Thread: [Mplayerxp-cvslog] SF.net SVN: mplayerxp:[390] mplayerxp/osdep
Brought to you by:
olov
Menu
▾
▴
mplayerxp-cvslog
|
From: <nic...@us...> - 2012-11-17 13:57:03
|
Revision: 390
http://mplayerxp.svn.sourceforge.net/mplayerxp/?rev=390&view=rev
Author: nickols_k
Date: 2012-11-17 13:56:54 +0000 (Sat, 17 Nov 2012)
Log Message:
-----------
more c++ sources
Modified Paths:
--------------
mplayerxp/osdep/Makefile
mplayerxp/osdep/fastmemcpy.h
Added Paths:
-----------
mplayerxp/osdep/aclib.cpp
mplayerxp/osdep/aclib_template.h
Removed Paths:
-------------
mplayerxp/osdep/aclib.c
mplayerxp/osdep/aclib_template.c
Modified: mplayerxp/osdep/Makefile
===================================================================
--- mplayerxp/osdep/Makefile 2012-11-17 10:42:51 UTC (rev 389)
+++ mplayerxp/osdep/Makefile 2012-11-17 13:56:54 UTC (rev 390)
@@ -3,8 +3,8 @@
LIBNAME = libosdep.a
-SRCS=aclib.c
-CXXSRCS=mplib.cpp mp_malloc.cpp getch2.cpp timer-lx.cpp timer.cpp cpudetect.cpp shmem.cpp get_path.cpp
+SRCS=
+CXXSRCS=mplib.cpp mp_malloc.cpp getch2.cpp timer-lx.cpp timer.cpp cpudetect.cpp shmem.cpp get_path.cpp aclib.cpp
OBJS=$(SRCS:.c=.o)
CXXOBJS=$(CXXSRCS:.cpp=.o)
Deleted: mplayerxp/osdep/aclib.c
===================================================================
--- mplayerxp/osdep/aclib.c 2012-11-17 10:42:51 UTC (rev 389)
+++ mplayerxp/osdep/aclib.c 2012-11-17 13:56:54 UTC (rev 390)
@@ -1,98 +0,0 @@
-#include <stdio.h>
-#include <pthread.h>
-#include "mp_config.h"
-#include "mplayerxp.h"
-#define MSGT_CLASS MSGT_GLOBAL
-#include "mp_msg.h"
-
-#if defined(USE_FASTMEMCPY)
-#include "fastmemcpy.h"
-#include "osdep/cpudetect.h"
-
-#define BLOCK_SIZE 4096
-#define CONFUSION_FACTOR 0
-
-#define PVECTOR_ACCEL_H "aclib_template.c"
-#include "pvector/pvector_inc.h"
-
-/*
- aclib - advanced C library ;)
- This file contains functions which improve and expand standard C-library
- see aclib_template.c ... this file only contains runtime cpu detection and config options stuff
- runtime cpu detection by michael niedermayer (mic...@gm...) is under GPL
-*/
-
-static any_t* init_fast_memcpy(any_t* to, const any_t* from, size_t len)
-{
-#ifdef __SSE2__
- if(gCpuCaps.hasSSE2)
- {
- MSG_V("Using SSE2 optimized memcpy\n");
- fast_memcpy_ptr = fast_memcpy_SSE2;
- }
- else
-#endif
-#ifndef __x86_64__
-#ifdef __SSE__
- if(gCpuCaps.hasMMX2)
- {
- MSG_V("Using MMX2 optimized memcpy\n");
- fast_memcpy_ptr = fast_memcpy_SSE;
- }
- else
-#endif
-//#ifdef __MMX__
-// if(gCpuCaps.hasMMX)
-// {
-// MSG_V("Using MMX optimized memcpy\n");
-// fast_memcpy_ptr = fast_memcpy_MMX;
-// }
-// else
-//#endif
-#endif
- {
- MSG_V("Using generic memcpy\n");
- fast_memcpy_ptr = memcpy; /* prior to mmx we use the standart memcpy */
- }
- return (*fast_memcpy_ptr)(to,from,len);
-}
-
-static any_t* init_stream_copy(any_t* to, const any_t* from, size_t len)
-{
-#ifdef __SSE2__
- if(gCpuCaps.hasSSE2)
- {
- MSG_V("Using SSE2 optimized agpcpy\n");
- fast_stream_copy_ptr = fast_stream_copy_SSE2;
- }
-#endif
-#ifndef __x86_64__
-#ifdef __SSE__
- if(gCpuCaps.hasMMX2)
- {
- MSG_V("Using MMX2 optimized agpcpy\n");
- fast_stream_copy_ptr = fast_stream_copy_SSE;
- }
- else
-#endif
-//#ifdef __MMX__
-// if(gCpuCaps.hasMMX)
-// {
-// MSG_V("Using MMX optimized agpcpy\n");
-// fast_stream_copy_ptr = fast_stream_copy_MMX;
-// }
-// else
-//#endif
-#endif
- {
- MSG_V("Using generic optimized agpcpy\n");
- fast_stream_copy_ptr = memcpy; /* prior to mmx we use the standart memcpy */
- }
- return (*fast_stream_copy_ptr)(to,from,len);
-}
-
-any_t*(*fast_memcpy_ptr)(any_t* to, const any_t* from, size_t len) = init_fast_memcpy;
-any_t*(*fast_stream_copy_ptr)(any_t* to, const any_t* from, size_t len) = init_stream_copy;
-
-#endif /* use fastmemcpy */
-
