generic-library/vpx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Revert "Specialize 4 to 3 scaling in vp9_scale_and_extend_frame_c()"

Browse Source 
This reverts commit afee58f2c4.

This causes ~8x slowdown in 4:3 in the C-code

Change-Id: I60a7ead12dc4ec1548b1b12cfe4b0be42ef04e0e
This commit is contained in:
James Zern
2017-09-14 13:06:40 -07:00
parent 5b558592f5
commit baf658ec4c
1 changed files with 17 additions and 41 deletions
Download Patch File Download Diff File Expand all files Collapse all files

28
vp9/encoder/vp9_frame_scale.c
View File

@@ -20,6 +20,8 @@ void vp9_scale_and_extend_frame_c(const YV12_BUFFER_CONFIG *src,
INTERP_FILTER filter_type, int phase_scaler) { INTERP_FILTER filter_type, int phase_scaler) {
const int src_w = src->y_crop_width; const int src_w = src->y_crop_width;
const int src_h = src->y_crop_height; const int src_h = src->y_crop_height;
const int dst_w = dst->y_crop_width;
const int dst_h = dst->y_crop_height;
const uint8_t *const srcs[3] = { src->y_buffer, src->u_buffer, const uint8_t *const srcs[3] = { src->y_buffer, src->u_buffer,
src->v_buffer }; src->v_buffer };
const int src_strides[3] = { src->y_stride, src->uv_stride, src->uv_stride }; const int src_strides[3] = { src->y_stride, src->uv_stride, src->uv_stride };
@@ -28,31 +30,6 @@ void vp9_scale_and_extend_frame_c(const YV12_BUFFER_CONFIG *src,
const InterpKernel *const kernel = vp9_filter_kernels[filter_type]; const InterpKernel *const kernel = vp9_filter_kernels[filter_type];
int x, y, i; int x, y, i;
if (4 * dst->y_crop_width == 3 * src_w &&
4 * dst->y_crop_height == 3 * src_h) {
// Specialize 4 to 3 scaling.
const int dst_ws[3] = { dst->y_crop_width, dst->uv_crop_width,
dst->uv_crop_width };
const int dst_hs[3] = { dst->y_crop_height, dst->uv_crop_height,
dst->uv_crop_height };
for (i = 0; i < MAX_MB_PLANE; ++i) {
const int dst_w = dst_ws[i];
const int dst_h = dst_hs[i];
const int src_stride = src_strides[i];
const int dst_stride = dst_strides[i];
for (y = 0; y < dst_h; y += 3) {
for (x = 0; x < dst_w; x += 3) {
const uint8_t *src_ptr = srcs[i] + 4 * y / 3 * src_stride + 4 * x / 3;
uint8_t *dst_ptr = dsts[i] + y * dst_stride + x;
vpx_scaled_2d_c(src_ptr, src_stride, dst_ptr, dst_stride, kernel,
phase_scaler, 64 / 3, phase_scaler, 64 / 3, 3, 3);
}
}
}
} else {
const int dst_w = dst->y_crop_width;
const int dst_h = dst->y_crop_height;
for (i = 0; i < MAX_MB_PLANE; ++i) { for (i = 0; i < MAX_MB_PLANE; ++i) {
const int factor = (i == 0 || i == 3 ? 1 : 2); const int factor = (i == 0 || i == 3 ? 1 : 2);
const int src_stride = src_strides[i]; const int src_stride = src_strides[i];
@@ -72,7 +49,6 @@ void vp9_scale_and_extend_frame_c(const YV12_BUFFER_CONFIG *src,
} }
} }
} }
}
vpx_extend_frame_borders(dst); vpx_extend_frame_borders(dst);
} }