mirrors/FreeRDP
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

libfreerdp-primitives: integrate H264 SSE3 color converter

Browse Source
This commit is contained in:
Marc-André Moreau 2014-09-09 19:15:07 -04:00
parent 666919d157
commit bcf1266f51
4 changed files with 225 additions and 268 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
include/freerdp/codec/h264.h
View File

@ -44,31 +44,12 @@ struct _H264_CONTEXT
{ {
BOOL Compressor; BOOL Compressor;
//BYTE* data;
//UINT32 size;
UINT32 width; UINT32 width;
UINT32 height; UINT32 height;
//int scanline;
BYTE* pYUVData[3];
int iStride[3]; int iStride[3];
/*
<<<<<<< HEAD
#ifdef WITH_OPENH264
ISVCDecoder* pDecoder;
BYTE* pYUVData[3]; BYTE* pYUVData[3];
int iStride[2];
#endif
#ifdef WITH_LIBAVCODEC
AVCodec* codec;
AVCodecContext* codecContext;
AVCodecParserContext* codecParser;
AVFrame* videoFrame;
#endif
=======
*/
void* pSystemData; void* pSystemData;
H264_CONTEXT_SUBSYSTEM* subsystem; H264_CONTEXT_SUBSYSTEM* subsystem;
}; };

55
libfreerdp/codec/h264.c
View File

@ -28,9 +28,6 @@
#include <freerdp/primitives.h> #include <freerdp/primitives.h>
#include <freerdp/codec/h264.h> #include <freerdp/codec/h264.h>
#include <sys/time.h>
/** /**
* Dummy subsystem * Dummy subsystem
*/ */
@ -87,8 +84,6 @@ static int openh264_decompress(H264_CONTEXT* h264, BYTE* pSrcData, UINT32 SrcSiz
SSysMEMBuffer* pSystemBuffer; SSysMEMBuffer* pSystemBuffer;
H264_CONTEXT_OPENH264* sys = (H264_CONTEXT_OPENH264*) h264->pSystemData; H264_CONTEXT_OPENH264* sys = (H264_CONTEXT_OPENH264*) h264->pSystemData;
struct timeval T1,T2;
if (!sys->pDecoder) if (!sys->pDecoder)
return -1; return -1;
@ -102,7 +97,6 @@ static int openh264_decompress(H264_CONTEXT* h264, BYTE* pSrcData, UINT32 SrcSiz
ZeroMemory(&sBufferInfo, sizeof(sBufferInfo)); ZeroMemory(&sBufferInfo, sizeof(sBufferInfo));
gettimeofday(&T1,NULL);
state = (*sys->pDecoder)->DecodeFrame2( state = (*sys->pDecoder)->DecodeFrame2(
sys->pDecoder, sys->pDecoder,
pSrcData, pSrcData,
@ -120,9 +114,6 @@ static int openh264_decompress(H264_CONTEXT* h264, BYTE* pSrcData, UINT32 SrcSiz
if (sBufferInfo.iBufferStatus != 1) if (sBufferInfo.iBufferStatus != 1)
state = (*sys->pDecoder)->DecodeFrame2(sys->pDecoder, NULL, 0, h264->pYUVData, &sBufferInfo); state = (*sys->pDecoder)->DecodeFrame2(sys->pDecoder, NULL, 0, h264->pYUVData, &sBufferInfo);
gettimeofday(&T2,NULL);
printf("OpenH264: decoding took: %u sec %u usec\n",(unsigned int)(T2.tv_sec-T1.tv_sec),(unsigned int)(T2.tv_usec-T1.tv_usec));
pSystemBuffer = &sBufferInfo.UsrData.sSystemBuffer; pSystemBuffer = &sBufferInfo.UsrData.sSystemBuffer;
#if 0 #if 0