Copied: mplayerxp/osdep/aclib.cpp (from rev 369, mplayerxp/osdep/aclib.c)
===================================================================
--- mplayerxp/osdep/aclib.cpp (rev 0)
+++ mplayerxp/osdep/aclib.cpp 2012-11-17 13:56:54 UTC (rev 390)
@@ -0,0 +1,98 @@
+#include <stdio.h>
+#include <pthread.h>
+#include "mp_config.h"
+#include "mplayerxp.h"
+#define MSGT_CLASS MSGT_GLOBAL
+#include "mp_msg.h"
+
+#if defined(USE_FASTMEMCPY)
+#include "fastmemcpy.h"
+#include "osdep/cpudetect.h"
+
+#define BLOCK_SIZE 4096
+#define CONFUSION_FACTOR 0
+
+#define PVECTOR_ACCEL_H "aclib_template.h"
+#include "pvector/pvector_inc.h"
+
+/*
+ aclib - advanced C library ;)
+ This file contains functions which improve and expand standard C-library
+ see aclib_template.c ... this file only contains runtime cpu detection and config options stuff
+ runtime cpu detection by michael niedermayer (mic...@gm...) is under GPL
+*/
+
+static any_t* init_fast_memcpy(any_t* to, const any_t* from, size_t len)
+{
+#ifdef __SSE2__
+ if(gCpuCaps.hasSSE2)
+ {
+ MSG_V("Using SSE2 optimized memcpy\n");
+ fast_memcpy_ptr = fast_memcpy_SSE2;
+ }
+ else
+#endif
+#ifndef __x86_64__
+#ifdef __SSE__
+ if(gCpuCaps.hasMMX2)
+ {
+ MSG_V("Using MMX2 optimized memcpy\n");
+ fast_memcpy_ptr = fast_memcpy_SSE;
+ }
+ else
+#endif
+//#ifdef __MMX__
+// if(gCpuCaps.hasMMX)
+// {
+// MSG_V("Using MMX optimized memcpy\n");
+// fast_memcpy_ptr = fast_memcpy_MMX;
+// }
+// else
+//#endif
+#endif
+ {
+ MSG_V("Using generic memcpy\n");
+ fast_memcpy_ptr = memcpy; /* prior to mmx we use the standart memcpy */
+ }
+ return (*fast_memcpy_ptr)(to,from,len);
+}
+
+static any_t* init_stream_copy(any_t* to, const any_t* from, size_t len)
+{
+#ifdef __SSE2__
+ if(gCpuCaps.hasSSE2)
+ {
+ MSG_V("Using SSE2 optimized agpcpy\n");
+ fast_stream_copy_ptr = fast_stream_copy_SSE2;
+ }
+#endif
+#ifndef __x86_64__
+#ifdef __SSE__
+ if(gCpuCaps.hasMMX2)
+ {
+ MSG_V("Using MMX2 optimized agpcpy\n");
+ fast_stream_copy_ptr = fast_stream_copy_SSE;
+ }
+ else
+#endif
+//#ifdef __MMX__
+// if(gCpuCaps.hasMMX)
+// {
+// MSG_V("Using MMX optimized agpcpy\n");
+// fast_stream_copy_ptr = fast_stream_copy_MMX;
+// }
+// else
+//#endif
+#endif
+ {
+ MSG_V("Using generic optimized agpcpy\n");
+ fast_stream_copy_ptr = memcpy; /* prior to mmx we use the standart memcpy */
+ }
+ return (*fast_stream_copy_ptr)(to,from,len);
+}
+
+any_t*(*fast_memcpy_ptr)(any_t* to, const any_t* from, size_t len) = init_fast_memcpy;
+any_t*(*fast_stream_copy_ptr)(any_t* to, const any_t* from, size_t len) = init_stream_copy;
+
+#endif /* use fastmemcpy */
+
Deleted: mplayerxp/osdep/aclib_template.c
===================================================================
--- mplayerxp/osdep/aclib_template.c 2012-11-17 10:42:51 UTC (rev 389)
+++ mplayerxp/osdep/aclib_template.c 2012-11-17 13:56:54 UTC (rev 390)
@@ -1,179 +0,0 @@
-/*
- aclib - advanced C library ;)
- This file contains functions which improve and expand standard C-library
-*/
-#include "pvector/pvector.h"
-
-#ifdef HAVE_INT_PVECTOR
-/* for small memory blocks (<256 bytes) this version is faster */
-#undef small_memcpy
-#ifdef __x86_64__
-#define small_memcpy(to,from,n)\
-{\
-register unsigned long int siz;\
-register unsigned long int dummy;\
- siz=n&0x7; n>>=3;\
- if(siz)\
-__asm__ __volatile__(\
- "rep; movsb"\
- :"=&D"(to), "=&S"(from), "=&c"(dummy)\
-/* It's most portable way to notify compiler */\
-/* that edi, esi and ecx are clobbered in asm block. */\
-/* Thanks to A'rpi for hint!!! */\
- :"0" (to), "1" (from),"2" (siz)\
- : "memory","cc");\
- if(n)\
-__asm__ __volatile__(\
- "rep; movsq"\
- :"=&D"(to), "=&S"(from), "=&c"(dummy)\
-/* It's most portable way to notify compiler */\
-/* that edi, esi and ecx are clobbered in asm block. */\
-/* Thanks to A'rpi for hint!!! */\
- :"0" (to), "1" (from),"2" (n)\
- : "memory","cc");\
-}
-#else
-#define small_memcpy(to,from,n)\
-{\
-register unsigned long int dummy;\
-__asm__ __volatile__(\
- "rep; movsb"\
- :"=&D"(to), "=&S"(from), "=&c"(dummy)\
-/* It's most portable way to notify compiler */\
-/* that edi, esi and ecx are clobbered in asm block. */\
-/* Thanks to A'rpi for hint!!! */\
- :"0" (to), "1" (from),"2" (n)\
- : "memory","cc");\
-}
-#endif
-
-#undef MIN_LEN
-#if defined( OPTIMIZE_MMX ) && !defined( OPTIMIZE_MMX2 )
-#define MIN_LEN 0x800 /* 2K blocks. Was found experimentally */
-#else
-#define MIN_LEN _ivec_size()*8
-#endif
-
-#undef FAST_MEMORY_COPY
-#define FAST_MEMORY_COPY(to,from, len)\
-{\
- any_t*retval;\
- const unsigned char *cfrom=from;\
- unsigned char *tto=to;\
- const unsigned block_size = _ivec_size()*8;\
- __ivec iarr[8];\
- size_t i;\
- retval = to;\
- if(!len) return retval;\
-\
- _ivec_prefetch(&cfrom[0]);\
- _ivec_prefetch(&cfrom[32]);\
- _ivec_prefetch(&cfrom[64]);\
- _ivec_prefetch(&cfrom[96]);\
- _ivec_prefetch(&cfrom[128]);\
- _ivec_prefetch(&cfrom[160]);\
- _ivec_prefetch(&cfrom[192]);\
- _ivec_prefetch(&cfrom[224]);\
- _ivec_prefetch(&cfrom[256]);\
-\
- if(len >= MIN_LEN)\
- {\
- register unsigned long int delta;\
- /* Align destinition to cache-line size -boundary */\
- delta = ((unsigned long int)tto)&(gCpuCaps.cl_size-1);\
- if(delta) {\
- delta=gCpuCaps.cl_size-delta;\
- len -= delta;\
- small_memcpy(tto, cfrom, delta);\
- }\
- i = len/block_size;\
- len&=(block_size-1);\
- /*\
- This algorithm is top effective when the code consequently\
- reads and writes blocks which have size of cache line.\
- Size of cache line is processor-dependent.\
- It will, however, be a minimum of 32 bytes on any processors.\
- It would be better to have a number of instructions which\
- perform reading and writing to be multiple to a number of\
- processor's decoders, but it's not always possible.\
- */\
- if(((unsigned long)cfrom) & 15)\
- /* if SRC is misaligned */\
- for(; i>0; i--)\
- {\
- _ivec_prefetch(&cfrom[_ivec_size()*8]);\
- _ivec_prefetch(&cfrom[_ivec_size()*8+32]);\
- iarr[0] = _ivec_loadu(&cfrom[_ivec_size()*0]);\
- iarr[1] = _ivec_loadu(&cfrom[_ivec_size()*1]);\
- iarr[2] = _ivec_loadu(&cfrom[_ivec_size()*2]);\
- iarr[3] = _ivec_loadu(&cfrom[_ivec_size()*3]);\
- iarr[4] = _ivec_loadu(&cfrom[_ivec_size()*4]);\
- iarr[5] = _ivec_loadu(&cfrom[_ivec_size()*5]);\
- iarr[6] = _ivec_loadu(&cfrom[_ivec_size()*6]);\
- iarr[7] = _ivec_loadu(&cfrom[_ivec_size()*7]);\
- MEM_STORE(&tto[_ivec_size()*0],iarr[0]);\
- MEM_STORE(&tto[_ivec_size()*1],iarr[1]);\
- MEM_STORE(&tto[_ivec_size()*2],iarr[2]);\
- MEM_STORE(&tto[_ivec_size()*3],iarr[3]);\
- MEM_STORE(&tto[_ivec_size()*4],iarr[4]);\
- MEM_STORE(&tto[_ivec_size()*5],iarr[5]);\
- MEM_STORE(&tto[_ivec_size()*6],iarr[6]);\
- MEM_STORE(&tto[_ivec_size()*7],iarr[7]);\
- cfrom+=block_size;\
- tto+=block_size;\
- }\
- else\
- /* if SRC is aligned */\
- for(; i>0; i--)\
- {\
- _ivec_prefetch(&cfrom[_ivec_size()*8]);\
- _ivec_prefetch(&cfrom[_ivec_size()*8+32]);\
- iarr[0] = _ivec_loada(&cfrom[_ivec_size()*0]);\
- iarr[1] = _ivec_loada(&cfrom[_ivec_size()*1]);\
- iarr[2] = _ivec_loada(&cfrom[_ivec_size()*2]);\
- iarr[3] = _ivec_loada(&cfrom[_ivec_size()*3]);\
- iarr[4] = _ivec_loada(&cfrom[_ivec_size()*4]);\
- iarr[5] = _ivec_loada(&cfrom[_ivec_size()*5]);\
- iarr[6] = _ivec_loada(&cfrom[_ivec_size()*6]);\
- iarr[7] = _ivec_loada(&cfrom[_ivec_size()*7]);\
- MEM_STORE(&tto[_ivec_size()*0],iarr[0]);\
- MEM_STORE(&tto[_ivec_size()*1],iarr[1]);\
- MEM_STORE(&tto[_ivec_size()*2],iarr[2]);\
- MEM_STORE(&tto[_ivec_size()*3],iarr[3]);\
- MEM_STORE(&tto[_ivec_size()*4],iarr[4]);\
- MEM_STORE(&tto[_ivec_size()*5],iarr[5]);\
