/* ------------------------------------------------------------------------- */
PRIM_STATIC pstatus_t general_set_32u(
UINT32 val,
UINT32 *pDst,
INT32 len)
{
UINT32 *dptr = (UINT32 *) pDst;
size_t span, remaining;
primitives_t *prims;
if (len < 256)
{
while (len--) *dptr++ = val;
return PRIMITIVES_SUCCESS;
}
/* else quadratic growth memcpy algorithm */
span = 1;
*dptr = val;
remaining = len - 1;
prims = primitives_get();
while (remaining)
{
size_t thiswidth = span;
if (thiswidth > remaining) thiswidth = remaining;
prims->copy_8u((BYTE *) dptr, (BYTE *) (dptr + span), thiswidth<<2);
remaining -= thiswidth;
span <<= 1;
}
return PRIMITIVES_SUCCESS;
}
/* stride is bytes between rows in the output buffer. */
static void rfx_decode_format_rgb(INT16* r_buf, INT16* g_buf, INT16* b_buf,
RDP_PIXEL_FORMAT pixel_format, BYTE* dst_buf, int stride)
{
primitives_t *prims = primitives_get();
INT16* r = r_buf;
INT16* g = g_buf;
INT16* b = b_buf;
INT16* pSrc[3];
static const prim_size_t roi_64x64 = { 64, 64 };
BYTE* dst = dst_buf;
int x, y;
switch (pixel_format)
{
case RDP_PIXEL_FORMAT_B8G8R8A8:
pSrc[0] = r; pSrc[1] = g; pSrc[2] = b;
prims->RGBToRGB_16s8u_P3AC4R(
(const INT16 **) pSrc, 64*sizeof(INT16),
dst, stride, &roi_64x64);
break;
case RDP_PIXEL_FORMAT_R8G8B8A8:
pSrc[0] = b; pSrc[1] = g; pSrc[2] = r;
prims->RGBToRGB_16s8u_P3AC4R(
(const INT16 **) pSrc, 64*sizeof(INT16),
dst, stride, &roi_64x64);
break;
case RDP_PIXEL_FORMAT_B8G8R8:
for (y=0; y<64; y++)
{
for (x=0; x<64; x++)
{
*dst++ = (BYTE) (*b++);
*dst++ = (BYTE) (*g++);
*dst++ = (BYTE) (*r++);
}
dst += stride - (64*3);
}
break;
case RDP_PIXEL_FORMAT_R8G8B8:
for (y=0; y<64; y++)
{
for (x=0; x<64; x++)
{
*dst++ = (BYTE) (*r++);
*dst++ = (BYTE) (*g++);
*dst++ = (BYTE) (*b++);
}
dst += stride - (64*3);
}
break;
default:
break;
}
}
/* ------------------------------------------------------------------------- */
int test_sign16s_func(void)
{
INT16 ALIGN(src[65535]), ALIGN(d1[65535]), ALIGN(d2[65535]);
int failed = 0;
int i;
UINT32 pflags = primitives_get_flags(primitives_get());
char testStr[256];
/* Test when we can reach 16-byte alignment */
testStr[0] = '\0';
get_random_data(src, sizeof(src));
general_sign_16s(src+1, d1+1, 65535);
#ifdef _M_IX86_AMD64
if (pflags & PRIM_X86_SSSE3_AVAILABLE)
{
strcat(testStr, " SSSE3");
ssse3_sign_16s(src+1, d2+1, 65535);
for (i=1; i<65535; ++i)
{
if (d1[i] != d2[i])
{
printf("SIGN16s-SSE-aligned FAIL[%d] of %d: want %d, got %d\n",
i, src[i], d1[i], d2[i]);
++failed;
}
}
}
#endif /* i386 */
/* Test when we cannot reach 16-byte alignment */
get_random_data(src, sizeof(src));
general_sign_16s(src+1, d1+2, 65535);
#ifdef _M_IX86_AMD64
if (pflags & PRIM_X86_SSSE3_AVAILABLE)
{
ssse3_sign_16s(src+1, d2+2, 65535);
for (i=2; i<65535; ++i)
{
if (d1[i] != d2[i])
{
printf("SIGN16s-SSE-unaligned FAIL[%d] of %d: want %d, got %d\n",
i, src[i-1], d1[i], d2[i]);
++failed;
}
}
}
#endif /* i386 */
if (!failed) printf("All sign16s tests passed (%s).\n", testStr);
return (failed > 0) ? FAILURE : SUCCESS;
}
void rfx_quantization_decode(INT16* buffer, const UINT32* quantVals)
{
const primitives_t* prims = primitives_get();
rfx_quantization_decode_block(prims, &buffer[0], 1024, quantVals[8] - 1); /* HL1 */
rfx_quantization_decode_block(prims, &buffer[1024], 1024, quantVals[7] - 1); /* LH1 */
rfx_quantization_decode_block(prims, &buffer[2048], 1024, quantVals[9] - 1); /* HH1 */
rfx_quantization_decode_block(prims, &buffer[3072], 256, quantVals[5] - 1); /* HL2 */
rfx_quantization_decode_block(prims, &buffer[3328], 256, quantVals[4] - 1); /* LH2 */
rfx_quantization_decode_block(prims, &buffer[3584], 256, quantVals[6] - 1); /* HH2 */
rfx_quantization_decode_block(prims, &buffer[3840], 64, quantVals[2] - 1); /* HL3 */
rfx_quantization_decode_block(prims, &buffer[3904], 64, quantVals[1] - 1); /* LH3 */
rfx_quantization_decode_block(prims, &buffer[3968], 64, quantVals[3] - 1); /* HH3 */
