Showing the changes to libavfilter (2 of 2 files). Show all
Unified · Split
| @@ -56,6 +56,7 @@ static const enum AVPixelFormat supported_formats[] = { · 55 unmodified lines | |||
| 56 | enum { | 56 | enum { |
| 57 | INTERP_ALGO_DEFAULT, | 57 | INTERP_ALGO_DEFAULT, |
| 58 | 58 | ||
| 59 | INTERP_ALGO_NEAREST, | ||
| 59 | INTERP_ALGO_BILINEAR, | 60 | INTERP_ALGO_BILINEAR, |
| 60 | INTERP_ALGO_BICUBIC, | 61 | INTERP_ALGO_BICUBIC, |
| 61 | 62 | ||
| @@ -273,6 +274,12 @@ static av_cold int cudascale_config_props(AVFilterLink *outlink) · 211 unmodified lines | |||
| 273 | extern char vf_scale_cuda_bicubic_ptx[]; | 274 | extern char vf_scale_cuda_bicubic_ptx[]; |
| 274 | 275 | ||
| 275 | switch(s->interp_algo) { | 276 | switch(s->interp_algo) { |
| 277 | case INTERP_ALGO_NEAREST: | ||
| 278 | scaler_ptx = vf_scale_cuda_ptx; | ||
| 279 | function_infix = "_Nearest"; | ||
| 280 | s->interp_use_linear = 0; | ||
| 281 | s->interp_as_integer = 1; | ||
| 282 | break; | ||
| 276 | case INTERP_ALGO_BILINEAR: | 283 | case INTERP_ALGO_BILINEAR: |
| 277 | scaler_ptx = vf_scale_cuda_ptx; | 284 | scaler_ptx = vf_scale_cuda_ptx; |
| 278 | function_infix = "_Bilinear"; | 285 | function_infix = "_Bilinear"; |
| @@ -591,6 +598,7 @@ static const AVOption options[] = { · 312 unmodified lines | |||
| 591 | { "w", "Output video width", OFFSET(w_expr), AV_OPT_TYPE_STRING, { .str = "iw" }, .flags = FLAGS }, | 598 | { "w", "Output video width", OFFSET(w_expr), AV_OPT_TYPE_STRING, { .str = "iw" }, .flags = FLAGS }, |
| 592 | { "h", "Output video height", OFFSET(h_expr), AV_OPT_TYPE_STRING, { .str = "ih" }, .flags = FLAGS }, | 599 | { "h", "Output video height", OFFSET(h_expr), AV_OPT_TYPE_STRING, { .str = "ih" }, .flags = FLAGS }, |
| 593 | { "interp_algo", "Interpolation algorithm used for resizing", OFFSET(interp_algo), AV_OPT_TYPE_INT, { .i64 = INTERP_ALGO_DEFAULT }, 0, INTERP_ALGO_COUNT - 1, FLAGS, "interp_algo" }, | 600 | { "interp_algo", "Interpolation algorithm used for resizing", OFFSET(interp_algo), AV_OPT_TYPE_INT, { .i64 = INTERP_ALGO_DEFAULT }, 0, INTERP_ALGO_COUNT - 1, FLAGS, "interp_algo" }, |
| 601 | { "nearest", "nearest neighbour", 0, AV_OPT_TYPE_CONST, { .i64 = INTERP_ALGO_NEAREST }, 0, 0, FLAGS, "interp_algo" }, | ||
| 594 | { "bilinear", "bilinear", 0, AV_OPT_TYPE_CONST, { .i64 = INTERP_ALGO_BILINEAR }, 0, 0, FLAGS, "interp_algo" }, | 602 | { "bilinear", "bilinear", 0, AV_OPT_TYPE_CONST, { .i64 = INTERP_ALGO_BILINEAR }, 0, 0, FLAGS, "interp_algo" }, |
| 595 | { "bicubic", "bicubic", 0, AV_OPT_TYPE_CONST, { .i64 = INTERP_ALGO_BICUBIC }, 0, 0, FLAGS, "interp_algo" }, | 603 | { "bicubic", "bicubic", 0, AV_OPT_TYPE_CONST, { .i64 = INTERP_ALGO_BICUBIC }, 0, 0, FLAGS, "interp_algo" }, |