- MEM_STORE(&tto[_ivec_size()*6],iarr[6]);\
- MEM_STORE(&tto[_ivec_size()*7],iarr[7]);\
- cfrom+=block_size;\
- tto+=block_size;\
- }\
- MEM_SFENCE\
- _ivec_empty();\
- }\
- /*\
- * Now do the tail of the block\
- */\
- if(len) small_memcpy(tto, cfrom, len);\
- return retval;\
-}
-
-#undef MEM_STORE
-#undef MEM_SFENCE
-#define MEM_STORE _ivec_stream
-#define MEM_SFENCE _ivec_sfence();
-static inline any_t* PVECTOR_RENAME(fast_stream_copy)(any_t* to, const any_t* from, size_t len)
-{
- MSG_DBG3("fast_stream_copy(%p, %p, %u) [cl_size=%u]\n",to,from,len,gCpuCaps.cl_size);
- FAST_MEMORY_COPY(to,from,len);
-}
-
-#undef MEM_STORE
-#undef MEM_SFENCE
-#define MEM_STORE _ivec_storea
-#define MEM_SFENCE
-static inline any_t* PVECTOR_RENAME(fast_memcpy)(any_t* to, const any_t* from, size_t len)
-{
- MSG_DBG3("fast_memcpy(%p, %p, %u) [cl_size=%u]\n",to,from,len,gCpuCaps.cl_size);
- FAST_MEMORY_COPY(to,from,len);
-}
-#endif
Copied: mplayerxp/osdep/aclib_template.h (from rev 369, mplayerxp/osdep/aclib_template.c)
===================================================================
--- mplayerxp/osdep/aclib_template.h (rev 0)
+++ mplayerxp/osdep/aclib_template.h 2012-11-17 13:56:54 UTC (rev 390)
@@ -0,0 +1,179 @@
+/*
+ aclib - advanced C library ;)
+ This file contains functions which improve and expand standard C-library
+*/
+#include "pvector/pvector.h"
+
+#ifdef HAVE_INT_PVECTOR
+/* for small memory blocks (<256 bytes) this version is faster */
+#undef small_memcpy
+#ifdef __x86_64__
+#define small_memcpy(to,from,n)\
+{\
+register unsigned long int siz;\
+register unsigned long int dummy;\
+ siz=n&0x7; n>>=3;\
+ if(siz)\
+__asm__ __volatile__(\
+ "rep; movsb"\
+ :"=&D"(to), "=&S"(from), "=&c"(dummy)\
+/* It's most portable way to notify compiler */\
+/* that edi, esi and ecx are clobbered in asm block. */\
+/* Thanks to A'rpi for hint!!! */\
+ :"0" (to), "1" (from),"2" (siz)\
+ : "memory","cc");\
+ if(n)\
+__asm__ __volatile__(\
+ "rep; movsq"\
+ :"=&D"(to), "=&S"(from), "=&c"(dummy)\
+/* It's most portable way to notify compiler */\
+/* that edi, esi and ecx are clobbered in asm block. */\
+/* Thanks to A'rpi for hint!!! */\
+ :"0" (to), "1" (from),"2" (n)\
+ : "memory","cc");\
+}
+#else
+#define small_memcpy(to,from,n)\
+{\
+register unsigned long int dummy;\
+__asm__ __volatile__(\
+ "rep; movsb"\
+ :"=&D"(to), "=&S"(from), "=&c"(dummy)\
+/* It's most portable way to notify compiler */\
+/* that edi, esi and ecx are clobbered in asm block. */\
+/* Thanks to A'rpi for hint!!! */\
+ :"0" (to), "1" (from),"2" (n)\
+ : "memory","cc");\
+}
+#endif
+
+#undef MIN_LEN
+#if defined( OPTIMIZE_MMX ) && !defined( OPTIMIZE_MMX2 )
+#define MIN_LEN 0x800 /* 2K blocks. Was found experimentally */
+#else
+#define MIN_LEN _ivec_size()*8
+#endif
+
+#undef FAST_MEMORY_COPY
+#define FAST_MEMORY_COPY(to,from, len)\
+{\
+ any_t*retval;\
+ const unsigned char *cfrom=(const unsigned char *)from;\
+ unsigned char *tto=(unsigned char *)to;\
+ const unsigned block_size = _ivec_size()*8;\
+ __ivec iarr[8];\
+ size_t i;\
+ retval = to;\
+ if(!len) return retval;\
+\
+ _ivec_prefetch(&cfrom[0]);\
+ _ivec_prefetch(&cfrom[32]);\
+ _ivec_prefetch(&cfrom[64]);\
+ _ivec_prefetch(&cfrom[96]);\
+ _ivec_prefetch(&cfrom[128]);\
+ _ivec_prefetch(&cfrom[160]);\
+ _ivec_prefetch(&cfrom[192]);\
+ _ivec_prefetch(&cfrom[224]);\
+ _ivec_prefetch(&cfrom[256]);\
+\
+ if(len >= MIN_LEN)\
+ {\
+ register unsigned long int delta;\
+ /* Align destinition to cache-line size -boundary */\
+ delta = ((unsigned long int)tto)&(gCpuCaps.cl_size-1);\
+ if(delta) {\
+ delta=gCpuCaps.cl_size-delta;\
+ len -= delta;\
+ small_memcpy(tto, cfrom, delta);\
+ }\
+ i = len/block_size;\
+ len&=(block_size-1);\
+ /*\
+ This algorithm is top effective when the code consequently\
