1
mirror of https://code.videolan.org/videolan/vlc synced 2024-09-04 09:11:33 +02:00

r�paration de mes betises sur la yuv (d�sol� tm), mais je remets ca bientot.

This commit is contained in:
Brieuc Jeunhomme 2000-06-14 21:37:36 +00:00
parent c2e97975f7
commit fe175c116a

View File

@ -12,15 +12,16 @@
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111, USA.
* You should have received a copy of the GNU General Public
* License along with this program; if not, write to the
* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
* Boston, MA 02111-1307, USA.
*****************************************************************************/
/*****************************************************************************
@ -72,26 +73,6 @@
#define V_RED_COEF ((int)(1.596 * (1<<SHIFT) / 1.164))
#define V_GREEN_COEF ((int)(-0.813 * (1<<SHIFT) / 1.164))
#ifdef HAVE_MMX
/* hope these constant values are cache line aligned */
static unsigned long long mmx_80w = 0x0080008000800080;
static unsigned long long mmx_10w = 0x1010101010101010;
static unsigned long long mmx_00ffw = 0x00ff00ff00ff00ff;
static unsigned long long mmx_Y_coeff = 0x253f253f253f253f;
/* hope these constant values are cache line aligned */
static unsigned long long mmx_U_green = 0xf37df37df37df37d;
static unsigned long long mmx_U_blue = 0x4093409340934093;
static unsigned long long mmx_V_red = 0x3312331233123312;
static unsigned long long mmx_V_green = 0xe5fce5fce5fce5fc;
/* hope these constant values are cache line aligned */
static unsigned long long mmx_redmask = 0xf8f8f8f8f8f8f8f8;
static unsigned long long mmx_grnmask = 0xfcfcfcfcfcfcfcfc;
static unsigned long long mmx_grnshift = 0x03;
static unsigned long long mmx_blueshift = 0x03;
#endif
/*****************************************************************************
* Local prototypes
*****************************************************************************/
@ -200,7 +181,7 @@ static void ConvertYUV444RGB32( p_vout_thread_t p_vout, u32 *p_pic, yuv_data
(((*p_y + dither10[i_real_y]) >> 4) << 7) \
+ ((*p_u + dither20[i_real_y]) >> 5) * 9 \
+ ((*p_v + dither20[i_real_y]) >> 5) ]; \
b_jump_uv += *p_offset; \
b_jump_uv = (b_jump_uv + *p_offset) & 0x1; \
p_y += *p_offset; \
p_u += *p_offset & b_jump_uv; \
p_v += *p_offset++ & b_jump_uv; \
@ -208,15 +189,15 @@ static void ConvertYUV444RGB32( p_vout_thread_t p_vout, u32 *p_pic, yuv_data
(((*p_y + dither11[i_real_y]) >> 4) << 7) \
+ ((*p_u + dither21[i_real_y]) >> 5) * 9 \
+ ((*p_v + dither21[i_real_y]) >> 5) ]; \
b_jump_uv += *p_offset; \
b_jump_uv = (b_jump_uv + *p_offset) & 0x1; \
p_y += *p_offset; \
p_u += *p_offset & b_jump_uv; \
p_v += *p_offset++ & b_jump_uv; \
*p_pic++ = p_lookup[ \
(((*p_y + dither12[i_real_y]) >> 4) << 7) \
+ ((*p_u + dither22[i_real_y]) >> 5) * 9 \
+ ((*p_v + dither22[i_real_y]) >> 5) ]; \
b_jump_uv += *p_offset; \
+ ((*p_v + dither22[i_real_y]) >> 5) ]; \
b_jump_uv = (b_jump_uv + *p_offset) & 0x1; \
p_y += *p_offset; \
p_u += *p_offset & b_jump_uv; \
p_v += *p_offset++ & b_jump_uv; \
@ -224,7 +205,7 @@ static void ConvertYUV444RGB32( p_vout_thread_t p_vout, u32 *p_pic, yuv_data
(((*p_y + dither13[i_real_y]) >> 4) << 7) \
+ ((*p_u + dither23[i_real_y]) >> 5) * 9 \
+ ((*p_v + dither23[i_real_y]) >> 5) ]; \
b_jump_uv += *p_offset; \
b_jump_uv = (b_jump_uv + *p_offset) & 0x1; \
p_y += *p_offset; \
p_u += *p_offset & b_jump_uv; \
p_v += *p_offset++ & b_jump_uv; \
@ -312,13 +293,12 @@ static void ConvertYUV444RGB32( p_vout_thread_t p_vout, u32 *p_pic, yuv_data
* and 4 Bpp.
*****************************************************************************/
#define SCALE_HEIGHT( CHROMA, BPP ) \
\
/* If line is odd, rewind 4:2:0 U and V samples */ \
/*if( ((CHROMA == 420) || (CHROMA == 422)) && !(i_y & 0x1) ) */\
/*{ */\
/* p_u -= i_chroma_width; */\
/* p_v -= i_chroma_width; */\
/*} */\
if( ((CHROMA == 420) || (CHROMA == 422)) && !(i_y & 0x1) ) \
{ \
p_u -= i_chroma_width; \
p_v -= i_chroma_width; \
} \
\
/* \
* Handle vertical scaling. The current line can be copied or next one \
@ -327,70 +307,24 @@ static void ConvertYUV444RGB32( p_vout_thread_t p_vout, u32 *p_pic, yuv_data
switch( i_vertical_scaling ) \
{ \