rfx_quantization_decode_block(prims, &buffer[4032], 64, quantVals[0] - 1); /* LL3 */
}
/* stride is bytes between rows in the output buffer. */
BOOL rfx_decode_rgb(RFX_CONTEXT* context, RFX_TILE* tile, BYTE* rgb_buffer, int stride)
{
INT16* pSrcDst[3];
UINT32 *y_quants, *cb_quants, *cr_quants;
static const prim_size_t roi_64x64 = { 64, 64 };
const primitives_t *prims = primitives_get();
PROFILER_ENTER(context->priv->prof_rfx_decode_rgb);
y_quants = context->quants + (tile->quantIdxY * 10);
cb_quants = context->quants + (tile->quantIdxCb * 10);
cr_quants = context->quants + (tile->quantIdxCr * 10);
pSrcDst[0] = (INT16*)((BYTE*)BufferPool_Take(context->priv->BufferPool, -1) + 16); /* y_r_buffer */
pSrcDst[1] = (INT16*)((BYTE*)BufferPool_Take(context->priv->BufferPool, -1) + 16); /* cb_g_buffer */
pSrcDst[2] = (INT16*)((BYTE*)BufferPool_Take(context->priv->BufferPool, -1) + 16); /* cr_b_buffer */
rfx_decode_component(context, y_quants, tile->YData, tile->YLen, pSrcDst[0]); /* YData */
rfx_decode_component(context, cb_quants, tile->CbData, tile->CbLen, pSrcDst[1]); /* CbData */
rfx_decode_component(context, cr_quants, tile->CrData, tile->CrLen, pSrcDst[2]); /* CrData */
PROFILER_ENTER(context->priv->prof_rfx_ycbcr_to_rgb);
prims->yCbCrToRGB_16s16s_P3P3((const INT16**) pSrcDst, 64 * sizeof(INT16),
pSrcDst, 64 * sizeof(INT16), &roi_64x64);
PROFILER_EXIT(context->priv->prof_rfx_ycbcr_to_rgb);
PROFILER_ENTER(context->priv->prof_rfx_decode_format_rgb);
rfx_decode_format_rgb(pSrcDst[0], pSrcDst[1], pSrcDst[2],
context->pixel_format, rgb_buffer, stride);
PROFILER_EXIT(context->priv->prof_rfx_decode_format_rgb);
PROFILER_EXIT(context->priv->prof_rfx_decode_rgb);
BufferPool_Return(context->priv->BufferPool, (BYTE*)pSrcDst[0] - 16);
BufferPool_Return(context->priv->BufferPool, (BYTE*)pSrcDst[1] - 16);
BufferPool_Return(context->priv->BufferPool, (BYTE*)pSrcDst[2] - 16);
return TRUE;
}
int h264_compress(H264_CONTEXT* h264, BYTE* pSrcData, DWORD SrcFormat,
int nSrcStep, int nSrcWidth, int nSrcHeight, UINT32 TargetFrameSizeInBits,
BYTE** ppDstData, UINT32* pDstSize)
{
int status;
prim_size_t roi;
int nWidth, nHeight;
primitives_t *prims = primitives_get();
if (!h264)
return -1;
if (!h264->subsystem->Compress)
return -1;
nWidth = (nSrcWidth + 1) & ~1;
nHeight = (nSrcHeight + 1) & ~1;
h264->pYUVData[0] = (BYTE*) malloc(nWidth * nHeight);
h264->iStride[0] = nWidth;
h264->pYUVData[1] = (BYTE*) malloc(nWidth * nHeight / 4);
h264->iStride[1] = nWidth / 2;
h264->pYUVData[2] = (BYTE*) malloc(nWidth * nHeight / 4);
h264->iStride[2] = nWidth / 2;
h264->width = nWidth;
h264->height = nHeight;
roi.width = nSrcWidth;
roi.height = nSrcHeight;
prims->RGBToYUV420_8u_P3AC4R(pSrcData, nSrcStep, h264->pYUVData, h264->iStride, &roi);
status = h264->subsystem->Compress(h264, TargetFrameSizeInBits, ppDstData, pDstSize);
free(h264->pYUVData[0]);
free(h264->pYUVData[1]);
free(h264->pYUVData[2]);
h264->pYUVData[0] = NULL;
h264->pYUVData[1] = NULL;
h264->pYUVData[2] = NULL;
return status;
}
/* ------------------------------------------------------------------------- */
int test_copy8u_func(void)
{
primitives_t *prims = primitives_get();
BYTE ALIGN(data[COPY_TESTSIZE+15]);
int i, soff;
int failed = 0;
char testStr[256];
BYTE ALIGN(dest[COPY_TESTSIZE+15]);
testStr[0] = '\0';
get_random_data(data, sizeof(data));
strcat(testStr, " ptr");
for (soff=0; soff<16; ++soff)
{
int doff;
for (doff=0; doff<16; ++doff)
{
int length;
for (length=1; length<=COPY_TESTSIZE-doff; ++length)
{
memset(dest, 0, sizeof(dest));
prims->copy_8u(data+soff, dest+doff, length);
for (i=0; i<length; ++i)
{
if (dest[i+doff] != data[i+soff])
{
printf("COPY8U FAIL: off=%d len=%d, dest[%d]=0x%02x"
"data[%d]=0x%02x\n",
doff, length, i+doff, dest[i+doff],
i+soff, data[i+soff]);
failed = 1;
}
}
}
}
}
if (!failed) printf("All copy8 tests passed (%s).\n", testStr);
return (failed > 0) ? FAILURE : SUCCESS;
}
/* Copy a block of pixels from one buffer to another.
