mirror of https://github.com/PCSX2/pcsx2.git
Merge branch 'alignment' of git://github.com/xsacha/pcsx2 into xsacha-alignment
Gregory: Get all changes but keep C11 code path which will be the future (if someone can find info on Visual Studio support) * Keep the old posix api to use address sanitizer on gcc 4.9 Conflicts: common/build/Utilities/utilities.vcxproj.filters common/build/Utilities/utilities_vs2012.vcxproj.filters common/build/Utilities/utilities_vs2013.vcxproj.filters common/include/Utilities/MemcpyFast.h common/include/Utilities/StringHelpers.h common/src/Utilities/AlignedMalloc.cpp common/src/Utilities/vssprintf.cpp plugins/GSdx/stdafx.cpp
This commit is contained in:
commit
4feeaac7d1
|
@ -235,7 +235,7 @@ namespace HostSys
|
|||
|
||||
extern void MemProtect( void* baseaddr, size_t size, const PageProtectionMode& mode );
|
||||
|
||||
extern void Munmap( void* base, size_t size );
|
||||
extern void Munmap( void* base, size_t size ) { Munmap( (uptr)base, size); }
|
||||
|
||||
template< uint size >
|
||||
void MemProtectStatic( u8 (&arr)[size], const PageProtectionMode& mode )
|
||||
|
|
|
@ -15,8 +15,6 @@
|
|||
|
||||
#pragma once
|
||||
|
||||
#include <wx/string.h>
|
||||
|
||||
namespace HashTools {
|
||||
|
||||
/// <summary>
|
||||
|
|
|
@ -31,7 +31,7 @@
|
|||
|
||||
#ifdef _WIN32
|
||||
|
||||
// Force availability of to WinNT APIs (change to 0x600 to enable XP-specific APIs)
|
||||
// Force availability of to WinNT APIs (change to 0x600 to disable XP-specific APIs)
|
||||
#ifndef WINVER
|
||||
#define WINVER 0x0501
|
||||
#define _WIN32_WINNT 0x0501
|
||||
|
|
|
@ -16,8 +16,7 @@
|
|||
#ifndef _LNX_MEMZERO_H_
|
||||
#define _LNX_MEMZERO_H_
|
||||
|
||||
// This header contains non-optimized implementation of memzero_ptr and memset8,
|
||||
// memset16, etc.
|
||||
// This header contains non-optimized implementation of memzero_ptr and memset8, etc
|
||||
|
||||
template< u32 data, typename T >
|
||||
static __fi void memset32( T& obj )
|
||||
|
@ -51,18 +50,6 @@ static __fi void memset8( T& obj )
|
|||
}
|
||||
}
|
||||
|
||||
template< u16 data, typename T >
|
||||
static __fi void memset16( T& obj )
|
||||
{
|
||||
if( (sizeof(T) & 0x3) != 0 )
|
||||
_memset16_unaligned( &obj, data, sizeof( T ) );
|
||||
else {
|
||||
const u32 data32 = data + (data<<16);
|
||||
memset32<data32>( obj );
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
// An optimized memset for 8 bit destination data.
|
||||
template< u8 data, size_t bytes >
|
||||
static __fi void memset_8( void *dest )
|
||||
|
|
|
@ -186,94 +186,6 @@ static __fi void memset_8( void *dest )
|
|||
}
|
||||
}
|
||||
|
||||
template< u16 data, size_t _bytes >
|
||||
static __fi void memset_16( void *dest )
|
||||
{
|
||||
if( MZFbytes == 0 ) return;
|
||||
|
||||
// Assertion: data length must be a multiple of 16 or 32 bits
|
||||
pxAssume( (MZFbytes & 0x1) == 0 );
|
||||
|
||||
if( (MZFbytes & 0x3) != 0 )
|
||||
{
|
||||
// Unaligned data length. No point in doing an optimized inline version (too complicated with
|
||||
// remainders and such).
|
||||
|
||||
_memset16_unaligned( dest, data, MZFbytes );
|
||||
return;
|
||||
}
|
||||
|
||||
//u64 _xmm_backup[2];
|
||||
|
||||
// This function only works on 32-bit alignments of data copied.