+ reads and writes blocks which have size of cache line.\
+ Size of cache line is processor-dependent.\
+ It will, however, be a minimum of 32 bytes on any processors.\
+ It would be better to have a number of instructions which\
+ perform reading and writing to be multiple to a number of\
+ processor's decoders, but it's not always possible.\
+ */\
+ if(((unsigned long)cfrom) & 15)\
+ /* if SRC is misaligned */\
+ for(; i>0; i--)\
+ {\
+ _ivec_prefetch(&cfrom[_ivec_size()*8]);\
+ _ivec_prefetch(&cfrom[_ivec_size()*8+32]);\
+ iarr[0] = _ivec_loadu(&cfrom[_ivec_size()*0]);\
+ iarr[1] = _ivec_loadu(&cfrom[_ivec_size()*1]);\
+ iarr[2] = _ivec_loadu(&cfrom[_ivec_size()*2]);\
+ iarr[3] = _ivec_loadu(&cfrom[_ivec_size()*3]);\
+ iarr[4] = _ivec_loadu(&cfrom[_ivec_size()*4]);\
+ iarr[5] = _ivec_loadu(&cfrom[_ivec_size()*5]);\
+ iarr[6] = _ivec_loadu(&cfrom[_ivec_size()*6]);\
+ iarr[7] = _ivec_loadu(&cfrom[_ivec_size()*7]);\
+ MEM_STORE(&tto[_ivec_size()*0],iarr[0]);\
+ MEM_STORE(&tto[_ivec_size()*1],iarr[1]);\
+ MEM_STORE(&tto[_ivec_size()*2],iarr[2]);\
+ MEM_STORE(&tto[_ivec_size()*3],iarr[3]);\
+ MEM_STORE(&tto[_ivec_size()*4],iarr[4]);\
+ MEM_STORE(&tto[_ivec_size()*5],iarr[5]);\
+ MEM_STORE(&tto[_ivec_size()*6],iarr[6]);\
+ MEM_STORE(&tto[_ivec_size()*7],iarr[7]);\
+ cfrom+=block_size;\
+ tto+=block_size;\
+ }\
+ else\
+ /* if SRC is aligned */\
+ for(; i>0; i--)\
+ {\
+ _ivec_prefetch(&cfrom[_ivec_size()*8]);\
+ _ivec_prefetch(&cfrom[_ivec_size()*8+32]);\
+ iarr[0] = _ivec_loada(&cfrom[_ivec_size()*0]);\
+ iarr[1] = _ivec_loada(&cfrom[_ivec_size()*1]);\
+ iarr[2] = _ivec_loada(&cfrom[_ivec_size()*2]);\
+ iarr[3] = _ivec_loada(&cfrom[_ivec_size()*3]);\
+ iarr[4] = _ivec_loada(&cfrom[_ivec_size()*4]);\
+ iarr[5] = _ivec_loada(&cfrom[_ivec_size()*5]);\
+ iarr[6] = _ivec_loada(&cfrom[_ivec_size()*6]);\
+ iarr[7] = _ivec_loada(&cfrom[_ivec_size()*7]);\
+ MEM_STORE(&tto[_ivec_size()*0],iarr[0]);\
+ MEM_STORE(&tto[_ivec_size()*1],iarr[1]);\
+ MEM_STORE(&tto[_ivec_size()*2],iarr[2]);\
+ MEM_STORE(&tto[_ivec_size()*3],iarr[3]);\
+ MEM_STORE(&tto[_ivec_size()*4],iarr[4]);\
+ MEM_STORE(&tto[_ivec_size()*5],iarr[5]);\
+ MEM_STORE(&tto[_ivec_size()*6],iarr[6]);\
+ MEM_STORE(&tto[_ivec_size()*7],iarr[7]);\
+ cfrom+=block_size;\
+ tto+=block_size;\
+ }\
+ MEM_SFENCE\
+ _ivec_empty();\
+ }\
+ /*\
+ * Now do the tail of the block\
+ */\
+ if(len) small_memcpy(tto, cfrom, len);\
+ return retval;\
+}
+
+#undef MEM_STORE
+#undef MEM_SFENCE
+#define MEM_STORE _ivec_stream
+#define MEM_SFENCE _ivec_sfence();
+static inline any_t* PVECTOR_RENAME(fast_stream_copy)(any_t* to, const any_t* from, size_t len)
+{
+ MSG_DBG3("fast_stream_copy(%p, %p, %u) [cl_size=%u]\n",to,from,len,gCpuCaps.cl_size);
+ FAST_MEMORY_COPY(to,from,len);
+}
+
+#undef MEM_STORE
+#undef MEM_SFENCE
+#define MEM_STORE _ivec_storea
+#define MEM_SFENCE
+static inline any_t* PVECTOR_RENAME(fast_memcpy)(any_t* to, const any_t* from, size_t len)
+{
+ MSG_DBG3("fast_memcpy(%p, %p, %u) [cl_size=%u]\n",to,from,len,gCpuCaps.cl_size);
+ FAST_MEMORY_COPY(to,from,len);
+}
+#endif
Modified: mplayerxp/osdep/fastmemcpy.h
===================================================================
--- mplayerxp/osdep/fastmemcpy.h 2012-11-17 10:42:51 UTC (rev 389)
+++ mplayerxp/osdep/fastmemcpy.h 2012-11-17 13:56:54 UTC (rev 390)
@@ -25,8 +25,8 @@
for(i=0; i<height; i++)
{
stream_copy(dst, src, bytesPerLine);
- src+= srcStride;
- dst+= dstStride;
+ src=(char *)src+ srcStride;
+ dst=(char *)dst+ dstStride;
}
}
@@ -44,8 +44,8 @@
for(i=0; i<height; i++)
{
memcpy(dst, src, bytesPerLine);
- src+= srcStride;
- dst+= dstStride;
+ src=(char *)src+ srcStride;
+ dst=(char *)dst+ dstStride;
}
}
This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.
