
#ifdef USE_FLOAT
ST0.f *= ST(nextop&7).f;
#else
ST0.d *= ST(nextop&7).d;
#endif
break;
case 0xD0:
case 0xD1:
@ -28,7 +36,11 @@
case 0xD5:
case 0xD6:
case 0xD7: /* FCOM */
#ifdef USE_FLOAT
fpu_fcom(emu, ST(nextop&7).f);
#else
fpu_fcom(emu, ST(nextop&7).d);
#endif
break;
case 0xD8:
case 0xD9:
@ -38,7 +50,11 @@
case 0xDD:
case 0xDE:
case 0xDF: /* FCOMP */
#ifdef USE_FLOAT
fpu_fcom(emu, ST(nextop&7).f);
#else
fpu_fcom(emu, ST(nextop&7).d);
#endif
fpu_do_pop(emu);
break;
case 0xE0:
@ -49,7 +65,11 @@
case 0xE5:
case 0xE6:
case 0xE7: /* FSUB */
#ifdef USE_FLOAT
ST0.f -= ST(nextop&7).f;
#else
ST0.d -= ST(nextop&7).d;
#endif
break;
case 0xE8:
case 0xE9:
@ -59,7 +79,11 @@
case 0xED:
case 0xEE:
case 0xEF: /* FSUBR */
#ifdef USE_FLOAT
ST0.f = ST(nextop&7).f - ST0.f;
#else
ST0.d = ST(nextop&7).d - ST0.d;
#endif
break;
case 0xF0:
case 0xF1:
@ -69,7 +93,11 @@
case 0xF5:
case 0xF6:
case 0xF7: /* FDIV */
#ifdef USE_FLOAT
ST0.f /= ST(nextop&7).f;
#else
ST0.d /= ST(nextop&7).d;
#endif
break;
case 0xF8:
case 0xF9:
@ -79,26 +107,46 @@
case 0xFD:
case 0xFE:
case 0xFF: /* FDIVR */
#ifdef USE_FLOAT
ST0.f = ST(nextop&7).f / ST0.f;
#else
ST0.d = ST(nextop&7).d / ST0.d;
#endif
break;
default:
switch((nextop>>3)&7) {
case 0: /* FADD ST0, float */
GET_ED;
if(!(((uintptr_t)ED)&3))
#ifdef USE_FLOAT
ST0.f += *(float*)ED;
#else
ST0.d += *(float*)ED;
#endif
else {
*(uint32_t*)&f = ED->dword[0];
#ifdef USE_FLOAT
ST0.f += f;
#else
ST0.d += f;
#endif
}
break;
case 1: /* FMUL ST0, float */
GET_ED;
if(!(((uintptr_t)ED)&3))
#ifdef USE_FLOAT
ST0.f *= *(float*)ED;
#else
ST0.d *= *(float*)ED;
#endif
else {
*(uint32_t*)&f = ED->dword[0];
#ifdef USE_FLOAT
ST0.f *= f;
#else
ST0.d *= f;
#endif
}
break;
case 2: /* FCOM ST0, float */
@ -123,37 +171,69 @@
case 4: /* FSUB ST0, float */
GET_ED;
if(!(((uintptr_t)ED)&3))
#ifdef USE_FLOAT
ST0.f -= *(float*)ED;
#else
ST0.d -= *(float*)ED;
#endif
else {
*(uint32_t*)&f = ED->dword[0];
#ifdef USE_FLOAT
ST0.f -= f;
#else
ST0.d -= f;
#endif
}
break;
case 5: /* FSUBR ST0, float */
GET_ED;
if(!(((uintptr_t)ED)&3))
#ifdef USE_FLOAT
ST0.f = *(float*)ED - ST0.f;
#else
ST0.d = *(float*)ED - ST0.d;
#endif
else {
*(uint32_t*)&f = ED->dword[0];
#ifdef USE_FLOAT
ST0.f = f - ST0.f;
#else
ST0.d = f - ST0.d;
#endif
}
break;
case 6: /* FDIV ST0, float */
GET_ED;
if(!(((uintptr_t)ED)&3))
#ifdef USE_FLOAT
ST0.f /= *(float*)ED;
#else
ST0.d /= *(float*)ED;
#endif
else {
*(uint32_t*)&f = ED->dword[0];
#ifdef USE_FLOAT
ST0.f /= f;
#else
ST0.d /= f;
#endif
}
break;
case 7: /* FDIVR ST0, float */
GET_ED;
if(!(((uintptr_t)ED)&3))
#ifdef USE_FLOAT
ST0.f = *(float*)ED / ST0.f;
#else
ST0.d = *(float*)ED / ST0.d;
#endif
else {
*(uint32_t*)&f = ED->dword[0];
#ifdef USE_FLOAT
ST0.f = f / ST0.f;
#else
ST0.d = f / ST0.d;
#endif
}
break;
default:

83
src/rund9.h
View File

@ -29,10 +29,18 @@
break;
case 0xE0: /* FCHS */
#ifdef USE_FLOAT
ST0.f = -ST0.f;
#else
ST0.d = -ST0.d;
#endif
break;
case 0xE1: /* FABS */
#ifdef USE_FLOAT
ST0.f = fabsf(ST0.f);
#else
ST0.d = fabs(ST0.d);
#endif
break;
case 0xE5: /* FXAM */
@ -41,46 +49,89 @@
case 0xE8: /* FLD1 */
fpu_do_push(emu);
#ifdef USE_FLOAT
ST0.f = 1.0f;
#else
ST0.d = 1.0;
#endif
break;
case 0xE9: /* FLDL2T */
fpu_do_push(emu);
#ifdef USE_FLOAT
ST0.f = L2T;
#else
ST0.d = L2T;
#endif
break;
case 0xEA: /* FLDL2E */
fpu_do_push(emu);
#ifdef USE_FLOAT
ST0.f = L2E;
#else
ST0.d = L2E;
#endif
break;
case 0xEB: /* FLDPI */
fpu_do_push(emu);
#ifdef USE_FLOAT
ST0.f = PI;
#else
ST0.d = PI;
#endif
break;
case 0xEC: /* FLDLG2 */
fpu_do_push(emu);
#ifdef USE_FLOAT
ST0.f = LG2;
#else
ST0.d = LG2;
#endif
break;
case 0xED: /* FLDLN2 */
fpu_do_push(emu);
#ifdef USE_FLOAT
ST0.f = LN2;
#else
ST0.d = LN2;
#endif
break;
case 0xEE: /* FLDZ */
fpu_do_push(emu);
#ifdef USE_FLOAT
ST0.f = 0.0f;
#else
ST0.d = 0.0;
#endif
break;
case 0xF2: /* FTAN */
#ifdef USE_FLOAT
ST0.f = tanf(ST0.f);
fpu_do_push(emu);
ST0.f = 1.0f;
#else
ST0.d = tan(ST0.d);
fpu_do_push(emu);
ST0.d = 1.0;
#endif
break;
case 0xF3: /* FPATAN */
#ifdef USE_FLOAT
ST1.f = atan2f(ST1.f, ST0.f);
#else
ST1.d = atan2(ST1.d, ST0.d);
#endif
fpu_do_pop(emu);
break;
case 0xF8: /* FPREM */
#ifdef USE_FLOAT
tmp32s = ST0.f / ST1.f;
ST0.f -= ST1.f * tmp32s;
#else
tmp32s = ST0.d / ST1.d;
ST0.d -= ST1.d * tmp32s;
#endif
emu->sw.f.F87_C2 = 0;
emu->sw.f.F87_C0 = (tmp32s&1);
emu->sw.f.F87_C3 = ((tmp32s>>1)&1);
@ -88,21 +139,41 @@
break;
case 0xFA: /* FSQRT */
#ifdef USE_FLOAT
ST0.f = sqrtf(ST0.f);
#else
ST0.d = sqrt(ST0.d);
#endif
break;
case 0xFB: /* FSINCOS */
fpu_do_push(emu);
#ifdef USE_FLOAT
sincosf(ST1.f, &ST1.f, &ST0.f);
#else
sincos(ST1.d, &ST1.d, &ST0.d);
#endif
break;
case 0xFC: /* FRNDINT */
#ifdef USE_FLOAT
ST0.f = fpu_round(emu, ST0.f);
#else
ST0.d = fpu_round(emu, ST0.d);
#endif
break;
case 0xFE: /* FSIN */
#ifdef USE_FLOAT
ST0.f = sinf(ST0.f);
#else
ST0.d = sin(ST0.d);
#endif
break;
case 0xFF: /* FCOS */
#ifdef USE_FLOAT
ST0.f = cosf(ST0.f);
#else
ST0.d = cos(ST0.d);
#endif
break;
@ -139,15 +210,27 @@
case 0: /* FLD ST0, Ed float */
GET_ED;
fpu_do_push(emu);
#ifdef USE_FLOAT
ST0.f = *(float*)ED;
#else
ST0.d = *(float*)ED;
#endif
break;
case 2: /* FST Ed, ST0 */
GET_ED;
#ifdef USE_FLOAT
*(float*)ED = ST0.f;
#else
*(float*)ED = ST0.d;
#endif
break;
case 3: /* FSTP Ed, ST0 */
GET_ED;
#ifdef USE_FLOAT
*(float*)ED = ST0.f;
#else
*(float*)ED = ST0.d;
#endif
fpu_do_pop(emu);
break;
case 4: /* FLDENV m */