|
||||
pxAssume( (MZFbytes & 0x3) == 0 );
|
||||
|
||||
enum
|
||||
{
|
||||
remdat = MZFbytes >> 2,
|
||||
data32 = data + (data<<16)
|
||||
};
|
||||
|
||||
// macro to execute the x86/32 "stosd" copies.
|
||||
switch( remdat )
|
||||
{
|
||||
case 1:
|
||||
*(u32*)dest = data32;
|
||||
return;
|
||||
|
||||
case 2:
|
||||
((u32*)dest)[0] = data32;
|
||||
((u32*)dest)[1] = data32;
|
||||
return;
|
||||
|
||||
case 3:
|
||||
__asm
|
||||
{
|
||||
mov edi, dest;
|
||||
mov eax, data32;
|
||||
stosd;
|
||||
stosd;
|
||||
stosd;
|
||||
}
|
||||
return;
|
||||
|
||||
case 4:
|
||||
__asm
|
||||
{
|
||||
mov edi, dest;
|
||||
mov eax, data32;
|
||||
stosd;
|
||||
stosd;
|
||||
stosd;
|
||||
stosd;
|
||||
}
|
||||
return;
|
||||
|
||||
case 5:
|
||||
__asm
|
||||
{
|
||||
mov edi, dest;
|
||||
mov eax, data32;
|
||||
stosd;
|
||||
stosd;
|
||||
stosd;
|
||||
stosd;
|
||||
stosd;
|
||||
}
|
||||
return;
|
||||
|
||||
default:
|
||||
__asm
|
||||
{
|
||||
mov ecx, remdat;
|
||||
mov edi, dest;
|
||||
mov eax, data32;
|
||||
rep stosd;
|
||||
}
|
||||
return
|
||||
}
|
||||
}
|
||||
|
||||
template< u32 data, size_t MZFbytes >
|
||||
static __fi void memset_32( void *dest )
|
||||
{
|
||||
|
@ -372,13 +284,6 @@ static __fi void memset8( T& object )
|
|||
memset_8<data, sizeof(T)>( &object );
|
||||
}
|
||||
|
||||
// This method clears an object with the given 16 bit value.
|
||||
template< u16 data, typename T >
|
||||
static __fi void memset16( T& object )
|
||||
{
|
||||
memset_16<data, sizeof(T)>( &object );
|
||||
}
|
||||
|
||||
// This method clears an object with the given 32 bit value.
|
||||
template< u32 data, typename T >
|
||||
static __fi void memset32( T& object )
|
||||
|
|
|
@ -33,19 +33,9 @@ void* __fastcall pcsx2_aligned_malloc(size_t size, size_t align)
|
|||
#if defined(__USE_ISOC11) && !defined(ASAN_WORKAROUND) // not supported yet on gcc 4.9
|
||||
return aligned_alloc(align, size);
|
||||
#else
|
||||
u8* p = (u8*)malloc(size+align+headsize);
|
||||
|
||||
// start alignment calculations from past the header.
|
||||
uptr pasthead = (uptr)(p+headsize);
|
||||
uptr aligned = (pasthead + align-1) & ~(align-1);
|
||||
|
||||
AlignedMallocHeader* header = (AlignedMallocHeader*)(aligned-headsize);
|
||||
pxAssert( (uptr)header >= (uptr)p );
|
||||
|
||||
header->baseptr = p;
|
||||
header->size = size;
|
||||
|
||||
return (void*)aligned;
|
||||
void *result=0;
|
||||
posix_memalign(&result, alignment, size);
|
||||
return result;
|
||||
#endif
|
||||
}
|
||||
|
||||
|
@ -57,45 +47,13 @@ void* __fastcall pcsx2_aligned_realloc(void* handle, size_t size, size_t align)
|
|||
|
||||
if( handle != NULL )
|
||||
{
|
||||
#if defined(__USE_ISOC11) && !defined(ASAN_WORKAROUND) // not supported yet on gcc 4.9
|
||||
memcpy_fast( newbuf, handle, size );
|
||||
free( handle );
|
||||
#else
|
||||
AlignedMallocHeader* header = (AlignedMallocHeader*)((uptr)handle - headsize);
|
||||
memcpy_fast( newbuf, handle, std::min( size, header->size ) );
|
||||
free( header->baseptr );
|
||||
#endif
|
||||
}
|
||||
return newbuf;
|
||||
}
|
||||
|
||||
__fi void pcsx2_aligned_free(void* pmem)
|
||||
{
|
||||
#if defined(__USE_ISOC11) && !defined(ASAN_WORKAROUND) // not supported yet on gcc 4.9
|
||||
free(pmem);
|
||||
#else
|
||||
if( pmem == NULL ) return;
|
||||
AlignedMallocHeader* header = (AlignedMallocHeader*)((uptr)pmem - headsize);
|
||||
free( header->baseptr );
|
||||
#endif
|
||||
}
|
||||
|
||||
// ----------------------------------------------------------------------------
|
||||
// And for lack of a better home ...