|
|
From: <nic...@us...> - 2012-11-21 12:54:49
|
Revision: 421
http://mplayerxp.svn.sourceforge.net/mplayerxp/?rev=421&view=rev
Author: nickols_k
Date: 2012-11-21 12:54:43 +0000 (Wed, 21 Nov 2012)
Log Message:
-----------
implement memaligned operator new
Modified Paths:
--------------
mplayerxp/osdep/mp_malloc.cpp
mplayerxp/osdep/mplib.h
Modified: mplayerxp/osdep/mp_malloc.cpp
===================================================================
--- mplayerxp/osdep/mp_malloc.cpp 2012-11-21 09:34:59 UTC (rev 420)
+++ mplayerxp/osdep/mp_malloc.cpp 2012-11-21 12:54:43 UTC (rev 421)
@@ -542,6 +542,16 @@
return ptr;
}
-void SECURE_NAME2(_mp_free)(any_t* ptr) {
+any_t* SECURE_NAME2(_mp_memalign)(size_t boundary,size_t size) {
+ any_t* ptr;
+ ptr = mp_memalign(boundary,size);
+ if(!ptr) {
+ std::bad_alloc ba;
+ throw ba;
+ }
+ return ptr;
+}
+
+void SECURE_NAME3(_mp_free)(any_t* ptr) {
mp_free(ptr);
}
Modified: mplayerxp/osdep/mplib.h
===================================================================
--- mplayerxp/osdep/mplib.h 2012-11-21 09:34:59 UTC (rev 420)
+++ mplayerxp/osdep/mplib.h 2012-11-21 12:54:43 UTC (rev 421)
@@ -77,7 +77,8 @@
extern any_t* __FASTCALL__ SECURE_NAME0(_mp_malloc)(size_t size);
extern any_t* __FASTCALL__ SECURE_NAME1(_mp_mallocz)(size_t size);
-extern void __FASTCALL__ SECURE_NAME2(_mp_free)(any_t* ptr);
+extern any_t* __FASTCALL__ SECURE_NAME2(_mp_memalign)(size_t boundary,size_t size);
+extern void __FASTCALL__ SECURE_NAME3(_mp_free)(any_t* ptr);
extern any_t* __FASTCALL__ SECURE_NAME9(rnd_fill)(any_t* buffer,size_t size);
#ifdef __cplusplus
}
@@ -85,14 +86,17 @@
extern "C++" {
#include <new>
enum zeromemory_t{ zeromem=0 };
+ enum alignedmemory_t{ alignmem=0 };
inline any_t *operator new(size_t size) { return SECURE_NAME0(_mp_malloc)(size); }
inline any_t *operator new(size_t size,const zeromemory_t&) { return SECURE_NAME1(_mp_mallocz)(size); }
+ inline any_t *operator new(size_t size,const alignedmemory_t&,size_t boundary) { return SECURE_NAME2(_mp_memalign)(boundary,size); }
inline any_t *operator new(size_t size,const std::nothrow_t&) { return mp_malloc(size); }
inline any_t *operator new[](size_t size) { return SECURE_NAME0(_mp_malloc)(size); }
inline any_t *operator new[](size_t size,const zeromemory_t&) { return SECURE_NAME1(_mp_mallocz)(size); }
+ inline any_t *operator new[](size_t size,const alignedmemory_t&,size_t boundary) { return SECURE_NAME2(_mp_memalign)(boundary,size); }
inline any_t *operator new[](size_t size,const std::nothrow_t&) { return mp_malloc(size); }
- inline void operator delete(any_t* p) { SECURE_NAME2(_mp_free)(p); }
- inline void operator delete[](any_t* p) { SECURE_NAME2(_mp_free)(p); }
+ inline void operator delete(any_t* p) { SECURE_NAME3(_mp_free)(p); }
+ inline void operator delete[](any_t* p) { SECURE_NAME3(_mp_free)(p); }
extern any_t* get_caller_address(unsigned num_caller=0);
}
This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.
|
|
From: <nic...@us...> - 2012-12-27 09:50:10
|
Revision: 605
http://mplayerxp.svn.sourceforge.net/mplayerxp/?rev=605&view=rev
Author: nickols_k
Date: 2012-12-27 09:50:00 +0000 (Thu, 27 Dec 2012)
Log Message:
-----------
fix pseudo-random filler
Modified Paths:
--------------
mplayerxp/osdep/get_path.cpp
mplayerxp/osdep/mplib.cpp
Modified: mplayerxp/osdep/get_path.cpp
===================================================================
--- mplayerxp/osdep/get_path.cpp 2012-12-27 08:38:19 UTC (rev 604)
+++ mplayerxp/osdep/get_path.cpp 2012-12-27 09:50:00 UTC (rev 605)
@@ -9,11 +9,11 @@
#include "osdep_msg.h"
namespace mpxp {
std::string get_path(const std::string& filename){
- char *homedir;
+ const char *homedir;
std::string rs;
std::string config_dir = std::string("/.")+PROGNAME;
- if ((homedir = ::getenv("HOME")) == NULL) return "";
+ if ((homedir = ::getenv("HOME")) == NULL) throw "No 'HOME' environment found";
rs=std::string(homedir)+config_dir;
if (!filename.empty()) rs+="/"+filename;
mpxp_v<<"get_path('"<<filename<<"') -> "<<rs<<std::endl;
Modified: mplayerxp/osdep/mplib.cpp
===================================================================
--- mplayerxp/osdep/mplib.cpp 2012-12-27 08:38:19 UTC (rev 604)
+++ mplayerxp/osdep/mplib.cpp 2012-12-27 09:50:00 UTC (rev 605)
@@ -35,7 +35,7 @@
long filler;
for(i=0;i<psize/sizeof(long);i++) {
filler=::rand()&lo_mask;
- filler=(reinterpret_cast<long>(buffer)&hi_mask)|lo_mask;
+ filler|=(reinterpret_cast<long>(buffer)&hi_mask);
((long *)buffer)[i]=::rand()%2?filler:0;
}
::memset(&((char *)buffer)[psize],0,size-psize);
This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.
