diff options
Diffstat (limited to 'vpx_dsp/vpx_convolve.c')
-rw-r--r-- | vpx_dsp/vpx_convolve.c | 43 |
1 files changed, 25 insertions, 18 deletions
diff --git a/vpx_dsp/vpx_convolve.c b/vpx_dsp/vpx_convolve.c index f17281b36..45097c8f4 100644 --- a/vpx_dsp/vpx_convolve.c +++ b/vpx_dsp/vpx_convolve.c @@ -25,6 +25,7 @@ static void convolve_horiz(const uint8_t *src, ptrdiff_t src_stride, int x_step_q4, int w, int h) { int x, y; src -= SUBPEL_TAPS / 2 - 1; + for (y = 0; y < h; ++y) { int x_q4 = x0_q4; for (x = 0; x < w; ++x) { @@ -46,6 +47,7 @@ static void convolve_avg_horiz(const uint8_t *src, ptrdiff_t src_stride, int x_step_q4, int w, int h) { int x, y; src -= SUBPEL_TAPS / 2 - 1; + for (y = 0; y < h; ++y) { int x_q4 = x0_q4; for (x = 0; x < w; ++x) { @@ -72,7 +74,7 @@ static void convolve_vert(const uint8_t *src, ptrdiff_t src_stride, for (x = 0; x < w; ++x) { int y_q4 = y0_q4; for (y = 0; y < h; ++y) { - const unsigned char *src_y = &src[(y_q4 >> SUBPEL_BITS) * src_stride]; + const uint8_t *src_y = &src[(y_q4 >> SUBPEL_BITS) * src_stride]; const int16_t *const y_filter = y_filters[y_q4 & SUBPEL_MASK]; int k, sum = 0; for (k = 0; k < SUBPEL_TAPS; ++k) @@ -95,7 +97,7 @@ static void convolve_avg_vert(const uint8_t *src, ptrdiff_t src_stride, for (x = 0; x < w; ++x) { int y_q4 = y0_q4; for (y = 0; y < h; ++y) { - const unsigned char *src_y = &src[(y_q4 >> SUBPEL_BITS) * src_stride]; + const uint8_t *src_y = &src[(y_q4 >> SUBPEL_BITS) * src_stride]; const int16_t *const y_filter = y_filters[y_q4 & SUBPEL_MASK]; int k, sum = 0; for (k = 0; k < SUBPEL_TAPS; ++k) @@ -128,7 +130,7 @@ static void convolve(const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, // --Must round-up because block may be located at sub-pixel position. // --Require an additional SUBPEL_TAPS rows for the 8-tap filter tails. // --((64 - 1) * 32 + 15) >> 4 + 8 = 135. - uint8_t temp[135 * 64]; + uint8_t temp[64 * 135]; int intermediate_height = (((h - 1) * y_step_q4 + y0_q4) >> SUBPEL_BITS) + SUBPEL_TAPS; @@ -219,7 +221,6 @@ void vpx_convolve8_c(const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, int w, int h) { const InterpKernel *const filters_x = get_filter_base(filter_x); const int x0_q4 = get_filter_offset(filter_x, filters_x); - const InterpKernel *const filters_y = get_filter_base(filter_y); const int y0_q4 = get_filter_offset(filter_y, filters_y); @@ -231,7 +232,7 @@ void vpx_convolve8_avg_c(const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h) { - /* Fixed size intermediate buffer places limits on parameters. */ + // Fixed size intermediate buffer places limits on parameters. DECLARE_ALIGNED(16, uint8_t, temp[64 * 64]); assert(w <= 64); assert(h <= 64); @@ -272,7 +273,6 @@ void vpx_convolve_avg_c(const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, for (y = 0; y < h; ++y) { for (x = 0; x < w; ++x) dst[x] = ROUND_POWER_OF_TWO(dst[x] + src[x], 1); - src += src_stride; dst += dst_stride; } @@ -334,9 +334,10 @@ static void highbd_convolve_horiz(const uint8_t *src8, ptrdiff_t src_stride, const InterpKernel *x_filters, int x0_q4, int x_step_q4, int w, int h, int bd) { int x, y; - uint16_t *src = CONVERT_TO_SHORTPTR(src8); + const uint16_t *src = CONVERT_TO_SHORTPTR(src8); uint16_t *dst = CONVERT_TO_SHORTPTR(dst8); src -= SUBPEL_TAPS / 2 - 1; + for (y = 0; y < h; ++y) { int x_q4 = x0_q4; for (x = 0; x < w; ++x) { @@ -357,9 +358,10 @@ static void highbd_convolve_avg_horiz(const uint8_t *src8, ptrdiff_t src_stride, const InterpKernel *x_filters, int x0_q4, int x_step_q4, int w, int h, int bd) { int x, y; - uint16_t *src = CONVERT_TO_SHORTPTR(src8); + const