28
src/runda.h
View File

@ -51,7 +51,11 @@
break;
case 0xE9: /* FUCOMPP */
#ifdef USE_FLOAT
fpu_fcom(emu, ST1.f);
#else
fpu_fcom(emu, ST1.d); // bad, should handle QNaN and IA interrupt
#endif
fpu_do_pop(emu);
fpu_do_pop(emu);
break;
@ -72,27 +76,51 @@
switch((nextop>>3)&7) {
case 0: /* FIADD ST0, Ed int */
GET_ED;
#ifdef USE_FLOAT
ST0.f += ED->sdword[0];
#else
ST0.d += ED->sdword[0];
#endif
break;
case 1: /* FIMUL ST0, Ed int */
GET_ED;
#ifdef USE_FLOAT
ST0.f *= ED->sdword[0];
#else
ST0.d *= ED->sdword[0];
#endif
break;
case 4: /* FISUB ST0, Ed int */
GET_ED;
#ifdef USE_FLOAT
ST0.f -= ED->sdword[0];
#else
ST0.d -= ED->sdword[0];
#endif
break;
case 5: /* FISUBR ST0, Ed int */
GET_ED;
#ifdef USE_FLOAT
ST0.f = ED->sdword[0] - ST0.f;
#else
ST0.d = ED->sdword[0] - ST0.d;
#endif
break;
case 6: /* FIDIV ST0, Ed int */
GET_ED;
#ifdef USE_FLOAT
ST0.f /= ED->sdword[0];
#else
ST0.d /= ED->sdword[0];
#endif
break;
case 7: /* FIDIVR ST0, Ed int */
GET_ED;
#ifdef USE_FLOAT
ST0.f = ED->sdword[0] / ST0.f;
#else
ST0.d = ED->sdword[0] / ST0.d;
#endif
break;
default:
goto _default;

69
src/rundb.h
View File

@ -73,7 +73,11 @@
case 0xED:
case 0xEE:
case 0xEF:
#ifdef USE_FLOAT
fpu_fcomi(emu, ST(nextop&7).f);
#else
fpu_fcomi(emu, ST(nextop&7).d); // bad, should handle QNaN and IA interrupt
#endif
break;
case 0xF0: /* FCOMI ST0, STx */
@ -84,7 +88,11 @@
case 0xF5:
case 0xF6:
case 0xF7:
#ifdef USE_FLOAT
fpu_fcomi(emu, ST(nextop&7).f);
#else
fpu_fcomi(emu, ST(nextop&7).d);
#endif
break;
case 0xE0:
case 0xE1:
@ -98,18 +106,48 @@
case 0: /* FILD ST0, Gd */
GET_ED;
fpu_do_push(emu);
#ifdef USE_FLOAT
ST0.f = ED->sdword[0];
#else
ST0.d = ED->sdword[0];
#endif
break;
case 1: /* FISTTP Ed, ST0 */
GET_ED;
#ifdef USE_FLOAT
tmp32s = ST0.f; // TODO: Handling of FPU Exception
if(tmp32s==0x7fffffff && isgreater(ST0.f, (float)(int32_t)0x7fffffff))
tmp32s = 0x80000000;
#else
tmp32s = ST0.d; // TODO: Handling of FPU Exception
if(tmp32s==0x7fffffff && isgreater(ST0.d, (double)(int32_t)0x7fffffff))
tmp32s = 0x80000000;
#endif
fpu_do_pop(emu);
ED->sdword[0] = tmp32s;
break;
case 2: /* FIST Ed, ST0 */
GET_ED;
#ifdef USE_FLOAT
if(isgreater(ST0.f, (float)(int32_t)0x7fffffff) || isless(ST0.f, -(float)(int32_t)0x7fffffff))
ED->sdword[0] = 0x80000000;
else {
switch(emu->round) {
case ROUND_Nearest:
ED->sdword[0] = floorf(ST0.f+0.5);
break;
case ROUND_Down:
ED->sdword[0] = floorf(ST0.f);
break;
case ROUND_Up:
ED->sdword[0] = ceilf(ST0.f);
break;
case ROUND_Chop:
ED->sdword[0] = ST0.f;
break;
}
}
#else
if(isgreater(ST0.d, (double)(int32_t)0x7fffffff) || isless(ST0.d, -(double)(int32_t)0x7fffffff))
ED->sdword[0] = 0x80000000;
else {
@ -128,9 +166,30 @@
break;
}
}
#endif
break;
case 3: /* FISTP Ed, ST0 */
GET_ED;
#ifdef USE_FLOAT
if(isgreater(ST0.f, (float)(int32_t)0x7fffffff) || isless(ST0.f, -(float)(int32_t)0x7fffffff))
ED->sdword[0] = 0x80000000;
else {
switch(emu->round) {
case ROUND_Nearest:
ED->sdword[0] = floorf(ST0.f+0.5);
break;
case ROUND_Down:
ED->sdword[0] = floorf(ST0.f);
break;
case ROUND_Up:
ED->sdword[0] = ceilf(ST0.f);
break;
case ROUND_Chop:
ED->sdword[0] = ST0.f;
break;
}
}
#else
if(isgreater(ST0.d, (double)(int32_t)0x7fffffff) || isless(ST0.d, -(double)(int32_t)0x7fffffff))
ED->sdword[0] = 0x80000000;
else {
@ -149,19 +208,29 @@
break;
}
}
#endif
fpu_do_pop(emu);
break;
case 5: /* FLD ST0, Gt */
GET_ED;
fpu_do_push(emu);
memcpy(&STld(0).ld, ED, 10);
#ifdef USE_FLOAT
LD2D(&STld(0), &d);
ST0.f = d;
#else
LD2D(&STld(0), &ST(0).d);
#endif
STld(0).ref = ST0.ll;
break;
case 7: /* FSTP tbyte */
GET_ED;
if(ST0.ll!=STld(0).ref)
#ifdef USE_FLOAT
{d = ST0.f; D2LD(&d, ED);}
#else
D2LD(&ST0.d, ED);
#endif
else
memcpy(ED, &STld(0).ld, 10);
fpu_do_pop(emu);