* The addresses are assumed to have been already offset to the upper-left
* corners of the source and destination region of interest.
*/
PRIM_STATIC pstatus_t general_copy_8u_AC4r(
const BYTE *pSrc, INT32 srcStep,
BYTE *pDst, INT32 dstStep,
INT32 width, INT32 height)
{
primitives_t *prims = primitives_get();
const BYTE *src = (const BYTE *) pSrc;
BYTE *dst = (BYTE *) pDst;
int rowbytes = width * sizeof(UINT32);
if ((width == 0) || (height == 0)) return PRIMITIVES_SUCCESS;
if (memory_regions_overlap_2d(pSrc, srcStep, sizeof(UINT32),
pDst, dstStep, sizeof(UINT32), width, height))
{
do {
prims->copy(src, dst, rowbytes);
src += srcStep;
dst += dstStep;
} while (--height);
}
else
{
/* TODO: do it in one operation when the rowdata is adjacent. */
do {
/* If we find a replacement for memcpy that is consistently
* faster, this could be replaced with that.
*/
memcpy(dst, src, rowbytes);
src += srcStep;
dst += dstStep;
} while (--height);
}
return PRIMITIVES_SUCCESS;
}
RFX_CONTEXT* rfx_context_new(void)
{
HKEY hKey;
LONG status;
DWORD dwType;
DWORD dwSize;
DWORD dwValue;
SYSTEM_INFO sysinfo;
RFX_CONTEXT* context;
context = (RFX_CONTEXT*) malloc(sizeof(RFX_CONTEXT));
ZeroMemory(context, sizeof(RFX_CONTEXT));
context->priv = (RFX_CONTEXT_PRIV*) malloc(sizeof(RFX_CONTEXT_PRIV));
ZeroMemory(context->priv, sizeof(RFX_CONTEXT_PRIV));
context->priv->TilePool = Queue_New(TRUE, -1, -1);
context->priv->TileQueue = Queue_New(TRUE, -1, -1);
/*
* align buffers to 16 byte boundary (needed for SSE/NEON instructions)
*
* y_r_buffer, cb_g_buffer, cr_b_buffer: 64 * 64 * 4 = 16384 (0x4000)
* dwt_buffer: 32 * 32 * 2 * 2 * 4 = 16384, maximum sub-band width is 32
*/
context->priv->BufferPool = BufferPool_New(TRUE, 16384, 16);
#ifdef _WIN32
{
BOOL isVistaOrLater;
OSVERSIONINFOA verinfo;
ZeroMemory(&verinfo, sizeof(OSVERSIONINFOA));
verinfo.dwOSVersionInfoSize = sizeof(OSVERSIONINFOA);
GetVersionExA(&verinfo);
isVistaOrLater = ((verinfo.dwMajorVersion >= 6) && (verinfo.dwMinorVersion >= 0)) ? TRUE : FALSE;
context->priv->UseThreads = isVistaOrLater;
}
#else
context->priv->UseThreads = TRUE;
#endif
GetNativeSystemInfo(&sysinfo);
context->priv->MinThreadCount = sysinfo.dwNumberOfProcessors;
context->priv->MaxThreadCount = 0;
status = RegOpenKeyEx(HKEY_LOCAL_MACHINE, _T("Software\\FreeRDP\\RemoteFX"), 0, KEY_READ | KEY_WOW64_64KEY, &hKey);
if (status == ERROR_SUCCESS)
{
if (RegQueryValueEx(hKey, _T("UseThreads"), NULL, &dwType, (BYTE*) &dwValue, &dwSize) == ERROR_SUCCESS)
context->priv->UseThreads = dwValue ? 1 : 0;
if (RegQueryValueEx(hKey, _T("MinThreadCount"), NULL, &dwType, (BYTE*) &dwValue, &dwSize) == ERROR_SUCCESS)
context->priv->MinThreadCount = dwValue;
if (RegQueryValueEx(hKey, _T("MaxThreadCount"), NULL, &dwType, (BYTE*) &dwValue, &dwSize) == ERROR_SUCCESS)
context->priv->MaxThreadCount = dwValue;
RegCloseKey(hKey);
}
if (context->priv->UseThreads)
{
/* Call primitives_get here in order to avoid race conditions when using primitives_get */
/* from multiple threads. This call will initialize all function pointers correctly */
/* before any decoding threads are started */
primitives_get();
context->priv->ThreadPool = CreateThreadpool(NULL);
InitializeThreadpoolEnvironment(&context->priv->ThreadPoolEnv);
SetThreadpoolCallbackPool(&context->priv->ThreadPoolEnv, context->priv->ThreadPool);
if (context->priv->MinThreadCount)