uint16_t *src = CONVERT_TO_SHORTPTR(src8); uint16_t *dst = CONVERT_TO_SHORTPTR(dst8); src -= SUBPEL_TAPS / 2 - 1; + for (y = 0; y < h; ++y) { int x_q4 = x0_q4; for (x = 0; x < w; ++x) { @@ -382,9 +384,10 @@ static void highbd_convolve_vert(const uint8_t *src8, ptrdiff_t src_stride, const InterpKernel *y_filters, int y0_q4, int y_step_q4, int w, int h, int bd) { int x, y; - uint16_t *src = CONVERT_TO_SHORTPTR(src8); + const uint16_t *src = CONVERT_TO_SHORTPTR(src8); uint16_t *dst = CONVERT_TO_SHORTPTR(dst8); src -= src_stride * (SUBPEL_TAPS / 2 - 1); + for (x = 0; x < w; ++x) { int y_q4 = y0_q4; for (y = 0; y < h; ++y) { @@ -407,9 +410,10 @@ static void highbd_convolve_avg_vert(const uint8_t *src8, ptrdiff_t src_stride, const InterpKernel *y_filters, int y0_q4, int y_step_q4, int w, int h, int bd) { int x, y; - uint16_t *src = CONVERT_TO_SHORTPTR(src8); + const uint16_t *src = CONVERT_TO_SHORTPTR(src8); uint16_t *dst = CONVERT_TO_SHORTPTR(dst8); src -= src_stride * (SUBPEL_TAPS / 2 - 1); + for (x = 0; x < w; ++x) { int y_q4 = y0_q4; for (y = 0; y < h; ++y) { @@ -470,6 +474,7 @@ void vpx_highbd_convolve8_horiz_c(const uint8_t *src, ptrdiff_t src_stride, int h, int bd) { const InterpKernel *const filters_x = get_filter_base(filter_x); const int x0_q4 = get_filter_offset(filter_x, filters_x); + (void)filter_y; (void)y_step_q4; @@ -484,6 +489,7 @@ void vpx_highbd_convolve8_avg_horiz_c(const uint8_t *src, ptrdiff_t src_stride, int w, int h, int bd) { const InterpKernel *const filters_x = get_filter_base(filter_x); const int x0_q4 = get_filter_offset(filter_x, filters_x); + (void)filter_y; (void)y_step_q4; @@ -498,6 +504,7 @@ void vpx_highbd_convolve8_vert_c(const uint8_t *src, ptrdiff_t src_stride, int h, int bd) { const InterpKernel *const filters_y = get_filter_base(filter_y); const int y0_q4 = get_filter_offset(filter_y, filters_y); + (void)filter_x; (void)x_step_q4; @@ -512,6 +519,7 @@ void vpx_highbd_convolve8_avg_vert_c(const uint8_t *src, ptrdiff_t src_stride, int w, int h, int bd) { const InterpKernel *const filters_y = get_filter_base(filter_y); const int y0_q4 = get_filter_offset(filter_y, filters_y); + (void)filter_x; (void)x_step_q4; @@ -526,7 +534,6 @@ void vpx_highbd_convolve8_c(const uint8_t *src, ptrdiff_t src_stride, int h, int bd) { const InterpKernel *const filters_x = get_filter_base(filter_x); const int x0_q4 = get_filter_offset(filter_x, filters_x); - const InterpKernel *const filters_y = get_filter_base(filter_y); const int y0_q4 = get_filter_offset(filter_y, filters_y); @@ -556,11 +563,12 @@ void vpx_highbd_convolve_copy_c(const uint8_t *src8, ptrdiff_t src_stride, const int16_t *filter_y, int filter_y_stride, int w, int h, int bd) { int r; - uint16_t *src = CONVERT_TO_SHORTPTR(src8); + const uint16_t *src = CONVERT_TO_SHORTPTR(src8); uint16_t *dst = CONVERT_TO_SHORTPTR(dst8); + (void)filter_x; - (void)filter_y; (void)filter_x_stride; + (void)filter_y; (void)filter_y_stride; (void)bd; @@ -577,18 +585,17 @@ void vpx_highbd_convolve_avg_c(const uint8_t *src8, ptrdiff_t src_stride, const int16_t *filter_y, int filter_y_stride, int w, int h, int bd) { int x, y; - uint16_t *src = CONVERT_TO_SHORTPTR(src8); + const uint16_t *src = CONVERT_TO_SHORTPTR(src8); uint16_t *dst = CONVERT_TO_SHORTPTR(dst8); + (void)filter_x; - (void)filter_y; (void)filter_x_stride; + (void)filter_y; (void)filter_y_stride; (void)bd; for (y = 0; y < h; ++y) { - for (x = 0; x < w; ++x) { - dst[x] = ROUND_POWER_OF_TWO(dst[x] + src[x], 1); - } + for (x = 0; x < w; ++x) dst[x] = ROUND_POWER_OF_TWO(dst[x] + src[x], 1); src += src_stride; dst += dst_stride; } |