Mercurial > libavcodec.hg
changeset 6499:c9416ce3c09f libavcodec
Move functions that are only used when libswscale is not compiled in below
the appropriate #ifdef. This avoids a bunch of "defined but not used" warnings
when compiling with libswscale enabled.
author | diego |
---|---|
date | Sun, 16 Mar 2008 17:34:31 +0000 |
parents | d9c48a85fd23 |
children | 4e37a1156d72 |
files | imgconvert.c |
diffstat | 1 files changed, 273 insertions(+), 273 deletions(-) [+] |
line wrap: on
line diff
--- a/imgconvert.c Sun Mar 16 13:15:42 2008 +0000 +++ b/imgconvert.c Sun Mar 16 17:34:31 2008 +0000 @@ -1269,140 +1269,6 @@ } } -static uint8_t y_ccir_to_jpeg[256]; -static uint8_t y_jpeg_to_ccir[256]; -static uint8_t c_ccir_to_jpeg[256]; -static uint8_t c_jpeg_to_ccir[256]; - -/* init various conversion tables */ -static void img_convert_init(void) -{ - int i; - uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; - - for(i = 0;i < 256; i++) { - y_ccir_to_jpeg[i] = Y_CCIR_TO_JPEG(i); - y_jpeg_to_ccir[i] = Y_JPEG_TO_CCIR(i); - c_ccir_to_jpeg[i] = C_CCIR_TO_JPEG(i); - c_jpeg_to_ccir[i] = C_JPEG_TO_CCIR(i); - } -} - -/* apply to each pixel the given table */ -static void img_apply_table(uint8_t *dst, int dst_wrap, - const uint8_t *src, int src_wrap, - int width, int height, const uint8_t *table1) -{ - int n; - const uint8_t *s; - uint8_t *d; - const uint8_t *table; - - table = table1; - for(;height > 0; height--) { - s = src; - d = dst; - n = width; - while (n >= 4) { - d[0] = table[s[0]]; - d[1] = table[s[1]]; - d[2] = table[s[2]]; - d[3] = table[s[3]]; - d += 4; - s += 4; - n -= 4; - } - while (n > 0) { - d[0] = table[s[0]]; - d++; - s++; - n--; - } - dst += dst_wrap; - src += src_wrap; - } -} - -/* XXX: use generic filter ? */ -/* XXX: in most cases, the sampling position is incorrect */ - -/* 4x1 -> 1x1 */ -static void shrink41(uint8_t *dst, int dst_wrap, - const uint8_t *src, int src_wrap, - int width, int height) -{ - int w; - const uint8_t *s; - uint8_t *d; - - for(;height > 0; height--) { - s = src; - d = dst; - for(w = width;w > 0; w--) { - d[0] = (s[0] + s[1] + s[2] + s[3] + 2) >> 2; - s += 4; - d++; - } - src += src_wrap; - dst += dst_wrap; - } -} - -/* 2x1 -> 1x1 */ -static void shrink21(uint8_t *dst, int dst_wrap, - const uint8_t *src, int src_wrap, - int width, int height) -{ - int w; - const uint8_t *s; - uint8_t *d; - - for(;height > 0; height--) { - s = src; - d = dst; - for(w = width;w > 0; w--) { - d[0] = (s[0] + s[1]) >> 1; - s += 2; - d++; - } - src += src_wrap; - dst += dst_wrap; - } -} - -/* 1x2 -> 1x1 */ -static void shrink12(uint8_t *dst, int dst_wrap, - const uint8_t *src, int src_wrap, - int width, int height) -{ - int w; - uint8_t *d; - const uint8_t *s1, *s2; - - for(;height > 0; height--) { - s1 = src; - s2 = s1 + src_wrap; - d = dst; - for(w = width;w >= 4; w-=4) { - d[0] = (s1[0] + s2[0]) >> 1; - d[1] = (s1[1] + s2[1]) >> 1; - d[2] = (s1[2] + s2[2]) >> 1; - d[3] = (s1[3] + s2[3]) >> 1; - s1 += 4; - s2 += 4; - d += 4; - } - for(;w > 0; w--) { - d[0] = (s1[0] + s2[0]) >> 1; - s1++; - s2++; - d++; - } - src += 2 * src_wrap; - dst += dst_wrap; - } -} - /* 2x2 -> 1x1 */ void ff_shrink22(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, @@ -1489,145 +1355,6 @@ } } -static void grow21_line(uint8_t *dst, const uint8_t *src, - int width) -{ - int w; - const uint8_t *s1; - uint8_t *d; - - s1 = src; - d = dst; - for(w = width;w >= 4; w-=4) { - d[1] = d[0] = s1[0]; - d[3] = d[2] = s1[1]; - s1 += 2; - d += 4; - } - for(;w >= 2; w -= 2) { - d[1] = d[0] = s1[0]; - s1 ++; - d += 2; - } - /* only needed if width is not a multiple of two */ - /* XXX: veryfy that */ - if (w) { - d[0] = s1[0]; - } -} - -static void grow41_line(uint8_t *dst, const uint8_t *src, - int width) -{ - int w, v; - const uint8_t *s1; - uint8_t *d; - - s1 = src; - d = dst; - for(w = width;w >= 4; w-=4) { - v = s1[0]; - d[0] = v; - d[1] = v; - d[2] = v; - d[3] = v; - s1 ++; - d += 4; - } -} - -/* 1x1 -> 2x1 */ -static void grow21(uint8_t *dst, int dst_wrap, - const uint8_t *src, int src_wrap, - int width, int height) -{ - for(;height > 0; height--) { - grow21_line(dst, src, width); - src += src_wrap; - dst += dst_wrap; - } -} - -/* 1x1 -> 1x2 */ -static void grow12(uint8_t *dst, int dst_wrap, - const uint8_t *src, int src_wrap, - int width, int height) -{ - for(;height > 0; height-=2) { - memcpy(dst, src, width); - dst += dst_wrap; - memcpy(dst, src, width); - dst += dst_wrap; - src += src_wrap; - } -} - -/* 1x1 -> 2x2 */ -static void grow22(uint8_t *dst, int dst_wrap, - const uint8_t *src, int src_wrap, - int width, int height) -{ - for(;height > 0; height--) { - grow21_line(dst, src, width); - if (height%2) - src += src_wrap; - dst += dst_wrap; - } -} - -/* 1x1 -> 4x1 */ -static void grow41(uint8_t *dst, int dst_wrap, - const uint8_t *src, int src_wrap, - int width, int height) -{ - for(;height > 0; height--) { - grow41_line(dst, src, width); - src += src_wrap; - dst += dst_wrap; - } -} - -/* 1x1 -> 4x4 */ -static void grow44(uint8_t *dst, int dst_wrap, - const uint8_t *src, int src_wrap, - int width, int height) -{ - for(;height > 0; height--) { - grow41_line(dst, src, width); - if ((height & 3) == 1) - src += src_wrap; - dst += dst_wrap; - } -} - -/* 1x2 -> 2x1 */ -static void conv411(uint8_t *dst, int dst_wrap, - const uint8_t *src, int src_wrap, - int width, int height) -{ - int w, c; - const uint8_t *s1, *s2; - uint8_t *d; - - width>>=1; - - for(;height > 0; height--) { - s1 = src; - s2 = src + src_wrap; - d = dst; - for(w = width;w > 0; w--) { - c = (s1[0] + s2[0]) >> 1; - d[0] = c; - d[1] = c; - s1++; - s2++; - d += 2; - } - src += src_wrap * 2; - dst += dst_wrap; - } -} - /* XXX: add jpeg quantize code */ #define TRANSP_INDEX (6*6*6) @@ -2381,6 +2108,279 @@ #endif #ifndef CONFIG_SWSCALE +static uint8_t y_ccir_to_jpeg[256]; +static uint8_t y_jpeg_to_ccir[256]; +static uint8_t c_ccir_to_jpeg[256]; +static uint8_t c_jpeg_to_ccir[256]; + +/* init various conversion tables */ +static void img_convert_init(void) +{ + int i; + uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; + + for(i = 0;i < 256; i++) { + y_ccir_to_jpeg[i] = Y_CCIR_TO_JPEG(i); + y_jpeg_to_ccir[i] = Y_JPEG_TO_CCIR(i); + c_ccir_to_jpeg[i] = C_CCIR_TO_JPEG(i); + c_jpeg_to_ccir[i] = C_JPEG_TO_CCIR(i); + } +} + +/* apply to each pixel the given table */ +static void img_apply_table(uint8_t *dst, int dst_wrap, + const uint8_t *src, int src_wrap, + int width, int height, const uint8_t *table1) +{ + int n; + const uint8_t *s; + uint8_t *d; + const uint8_t *table; + + table = table1; + for(;height > 0; height--) { + s = src; + d = dst; + n = width; + while (n >= 4) { + d[0] = table[s[0]]; + d[1] = table[s[1]]; + d[2] = table[s[2]]; + d[3] = table[s[3]]; + d += 4; + s += 4; + n -= 4; + } + while (n > 0) { + d[0] = table[s[0]]; + d++; + s++; + n--; + } + dst += dst_wrap; + src += src_wrap; + } +} + +/* XXX: use generic filter ? */ +/* XXX: in most cases, the sampling position is incorrect */ + +/* 4x1 -> 1x1 */ +static void shrink41(uint8_t *dst, int dst_wrap, + const uint8_t *src, int src_wrap, + int width, int height) +{ + int w; + const uint8_t *s; + uint8_t *d; + + for(;height > 0; height--) { + s = src; + d = dst; + for(w = width;w > 