|
||||
|
||||
|
||||
// Special unaligned memset used when all other optimized memsets fail (it's called from
|
||||
// memzero_obj and stuff).
|
||||
__fi void _memset16_unaligned( void* dest, u16 data, size_t size )
|
||||
{
|
||||
pxAssume( (size & 0x1) == 0 );
|
||||
|
||||
u16* dst = (u16*)dest;
|
||||
for(int i=size; i; --i, ++dst )
|
||||
*dst = data;
|
||||
}
|
||||
|
||||
__fi void HostSys::Munmap( void* base, size_t size )
|
||||
{
|
||||
Munmap( (uptr)base, size );
|
||||
}
|
||||
|
|
|
@ -54,11 +54,11 @@ endif(CMAKE_BUILD_TYPE STREQUAL Release)
|
|||
|
||||
# variable with all sources of this library
|
||||
set(UtilitiesSources
|
||||
AlignedMalloc.cpp
|
||||
../../include/Utilities/FixedPointTypes.inl
|
||||
../../include/Utilities/EventSource.inl
|
||||
../../include/Utilities/SafeArray.inl
|
||||
../../include/Utilities/TlsVariable.inl
|
||||
AlignedMalloc.cpp
|
||||
CheckedStaticBox.cpp
|
||||
Console.cpp
|
||||
EventSource.cpp
|
||||
|
|
|
@ -103,7 +103,7 @@ void vmfree(void* ptr, size_t size)
|
|||
|
||||
#endif
|
||||
|
||||
#if !defined(_MSC_VER) && !defined(HAVE_ALIGNED_MALLOC)
|
||||
#if !defined(_MSC_VER)
|
||||
|
||||
// declare linux equivalents (alignment must be power of 2 (1,2,4...2^15)
|
||||
|
||||
|
@ -111,18 +111,9 @@ void vmfree(void* ptr, size_t size)
|
|||
|
||||
void* _aligned_malloc(size_t size, size_t alignment)
|
||||
{
|
||||
ASSERT(alignment <= 0x8000);
|
||||
size_t r = (size_t)malloc(size + --alignment + 2);
|
||||
size_t off = (r + 2 + alignment) & ~(size_t)alignment;
|
||||
if(!r) return NULL;
|
||||
((uint16*)off)[-1] = (uint16)(off-r);
|
||||
return (void*)off;
|
||||
}
|
||||
|
||||
void _aligned_free(void* p)
|
||||
{
|
||||
if(!p) return;
|
||||
free((void*)((size_t)p-((uint16*)p)[-1]));
|
||||
void* ret;
|
||||
posix_memalign(&return, a, size);
|
||||
return ret;
|
||||
}
|
||||
|
||||
#endif
|
||||
|
|
|
@ -359,19 +359,20 @@ struct aligned_free_second {template<class T> void operator()(T& p) {_aligned_fr
|
|||
#undef abs
|
||||
|
||||
#if !defined(_MSC_VER)
|
||||
|
||||
#if defined(__USE_ISOC11) && !defined(ASAN_WORKAROUND) // not supported yet on gcc 4.9
|
||||
|
||||
#define _aligned_malloc(size, a) aligned_alloc(a, size)
|
||||
static inline void _aligned_free(void* p) { free(p); }
|
||||
|
||||
#elif !defined(HAVE_ALIGNED_MALLOC)
|
||||
#else
|
||||
|
||||
extern void* _aligned_malloc(size_t size, size_t alignment);
|
||||
extern void _aligned_free(void* p);
|
||||
|
||||
#endif
|
||||
|
||||
static inline void _aligned_free(void* p) {
|
||||
free(p);
|
||||
}
|
||||
|
||||
// http://svn.reactos.org/svn/reactos/trunk/reactos/include/crt/mingw32/intrin_x86.h?view=markup