SetThreadpoolThreadMinimum(context->priv->ThreadPool, context->priv->MinThreadCount);
if (context->priv->MaxThreadCount)
SetThreadpoolThreadMaximum(context->priv->ThreadPool, context->priv->MaxThreadCount);
}
/* initialize the default pixel format */
rfx_context_set_pixel_format(context, RDP_PIXEL_FORMAT_B8G8R8A8);
/* create profilers for default decoding routines */
rfx_profiler_create(context);
/* set up default routines */
context->quantization_decode = rfx_quantization_decode;
context->quantization_encode = rfx_quantization_encode;
context->dwt_2d_decode = rfx_dwt_2d_decode;
context->dwt_2d_encode = rfx_dwt_2d_encode;
RFX_INIT_SIMD(context);
return context;
}
static BOOL TestPrimitiveYUV(BOOL use444)
{
BOOL rc = FALSE;
UINT32 x, y;
UINT32 awidth, aheight;
BYTE* yuv[3] = {0};
UINT32 yuv_step[3];
prim_size_t roi;
BYTE* rgb = NULL;
BYTE* rgb_dst = NULL;
size_t size;
primitives_t* prims = primitives_get();
size_t uvsize, uvwidth;
size_t padding = 10000;
size_t stride;
get_size(&roi.width, &roi.height);
/* Buffers need to be 16x16 aligned. */
awidth = roi.width + 16 - roi.width % 16;
aheight = roi.height + 16 - roi.height % 16;
stride = awidth * sizeof(UINT32);
size = awidth * aheight;
if (use444)
{
uvwidth = awidth;
uvsize = size;
if (!prims || !prims->RGBToYUV444_8u_P3AC4R || !prims->YUV444ToRGB_8u_P3AC4R)
return FALSE;
}
else
{
uvwidth = (awidth + 1) / 2;
uvsize = (aheight + 1) / 2 * uvwidth;
if (!prims || !prims->RGBToYUV420_8u_P3AC4R || !prims->YUV420ToRGB_8u_P3AC4R)
return FALSE;
}
fprintf(stderr, "Running AVC%s on frame size %"PRIu32"x%"PRIu32"\n", use444 ? "444" : "420",
roi.width, roi.height);
/* Test RGB to YUV444 conversion and vice versa */
if (!(rgb = set_padding(size * sizeof(UINT32), padding)))
goto fail;
if (!(rgb_dst = set_padding(size * sizeof(UINT32), padding)))
goto fail;
if (!(yuv[0] = set_padding(size, padding)))
goto fail;
if (!(yuv[1] = set_padding(uvsize, padding)))
goto fail;
if (!(yuv[2] = set_padding(uvsize, padding)))
goto fail;
for (y = 0; y < roi.height; y++)
{
BYTE* line = &rgb[y * stride];
for (x = 0; x < roi.width; x++)
{
line[x * 4 + 0] = 0x81;
line[x * 4 + 1] = 0x33;
line[x * 4 + 2] = 0xAB;
line[x * 4 + 3] = 0xFF;
}
}
yuv_step[0] = awidth;
yuv_step[1] = uvwidth;
yuv_step[2] = uvwidth;
if (use444)
{
if (prims->RGBToYUV444_8u_P3AC4R(rgb, PIXEL_FORMAT_BGRA32,
stride, yuv, yuv_step,
&roi) != PRIMITIVES_SUCCESS)
goto fail;
}
else if (prims->RGBToYUV420_8u_P3AC4R(rgb, PIXEL_FORMAT_BGRA32,
stride, yuv, yuv_step,
&roi) != PRIMITIVES_SUCCESS)
goto fail;
if (!check_padding(rgb, size * sizeof(UINT32), padding, "rgb"))
goto fail;
if ((!check_padding(yuv[0], size, padding, "Y")) ||
(!check_padding(yuv[1], uvsize, padding, "U")) ||
(!check_padding(yuv[2], uvsize, padding, "V")))
goto fail;
if (use444)
{
if (prims->YUV444ToRGB_8u_P3AC4R((const BYTE**)yuv, yuv_step, rgb_dst, stride,
PIXEL_FORMAT_BGRA32,
&roi) != PRIMITIVES_SUCCESS)
goto fail;
}
else if (prims->YUV420ToRGB_8u_P3AC4R((const BYTE**)yuv, yuv_step, rgb_dst,
stride, PIXEL_FORMAT_BGRA32, &roi) != PRIMITIVES_SUCCESS)
goto fail;
if (!check_padding(rgb_dst, size * sizeof(UINT32), padding, "rgb dst"))
goto fail;
if ((!check_padding(yuv[0], size, padding, "Y")) ||
(!check_padding(yuv[1], uvsize, padding, "U")) ||
(!check_padding(yuv[2], uvsize, padding, "V")))
goto fail;
for (y = 0; y < roi.height; y++)
{
BYTE* srgb = &rgb[y * stride];
BYTE* drgb = &rgb_dst[y * stride];
