har0ke/FFmpeg
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

avfilter/scale_cuda: simplify linesize calculation

Browse Source
This commit is contained in:
Timo Rothenpieler 2020-11-04 20:00:35 +01:00
parent 94cb78ed94
commit 09e49ba31e
1 changed files with 19 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

36
libavfilter/vf_scale_cuda.c
View File

@ -406,6 +406,8 @@ static int call_resize_kernel(AVFilterContext *ctx, CUfunction func, int channel
&src_width, &src_height, &bit_depth, &s->param }; &src_width, &src_height, &bit_depth, &s->param };
int ret; int ret;
dst_pitch /= channels;
CUDA_TEXTURE_DESC tex_desc = { CUDA_TEXTURE_DESC tex_desc = {
.filterMode = s->interp_use_linear ? .filterMode = s->interp_use_linear ?
CU_TR_FILTER_MODE_LINEAR : CU_TR_FILTER_MODE_LINEAR :
@ -421,7 +423,7 @@ static int call_resize_kernel(AVFilterContext *ctx, CUfunction func, int channel
.res.pitch2D.numChannels = channels, .res.pitch2D.numChannels = channels,
.res.pitch2D.width = src_width, .res.pitch2D.width = src_width,
.res.pitch2D.height = src_height, .res.pitch2D.height = src_height,
.res.pitch2D.pitchInBytes = src_pitch * pixel_size, .res.pitch2D.pitchInBytes = src_pitch,
.res.pitch2D.devPtr = (CUdeviceptr)src_dptr, .res.pitch2D.devPtr = (CUdeviceptr)src_dptr,
}; };
@ -477,16 +479,16 @@ static int scalecuda_resize(AVFilterContext *ctx,
break; break;
case AV_PIX_FMT_YUV444P16: case AV_PIX_FMT_YUV444P16:
call_resize_kernel(ctx, s->cu_func_ushort, 1, call_resize_kernel(ctx, s->cu_func_ushort, 1,
in->data[0], in->width, in->height, in->linesize[0] / 2, in->data[0], in->width, in->height, in->linesize[0],
out->data[0], out->width, out->height, out->linesize[0] / 2, out->data[0], out->width, out->height, out->linesize[0],
2, 16); 2, 16);
call_resize_kernel(ctx, s->cu_func_ushort, 1, call_resize_kernel(ctx, s->cu_func_ushort, 1,
in->data[1], in->width, in->height, in->linesize[1] / 2, in->data[1], in->width, in->height, in->linesize[1],
out->data[1], out->width, out->height, out->linesize[1] / 2, out->data[1], out->width, out->height, out->linesize[1],
2, 16); 2, 16);
call_resize_kernel(ctx, s->cu_func_ushort, 1, call_resize_kernel(ctx, s->cu_func_ushort, 1,
in->data[2], in->width, in->height, in->linesize[2] / 2, in->data[2], in->width, in->height, in->linesize[2],
out->data[2], out->width, out->height, out->linesize[2] / 2, out->data[2], out->width, out->height, out->linesize[2],
2, 16); 2, 16);
break; break;
case AV_PIX_FMT_NV12: case AV_PIX_FMT_NV12:
@ -496,34 +498,34 @@ static int scalecuda_resize(AVFilterContext *ctx,
1, 8); 1, 8);
call_resize_kernel(ctx, s->cu_func_uchar2, 2, call_resize_kernel(ctx, s->cu_func_uchar2, 2,
in->data[1], in->width / 2, in->height / 2, in->linesize[1], in->data[1], in->width / 2, in->height / 2, in->linesize[1],
out->data[1], out->width / 2, out->height / 2, out->linesize[1] / 2, out->data[1], out->width / 2, out->height / 2, out->linesize[1],
1, 8); 1, 8);
break; break;
case AV_PIX_FMT_P010LE: case AV_PIX_FMT_P010LE:
call_resize_kernel(ctx, s->cu_func_ushort, 1, call_resize_kernel(ctx, s->cu_func_ushort, 1,
in->data[0], in->width, in->height, in->linesize[0] / 2, in->data[0], in->width, in->height, in->linesize[0],
out->data[0], out->width, out->height, out->linesize[0] / 2, out->data[0], out->width, out->height, out->linesize[0],
2, 10); 2, 10);
call_resize_kernel(ctx, s->cu_func_ushort2, 2, call_resize_kernel(ctx, s->cu_func_ushort2, 2,
in->data[1], in->width / 2, in->height / 2, in->linesize[1] / 2, in->data[1], in->width / 2, in->height / 2, in->linesize[1],
out->data[1], out->width / 2, out->height / 2, out->linesize[1] / 4, out->data[1], out->width / 2, out->height / 2, out->linesize[1],
2, 10); 2, 10);
break; break;
case AV_PIX_FMT_P016LE: case AV_PIX_FMT_P016LE:
call_resize_kernel(ctx, s->cu_func_ushort, 1, call_resize_kernel(ctx, s->cu_func_ushort, 1,
in->data[0], in->width, in->height, in->linesize[0] / 2, in->data[0], in->width, in->height, in->linesize[0],
out->data[0], out->width, out->height, out->linesize[0] / 2, out->data[0], out->width, out->height, out->linesize[0],
2, 16); 2, 16);
call_resize_kernel(ctx, s->cu_func_ushort2, 2, call_resize_kernel(ctx, s->cu_func_ushort2, 2,
in->data[1], in->width / 2, in->height / 2, in->linesize[1] / 2, in->data[1], in->width / 2, in->height / 2, in->linesize[1],
out->data[1], out->width / 2, out->height / 2, out->linesize[1] / 4, out->data[1], out->width / 2, out->height / 2, out->linesize[1],
2, 16); 2, 16);
break; break;
case AV_PIX_FMT_0RGB32: case AV_PIX_FMT_0RGB32:
case AV_PIX_FMT_0BGR32: case AV_PIX_FMT_0BGR32:
call_resize_kernel(ctx, s->cu_func_uchar4, 4, call_resize_kernel(ctx, s->cu_func_uchar4, 4,
in->data[0], in->width, in->height, in->linesize[0], in->data[0], in->width, in->height, in->linesize[0],
out->data[0], out->width, out->height, out->linesize[0] / 4, out->data[0], out->width, out->height, out->linesize[0],
1, 8); 1, 8);
break; break;
default: default: