mirror-github-bspeice/cuburn
1
0
Fork 0
mirror of https://github.com/stevenrobertson/cuburn.git synced 2025-02-05 11:40:04 -05:00
Code Issues Actions Packages Projects Releases Wiki Activity

Add YUV444P12 support

Browse Source
This commit is contained in:
Steven Robertson 2015-10-10 16:00:27 -07:00
parent 0ce1b51d16
commit f93b4dbf23
2 changed files with 38 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

32
cuburn/code/output.py
View File

@ -188,5 +188,37 @@ __global__ void f32_to_yuv420p10(
rctxs[rb_incr(rb->tail, tid)] = rctx; rctxs[rb_incr(rb->tail, tid)] = rctx;
} }
// Convert from rgb444 to planar YUV 10-bit, using JPEG full-range primaries.
// TODO(strobe): Share more code.
__global__ void f32_to_yuv444p12(
uint16_t *dst, const float4 *src,
int gutter, int dstride, int sstride, int height,
ringbuf *rb, mwc_st *rctxs)
{
int x = blockIdx.x * blockDim.x + threadIdx.x;
int y = blockIdx.y * blockDim.y + threadIdx.y;
if (x > dstride || y > height) return;
int isrc = sstride * (y + gutter) + x + gutter;
int tid = blockDim.x * threadIdx.y + threadIdx.x;
mwc_st rctx = rctxs[rb_incr(rb->head, tid)];
float4 in = src[isrc];
ushort3 out = make_ushort3(
dclampf(rctx, 4095.0f, 0.299f * in.x + 0.587f * in.y + 0.114f * in.z),
dclampf(rctx, 4095.0f, -0.168736f * in.x - 0.331264f * in.y + 0.5f * in.z + 0.5f),
dclampf(rctx, 4095.0f, 0.5f * in.x - 0.418688f * in.y - 0.081312f * in.z + 0.5f)
);
int idst = dstride * y + x;
dst[idst] = out.x;
idst += dstride * height;
dst[idst] = out.y;
idst += dstride * height;
dst[idst] = out.z;
rctxs[rb_incr(rb->tail, tid)] = rctx;
}
''') ''')

7
cuburn/output.py
View File

@ -260,6 +260,11 @@ class VPxOutput(Output, ClsMod):
self.out_filter = 'f32_to_yuv444p10' self.out_filter = 'f32_to_yuv444p10'
self.args += ['-b', '10', '--input-bit-depth=10', self.args += ['-b', '10', '--input-bit-depth=10',
'--profile=3', '--i444'] '--profile=3', '--i444']
elif pix_fmt == 'yuv444p12':
assert codec == 'vp9'
self.out_filter = 'f32_to_yuv444p12'
self.args += ['-b', '12', '--input-bit-depth=12',
'--profile=3', '--i444']
else: else:
raise ValueError('Invalid pix_fmt: ' + pix_fmt) raise ValueError('Invalid pix_fmt: ' + pix_fmt)
self.args += ['--codec=' + codec, '--cq-level=' + str(crf), '--fps=%d/1' % fps] self.args += ['--codec=' + codec, '--cq-level=' + str(crf), '--fps=%d/1' % fps]
@ -273,7 +278,7 @@ class VPxOutput(Output, ClsMod):
def copy(self, fb, dim, pool, stream=None): def copy(self, fb, dim, pool, stream=None):
fmt = 'u1' fmt = 'u1'
if self.pix_fmt in ('yuv444p10', 'yuv420p10'): if self.pix_fmt in ('yuv444p10', 'yuv420p10', 'yuv444p12'):
fmt = 'u2' fmt = 'u2'
dims = (3, dim.h, dim.w) dims = (3, dim.h, dim.w)
if self.pix_fmt == 'yuv420p10': if self.pix_fmt == 'yuv420p10':