stdlib: move all mslibc functions to SDL_mslibc.c

This allows disabling LTO on them by only specifying a single file.
This commit is contained in:
Anonymous Maarten
2022-07-26 21:47:52 +02:00
committed by Sam Lantinga
parent a169259f62
commit 5da85376b9
12 changed files with 829 additions and 894 deletions

View File

@@ -533,697 +533,114 @@ int SDL_toupper(int x) { return ((x) >= 'a') && ((x) <= 'z') ? ('A'+((x)-'a')) :
int SDL_tolower(int x) { return ((x) >= 'A') && ((x) <= 'Z') ? ('a'+((x)-'A')) : (x); }
#endif
/* This file contains a portable memcpy manipulation function for SDL */
void *
SDL_memcpy(SDL_OUT_BYTECAP(len) void *dst, SDL_IN_BYTECAP(len) const void *src, size_t len)
{
#ifdef __GNUC__
/* Presumably this is well tuned for speed.
On my machine this is twice as fast as the C code below.
*/
return __builtin_memcpy(dst, src, len);
#elif defined(HAVE_MEMCPY)
return memcpy(dst, src, len);
#elif defined(HAVE_BCOPY)
bcopy(src, dst, len);
return dst;
#else
/* GCC 4.9.0 with -O3 will generate movaps instructions with the loop
using Uint32* pointers, so we need to make sure the pointers are
aligned before we loop using them.
*/
if (((uintptr_t)src & 0x3) || ((uintptr_t)dst & 0x3)) {
/* Do an unaligned byte copy */
Uint8 *srcp1 = (Uint8 *)src;
Uint8 *dstp1 = (Uint8 *)dst;
while (len--) {
*dstp1++ = *srcp1++;
}
} else {
size_t left = (len % 4);
Uint32 *srcp4, *dstp4;
Uint8 *srcp1, *dstp1;
srcp4 = (Uint32 *) src;
dstp4 = (Uint32 *) dst;
len /= 4;
while (len--) {
*dstp4++ = *srcp4++;
}
srcp1 = (Uint8 *) srcp4;
dstp1 = (Uint8 *) dstp4;
switch (left) {
case 3:
*dstp1++ = *srcp1++;
case 2:
*dstp1++ = *srcp1++;
case 1:
*dstp1++ = *srcp1++;
}
}
return dst;
#endif /* __GNUC__ */
}
void *
SDL_memset(SDL_OUT_BYTECAP(len) void *dst, int c, size_t len)
{
#if defined(HAVE_MEMSET)
return memset(dst, c, len);
#else
size_t left;
Uint32 *dstp4;
Uint8 *dstp1 = (Uint8 *) dst;
Uint8 value1;
Uint32 value4;
/* The value used in memset() is a byte, passed as an int */
c &= 0xff;
/* The destination pointer needs to be aligned on a 4-byte boundary to
* execute a 32-bit set. Set first bytes manually if needed until it is
* aligned. */
value1 = (Uint8)c;
while ((uintptr_t)dstp1 & 0x3) {
if (len--) {
*dstp1++ = value1;
} else {
return dst;
}
}
value4 = ((Uint32)c | ((Uint32)c << 8) | ((Uint32)c << 16) | ((Uint32)c << 24));
dstp4 = (Uint32 *) dstp1;
left = (len % 4);
len /= 4;
while (len--) {
*dstp4++ = value4;
}
dstp1 = (Uint8 *) dstp4;
switch (left) {
case 3:
*dstp1++ = value1;
case 2:
*dstp1++ = value1;
case 1:
*dstp1++ = value1;
}
return dst;
#endif /* HAVE_MEMSET */
}
#if defined(HAVE_CTYPE_H) && defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
int SDL_isblank(int x) { return isblank(x); }
#else
int SDL_isblank(int x) { return ((x) == ' ') || ((x) == '\t'); }
#endif
#ifndef HAVE_LIBC
/* These are some C runtime intrinsics that need to be defined */
#if defined(_MSC_VER)
#ifndef __FLTUSED__
#define __FLTUSED__
__declspec(selectany) int _fltused = 1;
#endif
/* The optimizer on Visual Studio 2005 and later generates memcpy() and memset() calls.