88
src/rundc.h
View File

@ -8,7 +8,11 @@
case 0xC5:
case 0xC6:
case 0xC7: /* FADD */
#ifdef USE_FLOAT
ST(nextop&7).f += ST0.f;
#else
ST(nextop&7).d += ST0.d;
#endif
break;
case 0xC8:
case 0xC9:
@ -18,7 +22,11 @@
case 0xCD:
case 0xCE:
case 0xCF: /* FMUL */
#ifdef USE_FLOAT
ST(nextop&7).f *= ST0.f;
#else
ST(nextop&7).d *= ST0.d;
#endif
break;
case 0xD0:
case 0xD1:
@ -28,7 +36,11 @@
case 0xD5:
case 0xD6:
case 0xD7: /* FCOM */
fpu_fcom(emu, ST(nextop&7).d); // TODO: is this ok?
#ifdef USE_FLOAT
fpu_fcom(emu, ST(nextop&7).f);
#else
fpu_fcom(emu, ST(nextop&7).d);
#endif
break;
case 0xD8:
case 0xD9:
@ -38,7 +50,11 @@
case 0xDD:
case 0xDE:
case 0xDF: /* FCOMP */
#ifdef USE_FLOAT
fpu_fcom(emu, ST(nextop&7).f);
#else
fpu_fcom(emu, ST(nextop&7).d); // TODO: is this ok?
#endif
fpu_do_pop(emu);
break;
case 0xE0:
@ -49,7 +65,11 @@
case 0xE5:
case 0xE6:
case 0xE7: /* FSUBR */
#ifdef USE_FLOAT
ST(nextop&7).f = ST0.f -ST(nextop&7).f;
#else
ST(nextop&7).d = ST0.d -ST(nextop&7).d;
#endif
break;
case 0xE8:
case 0xE9:
@ -59,7 +79,11 @@
case 0xED:
case 0xEE:
case 0xEF: /* FSUB */
#ifdef USE_FLOAT
ST(nextop&7).f -= ST0.f;
#else
ST(nextop&7).d -= ST0.d;
#endif
break;
case 0xF0:
case 0xF1:
@ -69,7 +93,11 @@
case 0xF5:
case 0xF6:
case 0xF7: /* FDIVR */
#ifdef USE_FLOAT
ST(nextop&7).f = ST0.f / ST(nextop&7).f;
#else
ST(nextop&7).d = ST0.d / ST(nextop&7).d;
#endif
break;
case 0xF8:
case 0xF9:
@ -79,26 +107,48 @@
case 0xFD:
case 0xFE:
case 0xFF: /* FDIV */
#ifdef USE_FLOAT
ST(nextop&7).f /= ST0.f;
#else
ST(nextop&7).d /= ST0.d;
#endif
break;
default:
GET_ED;
switch((nextop>>3)&7) {
case 0: /* FADD ST0, double */
#ifdef USE_FLOAT
if(!(((uintptr_t)ED)&7))
ST0.f += *(double*)ED;
else {
*(uint64_t*)&d = *(uint64_t*)ED;
ST0.f += d;
}
#else
if(!(((uintptr_t)ED)&7))
ST0.d += *(double*)ED;
else {
*(uint64_t*)&d = *(uint64_t*)ED;
ST0.d += d;
}
#endif
break;
case 1: /* FMUL ST0, double */
#ifdef USE_FLOAT
if(!(((uintptr_t)ED)&7))
ST0.f *= *(double*)ED;
else {
*(uint64_t*)&d = *(uint64_t*)ED;
ST0.f *= d;
}
#else
if(!(((uintptr_t)ED)&7))
ST0.d *= *(double*)ED;
else {
*(uint64_t*)&d = *(uint64_t*)ED;
ST0.d *= d;
}
#endif
break;
case 2: /* FCOM ST0, double */
if(!(((uintptr_t)ED)&7))
@ -118,36 +168,72 @@
fpu_do_pop(emu);
break;
case 4: /* FSUB ST0, double */
#ifdef USE_FLOAT
if(!(((uintptr_t)ED)&7))
ST0.f -= *(double*)ED;
else {
*(uint64_t*)&d = *(uint64_t*)ED;
ST0.f -= d;
}
#else
if(!(((uintptr_t)ED)&7))
ST0.d -= *(double*)ED;
else {
*(uint64_t*)&d = *(uint64_t*)ED;
ST0.d -= d;
}
#endif
break;
case 5: /* FSUBR ST0, double */
#ifdef USE_FLOAT
if(!(((uintptr_t)ED)&7))
ST0.f = *(double*)ED - ST0.f;
else {
*(uint64_t*)&d = *(uint64_t*)ED;
ST0.f = d - ST0.f;
}
#else
if(!(((uintptr_t)ED)&7))
ST0.d = *(double*)ED - ST0.d;
else {
*(uint64_t*)&d = *(uint64_t*)ED;
ST0.d = d - ST0.d;
}
#endif
break;
case 6: /* FDIV ST0, double */
#ifdef USE_FLOAT
if(!(((uintptr_t)ED)&7))
ST0.f /= *(double*)ED;
else {
*(uint64_t*)&d = *(uint64_t*)ED;
ST0.f /= d;
}
#else
if(!(((uintptr_t)ED)&7))
ST0.d /= *(double*)ED;
else {
*(uint64_t*)&d = *(uint64_t*)ED;
ST0.d /= d;
}
#endif
break;
case 7: /* FDIVR ST0, double */
#ifdef USE_FLOAT
if(!(((uintptr_t)ED)&7))
ST0.f = *(double*)ED / ST0.f;
else {
*(uint64_t*)&d = *(uint64_t*)ED;
ST0.f = d / ST0.f;
}
#else
if(!(((uintptr_t)ED)&7))
ST0.d = *(double*)ED / ST0.d;
else {
*(uint64_t*)&d = *(uint64_t*)ED;
ST0.d = d / ST0.d;
}
#endif
break;
default:
goto _default;