if (!similar(srgb, drgb, roi.width * sizeof(UINT32)))
goto fail;
}
rc = TRUE;
fail:
free_padding(rgb, padding);
free_padding(rgb_dst, padding);
free_padding(yuv[0], padding);
free_padding(yuv[1], padding);
free_padding(yuv[2], padding);
return rc;
}
/* Create 2 pseudo YUV420 frames of same size.
* Combine them and check, if the data is at the expected position. */
static BOOL TestPrimitiveYUVCombine(void)
{
UINT32 x, y, i;
UINT32 awidth, aheight;
BOOL rc = FALSE;
BYTE* luma[3] = { 0 };
BYTE* chroma[3] = { 0 };
BYTE* yuv[3] = { 0 };
BYTE* pmain[3] = { 0 };
BYTE* paux[3] = { 0 };
UINT32 lumaStride[3];
UINT32 chromaStride[3];
UINT32 yuvStride[3];
size_t padding = 10000;
prim_size_t roi;
primitives_t* prims = primitives_get();
get_size(&roi.width, &roi.height);
awidth = roi.width + 16 - roi.width % 16;
aheight = roi.height + 16 - roi.height % 16;
fprintf(stderr, "Running YUVCombine on frame size %"PRIu32"x%"PRIu32" [%"PRIu32"x%"PRIu32"]\n",
roi.width, roi.height, awidth, aheight);
if (!prims || !prims->YUV420CombineToYUV444)
goto fail;
for (x = 0; x < 3; x++)
{
size_t halfStride = ((x > 0) ? awidth / 2 : awidth);
size_t size = aheight * awidth;
size_t halfSize = ((x > 0) ? halfStride * aheight / 2 : awidth * aheight);
yuvStride[x] = awidth;
if (!(yuv[x] = set_padding(size, padding)))
goto fail;
lumaStride[x] = halfStride;
if (!(luma[x] = set_padding(halfSize, padding)))
goto fail;
if (!(pmain[x] = set_padding(halfSize, padding)))
goto fail;
chromaStride[x] = halfStride;
if (!(chroma[x] = set_padding(halfSize, padding)))
goto fail;
if (!(paux[x] = set_padding(halfSize, padding)))
goto fail;
memset(luma[x], 0xAB + 3 * x, halfSize);
memset(chroma[x], 0x80 + 2 * x, halfSize);
if (!check_padding(luma[x], halfSize, padding, "luma"))
goto fail;
if (!check_padding(chroma[x], halfSize, padding, "chroma"))
goto fail;
if (!check_padding(pmain[x], halfSize, padding, "main"))
goto fail;
if (!check_padding(paux[x], halfSize, padding, "aux"))
goto fail;
if (!check_padding(yuv[x], size, padding, "yuv"))
goto fail;
}
if (prims->YUV420CombineToYUV444((const BYTE**)luma, lumaStride,
(const BYTE**)chroma, chromaStride,
yuv, yuvStride, &roi) != PRIMITIVES_SUCCESS)
goto fail;
for (x = 0; x < 3; x++)
{
size_t halfStride = ((x > 0) ? awidth / 2 : awidth);
size_t size = aheight * awidth;
size_t halfSize = ((x > 0) ? halfStride * aheight / 2 : awidth * aheight);
if (!check_padding(luma[x], halfSize, padding, "luma"))
goto fail;
if (!check_padding(chroma[x], halfSize, padding, "chroma"))
goto fail;
if (!check_padding(yuv[x], size, padding, "yuv"))
goto fail;
}
if (prims->YUV444SplitToYUV420((const BYTE**)yuv, yuvStride, pmain, lumaStride,
paux, chromaStride, &roi) != PRIMITIVES_SUCCESS)
goto fail;
for (x = 0; x < 3; x++)
{
size_t halfStride = ((x > 0) ? awidth / 2 : awidth);
size_t size = aheight * awidth;
size_t halfSize = ((x > 0) ? halfStride * aheight / 2 : awidth * aheight);
if (!check_padding(pmain[x], halfSize, padding, "main"))
goto fail;
if (!check_padding(paux[x], halfSize, padding, "aux"))
goto fail;
if (!check_padding(yuv[x], size, padding, "yuv"))
goto fail;
}
for (i = 0; i < 3; i++)
{
for (y = 0; y < roi.height; y++)
{
UINT32 w = roi.width;
UINT32 lstride = lumaStride[i];
UINT32 cstride = chromaStride[i];
if (i > 0)
{
w = (roi.width + 3) / 4;
if (roi.height > (roi.height + 1) / 2)
continue;
}
if (!similar(luma[i] + y * lstride,
pmain[i] + y * lstride,
w))