Always provide it for the SDL2 DLL, but skip it when building static lib w/ static runtime. */
#if (_MSC_VER >= 1400) && (!defined(_MT) || defined(DLL_EXPORT))
extern void *memcpy(void* dst, const void* src, size_t len);
#pragma intrinsic(memcpy)
#pragma function(memcpy)
void *
memcpy(void *dst, const void *src, size_t len)
{
return SDL_memcpy(dst, src, len);
}
extern void *memset(void* dst, int c, size_t len);
#pragma intrinsic(memset)
#pragma function(memset)
void *
memset(void *dst, int c, size_t len)
{
return SDL_memset(dst, c, len);
}
#endif /* (_MSC_VER >= 1400) && (!defined(_MT) || defined(DLL_EXPORT)) */
#ifdef _M_IX86
/* Float to long */
void
__declspec(naked)
_ftol()
{
/* *INDENT-OFF* */
__asm {
push ebp
mov ebp,esp
sub esp,20h
and esp,0FFFFFFF0h
fld st(0)
fst dword ptr [esp+18h]
fistp qword ptr [esp+10h]
fild qword ptr [esp+10h]
mov edx,dword ptr [esp+18h]
mov eax,dword ptr [esp+10h]
test eax,eax
je integer_QnaN_or_zero
arg_is_not_integer_QnaN:
fsubp st(1),st
test edx,edx
jns positive
fstp dword ptr [esp]
mov ecx,dword ptr [esp]
xor ecx,80000000h
add ecx,7FFFFFFFh
adc eax,0
mov edx,dword ptr [esp+14h]
adc edx,0
jmp localexit
positive:
fstp dword ptr [esp]
mov ecx,dword ptr [esp]
add ecx,7FFFFFFFh
sbb eax,0
mov edx,dword ptr [esp+14h]
sbb edx,0
jmp localexit
integer_QnaN_or_zero:
mov edx,dword ptr [esp+14h]
test edx,7FFFFFFFh
jne arg_is_not_integer_QnaN
fstp dword ptr [esp+18h]
fstp dword ptr [esp+18h]
localexit:
leave
ret
}
/* *INDENT-ON* */
}
void
_ftol2_sse()
{
_ftol();
}
/* 64-bit math operators for 32-bit systems */
void
__declspec(naked)
_allmul()
{
/* *INDENT-OFF* */
__asm {
mov eax, dword ptr[esp+8]
mov ecx, dword ptr[esp+10h]
or ecx, eax
mov ecx, dword ptr[esp+0Ch]
jne hard
mov eax, dword ptr[esp+4]
mul ecx
ret 10h
hard:
push ebx
mul ecx
mov ebx, eax
mov eax, dword ptr[esp+8]
mul dword ptr[esp+14h]
add ebx, eax
mov eax, dword ptr[esp+8]
mul ecx
add edx, ebx
pop ebx
ret 10h
}
/* *INDENT-ON* */
}
void
__declspec(naked)
_alldiv()
{
/* *INDENT-OFF* */
__asm {
push edi
push esi
push ebx
xor edi,edi
mov eax,dword ptr [esp+14h]
or eax,eax
jge L1
inc edi
mov edx,dword ptr [esp+10h]
neg eax
neg edx
sbb eax,0
mov dword ptr [esp+14h],eax
mov dword ptr [esp+10h],edx
L1:
mov eax,dword ptr [esp+1Ch]
or eax,eax
jge L2
inc edi
mov edx,dword ptr [esp+18h]
neg eax
neg edx
sbb eax,0
mov dword ptr [esp+1Ch],eax
mov dword ptr [esp+18h],edx
L2:
or eax,eax
jne L3
mov ecx,dword ptr [esp+18h]
mov eax,dword ptr [esp+14h]
xor edx,edx
div ecx
mov ebx,eax
mov eax,dword ptr [esp+10h]
div ecx
mov edx,ebx
jmp L4
L3:
mov ebx,eax
mov ecx,dword ptr [esp+18h]
mov edx,dword ptr [esp+14h]
mov eax,dword ptr [esp+10h]
L5:
shr ebx,1
rcr ecx,1
shr edx,1
rcr eax,1
or ebx,ebx
jne L5
div ecx
mov esi,eax
mul dword ptr [esp+1Ch]
mov ecx,eax
mov eax,dword ptr [esp+18h]
mul esi
add edx,ecx
jb L6
cmp edx,dword ptr [esp+14h]
ja L6
jb L7
cmp eax,dword ptr [esp+10h]
jbe L7
L6:
dec esi
L7:
xor edx,edx
mov eax,esi
L4:
dec edi
jne L8
neg edx
neg eax