37
src/rundd.h
View File

@ -40,7 +40,11 @@
case 0xE5:
case 0xE6:
case 0xE7:
#ifdef USE_FLOAT
fpu_fcom(emu, ST(nextop&7).f);
#else
fpu_fcom(emu, ST(nextop&7).d); // bad, should handle QNaN and IA interrupt
#endif
break;
case 0xE8: /* FUCOMP ST0, STx */
case 0xE9:
@ -50,7 +54,11 @@
case 0xED:
case 0xEE:
case 0xEF:
#ifdef USE_FLOAT
fpu_fcom(emu, ST(nextop&7).f);
#else
fpu_fcom(emu, ST(nextop&7).d); // bad, should handle QNaN and IA interrupt
#endif
fpu_do_pop(emu);
break;
@ -85,20 +93,39 @@
case 0: /* FLD double */
GET_ED;
fpu_do_push(emu);
#ifdef USE_FLOAT
*(uint64_t*)&d = *(int64_t*)ED;
ST0.f = d;
#else
ST0.ll = *(int64_t*)ED;
#endif
break;
case 1: /* FISTTP ED qword */
GET_ED;
#ifdef USE_FLOAT
*(int64_t*)ED = ST0.f;
#else
*(int64_t*)ED = ST0.d;
#endif
fpu_do_pop(emu);
break;
case 2: /* FST double */
GET_ED;
#ifdef USE_FLOAT
d = ST0.f;
*(int64_t*)ED = *(uint64_t*)&d;
#else
*(int64_t*)ED = ST0.ll;
#endif
break;
case 3: /* FSTP double */
GET_ED;
#ifdef USE_FLOAT
d = ST0.f;
*(int64_t*)ED = *(uint64_t*)&d;
#else
*(int64_t*)ED = ST0.ll;
#endif
fpu_do_pop(emu);
break;
case 4: /* FRSTOR m108byte */
@ -109,7 +136,12 @@
char* p =(char*)ED;
p += 28;
for (int i=0; i<8; ++i) {
#ifdef USE_FLOAT
d = ST(i).f;
LD2D(p, &d);
#else
LD2D(p, &ST(i).d);
#endif
p+=10;
}
}
@ -124,7 +156,12 @@
char* p =(char*)ED;
p += 28;
for (int i=0; i<8; ++i) {
#ifdef USE_FLOAT
D2LD(&d, p);
ST(i).f = d;
#else
D2LD(&ST(i).d, p);
#endif
p+=10;
}
}

71
src/runde.h
View File

@ -1,7 +1,11 @@
nextop = F8;
switch (nextop) {
case 0xC1: /* FADDP ST1, ST0 */
#ifdef USE_FLOAT
ST(1).f += ST0.f;
#else
ST(1).d += ST0.d;
#endif
fpu_do_pop(emu);
break;
case 0xC0: /* FADDP STx, ST0 */
@ -11,11 +15,19 @@
case 0xC5:
case 0xC6:
case 0xC7:
#ifdef USE_FLOAT
ST(nextop&7).f += ST0.f;
#else
ST(nextop&7).d += ST0.d;
#endif
fpu_do_pop(emu);
break;
case 0xC9: /* FMULP ST1, ST0 */
#ifdef USE_FLOAT
ST1.f *= ST0.f;
#else
ST1.d *= ST0.d;
#endif
fpu_do_pop(emu);
break;
case 0xC8: /* FMULP STx, ST0 */
@ -25,11 +37,19 @@
case 0xCD:
case 0xCE:
case 0xCF:
#ifdef USE_FLOAT
ST(nextop&7).f *= ST0.f;
#else
ST(nextop&7).d *= ST0.d;
#endif
fpu_do_pop(emu);
break;
case 0xD9: /* FCOMPP */
#ifdef USE_FLOAT
fpu_fcom(emu, ST1.f);
#else
fpu_fcom(emu, ST1.d);
#endif
fpu_do_pop(emu);
fpu_do_pop(emu);
break;
@ -45,11 +65,19 @@
case 0xE5:
case 0xE6:
case 0xE7:
#ifdef USE_FLOAT
ST(nextop&7).f = ST0.f - ST(nextop&7).f;
#else
ST(nextop&7).d = ST0.d - ST(nextop&7).d;
#endif
fpu_do_pop(emu);
break;
case 0xE9: /* FSUBP ST1, ST0 */
#ifdef USE_FLOAT
ST1.f -= ST0.f;
#else
ST1.d -= ST0.d;
#endif
fpu_do_pop(emu);
break;
case 0xE8: /* FSUBP STx, ST0 */
@ -59,12 +87,13 @@
case 0xED:
case 0xEE:
case 0xEF:
#ifdef USE_FLOAT
ST(nextop&7).f -= ST0.f;
#else
ST(nextop&7).d -= ST0.d;
#endif
fpu_do_pop(emu);
break;
/*ST(1).d = ST0.d / ST(1).d;
fpu_do_pop(emu);
break;*/
case 0xF0: /* FDIVRP STx, ST0 */
case 0xF1: /* FDIVRP ST1, ST0 */
case 0xF2:
@ -73,13 +102,13 @@
case 0xF5:
case 0xF6:
case 0xF7:
#ifdef USE_FLOAT
ST(nextop&7).f = ST0.f / ST(nextop&7).f;
#else
ST(nextop&7).d = ST0.d / ST(nextop&7).d;
#endif
fpu_do_pop(emu);
break;
/*d = ST0.d;
fpu_do_pop(emu);
ST0.d /= d;
break;*/
case 0xF8: /* FDIVP STx, ST0 */
case 0xF9: /* FDIVP ST1, ST0 */
case 0xFA:
@ -88,7 +117,11 @@
case 0xFD:
case 0xFE:
case 0xFF:
#ifdef USE_FLOAT
ST(nextop&7).f /= ST0.f;
#else
ST(nextop&7).d /= ST0.d;
#endif
fpu_do_pop(emu);
break;
@ -113,27 +146,51 @@
switch((nextop>>3)&7) {
case 0: /* FIADD ST0, Ew int */
GET_EW;
#ifdef USE_FLOAT
ST0.f += EW->sword[0];
#else
ST0.d += EW->sword[0];
#endif
break;
case 1: /* FIMUL ST0, Ew int */
GET_EW;
#ifdef USE_FLOAT
ST0.f *= EW->sword[0];
#else
ST0.d *= EW->sword[0];
#endif
break;
case 4: /* FISUB ST0, Ew int */
GET_EW;
#ifdef USE_FLOAT
ST0.f -= EW->sword[0];
#else
ST0.d -= EW->sword[0];
#endif
break;
case 5: /* FISUBR ST0, Ew int */
GET_EW;
#ifdef USE_FLOAT
ST0.f = EW->sword[0] - ST0.f;
#else
ST0.d = EW->sword[0] - ST0.d;
#endif
break;
case 6: /* FIDIV ST0, Ew int */
GET_EW;
#ifdef USE_FLOAT
ST0.f /= EW->sword[0];
#else
ST0.d /= EW->sword[0];
#endif
break;
case 7: /* FIDIVR ST0, Ew int */
GET_EW;
#ifdef USE_FLOAT
ST0.f = EW->sword[0] / ST0.f;
#else
ST0.d = EW->sword[0] / ST0.d;
#endif
break;
default:
goto _default;