goto fail;
/* Need to ignore lines of destination Y plane,
* if the lines are not a multiple of 16
* as the UV planes are packed in 8 line stripes. */
if (i == 0)
{
/* TODO: This check is not perfect, it does not
* include the last V lines packed to the Y
* frame. */
UINT32 rem = roi.height % 16;
if (y > roi.height - rem)
continue;
}
if (!similar(chroma[i] + y * cstride,
paux[i] + y * cstride,
w))
goto fail;
}
}
rc = TRUE;
fail:
for (x = 0; x < 3; x++)
{
free_padding(yuv[x], padding);
free_padding(luma[x], padding);
free_padding(chroma[x], padding);
free_padding(pmain[x], padding);
free_padding(paux[x], padding);
}
return rc;
}
int h264_decompress(H264_CONTEXT* h264, BYTE* pSrcData, UINT32 SrcSize,
BYTE** ppDstData, DWORD DstFormat, int nDstStep, int nDstWidth,
int nDstHeight, RDPGFX_RECT16* regionRects, int numRegionRects)
{
int index;
int status;
int* iStride;
BYTE* pDstData;
BYTE* pDstPoint;
prim_size_t roi;
BYTE** pYUVData;
int width, height;
BYTE* pYUVPoint[3];
RDPGFX_RECT16* rect;
primitives_t *prims = primitives_get();
if (!h264)
return -1;
#if 0
WLog_INFO(TAG, "h264_decompress: pSrcData=%p, SrcSize=%u, pDstData=%p, nDstStep=%d, nDstHeight=%d, numRegionRects=%d",
pSrcData, SrcSize, *ppDstData, nDstStep, nDstHeight, numRegionRects);
#endif
if (!(pDstData = *ppDstData))
return -1;
if ((status = h264->subsystem->Decompress(h264, pSrcData, SrcSize)) < 0)
return status;
pYUVData = h264->pYUVData;
iStride = h264->iStride;
for (index = 0; index < numRegionRects; index++)
{
rect = &(regionRects[index]);
/* Check, if the ouput rectangle is valid in decoded h264 frame. */
if ((rect->right > h264->width) || (rect->left > h264->width))
return -1;
if ((rect->top > h264->height) || (rect->bottom > h264->height))
return -1;
/* Check, if the output rectangle is valid in destination buffer. */
if ((rect->right > nDstWidth) || (rect->left > nDstWidth))
return -1;
if ((rect->bottom > nDstHeight) || (rect->top > nDstHeight))
return -1;
width = rect->right - rect->left;
height = rect->bottom - rect->top;
pDstPoint = pDstData + rect->top * nDstStep + rect->left * 4;
pYUVPoint[0] = pYUVData[0] + rect->top * iStride[0] + rect->left;
pYUVPoint[1] = pYUVData[1] + rect->top/2 * iStride[1] + rect->left/2;
pYUVPoint[2] = pYUVData[2] + rect->top/2 * iStride[2] + rect->left/2;
#if 0
WLog_INFO(TAG, "regionRect: x: %d y: %d width: %d height: %d",
rect->left, rect->top, width, height);
#endif
roi.width = width;
roi.height = height;
prims->YUV420ToRGB_8u_P3AC4R((const BYTE**) pYUVPoint, iStride, pDstPoint, nDstStep, &roi);
}
return 1;
}
RFX_CONTEXT* rfx_context_new(BOOL encoder)
{
HKEY hKey;
LONG status;
DWORD dwType;
DWORD dwSize;
DWORD dwValue;
SYSTEM_INFO sysinfo;
RFX_CONTEXT* context;
wObject *pool;
RFX_CONTEXT_PRIV *priv;
context = (RFX_CONTEXT*)calloc(1, sizeof(RFX_CONTEXT));
if (!context)
return NULL;
context->encoder = encoder;
context->priv = priv = (RFX_CONTEXT_PRIV *)calloc(1, sizeof(RFX_CONTEXT_PRIV) );
if (!priv)
goto error_priv;
WLog_Init();
priv->log = WLog_Get("com.freerdp.codec.rfx");
WLog_OpenAppender(priv->log);
#ifdef WITH_DEBUG_RFX
WLog_SetLogLevel(priv->log, WLOG_DEBUG);
#endif
priv->TilePool = ObjectPool_New(TRUE);
if (!priv->TilePool)
goto error_tilePool;
pool = ObjectPool_Object(priv->TilePool);
pool->fnObjectInit = (OBJECT_INIT_FN) rfx_tile_init;
if (context->encoder)
{
pool->fnObjectNew = (OBJECT_NEW_FN) rfx_encoder_tile_new;