@ -285,18 +276,12 @@ static int libavcodec_decompress(H264_CONTEXT* h264, BYTE* pSrcData, UINT32 SrcS
AVPacket packet; AVPacket packet;
H264_CONTEXT_LIBAVCODEC* sys = (H264_CONTEXT_LIBAVCODEC*) h264->pSystemData; H264_CONTEXT_LIBAVCODEC* sys = (H264_CONTEXT_LIBAVCODEC*) h264->pSystemData;
struct timeval T1,T2;
av_init_packet(&packet); av_init_packet(&packet);
packet.data = pSrcData; packet.data = pSrcData;
packet.size = SrcSize; packet.size = SrcSize;
gettimeofday(&T1,NULL);
status = avcodec_decode_video2(sys->codecContext, sys->videoFrame, &gotFrame, &packet); status = avcodec_decode_video2(sys->codecContext, sys->videoFrame, &gotFrame, &packet);
gettimeofday(&T2,NULL);
printf("libavcodec: decoding took: %u sec %u usec\n",(unsigned int)(T2.tv_sec-T1.tv_sec),(unsigned int)(T2.tv_usec-T1.tv_usec));
if (status < 0) if (status < 0)
{ {
@ -437,20 +422,18 @@ static H264_CONTEXT_SUBSYSTEM g_Subsystem_libavcodec =
int h264_decompress(H264_CONTEXT* h264, BYTE* pSrcData, UINT32 SrcSize, int h264_decompress(H264_CONTEXT* h264, BYTE* pSrcData, UINT32 SrcSize,
BYTE** ppDstData, DWORD DstFormat, int nDstStep, int nDstHeight, RDPGFX_RECT16* regionRects, int numRegionRects) BYTE** ppDstData, DWORD DstFormat, int nDstStep, int nDstHeight, RDPGFX_RECT16* regionRects, int numRegionRects)
{ {
int index;
int status;
int* iStride;
BYTE* pDstData; BYTE* pDstData;
BYTE* pDstPoint; BYTE* pDstPoint;
prim_size_t roi;
BYTE** pYUVData; BYTE** pYUVData;
int width, height;
BYTE* pYUVPoint[3]; BYTE* pYUVPoint[3];
RDPGFX_RECT16* rect; RDPGFX_RECT16* rect;
int* iStride;
int ret, i, cx, cy;
int UncompressedSize; int UncompressedSize;
primitives_t *prims = primitives_get(); primitives_t *prims = primitives_get();
prim_size_t roi;
struct timeval T1,T2;
if (!h264) if (!h264)
return -1; return -1;
@ -463,23 +446,23 @@ int h264_decompress(H264_CONTEXT* h264, BYTE* pSrcData, UINT32 SrcSize,
if (!(pDstData = *ppDstData)) if (!(pDstData = *ppDstData))
return -1; return -1;
if ((status = h264->subsystem->Decompress(h264, pSrcData, SrcSize)) < 0)
if ((ret = h264->subsystem->Decompress(h264, pSrcData, SrcSize)) < 0) return status;
return ret;
UncompressedSize = h264->width * h264->height * 4; UncompressedSize = h264->width * h264->height * 4;
if (UncompressedSize > (nDstStep * nDstHeight)) if (UncompressedSize > (nDstStep * nDstHeight))
return -1; return -1;
pYUVData = h264->pYUVData; pYUVData = h264->pYUVData;
iStride = h264->iStride; iStride = h264->iStride;
gettimeofday(&T1,NULL); for (index = 0; index < numRegionRects; index++)
for (i = 0; i < numRegionRects; i++){ {
rect = &(regionRects[i]); rect = &(regionRects[index]);