69
src/rundf.h
View File

@ -13,7 +13,11 @@
case 0xED:
case 0xEE:
case 0xEF:
#ifdef USE_FLOAT
fpu_fcomi(emu, ST(nextop&7).f);
#else
fpu_fcomi(emu, ST(nextop&7).d); // bad, should handle QNaN and IA interrupt
#endif
fpu_do_pop(emu);
break;
@ -25,7 +29,11 @@
case 0xF5:
case 0xF6:
case 0xF7:
#ifdef USE_FLOAT
fpu_fcomi(emu, ST(nextop&7).f);
#else
fpu_fcomi(emu, ST(nextop&7).d);
#endif
fpu_do_pop(emu);
break;
@ -84,16 +92,44 @@
GET_EW;
tmp16s = EW->sword[0];
fpu_do_push(emu);
#ifdef USE_FLOAT
ST0.f = tmp16s;
#else
ST0.d = tmp16s;
#endif
break;
case 1: /* FISTTP Ew, ST0 */
GET_EW;
#ifdef USE_FLOAT
tmp16s = ST0.f;
#else
tmp16s = ST0.d;
#endif
EW->sword[0] = tmp16s;
fpu_do_pop(emu);
break;
case 2: /* FIST Ew, ST0 */
GET_EW;
#ifdef USE_FLOAT
if(isgreater(ST0.f, (float)(int32_t)0x7fff) || isless(ST0.f, -(float)(int32_t)0x7fff))
EW->sword[0] = 0x8000;
else {
switch(emu->round) {
case ROUND_Nearest:
EW->sword[0] = floorf(ST0.f+0.5);
break;
case ROUND_Down:
EW->sword[0] = floorf(ST0.f);
break;
case ROUND_Up:
EW->sword[0] = ceilf(ST0.f);
break;
case ROUND_Chop:
EW->sword[0] = ST0.f;
break;
}
}
#else
if(isgreater(ST0.d, (double)(int32_t)0x7fff) || isless(ST0.d, -(double)(int32_t)0x7fff))
EW->sword[0] = 0x8000;
else {
@ -112,9 +148,30 @@
break;
}
}
#endif
break;
case 3: /* FISTP Ew, ST0 */
GET_EW;
#ifdef USE_FLOAT
if(isgreater(ST0.f, (float)(int32_t)0x7fff) || isless(ST0.f, -(float)(int32_t)0x7fff))
EW->sword[0] = 0x8000;
else {
switch(emu->round) {
case ROUND_Nearest:
EW->sword[0] = floorf(ST0.f+0.5);
break;
case ROUND_Down:
EW->sword[0] = floorf(ST0.f);
break;
case ROUND_Up:
EW->sword[0] = ceilf(ST0.f);
break;
case ROUND_Chop:
EW->sword[0] = ST0.f;
break;
}
}
#else
if(isgreater(ST0.d, (double)(int32_t)0x7fff) || isless(ST0.d, -(double)(int32_t)0x7fff))
EW->sword[0] = 0x8000;
else {
@ -133,6 +190,7 @@
break;
}
}
#endif
fpu_do_pop(emu);
break;
case 4: /* FBLD ST0, tbytes */
@ -144,7 +202,11 @@
GET_ED;
tmp64s = *(int64_t*)ED;
fpu_do_push(emu);
#ifdef USE_FLOAT
ST0.f = tmp64s;
#else
ST0.d = tmp64s;
#endif
STll(0).ll = tmp64s;
STll(0).ref = ST0.ll;
break;
@ -158,10 +220,17 @@
if(STll(0).ref==ST(0).ll) {
*(int64_t*)ED = STll(0).ll;
} else {
#ifdef USE_FLOAT
if(isgreater(ST0.f, (float)(int64_t)0x7fffffffffffffffLL) || isless(ST0.f, (float)(int64_t)0x8000000000000000LL))
*(int64_t*)ED = 0x8000000000000000LL;
else
*(int64_t*)ED = (int64_t)ST0.f;
#else
if(isgreater(ST0.d, (double)(int64_t)0x7fffffffffffffffLL) || isless(ST0.d, (double)(int64_t)0x8000000000000000LL))
*(int64_t*)ED = 0x8000000000000000LL;
else
*(int64_t*)ED = (int64_t)ST0.d;
#endif
}
fpu_do_pop(emu);
break;