sbb edx,0
L8:
pop ebx
pop esi
pop edi
ret 10h
}
/* *INDENT-ON* */
}
void
__declspec(naked)
_aulldiv()
{
/* *INDENT-OFF* */
__asm {
push ebx
push esi
mov eax,dword ptr [esp+18h]
or eax,eax
jne L1
mov ecx,dword ptr [esp+14h]
mov eax,dword ptr [esp+10h]
xor edx,edx
div ecx
mov ebx,eax
mov eax,dword ptr [esp+0Ch]
div ecx
mov edx,ebx
jmp L2
L1:
mov ecx,eax
mov ebx,dword ptr [esp+14h]
mov edx,dword ptr [esp+10h]
mov eax,dword ptr [esp+0Ch]
L3:
shr ecx,1
rcr ebx,1
shr edx,1
rcr eax,1
or ecx,ecx
jne L3
div ebx
mov esi,eax
mul dword ptr [esp+18h]
mov ecx,eax
mov eax,dword ptr [esp+14h]
mul esi
add edx,ecx
jb L4
cmp edx,dword ptr [esp+10h]
ja L4
jb L5
cmp eax,dword ptr [esp+0Ch]
jbe L5
L4:
dec esi
L5:
xor edx,edx
mov eax,esi
L2:
pop esi
pop ebx
ret 10h
}
/* *INDENT-ON* */
}
void
__declspec(naked)
_allrem()
{
/* *INDENT-OFF* */
__asm {
push ebx
push edi
xor edi,edi
mov eax,dword ptr [esp+10h]
or eax,eax
jge L1
inc edi
mov edx,dword ptr [esp+0Ch]
neg eax
neg edx
sbb eax,0
mov dword ptr [esp+10h],eax
mov dword ptr [esp+0Ch],edx
L1:
mov eax,dword ptr [esp+18h]
or eax,eax
jge L2
mov edx,dword ptr [esp+14h]
neg eax
neg edx
sbb eax,0
mov dword ptr [esp+18h],eax
mov dword ptr [esp+14h],edx
L2:
or eax,eax
jne L3
mov ecx,dword ptr [esp+14h]
mov eax,dword ptr [esp+10h]
xor edx,edx
div ecx
mov eax,dword ptr [esp+0Ch]
div ecx
mov eax,edx
xor edx,edx
dec edi
jns L4
jmp L8
L3:
mov ebx,eax
mov ecx,dword ptr [esp+14h]
mov edx,dword ptr [esp+10h]
mov eax,dword ptr [esp+0Ch]
L5:
shr ebx,1
rcr ecx,1
shr edx,1
rcr eax,1
or ebx,ebx
jne L5
div ecx
mov ecx,eax
mul dword ptr [esp+18h]
xchg eax,ecx
mul dword ptr [esp+14h]
add edx,ecx
jb L6
cmp edx,dword ptr [esp+10h]
ja L6
jb L7
cmp eax,dword ptr [esp+0Ch]
jbe L7
L6:
sub eax,dword ptr [esp+14h]
sbb edx,dword ptr [esp+18h]
L7:
sub eax,dword ptr [esp+0Ch]
sbb edx,dword ptr [esp+10h]
dec edi
jns L8
L4:
neg edx
neg eax
sbb edx,0
L8:
pop edi
pop ebx
ret 10h
}
/* *INDENT-ON* */
}
void
__declspec(naked)
_aullrem()
{
/* *INDENT-OFF* */
__asm {
push ebx
mov eax,dword ptr [esp+14h]
or eax,eax
jne L1
mov ecx,dword ptr [esp+10h]
mov eax,dword ptr [esp+0Ch]
xor edx,edx
div ecx
mov eax,dword ptr [esp+8]
div ecx
mov eax,edx
xor edx,edx
jmp L2
L1:
mov ecx,eax
mov ebx,dword ptr [esp+10h]
mov edx,dword ptr [esp+0Ch]
mov eax,dword ptr [esp+8]
L3:
shr ecx,1
rcr ebx,1
shr edx,1
rcr eax,1
or ecx,ecx
jne L3
div ebx
mov ecx,eax
mul dword ptr [esp+14h]
xchg eax,ecx
mul dword ptr [esp+10h]
add edx,ecx
jb L4
cmp edx,dword ptr [esp+0Ch]
ja L4
jb L5
cmp eax,dword ptr [esp+8]
jbe L5
L4:
sub eax,dword ptr [esp+10h]
sbb edx,dword ptr [esp+14h]
L5:
sub eax,dword ptr [esp+8]
sbb edx,dword ptr [esp+0Ch]
neg edx
neg eax
sbb edx,0
L2:
pop ebx
ret 10h
}
/* *INDENT-ON* */
}
void
__declspec(naked)
_alldvrm()
{
/* *INDENT-OFF* */
__asm {
push edi
push esi
push ebp
xor edi,edi
xor ebp,ebp
mov eax,dword ptr [esp+14h]
or eax,eax
jge L1
inc edi
inc ebp
mov edx,dword ptr [esp+10h]
neg eax
neg edx
sbb eax,0
mov dword ptr [esp+14h],eax