cx = rect->right - rect->left;
cy = rect->bottom - rect->top; width = rect->right - rect->left;
height = rect->bottom - rect->top;
pDstPoint = pDstData + rect->top * nDstStep + rect->left * 4; pDstPoint = pDstData + rect->top * nDstStep + rect->left * 4;
pYUVPoint[0] = pYUVData[0] + rect->top * iStride[0] + rect->left; pYUVPoint[0] = pYUVData[0] + rect->top * iStride[0] + rect->left;
@ -488,17 +471,15 @@ int h264_decompress(H264_CONTEXT* h264, BYTE* pSrcData, UINT32 SrcSize,
pYUVPoint[2] = pYUVData[2] + rect->top/2 * iStride[2] + rect->left/2; pYUVPoint[2] = pYUVData[2] + rect->top/2 * iStride[2] + rect->left/2;
#if 0 #if 0
printf("regionRect: x: %d, y: %d, cx: %d, cy: %d\n", printf("regionRect: x: %d y: %d width: %d height: %d\n",
rect->left, rect->top, cx, cy); rect->left, rect->top, width, height);
#endif #endif
roi.width = cx; roi.width = width;
roi.height = cy; roi.height = height;
prims->YUV420ToRGB_8u_P3AC4R((const BYTE**) pYUVPoint, iStride, pDstPoint, nDstStep, &roi); prims->YUV420ToRGB_8u_P3AC4R((const BYTE**) pYUVPoint, iStride, pDstPoint, nDstStep, &roi);
} }
gettimeofday(&T2,NULL);
printf("converting took %u sec %u usec\n",(unsigned int)(T2.tv_sec-T1.tv_sec),(unsigned int)(T2.tv_usec-T1.tv_usec));
return 1; return 1;
} }

39
libfreerdp/primitives/prim_YUV.c
View File

@ -27,6 +27,16 @@
#include "prim_internal.h" #include "prim_internal.h"
#include "prim_YUV.h" #include "prim_YUV.h"
/**
* | R | ( | 256 0 403 | | Y | )
* | G | = ( | 256 -48 -120 | | U - 128 | ) >> 8
* | B | ( | 256 475 0 | | V - 128 | )
*
* | Y | ( | 54 183 18 | | R | ) | 0 |
* | U | = ( | -29 -99 128 | | G | ) >> 8 + | 128 |
* | V | ( | 128 -116 -12 | | B | ) | 128 |
*/
pstatus_t general_YUV420ToRGB_8u_P3AC4R(const BYTE* pSrc[3], int srcStep[3], pstatus_t general_YUV420ToRGB_8u_P3AC4R(const BYTE* pSrc[3], int srcStep[3],
BYTE* pDst, int dstStep, const prim_size_t* roi) BYTE* pDst, int dstStep, const prim_size_t* roi)
{ {
@ -45,14 +55,14 @@ pstatus_t general_YUV420ToRGB_8u_P3AC4R(const BYTE* pSrc[3], int srcStep[3],
int Vp403, Vp120; int Vp403, Vp120;
BYTE* pRGB = pDst; BYTE* pRGB = pDst;
int nWidth, nHeight; int nWidth, nHeight;
int last_line, last_column; int lastRow, lastCol;
pY = pSrc[0]; pY = pSrc[0];
pU = pSrc[1]; pU = pSrc[1];
pV = pSrc[2]; pV = pSrc[2];
last_column = roi->width & 0x01; lastCol = roi->width & 0x01;
last_line = roi->height & 0x01; lastRow = roi->height & 0x01;
nWidth = (roi->width + 1) & ~0x0001; nWidth = (roi->width + 1) & ~0x0001;
nHeight = (roi->height + 1) & ~0x0001; nHeight = (roi->height + 1) & ~0x0001;
@ -68,15 +78,13 @@ pstatus_t general_YUV420ToRGB_8u_P3AC4R(const BYTE* pSrc[3], int srcStep[3],