54
src/wrapper.c
View File

@ -17,6 +17,12 @@ typedef union ui64_s {
uint32_t d[2];
} ui64_t;
#ifdef USE_FLOAT
#define ST0val ST0.f
#else
#define ST0val ST0.d
#endif
typedef int32_t (*iF_t)();
typedef void (*vFE_t)(x86emu_t*);
typedef void (*vFv_t)(void);
@ -818,11 +824,11 @@ void uFu(x86emu_t *emu, uintptr_t fcn) { uFu_t fn = (uFu_t)fcn; R_EAX=(uint32_t)
void uFp(x86emu_t *emu, uintptr_t fcn) { uFp_t fn = (uFp_t)fcn; R_EAX=(uint32_t)fn(*(void**)(R_ESP + 4)); }
void UFv(x86emu_t *emu, uintptr_t fcn) { UFv_t fn = (UFv_t)fcn; ui64_t r; r.u=(uint64_t)fn(); R_EAX=r.d[0]; R_EDX=r.d[1]; }
void UFp(x86emu_t *emu, uintptr_t fcn) { UFp_t fn = (UFp_t)fcn; ui64_t r; r.u=(uint64_t)fn(*(void**)(R_ESP + 4)); R_EAX=r.d[0]; R_EDX=r.d[1]; }
void fFi(x86emu_t *emu, uintptr_t fcn) { fFi_t fn = (fFi_t)fcn; float fl=fn(*(int32_t*)(R_ESP + 4)); fpu_do_push(emu); ST0.d = fl; }
void fFf(x86emu_t *emu, uintptr_t fcn) { fFf_t fn = (fFf_t)fcn; float fl=fn(*(float*)(R_ESP + 4)); fpu_do_push(emu); ST0.d = fl; }
void dFi(x86emu_t *emu, uintptr_t fcn) { dFi_t fn = (dFi_t)fcn; double db=fn(*(int32_t*)(R_ESP + 4)); fpu_do_push(emu); ST0.d = db; }
void dFd(x86emu_t *emu, uintptr_t fcn) { dFd_t fn = (dFd_t)fcn; double db=fn(*(double*)(R_ESP + 4)); fpu_do_push(emu); ST0.d = db; }
void dFp(x86emu_t *emu, uintptr_t fcn) { dFp_t fn = (dFp_t)fcn; double db=fn(*(void**)(R_ESP + 4)); fpu_do_push(emu); ST0.d = db; }
void fFi(x86emu_t *emu, uintptr_t fcn) { fFi_t fn = (fFi_t)fcn; float fl=fn(*(int32_t*)(R_ESP + 4)); fpu_do_push(emu); ST0val = fl; }
void fFf(x86emu_t *emu, uintptr_t fcn) { fFf_t fn = (fFf_t)fcn; float fl=fn(*(float*)(R_ESP + 4)); fpu_do_push(emu); ST0val = fl; }
void dFi(x86emu_t *emu, uintptr_t fcn) { dFi_t fn = (dFi_t)fcn; double db=fn(*(int32_t*)(R_ESP + 4)); fpu_do_push(emu); ST0val = db; }
void dFd(x86emu_t *emu, uintptr_t fcn) { dFd_t fn = (dFd_t)fcn; double db=fn(*(double*)(R_ESP + 4)); fpu_do_push(emu); ST0val = db; }
void dFp(x86emu_t *emu, uintptr_t fcn) { dFp_t fn = (dFp_t)fcn; double db=fn(*(void**)(R_ESP + 4)); fpu_do_push(emu); ST0val = db; }
void pFE(x86emu_t *emu, uintptr_t fcn) { pFE_t fn = (pFE_t)fcn; R_EAX=(uintptr_t)fn(emu); }
void pFv(x86emu_t *emu, uintptr_t fcn) { pFv_t fn = (pFv_t)fcn; R_EAX=(uintptr_t)fn(); }
void pFi(x86emu_t *emu, uintptr_t fcn) { pFi_t fn = (pFi_t)fcn; R_EAX=(uintptr_t)fn(*(int32_t*)(R_ESP + 4)); }
@ -882,18 +888,18 @@ void uFpi(x86emu_t *emu, uintptr_t fcn) { uFpi_t fn = (uFpi_t)fcn; R_EAX=(uint32
void uFpu(x86emu_t *emu, uintptr_t fcn) { uFpu_t fn = (uFpu_t)fcn; R_EAX=(uint32_t)fn(*(void**)(R_ESP + 4), *(uint32_t*)(R_ESP + 8)); }
void uFpp(x86emu_t *emu, uintptr_t fcn) { uFpp_t fn = (uFpp_t)fcn; R_EAX=(uint32_t)fn(*(void**)(R_ESP + 4), *(void**)(R_ESP + 8)); }
void UFEp(x86emu_t *emu, uintptr_t fcn) { UFEp_t fn = (UFEp_t)fcn; ui64_t r; r.u=(uint64_t)fn(emu, *(void**)(R_ESP + 4)); R_EAX=r.d[0]; R_EDX=r.d[1]; }
void fFfi(x86emu_t *emu, uintptr_t fcn) { fFfi_t fn = (fFfi_t)fcn; float fl=fn(*(float*)(R_ESP + 4), *(int32_t*)(R_ESP + 8)); fpu_do_push(emu); ST0.d = fl; }
void fFff(x86emu_t *emu, uintptr_t fcn) { fFff_t fn = (fFff_t)fcn; float fl=fn(*(float*)(R_ESP + 4), *(float*)(R_ESP + 8)); fpu_do_push(emu); ST0.d = fl; }
void fFfp(x86emu_t *emu, uintptr_t fcn) { fFfp_t fn = (fFfp_t)fcn; float fl=fn(*(float*)(R_ESP + 4), *(void**)(R_ESP + 8)); fpu_do_push(emu); ST0.d = fl; }
void fFpp(x86emu_t *emu, uintptr_t fcn) { fFpp_t fn = (fFpp_t)fcn; float fl=fn(*(void**)(R_ESP + 4), *(void**)(R_ESP + 8)); fpu_do_push(emu); ST0.d = fl; }
void dFEp(x86emu_t *emu, uintptr_t fcn) { dFEp_t fn = (dFEp_t)fcn; double db=fn(emu, *(void**)(R_ESP + 4)); fpu_do_push(emu); ST0.d = db; }
void dFuu(x86emu_t *emu, uintptr_t fcn) { dFuu_t fn = (dFuu_t)fcn; double db=fn(*(uint32_t*)(R_ESP + 4), *(uint32_t*)(R_ESP + 8)); fpu_do_push(emu); ST0.d = db; }
void dFdi(x86emu_t *emu, uintptr_t fcn) { dFdi_t fn = (dFdi_t)fcn; double db=fn(*(double*)(R_ESP + 4), *(int32_t*)(R_ESP + 12)); fpu_do_push(emu); ST0.d = db; }
void dFdd(x86emu_t *emu, uintptr_t fcn) { dFdd_t fn = (dFdd_t)fcn; double db=fn(*(double*)(R_ESP + 4), *(double*)(R_ESP + 12)); fpu_do_push(emu); ST0.d = db; }
void dFdp(x86emu_t *emu, uintptr_t fcn) { dFdp_t fn = (dFdp_t)fcn; double db=fn(*(double*)(R_ESP + 4), *(void**)(R_ESP + 12)); fpu_do_push(emu); ST0.d = db; }
void dFpi(x86emu_t *emu, uintptr_t fcn) { dFpi_t fn = (dFpi_t)fcn; double db=fn(*(void**)(R_ESP + 4), *(int32_t*)(R_ESP + 8)); fpu_do_push(emu); ST0.d = db; }
void dFpp(x86emu_t *emu, uintptr_t fcn) { dFpp_t fn = (dFpp_t)fcn; double db=fn(*(void**)(R_ESP + 4), *(void**)(R_ESP + 8)); fpu_do_push(emu); ST0.d = db; }
void LFLp(x86emu_t *emu, uintptr_t fcn) { LFLp_t fn = (LFLp_t)fcn; double db=fn(FromLD((void*)(R_ESP + 4)), *(void**)(R_ESP + 16)); fpu_do_push(emu); ST0.d = db; }
void fFfi(x86emu_t *emu, uintptr_t fcn) { fFfi_t fn = (fFfi_t)fcn; float fl=fn(*(float*)(R_ESP + 4), *(int32_t*)(R_ESP + 8)); fpu_do_push(emu); ST0val = fl; }
void fFff(x86emu_t *emu, uintptr_t fcn) { fFff_t fn = (fFff_t)fcn; float fl=fn(*(float*)(R_ESP + 4), *(float*)(R_ESP + 8)); fpu_do_push(emu); ST0val = fl; }
void fFfp(x86emu_t *emu, uintptr_t fcn) { fFfp_t fn = (fFfp_t)fcn; float fl=fn(*(float*)(R_ESP + 4), *(void**)(R_ESP + 8)); fpu_do_push(emu); ST0val = fl; }
void fFpp(x86emu_t *emu, uintptr_t fcn) { fFpp_t fn = (fFpp_t)fcn; float fl=fn(*(void**)(R_ESP + 4), *(void**)(R_ESP + 8)); fpu_do_push(emu); ST0val = fl; }