|
|
From: <nic...@us...> - 2013-04-23 12:51:14
|
Revision: 631
http://sourceforge.net/p/mplayerxp/code/631
Author: nickols_k
Date: 2013-04-23 12:51:09 +0000 (Tue, 23 Apr 2013)
Log Message:
-----------
move mallocs into separate header
Modified Paths:
--------------
mplayerxp/osdep/mplib.h
Added Paths:
-----------
mplayerxp/osdep/mp_malloc.h
Added: mplayerxp/osdep/mp_malloc.h
===================================================================
--- mplayerxp/osdep/mp_malloc.h (rev 0)
+++ mplayerxp/osdep/mp_malloc.h 2013-04-23 12:51:09 UTC (rev 631)
@@ -0,0 +1,98 @@
+/*
+ * mp_malloc.h - Memory Protected malloc
+ *
+ * You can redistribute this file under terms and conditions
+ * of GNU General Public licence v3.
+ */
+#ifndef __MP_MALLOC_H_INCLUDED__
+#define __MP_MALLOC_H_INCLUDED__ 1
+
+#include <string>
+
+#include <execinfo.h>
+#include <stddef.h>
+#include <sys/mman.h>
+#include "mpxp_config.h"
+
+namespace mpxp {
+
+ extern volatile unsigned long long int my_profile_start,my_profile_end,my_profile_total;
+
+#if defined ( ENABLE_PROFILE ) && (defined ( ARCH_X86 ) || defined( ARCH_X86_64 ))
+ inline unsigned long long int read_tsc( void ) {
+ unsigned long long int retval;
+ __asm __volatile ("rdtsc":"=A"(retval)::"memory");
+ return retval;
+ }
+#define PROFILE_RESET() (my_profile_total=0ULL)
+#define PROFILE_START() { static int inited=0; if(!inited) { inited=1; my_profile_total=0ULL; } my_profile_start=read_tsc(); }
+#define PROFILE_END(your_message) { my_profile_end=read_tsc(); my_profile_total+=(my_profile_end-my_profile_start); printf(your_message" current=%llu total=%llu\n\t",(my_profile_end-my_profile_start),my_profile_total); }
+#else
+#define PROFILE_RESET()
+#define PROFILE_START()
+#define PROFILE_END(your_message)
+#endif
+/** Initializes randomizer for malloc.
+ * @param rnd_limit upper limit of random generator (recommened: 1000)
+ * @param every_nth_call how often call randimzer (recommened: 10)
+ * @note Pseudo-randomizing memory objects makes memory
+ * exploits harder
+*/
+ enum mp_malloc_e {
+ MPA_FLG_RANDOMIZER = 0x00000000,
+ MPA_FLG_BOUNDS_CHECK = 0x00000001,
+ MPA_FLG_BEFORE_CHECK = 0x00000002,
+ MPA_FLG_BACKTRACE = 0x00000004
+ };
+ void __FASTCALL__ mp_init_malloc(const std::string& argv0,unsigned rnd_limit,unsigned every_nth_call,enum mp_malloc_e flags);
+ void __FASTCALL__ mp_uninit_malloc(int verbose);
+
+ any_t* __FASTCALL__ mp_malloc(size_t __size);
+ any_t* __FASTCALL__ mp_mallocz(size_t __size);
+ any_t* __FASTCALL__ mp_realloc(any_t*__ptr, size_t __size);
+ inline any_t* mp_calloc (size_t __nelem, size_t __size) { return mp_mallocz(__nelem*__size); }
+ any_t* __FASTCALL__ mp_memalign (size_t boundary, size_t __size);
+ void __FASTCALL__ mp_free(any_t*__ptr);
+ char * __FASTCALL__ mp_strdup(const char *src);
+
+ /* flags: PROT_NONE, PROT_READ, PROT_WRITE, PROT_EXEC */
+ enum mp_prot_e {
+ MP_PROT_READ =0x1, /* Page can be read. */
+ MP_PROT_WRITE =0x2, /* Page can be written. */
+ MP_PROT_EXEC =0x4, /* Page can be executed. */
+ MP_DENY_ALL =0x0, /* Page can not be accessed. */
+ };
+ int __FASTCALL__ mp_mprotect(any_t* addr,size_t len,enum mp_prot_e flags);
+ void print_backtrace(const std::string& why,any_t** stack,unsigned num);
+
+ inline void show_backtrace(const std::string& why,unsigned num_calls) {
+ any_t* stack[num_calls];
+ unsigned ncalls;
+ ncalls=backtrace(stack,num_calls);
+ print_backtrace(why,stack,ncalls);
+ }
+ any_t* __FASTCALL__ rnd_fill(any_t* buffer,size_t size);
+ any_t* __FASTCALL__ make_false_pointer(any_t* tmplt);
+ any_t* __FASTCALL__ make_false_pointer_to(any_t* tmplt,unsigned size);
+ any_t* __FASTCALL__ fill_false_pointers(any_t* buffer,size_t size);
+ any_t* get_caller_address(unsigned num_caller=0);
+
+} // namespace mpxp
+
+/* Note: it seems that compiler cannot distinguish which version of operators
+ new and delete to use: from global namespace or from user specified */
+#include <new>
+enum zeromemory_t{ zeromem=0 };
+enum alignedmemory_t{ alignmem=0 };
+any_t* operator new(size_t size);
+any_t* operator new(size_t size,const zeromemory_t&);
+any_t* operator new(size_t size,const alignedmemory_t&,size_t boundary);
+any_t* operator new(size_t size,const std::nothrow_t&);
+any_t* operator new[](size_t size);
+any_t* operator new[](size_t size,const zeromemory_t&);
+any_t* operator new[](size_t size,const alignedmemory_t&,size_t boundary);
+any_t* operator new[](size_t size,const std::nothrow_t&);