pool->fnObjectFree = (OBJECT_FREE_FN) rfx_encoder_tile_free;
}
else
{
pool->fnObjectNew = (OBJECT_NEW_FN) rfx_decoder_tile_new;
pool->fnObjectFree = (OBJECT_FREE_FN) rfx_decoder_tile_free;
}
/*
* align buffers to 16 byte boundary (needed for SSE/NEON instructions)
*
* y_r_buffer, cb_g_buffer, cr_b_buffer: 64 * 64 * sizeof(INT16) = 8192 (0x2000)
* dwt_buffer: 32 * 32 * 2 * 2 * sizeof(INT16) = 8192, maximum sub-band width is 32
*
* Additionally we add 32 bytes (16 in front and 16 at the back of the buffer)
* in order to allow optimized functions (SEE, NEON) to read from positions
* that are actually in front/beyond the buffer. Offset calculations are
* performed at the BufferPool_Take function calls in rfx_encode/decode.c.
*
* We then multiply by 3 to use a single, partioned buffer for all 3 channels.
*/
priv->BufferPool = BufferPool_New(TRUE, (8192 + 32) * 3, 16);
if (!priv->BufferPool)
goto error_BufferPool;
#ifdef _WIN32
{
BOOL isVistaOrLater;
OSVERSIONINFOA verinfo;
ZeroMemory(&verinfo, sizeof(OSVERSIONINFOA));
verinfo.dwOSVersionInfoSize = sizeof(OSVERSIONINFOA);
GetVersionExA(&verinfo);
isVistaOrLater = ((verinfo.dwMajorVersion >= 6) && (verinfo.dwMinorVersion >= 0)) ? TRUE : FALSE;
priv->UseThreads = isVistaOrLater;
}
#else
priv->UseThreads = TRUE;
#endif
GetNativeSystemInfo(&sysinfo);
priv->MinThreadCount = sysinfo.dwNumberOfProcessors;
priv->MaxThreadCount = 0;
status = RegOpenKeyEx(HKEY_LOCAL_MACHINE, _T("Software\\FreeRDP\\RemoteFX"), 0, KEY_READ | KEY_WOW64_64KEY, &hKey);
if (status == ERROR_SUCCESS)
{
dwSize = sizeof(dwValue);
if (RegQueryValueEx(hKey, _T("UseThreads"), NULL, &dwType, (BYTE*) &dwValue, &dwSize) == ERROR_SUCCESS)
priv->UseThreads = dwValue ? 1 : 0;
if (RegQueryValueEx(hKey, _T("MinThreadCount"), NULL, &dwType, (BYTE*) &dwValue, &dwSize) == ERROR_SUCCESS)
priv->MinThreadCount = dwValue;
if (RegQueryValueEx(hKey, _T("MaxThreadCount"), NULL, &dwType, (BYTE*) &dwValue, &dwSize) == ERROR_SUCCESS)
priv->MaxThreadCount = dwValue;
RegCloseKey(hKey);
}
if (priv->UseThreads)
{
/* Call primitives_get here in order to avoid race conditions when using primitives_get */
/* from multiple threads. This call will initialize all function pointers correctly */
/* before any decoding threads are started */
primitives_get();
priv->ThreadPool = CreateThreadpool(NULL);
if (!priv->ThreadPool)
goto error_threadPool;
InitializeThreadpoolEnvironment(&priv->ThreadPoolEnv);
SetThreadpoolCallbackPool(&priv->ThreadPoolEnv, priv->ThreadPool);
if (priv->MinThreadCount)
if (!SetThreadpoolThreadMinimum(priv->ThreadPool, priv->MinThreadCount))
goto error_threadPool_minimum;
if (priv->MaxThreadCount)
SetThreadpoolThreadMaximum(priv->ThreadPool, priv->MaxThreadCount);
}
/* initialize the default pixel format */
rfx_context_set_pixel_format(context, RDP_PIXEL_FORMAT_B8G8R8A8);
/* create profilers for default decoding routines */
rfx_profiler_create(context);
/* set up default routines */
context->quantization_decode = rfx_quantization_decode;
context->quantization_encode = rfx_quantization_encode;
context->dwt_2d_decode = rfx_dwt_2d_decode;
context->dwt_2d_encode = rfx_dwt_2d_encode;
RFX_INIT_SIMD(context);
context->state = RFX_STATE_SEND_HEADERS;
return context;
error_threadPool_minimum:
CloseThreadpool(priv->ThreadPool);
error_threadPool:
BufferPool_Free(priv->BufferPool);
error_BufferPool:
ObjectPool_Free(priv->TilePool);
error_tilePool:
