void
wavelet_speed_cuda (SchroFrame *frame, int filter, int width, int height)
{
int i;
int j;
double start, stop;
SchroParams params;
params.transform_depth = 4;
params.iwt_luma_width = ROUND_UP_POW2(width, params.transform_depth);
params.iwt_luma_height = ROUND_UP_POW2(height, params.transform_depth);
params.iwt_chroma_width = params.iwt_luma_width/2;
params.iwt_chroma_height = params.iwt_luma_height;
params.wavelet_filter_index = filter;
for(j=0; j<10; j++) {
start = gettime();
for(i=0; i<10; i++) {
schro_gpuframe_iwt_transform (frame, ¶ms);
}
stop = gettime();
printf("time %g, %g fps\n", (stop - start)/10.0, 10.0/(stop-start));
}
}
/**
* schro_params_calculate_iwt_sizes:
* @params: pointer to @SchroParams structure
*
* Calculates the size of the array used for wavelet transformation
* using the current video format and transformation depth in the
* @params structure. The @params structure is updated with the new
* values.
*
* The structure fields changed are: iwt_chroma_width, iwt_chroma_height,
* iwt_luma_width, iwt_luma_height.
*/
void
schro_params_calculate_iwt_sizes (SchroParams * params)
{
SchroVideoFormat *video_format = params->video_format;
int picture_luma_width, picture_luma_height;
int picture_chroma_width, picture_chroma_height;
schro_video_format_get_picture_luma_size (video_format,
&picture_luma_width, &picture_luma_height);
params->iwt_luma_width =
ROUND_UP_POW2 (picture_luma_width, params->transform_depth);
params->iwt_luma_height =
ROUND_UP_POW2 (picture_luma_height, params->transform_depth);
schro_video_format_get_picture_chroma_size (video_format,
&picture_chroma_width, &picture_chroma_height);
params->iwt_chroma_width =
ROUND_UP_POW2 (picture_chroma_width, params->transform_depth);
params->iwt_chroma_height =
ROUND_UP_POW2 (picture_chroma_height, params->transform_depth);
SCHRO_DEBUG ("iwt chroma size %d x %d", params->iwt_chroma_width,
params->iwt_chroma_height);
SCHRO_DEBUG ("iwt luma size %d x %d", params->iwt_luma_width,
params->iwt_luma_height);
}
static int
get_buffer_size (GstSchroDec *schro_dec)
{
int size;
int width = schro_dec->width;
int height = schro_dec->height;
switch (schro_dec->fourcc) {
case GST_MAKE_FOURCC('I','4','2','0'):
size = ROUND_UP_POW2(width,2) * ROUND_UP_POW2(height,1);
size += 2 * ROUND_UP_POW2(ROUND_UP_SHIFT(width,1),2) *
ROUND_UP_SHIFT(height,1);
break;
case GST_MAKE_FOURCC('Y','U','Y','2'):
size = ROUND_UP_POW2(width,2) * 2 * height;
break;
case GST_MAKE_FOURCC('A','Y','U','V'):
size = width * 4 * height;
break;
default:
g_assert_not_reached();
break;
}
return size;
}
StreamBuffer::StreamBuffer(u32 type, u32 size)
: m_buffer(GenBuffer()), m_buffertype(type), m_size(ROUND_UP_POW2(size)),
m_bit_per_slot(IntLog2(ROUND_UP_POW2(size) / SYNC_POINTS))
{
m_iterator = 0;
m_used_iterator = 0;
m_free_iterator = 0;
}
StreamBuffer::StreamBuffer(u32 type, u32 size, u32 align_size, bool need_cpu_buffer)
: m_buffer(GenBuffer()),
m_buffertype(type),
m_size(Common::AlignUpSizePow2(ROUND_UP_POW2(size), align_size)),
m_bit_per_slot(IntLog2(Common::AlignUpSizePow2(ROUND_UP_POW2(size), align_size) / SYNC_POINTS))
{
m_iterator = 0;
m_used_iterator = 0;
m_free_iterator = 0;
for (int i = 0; i < SYNC_POINTS; i++)
{
m_fences[i] = 0;
}
}
// returns true if size was changed
bool FixTextureSize(int& width, int& height)
{
int oldw = width, oldh = height;
// conditional nonpow2 support should work fine for us
if ((caps.TextureCaps & D3DPTEXTURECAPS_POW2) && !(caps.TextureCaps & D3DPTEXTURECAPS_NONPOW2CONDITIONAL))
{
// all texture dimensions need to be powers of two
width = (int)ROUND_UP_POW2((u32)width);
height = (int)ROUND_UP_POW2((u32)height);
}
if (caps.TextureCaps & D3DPTEXTURECAPS_SQUAREONLY)
{
width = height = std::max(width, height);
}
width = std::min(width, (int)caps.MaxTextureWidth);
height = std::min(height, (int)caps.MaxTextureHeight);
return (width != oldw) || (height != oldh);
}