mov dword ptr [esp+10h],edx
L1:
mov eax,dword ptr [esp+1Ch]
or eax,eax
jge L2
inc edi
mov edx,dword ptr [esp+18h]
neg eax
neg edx
sbb eax,0
mov dword ptr [esp+1Ch],eax
mov dword ptr [esp+18h],edx
L2:
or eax,eax
jne L3
mov ecx,dword ptr [esp+18h]
mov eax,dword ptr [esp+14h]
xor edx,edx
div ecx
mov ebx,eax
mov eax,dword ptr [esp+10h]
div ecx
mov esi,eax
mov eax,ebx
mul dword ptr [esp+18h]
mov ecx,eax
mov eax,esi
mul dword ptr [esp+18h]
add edx,ecx
jmp L4
L3:
mov ebx,eax
mov ecx,dword ptr [esp+18h]
mov edx,dword ptr [esp+14h]
mov eax,dword ptr [esp+10h]
L5:
shr ebx,1
rcr ecx,1
shr edx,1
rcr eax,1
or ebx,ebx
jne L5
div ecx
mov esi,eax
mul dword ptr [esp+1Ch]
mov ecx,eax
mov eax,dword ptr [esp+18h]
mul esi
add edx,ecx
jb L6
cmp edx,dword ptr [esp+14h]
ja L6
jb L7
cmp eax,dword ptr [esp+10h]
jbe L7
L6:
dec esi
sub eax,dword ptr [esp+18h]
sbb edx,dword ptr [esp+1Ch]
L7:
xor ebx,ebx
L4:
sub eax,dword ptr [esp+10h]
sbb edx,dword ptr [esp+14h]
dec ebp
jns L9
neg edx
neg eax
sbb edx,0
L9:
mov ecx,edx
mov edx,ebx
mov ebx,ecx
mov ecx,eax
mov eax,esi
dec edi
jne L8
neg edx
neg eax
sbb edx,0
L8:
pop ebp
pop esi
pop edi
ret 10h
}
/* *INDENT-ON* */
}
void
__declspec(naked)
_aulldvrm()
{
/* *INDENT-OFF* */
__asm {
push esi
mov eax,dword ptr [esp+14h]
or eax,eax
jne L1
mov ecx,dword ptr [esp+10h]
mov eax,dword ptr [esp+0Ch]
xor edx,edx
div ecx
mov ebx,eax
mov eax,dword ptr [esp+8]
div ecx
mov esi,eax
mov eax,ebx
mul dword ptr [esp+10h]
mov ecx,eax
mov eax,esi
mul dword ptr [esp+10h]
add edx,ecx
jmp L2
L1:
mov ecx,eax
mov ebx,dword ptr [esp+10h]
mov edx,dword ptr [esp+0Ch]
mov eax,dword ptr [esp+8]
L3:
shr ecx,1
rcr ebx,1
shr edx,1
rcr eax,1
or ecx,ecx
jne L3
div ebx
mov esi,eax
mul dword ptr [esp+14h]
mov ecx,eax
mov eax,dword ptr [esp+10h]
mul esi
add edx,ecx
jb L4
cmp edx,dword ptr [esp+0Ch]
ja L4
jb L5
cmp eax,dword ptr [esp+8]
jbe L5
L4:
dec esi
sub eax,dword ptr [esp+10h]
sbb edx,dword ptr [esp+14h]
L5:
xor ebx,ebx
L2:
sub eax,dword ptr [esp+8]
sbb edx,dword ptr [esp+0Ch]
neg edx
neg eax
sbb edx,0
mov ecx,edx
mov edx,ebx
mov ebx,ecx
mov ecx,eax
mov eax,esi
pop esi
ret 10h
}
/* *INDENT-ON* */
}
void
__declspec(naked)
_allshl()
{
/* *INDENT-OFF* */
__asm {
cmp cl,40h
jae RETZERO
cmp cl,20h
jae MORE32
shld edx,eax,cl
shl eax,cl
ret
MORE32:
mov edx,eax
xor eax,eax
and cl,1Fh
shl edx,cl
ret
RETZERO:
xor eax,eax
xor edx,edx
ret
}
/* *INDENT-ON* */
}
void
__declspec(naked)
_allshr()
{
/* *INDENT-OFF* */
__asm {
cmp cl,3Fh
jae RETSIGN
cmp cl,20h
jae MORE32
shrd eax,edx,cl
sar edx,cl
ret
MORE32:
mov eax,edx
sar edx,1Fh
and cl,1Fh
sar eax,cl
ret
RETSIGN:
sar edx,1Fh
mov eax,edx
ret
}
/* *INDENT-ON* */
}
void
__declspec(naked)
_aullshr()
{
/* *INDENT-OFF* */
__asm {
cmp cl,40h
jae RETZERO
cmp cl,20h
jae MORE32
shrd eax,edx,cl
shr edx,cl
ret
MORE32:
mov eax,edx
xor edx,edx
and cl,1Fh
shr eax,cl
ret
RETZERO:
xor eax,eax
xor edx,edx
ret
}
/* *INDENT-ON* */
}
#endif /* _M_IX86 */
#endif /* MSC_VER */
#endif /* !HAVE_LIBC */
/* vi: set ts=4 sw=4 expandtab: */