+void operator delete(any_t* p);
+void operator delete[](any_t* p);
+
+#endif
Modified: mplayerxp/osdep/mplib.h
===================================================================
--- mplayerxp/osdep/mplib.h 2013-04-17 14:04:33 UTC (rev 630)
+++ mplayerxp/osdep/mplib.h 2013-04-23 12:51:09 UTC (rev 631)
@@ -15,70 +15,10 @@
#include <stddef.h>
#include <sys/mman.h>
#include "mpxp_config.h"
+#include "mp_malloc.h"
namespace mpxp {
- extern volatile unsigned long long int my_profile_start,my_profile_end,my_profile_total;
-
-#if defined ( ENABLE_PROFILE ) && (defined ( ARCH_X86 ) || defined( ARCH_X86_64 ))
- inline unsigned long long int read_tsc( void ) {
- unsigned long long int retval;
- __asm __volatile ("rdtsc":"=A"(retval)::"memory");
- return retval;
- }
-#define PROFILE_RESET() (my_profile_total=0ULL)
-#define PROFILE_START() { static int inited=0; if(!inited) { inited=1; my_profile_total=0ULL; } my_profile_start=read_tsc(); }
-#define PROFILE_END(your_message) { my_profile_end=read_tsc(); my_profile_total+=(my_profile_end-my_profile_start); printf(your_message" current=%llu total=%llu\n\t",(my_profile_end-my_profile_start),my_profile_total); }
-#else
-#define PROFILE_RESET()
-#define PROFILE_START()
-#define PROFILE_END(your_message)
-#endif
-/** Initializes randomizer for malloc.
- * @param rnd_limit upper limit of random generator (recommened: 1000)
- * @param every_nth_call how often call randimzer (recommened: 10)
- * @note Pseudo-randomizing memory objects makes memory
- * exploits harder
-*/
- enum mp_malloc_e {
- MPA_FLG_RANDOMIZER = 0x00000000,
- MPA_FLG_BOUNDS_CHECK = 0x00000001,
- MPA_FLG_BEFORE_CHECK = 0x00000002,
- MPA_FLG_BACKTRACE = 0x00000004
- };
- void __FASTCALL__ mp_init_malloc(const std::string& argv0,unsigned rnd_limit,unsigned every_nth_call,enum mp_malloc_e flags);
- void __FASTCALL__ mp_uninit_malloc(int verbose);
-
- any_t* __FASTCALL__ mp_malloc(size_t __size);
- any_t* __FASTCALL__ mp_mallocz(size_t __size);
- any_t* __FASTCALL__ mp_realloc(any_t*__ptr, size_t __size);
- inline any_t* mp_calloc (size_t __nelem, size_t __size) { return mp_mallocz(__nelem*__size); }
- any_t* __FASTCALL__ mp_memalign (size_t boundary, size_t __size);
- void __FASTCALL__ mp_free(any_t*__ptr);
- char * __FASTCALL__ mp_strdup(const char *src);
-
- /* flags: PROT_NONE, PROT_READ, PROT_WRITE, PROT_EXEC */
- enum mp_prot_e {
- MP_PROT_READ =0x1, /* Page can be read. */
- MP_PROT_WRITE =0x2, /* Page can be written. */
- MP_PROT_EXEC =0x4, /* Page can be executed. */
- MP_DENY_ALL =0x0, /* Page can not be accessed. */
- };
- int __FASTCALL__ mp_mprotect(any_t* addr,size_t len,enum mp_prot_e flags);
- void print_backtrace(const std::string& why,any_t** stack,unsigned num);
-
- inline void show_backtrace(const std::string& why,unsigned num_calls) {
- any_t* stack[num_calls];
- unsigned ncalls;
- ncalls=backtrace(stack,num_calls);
- print_backtrace(why,stack,ncalls);
- }
- any_t* __FASTCALL__ rnd_fill(any_t* buffer,size_t size);
- any_t* __FASTCALL__ make_false_pointer(any_t* tmplt);
- any_t* __FASTCALL__ make_false_pointer_to(any_t* tmplt,unsigned size);
- any_t* __FASTCALL__ fill_false_pointers(any_t* buffer,size_t size);
- any_t* get_caller_address(unsigned num_caller=0);
-
template <typename T> class LocalPtr {
public:
LocalPtr(T* value):ptr(value) {}
@@ -102,21 +42,4 @@
any_t* unusable;
};
} // namespace mpxp
-
-/* Note: it seems that compiler cannot distinguish which version of operators
- new and delete to use: from global namespace or from user specified */
-#include <new>
-enum zeromemory_t{ zeromem=0 };
-enum alignedmemory_t{ alignmem=0 };
-any_t* operator new(size_t size);
-any_t* operator new(size_t size,const zeromemory_t&);
-any_t* operator new(size_t size,const alignedmemory_t&,size_t boundary);
-any_t* operator new(size_t size,const std::nothrow_t&);
-any_t* operator new[](size_t size);
-any_t* operator new[](size_t size,const zeromemory_t&);
-any_t* operator new[](size_t size,const alignedmemory_t&,size_t boundary);
-any_t* operator new[](size_t size,const std::nothrow_t&);
-void operator delete(any_t* p);
-void operator delete[](any_t* p);
-
#endif
This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.
|
×
Want the latest updates on software, tech news, and AI?
Get latest updates about software, tech news, and AI from SourceForge directly in your inbox once a month.
Thanks for helping keep SourceForge clean.
X