for (y = 0; y < halfHeight; ) for (y = 0; y < halfHeight; )
{ {
y++; if (++y == halfHeight)
if (y == halfHeight) lastRow <<= 1;
last_line = last_line << 1;
for (x = 0; x < halfWidth; ) for (x = 0; x < halfWidth; )
{ {
x++; if (++x == halfWidth)
if (x == halfWidth) lastCol <<= 1;
last_column = last_column << 1;
U = *pU++; U = *pU++;
V = *pV++; V = *pV++;
@ -121,7 +129,7 @@ pstatus_t general_YUV420ToRGB_8u_P3AC4R(const BYTE* pSrc[3], int srcStep[3],
/* 2nd pixel */ /* 2nd pixel */
if (!(last_column & 0x02)) if (!(lastCol & 0x02))
{ {
Y = *pY++; Y = *pY++;
Yp = Y << 8; Yp = Y << 8;
@ -154,7 +162,7 @@ pstatus_t general_YUV420ToRGB_8u_P3AC4R(const BYTE* pSrc[3], int srcStep[3],
{ {
pY++; pY++;
pRGB += 4; pRGB += 4;
last_column = last_column >> 1; lastCol >>= 1;
} }
} }
@ -165,9 +173,8 @@ pstatus_t general_YUV420ToRGB_8u_P3AC4R(const BYTE* pSrc[3], int srcStep[3],
for (x = 0; x < halfWidth; ) for (x = 0; x < halfWidth; )
{ {
x++; if (++x == halfWidth)
if (x == halfWidth) lastCol <<= 1;
last_column = last_column << 1;
U = *pU++; U = *pU++;
V = *pV++; V = *pV++;
@ -212,7 +219,7 @@ pstatus_t general_YUV420ToRGB_8u_P3AC4R(const BYTE* pSrc[3], int srcStep[3],
/* 4th pixel */ /* 4th pixel */
if(!(last_column & 0x02)) if (!(lastCol & 0x02))
{ {
Y = *pY++; Y = *pY++;
Yp = Y << 8; Yp = Y << 8;
@ -245,7 +252,7 @@ pstatus_t general_YUV420ToRGB_8u_P3AC4R(const BYTE* pSrc[3], int srcStep[3],
{ {
pY++; pY++;
pRGB += 4; pRGB += 4;
last_column = last_column >> 1; lastCol >>= 1;
} }
} }

96
libfreerdp/primitives/prim_YUV_opt.c
View File

@ -25,21 +25,17 @@
pstatus_t ssse3_YUV420ToRGB_8u_P3AC4R(const BYTE **pSrc, int *srcStep, pstatus_t ssse3_YUV420ToRGB_8u_P3AC4R(const BYTE **pSrc, int *srcStep,
BYTE *pDst, int dstStep, const prim_size_t *roi) BYTE *pDst, int dstStep, const prim_size_t *roi)
{ {
char last_line,last_column; int lastRow, lastCol;
/* last_line: if the last (U,V doubled) line should be skipped, set to 10B
* last_column: if it's the last column in a line, set to 10B (for handling line-endings not multiple by four) */
int i,nWidth,nHeight,VaddDst,VaddY,VaddU,VaddV;
BYTE *UData,*VData,*YData; BYTE *UData,*VData,*YData;
int i,nWidth,nHeight,VaddDst,VaddY,VaddU,VaddV;
__m128i r0,r1,r2,r3,r4,r5,r6,r7; __m128i r0,r1,r2,r3,r4,r5,r6,r7;
__m128i *buffer; __m128i *buffer;
/* last_line: if the last (U,V doubled) line should be skipped, set to 10B
* last_column: if it's the last column in a line, set to 10B (for handling line-endings not multiple by four) */
buffer = _aligned_malloc(4 * 16, 16); buffer = _aligned_malloc(4 * 16, 16);
YData = (BYTE*) pSrc[0]; YData = (BYTE*) pSrc[0];