| 596 | { "passthrough", "Do not process frames at all if parameters match", OFFSET(passthrough), AV_OPT_TYPE_BOOL, { .i64 = 1 }, 0, 1, FLAGS }, | 604 | { "passthrough", "Do not process frames at all if parameters match", OFFSET(passthrough), AV_OPT_TYPE_BOOL, { .i64 = 1 }, 0, 1, FLAGS }, |
| @@ -22,6 +22,27 @@ · 21 unmodified lines | |||
| 22 | 22 | ||
| 23 | #include "cuda/vector_helpers.cuh" | 23 | #include "cuda/vector_helpers.cuh" |
| 24 | 24 | ||
| 25 | template<typename T> | ||
| 26 | __device__ inline void Subsample_Nearest(cudaTextureObject_t tex, | ||
| 27 | T *dst, | ||
| 28 | int dst_width, int dst_height, int dst_pitch, | ||
| 29 | int src_width, int src_height, | ||
| 30 | int bit_depth) | ||
| 31 | { | ||
| 32 | int xo = blockIdx.x * blockDim.x + threadIdx.x; | ||
| 33 | int yo = blockIdx.y * blockDim.y + threadIdx.y; | ||
| 34 | |||
| 35 | if (yo < dst_height && xo < dst_width) | ||
| 36 | { | ||
| 37 | float hscale = (float)src_width / (float)dst_width; | ||
| 38 | float vscale = (float)src_height / (float)dst_height; | ||
| 39 | float xi = (xo + 0.5f) * hscale; | ||
| 40 | float yi = (yo + 0.5f) * vscale; | ||
| 41 | |||
| 42 | dst[yo*dst_pitch+xo] = tex2D<T>(tex, xi, yi); | ||
| 43 | } | ||
| 44 | } | ||
| 45 | |||
| 25 | template<typename T> | 46 | template<typename T> |
| 26 | __device__ inline void Subsample_Bilinear(cudaTextureObject_t tex, | 47 | __device__ inline void Subsample_Bilinear(cudaTextureObject_t tex, |
| 27 | T *dst, | 48 | T *dst, |
| @@ -57,6 +78,27 @@ __device__ inline void Subsample_Bilinear(cudaTextureObject_t tex, · 29 unmodified lines | |||
| 57 | 78 | ||
| 58 | extern "C" { | 79 | extern "C" { |
| 59 | 80 | ||
| 81 | #define NEAREST_KERNEL(T) \ | ||
| 82 | __global__ void Subsample_Nearest_ ## T(cudaTextureObject_t src_tex, \ | ||
| 83 | T *dst, \ | ||
| 84 | int dst_width, int dst_height, int dst_pitch, \ | ||
| 85 | int src_width, int src_height, \ | ||
| 86 | int bit_depth) \ | ||
| 87 | { \ | ||
| 88 | Subsample_Nearest<T>(src_tex, dst, \ | ||
| 89 | dst_width, dst_height, dst_pitch, \ | ||
| 90 | src_width, src_height, \ | ||
| 91 | bit_depth); \ | ||
| 92 | } | ||
| 93 | |||
| 94 | NEAREST_KERNEL(uchar) | ||
| 95 | NEAREST_KERNEL(uchar2) | ||
| 96 | NEAREST_KERNEL(uchar4) | ||
| 97 | |||
| 98 | NEAREST_KERNEL(ushort) | ||
| 99 | NEAREST_KERNEL(ushort2) | ||
| 100 | NEAREST_KERNEL(ushort4) | ||
| 101 | |||
| 60 | #define BILINEAR_KERNEL(T) \ | 102 | #define BILINEAR_KERNEL(T) \ |
| 61 | __global__ void Subsample_Bilinear_ ## T(cudaTextureObject_t src_tex, \ | 103 | __global__ void Subsample_Bilinear_ ## T(cudaTextureObject_t src_tex, \ |
| 62 | T *dst, \ | 104 | T *dst, \ |