void dFEp(x86emu_t *emu, uintptr_t fcn) { dFEp_t fn = (dFEp_t)fcn; double db=fn(emu, *(void**)(R_ESP + 4)); fpu_do_push(emu); ST0val = db; }
void dFuu(x86emu_t *emu, uintptr_t fcn) { dFuu_t fn = (dFuu_t)fcn; double db=fn(*(uint32_t*)(R_ESP + 4), *(uint32_t*)(R_ESP + 8)); fpu_do_push(emu); ST0val = db; }
void dFdi(x86emu_t *emu, uintptr_t fcn) { dFdi_t fn = (dFdi_t)fcn; double db=fn(*(double*)(R_ESP + 4), *(int32_t*)(R_ESP + 12)); fpu_do_push(emu); ST0val = db; }
void dFdd(x86emu_t *emu, uintptr_t fcn) { dFdd_t fn = (dFdd_t)fcn; double db=fn(*(double*)(R_ESP + 4), *(double*)(R_ESP + 12)); fpu_do_push(emu); ST0val = db; }
void dFdp(x86emu_t *emu, uintptr_t fcn) { dFdp_t fn = (dFdp_t)fcn; double db=fn(*(double*)(R_ESP + 4), *(void**)(R_ESP + 12)); fpu_do_push(emu); ST0val = db; }
void dFpi(x86emu_t *emu, uintptr_t fcn) { dFpi_t fn = (dFpi_t)fcn; double db=fn(*(void**)(R_ESP + 4), *(int32_t*)(R_ESP + 8)); fpu_do_push(emu); ST0val = db; }
void dFpp(x86emu_t *emu, uintptr_t fcn) { dFpp_t fn = (dFpp_t)fcn; double db=fn(*(void**)(R_ESP + 4), *(void**)(R_ESP + 8)); fpu_do_push(emu); ST0val = db; }
void LFLp(x86emu_t *emu, uintptr_t fcn) { LFLp_t fn = (LFLp_t)fcn; double db=fn(FromLD((void*)(R_ESP + 4)), *(void**)(R_ESP + 16)); fpu_do_push(emu); ST0val = db; }
void pFEp(x86emu_t *emu, uintptr_t fcn) { pFEp_t fn = (pFEp_t)fcn; R_EAX=(uintptr_t)fn(emu, *(void**)(R_ESP + 4)); }
void pFii(x86emu_t *emu, uintptr_t fcn) { pFii_t fn = (pFii_t)fcn; R_EAX=(uintptr_t)fn(*(int32_t*)(R_ESP + 4), *(int32_t*)(R_ESP + 8)); }
void pFip(x86emu_t *emu, uintptr_t fcn) { pFip_t fn = (pFip_t)fcn; R_EAX=(uintptr_t)fn(*(int32_t*)(R_ESP + 4), *(void**)(R_ESP + 8)); }
@ -1005,13 +1011,13 @@ void uFppi(x86emu_t *emu, uintptr_t fcn) { uFppi_t fn = (uFppi_t)fcn; R_EAX=(uin
void uFppu(x86emu_t *emu, uintptr_t fcn) { uFppu_t fn = (uFppu_t)fcn; R_EAX=(uint32_t)fn(*(void**)(R_ESP + 4), *(void**)(R_ESP + 8), *(uint32_t*)(R_ESP + 12)); }
void uFppp(x86emu_t *emu, uintptr_t fcn) { uFppp_t fn = (uFppp_t)fcn; R_EAX=(uint32_t)fn(*(void**)(R_ESP + 4), *(void**)(R_ESP + 8), *(void**)(R_ESP + 12)); }
void UFppi(x86emu_t *emu, uintptr_t fcn) { UFppi_t fn = (UFppi_t)fcn; ui64_t r; r.u=(uint64_t)fn(*(void**)(R_ESP + 4), *(void**)(R_ESP + 8), *(int32_t*)(R_ESP + 12)); R_EAX=r.d[0]; R_EDX=r.d[1]; }
void fFuii(x86emu_t *emu, uintptr_t fcn) { fFuii_t fn = (fFuii_t)fcn; float fl=fn(*(uint32_t*)(R_ESP + 4), *(int32_t*)(R_ESP + 8), *(int32_t*)(R_ESP + 12)); fpu_do_push(emu); ST0.d = fl; }
void fFppu(x86emu_t *emu, uintptr_t fcn) { fFppu_t fn = (fFppu_t)fcn; float fl=fn(*(void**)(R_ESP + 4), *(void**)(R_ESP + 8), *(uint32_t*)(R_ESP + 12)); fpu_do_push(emu); ST0.d = fl; }
void fFppp(x86emu_t *emu, uintptr_t fcn) { fFppp_t fn = (fFppp_t)fcn; float fl=fn(*(void**)(R_ESP + 4), *(void**)(R_ESP + 8), *(void**)(R_ESP + 12)); fpu_do_push(emu); ST0.d = fl; }
void dFEpi(x86emu_t *emu, uintptr_t fcn) { dFEpi_t fn = (dFEpi_t)fcn; double db=fn(emu, *(void**)(R_ESP + 4), *(int32_t*)(R_ESP + 8)); fpu_do_push(emu); ST0.d = db; }
void dFppu(x86emu_t *emu, uintptr_t fcn) { dFppu_t fn = (dFppu_t)fcn; double db=fn(*(void**)(R_ESP + 4), *(void**)(R_ESP + 8), *(uint32_t*)(R_ESP + 12)); fpu_do_push(emu); ST0.d = db; }
void dFppp(x86emu_t *emu, uintptr_t fcn) { dFppp_t fn = (dFppp_t)fcn; double db=fn(*(void**)(R_ESP + 4), *(void**)(R_ESP + 8), *(void**)(R_ESP + 12)); fpu_do_push(emu); ST0.d = db; }
void DFppu(x86emu_t *emu, uintptr_t fcn) { DFppu_t fn = (DFppu_t)fcn; long double ld=fn(*(void**)(R_ESP + 4), *(void**)(R_ESP + 8), *(uint32_t*)(R_ESP + 12)); fpu_do_push(emu); ST0.d = ld; }
void fFuii(x86emu_t *emu, uintptr_t fcn) { fFuii_t fn = (fFuii_t)fcn; float fl=fn(*(uint32_t*)(R_ESP + 4), *(int32_t*)(R_ESP + 8), *(int32_t*)(R_ESP + 12)); fpu_do_push(emu); ST0val = fl; }
void fFppu(x86emu_t *emu, uintptr_t fcn) { fFppu_t fn = (fFppu_t)fcn; float fl=fn(*(void**)(R_ESP + 4), *(void**)(R_ESP + 8), *(uint32_t*)(R_ESP + 12)); fpu_do_push(emu); ST0val = fl; }
void fFppp(x86emu_t *emu, uintptr_t fcn) { fFppp_t fn = (fFppp_t)fcn; float fl=fn(*(void**)(R_ESP + 4), *(void**)(R_ESP + 8), *(void**)(R_ESP + 12)); fpu_do_push(emu); ST0val = fl; }
void dFEpi(x86emu_t *emu, uintptr_t fcn) { dFEpi_t fn = (dFEpi_t)fcn; double db=fn(emu, *(void**)(R_ESP + 4), *(int32_t*)(R_ESP + 8)); fpu_do_push(emu); ST0val = db; }
void dFppu(x86emu_t *emu, uintptr_t fcn) { dFppu_t fn = (dFppu_t)fcn; double db=fn(*(void**)(R_ESP + 4), *(void**)(R_ESP + 8), *(uint32_t*)(R_ESP + 12)); fpu_do_push(emu); ST0val = db; }
void dFppp(x86emu_t *emu, uintptr_t fcn) { dFppp_t fn = (dFppp_t)fcn; double db=fn(*(void**)(R_ESP + 4), *(void**)(R_ESP + 8), *(void**)(R_ESP + 12)); fpu_do_push(emu); ST0val = db; }
void DFppu(x86emu_t *emu, uintptr_t fcn) { DFppu_t fn = (DFppu_t)fcn; long double ld=fn(*(void**)(R_ESP + 4), *(void**)(R_ESP + 8), *(uint32_t*)(R_ESP + 12)); fpu_do_push(emu); ST0val = ld; }
void pFEip(x86emu_t *emu, uintptr_t fcn) { pFEip_t fn = (pFEip_t)fcn; R_EAX=(uintptr_t)fn(emu, *(int32_t*)(R_ESP + 4), *(void**)(R_ESP + 8)); }
void pFEpi(x86emu_t *emu, uintptr_t fcn) { pFEpi_t fn = (pFEpi_t)fcn; R_EAX=(uintptr_t)fn(emu, *(void**)(R_ESP + 4), *(int32_t*)(R_ESP + 8)); }
void pFEpp(x86emu_t *emu, uintptr_t fcn) { pFEpp_t fn = (pFEpp_t)fcn; R_EAX=(uintptr_t)fn(emu, *(void**)(R_ESP + 4), *(void**)(R_ESP + 8)); }

4
src/x86emu.c
View File

@ -275,7 +275,11 @@ const char* DumpCPURegs(x86emu_t* emu, uintptr_t ip)
// start with FPU regs...
if(emu->fpu_stack) {
for (int i=0; i<emu->fpu_stack; i++) {
#ifdef USE_FLOAT
sprintf(tmp, "ST%d=%f", i, emu->fpu[(emu->top+i)&7].f);
#else
sprintf(tmp, "ST%d=%f", i, emu->fpu[(emu->top+i)&7].d);
#endif
strcat(buff, tmp);
int c = 10-strlen(tmp);
if(c<1) c=1;

41
src/x87emu_private.c
View File

@ -23,8 +23,14 @@ void fpu_fbst(x86emu_t* emu, uint8_t* d) {
// very aproximative... but should not be much used...
uint8_t p;
uint8_t sign = 0x00;
#ifdef USE_FLOAT
double tmp, v = ST0.f;
if(ST0.f<0.0f)
#else
double tmp, v = ST0.d;
if(ST0.d<0.0) {
if(ST0.d<0.0)
#endif
{
sign = 0x80;
v = -v;
}
@ -56,23 +62,47 @@ void fpu_fbld(x86emu_t* emu, uint8_t* s) {
tmp += m * ((p>>4)&0x0f);
m *= 10;
}
#ifdef USE_FLOAT
ST0.f = tmp;
p =*(s++);
ST0.f += m * (p&0x0f);
if(p&0x80)
ST0.f = -ST0.f;
#else
ST0.d = tmp;
p =*(s++);
ST0.d += m * (p&0x0f);
if(p&0x80)
ST0.d = -ST0.d;
#endif
}
#ifdef USE_FLOAT
typedef union {
double d;
struct {
uint32_t lower;
uint32_t upper;
} l;
struct {
float lower;
float upper;
} f;
int64_t ll;
} FPU_t;
#else
#define fpu_reg_t FPU_t
#endif
#define BIAS80 16383
#define BIAS64 1023
// long double (80bits) -> double (64bits)
void LD2D(void* ld, void* d)
{
fpu_reg_t result;
FPU_t result;
#pragma pack(push, 1)
struct {
fpu_reg_t f;
FPU_t f;
int16_t b;
} val;
#pragma pack(pop)
@ -138,11 +168,11 @@ void D2LD(void* d, void* ld)
{
#pragma pack(push, 1)
struct {
fpu_reg_t f;
FPU_t f;
int16_t b;
} val;
#pragma pack(pop)
fpu_reg_t s;
FPU_t s;
s.ll = *(uint64_t*)d; // use memcpy to avoid risk of Bus Error?
// do special value first
if((s.ll&0x7fffffffffffffffLL)==0) {
@ -188,7 +218,6 @@ double FromLD(void* ld)
return ret;
}
void fpu_loadenv(x86emu_t* emu, char* p, int b16)
{
emu->cw = *(uint16_t*)p;

88
src/x87emu_private.h
View File

@ -50,9 +50,28 @@ static inline void fpu_do_pop(x86emu_t* emu)
void reset_fpu(x86emu_t* emu);
#ifdef USE_FLOAT
static inline void fpu_fcom(x86emu_t* emu, float b)
#else
static inline void fpu_fcom(x86emu_t* emu, double b)
#endif
{
emu->sw.f.F87_C1 = 0;
#ifdef USE_FLOAT
if(isnan(ST0.f) || isnan(b)) {
emu->sw.f.F87_C0 = 1;
emu->sw.f.F87_C2 = 1;
emu->sw.f.F87_C3 = 1;
} else if (isgreater(ST0.f, b)) {
emu->sw.f.F87_C0 = 0;
emu->sw.f.F87_C2 = 0;
emu->sw.f.F87_C3 = 0;
} else if (isless(ST0.f, b)) {
emu->sw.f.F87_C0 = 1;
emu->sw.f.F87_C2 = 0;
emu->sw.f.F87_C3 = 0;
}
#else
if(isnan(ST0.d) || isnan(b)) {
emu->sw.f.F87_C0 = 1;
emu->sw.f.F87_C2 = 1;
@ -65,17 +84,38 @@ static inline void fpu_fcom(x86emu_t* emu, double b)
emu->sw.f.F87_C0 = 1;
emu->sw.f.F87_C2 = 0;
emu->sw.f.F87_C3 = 0;
} else {
}
#endif
else {
emu->sw.f.F87_C0 = 0;
emu->sw.f.F87_C2 = 0;
emu->sw.f.F87_C3 = 1;
}
}
#ifdef USE_FLOAT
static inline void fpu_fcomi(x86emu_t* emu, float b)
#else
static inline void fpu_fcomi(x86emu_t* emu, double b)
#endif
{
RESET_FLAGS(emu);
emu->sw.f.F87_C1 = 0;
#ifdef USE_FLOAT
if(isnan(ST0.f) || isnan(b)) {
emu->flags[F_CF] = 1;
emu->flags[F_PF] = 1;
emu->flags[F_ZF] = 1;
} else if (isgreater(ST0.f, b)) {
emu->flags[F_CF] = 0;
emu->flags[F_PF] = 0;
emu->flags[F_ZF] = 0;
} else if (isless(ST0.f, b)) {
emu->flags[F_CF] = 1;
emu->flags[F_PF] = 0;
emu->flags[F_ZF] = 0;
}
#else
if(isnan(ST0.d) || isnan(b)) {
emu->flags[F_CF] = 1;
emu->flags[F_PF] = 1;
@ -88,13 +128,31 @@ static inline void fpu_fcomi(x86emu_t* emu, double b)
emu->flags[F_CF] = 1;
emu->flags[F_PF] = 0;
emu->flags[F_ZF] = 0;
} else {
}
#endif
else {
emu->flags[F_CF] = 0;
emu->flags[F_PF] = 0;
emu->flags[F_ZF] = 1;
}
}
#ifdef USE_FLOAT
static inline float fpu_round(x86emu_t* emu, float d) {
if (!isfinite(d))
return d;
switch(emu->round) {
case ROUND_Nearest:
return floorf(d+0.5);
case ROUND_Down:
return floorf(d);
case ROUND_Up:
return ceilf(d);
case ROUND_Chop:
return roundf(d);
}
}
#else
static inline double fpu_round(x86emu_t* emu, double d) {
if (!isfinite(d))
return d;
@ -109,28 +167,48 @@ static inline double fpu_round(x86emu_t* emu, double d) {
return round(d);
}
}
#endif
static inline void fpu_fxam(x86emu_t* emu) {
#ifdef USE_FLOAT
emu->sw.f.F87_C1 = (ST0.ll&0x8000)?1:0;
#else
emu->sw.f.F87_C1 = (ST0.l.upper&0x80000000)?1:0;
#endif
if(!emu->fpu_stack) {
emu->sw.f.F87_C3 = 1;
emu->sw.f.F87_C2 = 0;
emu->sw.f.F87_C0 = 1;
return;
}
if(isinf(ST0.d)) { // TODO: Unsuported and denormal not analysed...
#ifdef USE_FLOAT
if(isinf(ST0.f))
#else
if(isinf(ST0.d))
#endif
{ // TODO: Unsuported and denormal not analysed...
emu->sw.f.F87_C3 = 0;
emu->sw.f.F87_C2 = 1;
emu->sw.f.F87_C0 = 1;
return;
}
if(isnan(ST0.d)) { // TODO: Unsuported and denormal not analysed...
#ifdef USE_FLOAT
if(isnan(ST0.f))
#else
if(isnan(ST0.d))
#endif
{ // TODO: Unsuported and denormal not analysed...
emu->sw.f.F87_C3 = 0;
emu->sw.f.F87_C2 = 0;
emu->sw.f.F87_C0 = 1;
return;
}
if(ST0.d==0.0) {
#ifdef USE_FLOAT
if(ST0.f==0.0f)
#else
if(ST0.d==0.0)
#endif
{
emu->sw.f.F87_C3 = 1;
emu->sw.f.F87_C2 = 0;
emu->sw.f.F87_C0 = 0;

10
src/x87run.c
View File

@ -159,7 +159,12 @@ void Run66DD(x86emu_t *emu)
char* p =(char*)op1;
p += 14;
for (int i=0; i<8; ++i) {
#ifdef USE_FLOAT
LD2D(p, &d);
ST(i).f = d;
#else
LD2D(p, &ST(i).d);
#endif
p+=10;
}
}
@ -173,7 +178,12 @@ void Run66DD(x86emu_t *emu)
char* p =(char*)op1;
p += 14;
for (int i=0; i<8; ++i) {
#ifdef USE_FLOAT
d = ST(i).f;
D2LD(&d, p);
#else
D2LD(&ST(i).d, p);
#endif
p+=10;
}
}