UData = (BYTE*) pSrc[1]; UData = (BYTE*) pSrc[1];
VData = (BYTE*) pSrc[2]; VData = (BYTE*) pSrc[2];
@ -47,51 +43,50 @@ pstatus_t ssse3_YUV420ToRGB_8u_P3AC4R(const BYTE **pSrc, int *srcStep,
nWidth = roi->width; nWidth = roi->width;
nHeight = roi->height; nHeight = roi->height;
if ((lastCol = (nWidth & 3)))
{
switch (lastCol)
{
case 1:
r7 = _mm_set_epi32(0,0,0,0xFFFFFFFF);
break;
if((last_column=nWidth&3)){ case 2:
switch(last_column){ r7 = _mm_set_epi32(0,0,0xFFFFFFFF,0xFFFFFFFF);
case 1: r7=_mm_set_epi32(0,0,0,0xFFFFFFFF); break; break;
case 2: r7=_mm_set_epi32(0,0,0xFFFFFFFF,0xFFFFFFFF); break;
case 3: r7=_mm_set_epi32(0,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF); break; case 3:
r7 = _mm_set_epi32(0,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF);
break;
} }
_mm_store_si128(buffer+3,r7); _mm_store_si128(buffer+3,r7);
last_column=1; lastCol = 1;
} }
nWidth += 3; nWidth += 3;
nWidth = nWidth >> 2; nWidth = nWidth >> 2;
lastRow = nHeight & 1;
last_line=nHeight&1;
nHeight++; nHeight++;
nHeight = nHeight >> 1; nHeight = nHeight >> 1;
VaddDst = (dstStep << 1) - (nWidth << 4); VaddDst = (dstStep << 1) - (nWidth << 4);
VaddY = (srcStep[0] << 1) - (nWidth << 2); VaddY = (srcStep[0] << 1) - (nWidth << 2);
VaddU = srcStep[1] - (((nWidth << 1) + 2) & 0xFFFC); VaddU = srcStep[1] - (((nWidth << 1) + 2) & 0xFFFC);
VaddV = srcStep[2] - (((nWidth << 1) + 2) & 0xFFFC); VaddV = srcStep[2] - (((nWidth << 1) + 2) & 0xFFFC);
while (nHeight-- > 0)
while(nHeight-- >0){ {
if(nHeight==0){ if (nHeight == 0)
last_line=last_line<<1; lastRow <<= 1;
}
i = 0; i = 0;
do{
/*
* Well, in the end it should look like this:
* C = Y;
* D = U - 128;
* E = V - 128;
*
* R = clip(( 256 * C + 403 * E + 128) >> 8);
* G = clip(( 256 * C - 48 * D - 120 * E + 128) >> 8);
* B = clip(( 256 * C + 475 * D + 128) >> 8);
*/
if(!(i&0x01)){
do
{
if (!(i & 0x01))
{
/* Y-, U- and V-data is stored in different arrays. /* Y-, U- and V-data is stored in different arrays.
* We start with processing U-data. * We start with processing U-data.
* *
@ -122,7 +117,6 @@ pstatus_t ssse3_YUV420ToRGB_8u_P3AC4R(const BYTE **pSrc, int *srcStep,
r0 = _mm_unpacklo_epi16(r0,r4); r0 = _mm_unpacklo_epi16(r0,r4);
r4 = _mm_unpackhi_epi16(r7,r4); r4 = _mm_unpackhi_epi16(r7,r4);
/* to complete this step, add (?) 128 to each value (rounding ?!) /* to complete this step, add (?) 128 to each value (rounding ?!)
* yeah, add. in the end this will be subtracted from something, * yeah, add. in the end this will be subtracted from something,
* because it's part of G: 256*C - (48*D + 120*E - 128), 48*D-128 ! * because it's part of G: 256*C - (48*D + 120*E - 128), 48*D-128 !