0; w--) { + d[0] = (s[0] + s[1] + s[2] + s[3] + 2) >> 2; + s += 4; + d++; + } + src += src_wrap; + dst += dst_wrap; + } +} + +/* 2x1 -> 1x1 */ +static void shrink21(uint8_t *dst, int dst_wrap, + const uint8_t *src, int src_wrap, + int width, int height) +{ + int w; + const uint8_t *s; + uint8_t *d; + + for(;height > 0; height--) { + s = src; + d = dst; + for(w = width;w > 0; w--) { + d[0] = (s[0] + s[1]) >> 1; + s += 2; + d++; + } + src += src_wrap; + dst += dst_wrap; + } +} + +/* 1x2 -> 1x1 */ +static void shrink12(uint8_t *dst, int dst_wrap, + const uint8_t *src, int src_wrap, + int width, int height) +{ + int w; + uint8_t *d; + const uint8_t *s1, *s2; + + for(;height > 0; height--) { + s1 = src; + s2 = s1 + src_wrap; + d = dst; + for(w = width;w >= 4; w-=4) { + d[0] = (s1[0] + s2[0]) >> 1; + d[1] = (s1[1] + s2[1]) >> 1; + d[2] = (s1[2] + s2[2]) >> 1; + d[3] = (s1[3] + s2[3]) >> 1; + s1 += 4; + s2 += 4; + d += 4; + } + for(;w > 0; w--) { + d[0] = (s1[0] + s2[0]) >> 1; + s1++; + s2++; + d++; + } + src += 2 * src_wrap; + dst += dst_wrap; + } +} + +static void grow21_line(uint8_t *dst, const uint8_t *src, + int width) +{ + int w; + const uint8_t *s1; + uint8_t *d; + + s1 = src; + d = dst; + for(w = width;w >= 4; w-=4) { + d[1] = d[0] = s1[0]; + d[3] = d[2] = s1[1]; + s1 += 2; + d += 4; + } + for(;w >= 2; w -= 2) { + d[1] = d[0] = s1[0]; + s1 ++; + d += 2; + } + /* only needed if width is not a multiple of two */ + /* XXX: veryfy that */ + if (w) { + d[0] = s1[0]; + } +} + +static void grow41_line(uint8_t *dst, const uint8_t *src, + int width) +{ + int w, v; + const uint8_t *s1; + uint8_t *d; + + s1 = src; + d = dst; + for(w = width;w >= 4; w-=4) { + v = s1[0]; + d[0] = v; + d[1] = v; + d[2] = v; + d[3] = v; + s1 ++; + d += 4; + } +} + +/* 1x1 -> 2x1 */ +static void grow21(uint8_t *dst, int dst_wrap, + const uint8_t *src, int src_wrap, + int width, int height) +{ + for(;height > 0; height--) { + grow21_line(dst, src, width); + src += src_wrap; + dst += dst_wrap; + } +} + +/* 1x1 -> 1x2 */ +static void grow12(uint8_t *dst, int dst_wrap, + const uint8_t *src, int src_wrap, + int width, int height) +{ + for(;height > 0; height-=2) { + memcpy(dst, src, width); + dst += dst_wrap; + memcpy(dst, src, width); + dst += dst_wrap; + src += src_wrap; + } +} + +/* 1x1 -> 2x2 */ +static void grow22(uint8_t *dst, int dst_wrap, + const uint8_t *src, int src_wrap, + int width, int height) +{ + for(;height > 0; height--) { + grow21_line(dst, src, width); + if (height%2) + src += src_wrap; + dst += dst_wrap; + } +} + +/* 1x1 -> 4x1 */ +static void grow41(uint8_t *dst, int dst_wrap, + const uint8_t *src, int src_wrap, + int width, int height) +{ + for(;height > 0; height--) { + grow41_line(dst, src, width); + src += src_wrap; + dst += dst_wrap; + } +} + +/* 1x1 -> 4x4 */ +static void grow44(uint8_t *dst, int dst_wrap, + const uint8_t *src, int src_wrap, + int width, int height) +{ + for(;height > 0; height--) { + grow41_line(dst, src, width); + if ((height & 3) == 1) + src += src_wrap; + dst += dst_wrap; + } +} + +/* 1x2 -> 2x1 */ +static void conv411(uint8_t *dst, int dst_wrap, + const uint8_t *src, int src_wrap, + int width, int height) +{ + int w, c; + const uint8_t *s1, *s2; + uint8_t *d; + + width>>=1; + + for(;height > 0; height--) { + s1 = src; + s2 = src + src_wrap; + d = dst; + for(w = width;w > 0; w--) { + c = (s1[0] + s2[0]) >> 1; + d[0] = c; + d[1] = c; + s1++; + s2++; + d += 2; + } + src += src_wrap * 2; + dst += dst_wrap; + } +} + /* XXX: always use linesize. Return -1 if not supported */ int img_convert(AVPicture *dst, int dst_pix_fmt, const AVPicture *src, int src_pix_fmt,