|
||||
|
||||
__forceinline unsigned char _BitScanForward(unsigned long* const Index, const unsigned long Mask)
|
||||
|
|
|
@ -69,32 +69,17 @@ extern std::string s_strIniPath;
|
|||
extern u32 THR_KeyEvent; // value for passing out key events beetwen threads
|
||||
extern bool THR_bShift;
|
||||
|
||||
#if !defined(_MSC_VER) && !defined(HAVE_ALIGNED_MALLOC)
|
||||
|
||||
// declare linux equivalents
|
||||
static __forceinline void* pcsx2_aligned_malloc(size_t size, size_t align)
|
||||
{
|
||||
assert( align < 0x10000 );
|
||||
char* p = (char*)malloc(size+align);
|
||||
int off = 2+align - ((int)(uptr)(p+2) % align);
|
||||
|
||||
p += off;
|
||||
*(u16*)(p-2) = off;
|
||||
|
||||
return p;
|
||||
#if !defined(_MSC_VER)
|
||||
// declare linux equivalents (alignment must be power of 2 (1,2,4...2^15)
|
||||
static __forceinline void* _aligned_malloc(size_t size, size_t alignment) {
|
||||
void *result=0;
|
||||
posix_memalign(&result, alignment, size);
|
||||
return result;
|
||||
}
|
||||
|
||||
static __forceinline void pcsx2_aligned_free(void* pmem)
|
||||
{
|
||||
if( pmem != NULL ) {
|
||||
char* p = (char*)pmem;
|
||||
free(p - (int)*(u16*)(p-2));
|
||||
static __forceinline void _aligned_free(void* p) {
|
||||
free(p);
|
||||
}
|
||||
}
|
||||
|
||||
#define _aligned_malloc pcsx2_aligned_malloc
|
||||
#define _aligned_free pcsx2_aligned_free
|
||||
|
||||
#endif
|
||||
|
||||
#include <sys/timeb.h> // ftime(), struct timeb
|
||||
|
|
|
@ -63,26 +63,15 @@ inline u64 GetMicroTime()
|
|||
#endif
|
||||
}
|
||||
|
||||
#if !defined(_MSC_VER) && !defined(HAVE_ALIGNED_MALLOC)
|
||||
|
||||
#include <assert.h>
|
||||
|
||||
#if !defined(_MSC_VER)
|
||||
// declare linux equivalents (alignment must be power of 2 (1,2,4...2^15)
|
||||
static __forceinline void* pcsx2_aligned_malloc(size_t size, size_t alignment) {
|
||||
assert(alignment <= 0x8000);
|
||||
uptr r = (uptr)malloc(size + --alignment + 2);
|
||||
uptr o = (r + 2 + alignment) & ~(uptr)alignment;
|
||||
if (!r) return NULL;
|
||||
((u16*)o)[-1] = (u16)(o-r);
|
||||
return (void*)o;
|
||||
static __forceinline void* _aligned_malloc(size_t size, size_t alignment) {
|
||||
void *result=0;
|
||||
posix_memalign(&result, alignment, size);
|
||||
return result;
|
||||
}
|
||||
|
||||
static __forceinline void pcsx2_aligned_free(void* p) {
|
||||
if (!p) return;
|
||||
free((void*)((uptr)p-((u16*)p)[-1]));
|
||||
static __forceinline void _aligned_free(void* p) {
|
||||
free(p);
|
||||
}
|
||||
|
||||
|
||||
#define _aligned_malloc pcsx2_aligned_malloc
|
||||
#define _aligned_free pcsx2_aligned_free
|
||||
#endif
|
||||
|
|
Loading…
Reference in New Issue