@ -131,7 +125,6 @@ pstatus_t ssse3_YUV420ToRGB_8u_P3AC4R(const BYTE **pSrc, int *srcStep,
r0 = _mm_sub_epi32(r0,r6); r0 = _mm_sub_epi32(r0,r6);
r4 = _mm_sub_epi32(r4,r6); r4 = _mm_sub_epi32(r4,r6);
/* to get B data, we need to prepare a secound value, D*475+128 */ /* to get B data, we need to prepare a secound value, D*475+128 */
r1 = r2; r1 = r2;
r7 = _mm_set_epi16(475,475,475,475,475,475,475,475); r7 = _mm_set_epi16(475,475,475,475,475,475,475,475);
@ -160,7 +153,6 @@ pstatus_t ssse3_YUV420ToRGB_8u_P3AC4R(const BYTE **pSrc, int *srcStep,
r5 = r2; r5 = r2;
/* this is also known as E*403+128, we need it to convert R data */ /* this is also known as E*403+128, we need it to convert R data */
r3 = r2; r3 = r2;
r7 = _mm_set_epi16(403,403,403,403,403,403,403,403); r7 = _mm_set_epi16(403,403,403,403,403,403,403,403);
@ -176,8 +168,6 @@ pstatus_t ssse3_YUV420ToRGB_8u_P3AC4R(const BYTE **pSrc, int *srcStep,
/* and preserve upper four values for future ... */ /* and preserve upper four values for future ... */
_mm_store_si128(buffer+2,r7); _mm_store_si128(buffer+2,r7);
/* doing this step: E*120 */ /* doing this step: E*120 */
r3 = r5; r3 = r5;
r7 = _mm_set_epi16(120,120,120,120,120,120,120,120); r7 = _mm_set_epi16(120,120,120,120,120,120,120,120);
@ -194,7 +184,9 @@ pstatus_t ssse3_YUV420ToRGB_8u_P3AC4R(const BYTE **pSrc, int *srcStep,
/* and store to memory ! */ /* and store to memory ! */
_mm_store_si128(buffer,r4); _mm_store_si128(buffer,r4);
}else{ }
else
{
/* maybe you've wondered about the conditional above ? /* maybe you've wondered about the conditional above ?
* Well, we prepared UV data for eight pixel in each line, but can only process four * Well, we prepared UV data for eight pixel in each line, but can only process four
* per loop. So we need to load the upper four pixel data from memory each secound loop! */ * per loop. So we need to load the upper four pixel data from memory each secound loop! */
@ -204,7 +196,7 @@ pstatus_t ssse3_YUV420ToRGB_8u_P3AC4R(const BYTE **pSrc, int *srcStep,
} }
if (++i == nWidth) if (++i == nWidth)
last_column=last_column<<1; lastCol <<= 1;
/* We didn't produce any output yet, so let's do so! /* We didn't produce any output yet, so let's do so!
* Ok, fetch four pixel from the Y-data array and shuffle them like this: * Ok, fetch four pixel from the Y-data array and shuffle them like this:
@ -221,7 +213,6 @@ pstatus_t ssse3_YUV420ToRGB_8u_P3AC4R(const BYTE **pSrc, int *srcStep,
r5 = _mm_sub_epi32(r5,r0); r5 = _mm_sub_epi32(r5,r0);
r6 = _mm_add_epi32(r6,r1); r6 = _mm_add_epi32(r6,r1);
/* in the end, we only need bytes for RGB values. /* in the end, we only need bytes for RGB values.
* So, what do we do? right! shifting left makes values bigger and thats always good. * So, what do we do? right! shifting left makes values bigger and thats always good.
* before we had dwords of data, and by shifting left and treating the result * before we had dwords of data, and by shifting left and treating the result
@ -266,16 +257,15 @@ pstatus_t ssse3_YUV420ToRGB_8u_P3AC4R(const BYTE **pSrc, int *srcStep,
r7 = _mm_set_epi32(0x8080800E,0x8080800A,0x80808006,0x80808002); r7 = _mm_set_epi32(0x8080800E,0x8080800A,0x80808006,0x80808002);
r6 = _mm_shuffle_epi8(r6,r7); r6 = _mm_shuffle_epi8(r6,r7);
/* and at last we or it together and get this one: /* and at last we or it together and get this one:
* xrgb xrgb xrgb xrgb */ * xrgb xrgb xrgb xrgb */
r4 = _mm_or_si128(r4,r5); r4 = _mm_or_si128(r4,r5);
r4 = _mm_or_si128(r4,r6); r4 = _mm_or_si128(r4,r6);
/* Only thing to do know is writing data to memory, but this gets a bit more /* Only thing to do know is writing data to memory, but this gets a bit more
* complicated if the width is not a multiple of four and it is the last column in line. */ * complicated if the width is not a multiple of four and it is the last column in line. */
if(last_column&0x02){ if (lastCol & 0x02)
{
/* let's say, we need to only convert six pixel in width /* let's say, we need to only convert six pixel in width
* Ok, the first 4 pixel will be converted just like every 4 pixel else, but * Ok, the first 4 pixel will be converted just like every 4 pixel else, but
* if it's the last loop in line, last_column is shifted left by one (curious? have a look above), * if it's the last loop in line, last_column is shifted left by one (curious? have a look above),
@ -294,8 +284,8 @@ pstatus_t ssse3_YUV420ToRGB_8u_P3AC4R(const BYTE **pSrc, int *srcStep,
} }
_mm_storeu_si128((__m128i *)pDst,r4); _mm_storeu_si128((__m128i *)pDst,r4);
if (!(lastRow & 0x02))
if(!(last_line&0x02)){ {
/* Because UV data is the same for two lines, we can process the secound line just here, /* Because UV data is the same for two lines, we can process the secound line just here,
* in the same loop. Only thing we need to do is to add some offsets to the Y- and destination * in the same loop. Only thing we need to do is to add some offsets to the Y- and destination
* pointer. These offsets are iStride[0] and the target scanline. * pointer. These offsets are iStride[0] and the target scanline.