free(priv);
error_priv:
free(context);
return NULL;
}
/* stride is bytes between rows in the output buffer. */
BOOL rfx_decode_rgb(RFX_CONTEXT* context, wStream* data_in,
int y_size, const UINT32* y_quants,
int cb_size, const UINT32* cb_quants,
int cr_size, const UINT32* cr_quants, BYTE* rgb_buffer, int stride)
{
INT16* pSrcDst[3];
static const prim_size_t roi_64x64 = { 64, 64 };
const primitives_t *prims = primitives_get();
PROFILER_ENTER(context->priv->prof_rfx_decode_rgb);
pSrcDst[0] = (INT16*)((BYTE*)BufferPool_Take(context->priv->BufferPool, -1) + 16); /* y_r_buffer */
pSrcDst[1] = (INT16*)((BYTE*)BufferPool_Take(context->priv->BufferPool, -1) + 16); /* cb_g_buffer */
pSrcDst[2] = (INT16*)((BYTE*)BufferPool_Take(context->priv->BufferPool, -1) + 16); /* cr_b_buffer */
#if 0
if (context->priv->UseThreads)
{
PTP_WORK work_objects[3];
RFX_COMPONENT_WORK_PARAM params[3];
params[0].context = context;
params[0].quantization_values = y_quants;
params[0].buffer = stream_get_tail(data_in);
params[0].capacity = y_size;
params[0].buffer = pSrcDst[0];
stream_seek(data_in, y_size);
params[1].context = context;
params[1].quantization_values = cb_quants;
params[1].buffer = stream_get_tail(data_in);
params[1].capacity = cb_size;
params[1].buffer = pSrcDst[1];
stream_seek(data_in, cb_size);
params[2].context = context;
params[2].quantization_values = cr_quants;
params[2].buffer = stream_get_tail(data_in);
params[2].capacity = cr_size;
params[2].buffer = pSrcDst[2];
stream_seek(data_in, cr_size);
work_objects[0] = CreateThreadpoolWork((PTP_WORK_CALLBACK) rfx_decode_component_work_callback,
(void*) ¶ms[0], &context->priv->ThreadPoolEnv);
work_objects[1] = CreateThreadpoolWork((PTP_WORK_CALLBACK) rfx_decode_component_work_callback,
(void*) ¶ms[1], &context->priv->ThreadPoolEnv);
work_objects[2] = CreateThreadpoolWork((PTP_WORK_CALLBACK) rfx_decode_component_work_callback,
(void*) ¶ms[2], &context->priv->ThreadPoolEnv);
SubmitThreadpoolWork(work_objects[0]);
SubmitThreadpoolWork(work_objects[1]);
SubmitThreadpoolWork(work_objects[2]);
WaitForThreadpoolWorkCallbacks(work_objects[0], FALSE);
WaitForThreadpoolWorkCallbacks(work_objects[1], FALSE);
WaitForThreadpoolWorkCallbacks(work_objects[2], FALSE);
}
else
#endif
{
if (stream_get_left(data_in) < y_size+cb_size+cr_size)
{
DEBUG_WARN("rfx_decode_rgb: packet too small for y_size+cb_size+cr_size");
return FALSE;
}
rfx_decode_component(context, y_quants, stream_get_tail(data_in), y_size, pSrcDst[0]); /* YData */
stream_seek(data_in, y_size);
rfx_decode_component(context, cb_quants, stream_get_tail(data_in), cb_size, pSrcDst[1]); /* CbData */
stream_seek(data_in, cb_size);
rfx_decode_component(context, cr_quants, stream_get_tail(data_in), cr_size, pSrcDst[2]); /* CrData */
stream_seek(data_in, cr_size);
}
prims->yCbCrToRGB_16s16s_P3P3((const INT16**) pSrcDst, 64 * sizeof(INT16),
pSrcDst, 64 * sizeof(INT16), &roi_64x64);
PROFILER_ENTER(context->priv->prof_rfx_decode_format_rgb);
rfx_decode_format_rgb(pSrcDst[0], pSrcDst[1], pSrcDst[2],
context->pixel_format, rgb_buffer, stride);
PROFILER_EXIT(context->priv->prof_rfx_decode_format_rgb);
PROFILER_EXIT(context->priv->prof_rfx_decode_rgb);
BufferPool_Return(context->priv->BufferPool, (BYTE*)pSrcDst[0] - 16);
BufferPool_Return(context->priv->BufferPool, (BYTE*)pSrcDst[1] - 16);
BufferPool_Return(context->priv->BufferPool, (BYTE*)pSrcDst[2] - 16);
return TRUE;
}
