static void rfx_decode_component(RFX_CONTEXT* context, const UINT32* quantization_values,
const BYTE* data, int size, INT16* buffer)
{
INT16* dwt_buffer;
dwt_buffer = BufferPool_Take(context->priv->BufferPool, -1); /* dwt_buffer */
PROFILER_ENTER(context->priv->prof_rfx_decode_component);
PROFILER_ENTER(context->priv->prof_rfx_rlgr_decode);
context->rlgr_decode(context->mode, data, size, buffer, 4096);
PROFILER_EXIT(context->priv->prof_rfx_rlgr_decode);
PROFILER_ENTER(context->priv->prof_rfx_differential_decode);
rfx_differential_decode(buffer + 4032, 64);
PROFILER_EXIT(context->priv->prof_rfx_differential_decode);
PROFILER_ENTER(context->priv->prof_rfx_quantization_decode);
context->quantization_decode(buffer, quantization_values);
PROFILER_EXIT(context->priv->prof_rfx_quantization_decode);
PROFILER_ENTER(context->priv->prof_rfx_dwt_2d_decode);
context->dwt_2d_decode(buffer, dwt_buffer);
PROFILER_EXIT(context->priv->prof_rfx_dwt_2d_decode);
PROFILER_EXIT(context->priv->prof_rfx_decode_component);
BufferPool_Return(context->priv->BufferPool, dwt_buffer);
}
void rfx_decode_rgb(RFX_CONTEXT* context, STREAM* data_in,
int y_size, const uint32 * y_quants,
int cb_size, const uint32 * cb_quants,
int cr_size, const uint32 * cr_quants, uint8* rgb_buffer)
{
PROFILER_ENTER(context->priv->prof_rfx_decode_rgb);
rfx_decode_component(context, y_quants, stream_get_tail(data_in), y_size, context->priv->y_r_buffer); /* YData */
stream_seek(data_in, y_size);
rfx_decode_component(context, cb_quants, stream_get_tail(data_in), cb_size, context->priv->cb_g_buffer); /* CbData */
stream_seek(data_in, cb_size);
rfx_decode_component(context, cr_quants, stream_get_tail(data_in), cr_size, context->priv->cr_b_buffer); /* CrData */
stream_seek(data_in, cr_size);
PROFILER_ENTER(context->priv->prof_rfx_decode_ycbcr_to_rgb);
context->decode_ycbcr_to_rgb(context->priv->y_r_buffer, context->priv->cb_g_buffer, context->priv->cr_b_buffer);
PROFILER_EXIT(context->priv->prof_rfx_decode_ycbcr_to_rgb);
PROFILER_ENTER(context->priv->prof_rfx_decode_format_rgb);
rfx_decode_format_rgb(context->priv->y_r_buffer, context->priv->cb_g_buffer, context->priv->cr_b_buffer,
context->pixel_format, rgb_buffer);
PROFILER_EXIT(context->priv->prof_rfx_decode_format_rgb);
PROFILER_EXIT(context->priv->prof_rfx_decode_rgb);
}
/* 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;
}
static void rfx_decode_component(RFX_CONTEXT* context, const uint32* quantization_values,
const uint8* data, int size, sint16* buffer)
{
PROFILER_ENTER(context->priv->prof_rfx_decode_component);
PROFILER_ENTER(context->priv->prof_rfx_rlgr_decode);
rfx_rlgr_decode(context->mode, data, size, buffer, 4096);
PROFILER_EXIT(context->priv->prof_rfx_rlgr_decode);
PROFILER_ENTER(context->priv->prof_rfx_differential_decode);
rfx_differential_decode(buffer + 4032, 64);
PROFILER_EXIT(context->priv->prof_rfx_differential_decode);
PROFILER_ENTER(context->priv->prof_rfx_quantization_decode);
context->quantization_decode(buffer, quantization_values);
PROFILER_EXIT(context->priv->prof_rfx_quantization_decode);
PROFILER_ENTER(context->priv->prof_rfx_dwt_2d_decode);
context->dwt_2d_decode(buffer, context->priv->dwt_buffer);
PROFILER_EXIT(context->priv->prof_rfx_dwt_2d_decode);
PROFILER_EXIT(context->priv->prof_rfx_decode_component);
}
/* ------------------------------------------------------------------------- */
static BOOL test_YCoCgRToRGB_8u_AC4R_func(UINT32 width, UINT32 height)
{
pstatus_t status = -1;
BYTE* out_sse = NULL;
BYTE* in = NULL;
BYTE* out_c = NULL;
UINT32 i, x;
const UINT32 srcStride = width * 4;
const UINT32 size = srcStride * height;
const UINT32 formats[] =
{
PIXEL_FORMAT_ARGB32,
PIXEL_FORMAT_ABGR32,
PIXEL_FORMAT_RGBA32,
PIXEL_FORMAT_RGBX32,
PIXEL_FORMAT_BGRA32,
PIXEL_FORMAT_BGRX32
};
PROFILER_DEFINE(genericProf);
PROFILER_DEFINE(optProf);
in = _aligned_malloc(size, 16);
out_c = _aligned_malloc(size, 16);
out_sse = _aligned_malloc(size, 16);
if (!in || !out_c || !out_sse)
goto fail;
winpr_RAND(in, size);
for (x = 0; x < sizeof(formats) / sizeof(formats[0]); x++)
{
const UINT32 format = formats[x];
const UINT32 dstStride = width * GetBytesPerPixel(format);
const char* formatName = GetColorFormatName(format);
PROFILER_CREATE(genericProf, "YCoCgRToRGB_8u_AC4R-GENERIC");
PROFILER_CREATE(optProf, "YCoCgRToRGB_8u_AC4R-OPT");
PROFILER_ENTER(genericProf);
status = generic->YCoCgToRGB_8u_AC4R(
in, srcStride,
out_c, format, dstStride, width, height, 2, TRUE);
PROFILER_EXIT(genericProf);
if (status != PRIMITIVES_SUCCESS)
goto loop_fail;
PROFILER_ENTER(optProf);
status = optimized->YCoCgToRGB_8u_AC4R(
in, srcStride,
out_sse, format, dstStride, width, height, 2, TRUE);
PROFILER_EXIT(optProf);
if (status != PRIMITIVES_SUCCESS)
goto loop_fail;
if (memcmp(out_c, out_sse, dstStride * height) != 0)
{
for (i = 0; i < width * height; ++i)
{
const UINT32 c = ReadColor(out_c + 4 * i, format);
const UINT32 sse = ReadColor(out_sse + 4 * i, format);
if (c != sse)
{
printf("optimized->YCoCgRToRGB FAIL[%s] [%"PRIu32"]: 0x%08"PRIx32" -> C 0x%08"PRIx32" vs optimized 0x%08"PRIx32"\n",
formatName, i, in[i + 1], c, sse);
status = -1;
}
}
}
printf("--------------------------- [%s] [%"PRIu32"x%"PRIu32"] ---------------------------\n",
formatName, width, height);
PROFILER_PRINT_HEADER;
PROFILER_PRINT(genericProf);
PROFILER_PRINT(optProf);
PROFILER_PRINT_FOOTER;
loop_fail:
PROFILER_FREE(genericProf);
PROFILER_FREE(optProf);
if (status != PRIMITIVES_SUCCESS)
goto fail;
}
fail:
_aligned_free(in);
_aligned_free(out_c);
_aligned_free(out_sse);
return status == PRIMITIVES_SUCCESS;
}
/* 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;
}