@ -312,7 +302,6 @@ pstatus_t ssse3_YUV420ToRGB_8u_P3AC4R(const BYTE **pSrc, int *srcStep,
r5 = _mm_sub_epi32(r5,r0); r5 = _mm_sub_epi32(r5,r0);
r6 = _mm_add_epi32(r6,r1); r6 = _mm_add_epi32(r6,r1);
r4 = _mm_slli_epi32(r4,8); r4 = _mm_slli_epi32(r4,8);
r5 = _mm_slli_epi32(r5,8); r5 = _mm_slli_epi32(r5,8);
r6 = _mm_slli_epi32(r6,8); r6 = _mm_slli_epi32(r6,8);
@ -336,12 +325,11 @@ pstatus_t ssse3_YUV420ToRGB_8u_P3AC4R(const BYTE **pSrc, int *srcStep,
r7 = _mm_set_epi32(0x8080800E,0x8080800A,0x80808006,0x80808002); r7 = _mm_set_epi32(0x8080800E,0x8080800A,0x80808006,0x80808002);
r6 = _mm_shuffle_epi8(r6,r7); r6 = _mm_shuffle_epi8(r6,r7);
r4 = _mm_or_si128(r4,r5); r4 = _mm_or_si128(r4,r5);
r4 = _mm_or_si128(r4,r6); r4 = _mm_or_si128(r4,r6);
if (lastCol & 0x02)
if(last_column&0x02){ {
r6 = _mm_load_si128(buffer+3); r6 = _mm_load_si128(buffer+3);
r4 = _mm_and_si128(r4,r6); r4 = _mm_and_si128(r4,r6);
r5 = _mm_lddqu_si128((__m128i *)(pDst+dstStep)); r5 = _mm_lddqu_si128((__m128i *)(pDst+dstStep));
@ -350,7 +338,7 @@ pstatus_t ssse3_YUV420ToRGB_8u_P3AC4R(const BYTE **pSrc, int *srcStep,
/* only thing is, we should shift [rbp-42] back here, because we have processed the last column, /* only thing is, we should shift [rbp-42] back here, because we have processed the last column,
* and this "special condition" can be released */ * and this "special condition" can be released */
last_column=last_column>>1; lastCol >>= 1;
} }
_mm_storeu_si128((__m128i *)(pDst+dstStep),r4); _mm_storeu_si128((__m128i *)(pDst+dstStep),r4);
} }
@ -358,8 +346,8 @@ pstatus_t ssse3_YUV420ToRGB_8u_P3AC4R(const BYTE **pSrc, int *srcStep,
/* after all we have to increase the destination- and Y-data pointer by four pixel */ /* after all we have to increase the destination- and Y-data pointer by four pixel */
pDst += 16; pDst += 16;
YData += 4; YData += 4;
}
}while(i<nWidth); while (i < nWidth);
/* after each line we have to add the scanline to the destination pointer, because /* after each line we have to add the scanline to the destination pointer, because
* we are processing two lines at once, but only increasing the destination pointer * we are processing two lines at once, but only increasing the destination pointer