case -1: /* vertical scaling factor is < 1 */ \
if( i_y & 0x1 ) \
while( (i_scale_count -= i_pic_height) >= 0 ) \
{ \
while( (i_scale_count -= i_pic_height) >= 0 ) \
{ \
/* Height reduction: skip next source line */ \
p_y += i_width; \
if( (CHROMA == 420) || (CHROMA == 422) ) \
{ \
if( (i_scale_count -= i_pic_height) >= 0 ) \
{ \
p_y += i_width; \
i_y += 2; \
p_u += i_chroma_width; \
p_v += i_chroma_width; \
continue; \
} \
else \
{ \
i_y++; \
break; \
} \
} \
else if( CHROMA == 444 ) \
{ \
i_y++; \
p_u += i_width; \
p_v += i_width; \
} \
} \
} \
else \
{ \
if( CHROMA == 420 || CHROMA == 422 ) \
p_y += i_width; \
i_y++; \
if( (CHROMA == 420) || (CHROMA == 422) ) \
{ \
p_u -= i_chroma_width; \
p_v -= i_chroma_width; \
} \
while( (i_scale_count -= i_pic_height) >= 0 ) \
{ \
/* Height reduction: skip next source line */ \
p_y += i_width; \
if( (CHROMA == 420) || (CHROMA == 422) ) \
if( i_y & 0x1 ) \
{ \
p_u += i_chroma_width; \
p_v += i_chroma_width; \
if( (i_scale_count -= i_pic_height) >= 0 ) \
{ \
p_y += i_width; \
i_y+=2; \
continue; \
} \
else \
{ \
i_y++; \
break; \
} \
} \
else if( CHROMA == 444 ) \
{ \
i_y++; \
p_u += i_width; \
p_v += i_width; \
} \
} \
else if( CHROMA == 444 ) \
{ \
p_u += i_width; \
p_v += i_width; \
} \
} \
i_scale_count += i_height; \
break; \
@ -398,7 +332,7 @@ static void ConvertYUV444RGB32( p_vout_thread_t p_vout, u32 *p_pic, yuv_data
while( (i_scale_count -= i_height) > 0 ) \
{ \
/* Height increment: copy previous picture line */ \
for( i_x = i_pic_width >> 4; i_x--; ) \
for( i_x = i_pic_width / 16; i_x--; ) \
{ \
*(((u64 *) p_pic)++) = *(((u64 *) p_pic_start)++ ); \
*(((u64 *) p_pic)++) = *(((u64 *) p_pic_start)++ ); \
@ -444,6 +378,7 @@ static void ConvertYUV444RGB32( p_vout_thread_t p_vout, u32 *p_pic, yuv_data
* Handle vertical scaling. The current line can be copied or next one \
* can be ignored. \
*/ \
\
switch( i_vertical_scaling ) \
{ \
case -1: /* vertical scaling factor is < 1 */ \
@ -1168,16 +1103,15 @@ static void ConvertYUV420RGB8( p_vout_thread_t p_vout, u8 *p_pic, yuv_data_t *p_
int dither22[4] = { 0x6, 0x16, 0x2, 0x12 };
int dither23[4] = { 0x1e, 0xe, 0x1a, 0xa };
#if 0
/* other matrices that can be interesting, either for debugging or for
* various effects */
/* other matrices that can be interesting, either for debugging or for effects */
#if 0
int dither[4][4] = { { 0, 8, 2, 10 }, { 12, 4, 14, 16 }, { 3, 11, 1, 9}, {15, 7, 13, 5} };
int dither[4][4] = { { 7, 8, 0, 15 }, { 0, 15, 8, 7 }, { 7, 0, 15, 8 }, { 15, 7, 8, 0 } };
int dither[4][4] = { { 0, 15, 0, 15 }, { 15, 0, 15, 0 }, { 0, 15, 0, 15 }, { 15, 0, 15, 0 } };
int dither[4][4] = { { 15, 15, 0, 0 }, { 15, 15, 0, 0 }, { 0, 0, 15, 15 }, { 0, 0, 15, 15 } };
int dither[4][4] = { { 8, 8, 8, 8 }, { 8, 8, 8, 8 }, { 8, 8, 8, 8 }, { 8, 8, 8, 8 } };
int dither[4][4] = { { 0, 1, 2, 3 }, { 4, 5, 6, 7 }, { 8, 9, 10, 11 }, { 12, 13, 14, 15 } };
#endif
#endif
/*
* Initialize some values - i_pic_line_width will store the line skip
@ -1352,15 +1286,11 @@ static void ConvertYUV420RGB16( p_vout_thread_t p_vout, u16 *p_pic, yuv_data_t *
int i_vertical_scaling; /* vertical scaling type */
int i_x, i_y; /* horizontal and vertical indexes */
int i_scale_count; /* scale modulo counter */
#ifndef HAVE_MMX
int i_uval, i_vval; /* U and V samples */
int i_red, i_green, i_blue; /* U and V modified samples */
#endif
int i_chroma_width; /* chroma width */
u16 * p_yuv; /* base conversion table */
#ifndef HAVE_MMX
u16 * p_ybase; /* Y dependant conversion table */
#endif
u16 * p_pic_start; /* beginning of the current line for copy */
u16 * p_buffer_start; /* conversion buffer start */
u16 * p_buffer; /* conversion buffer pointer */
@ -1389,9 +1319,6 @@ static void ConvertYUV420RGB16( p_vout_thread_t p_vout, u16 *p_pic, yuv_data_t *
p_pic_start = p_pic;
p_buffer = b_horizontal_scaling ? p_buffer_start : p_pic;
#ifndef HAVE_MMX
/* Do YUV conversion to buffer - YUV picture is always formed of 16
* pixels wide blocks */
for( i_x = i_width / 16; i_x--; )
@ -1405,112 +1332,11 @@ static void ConvertYUV420RGB16( p_vout_thread_t p_vout, u16 *p_pic, yuv_data_t *
CONVERT_YUV_PIXEL(2); CONVERT_Y_PIXEL(2);
CONVERT_YUV_PIXEL(2); CONVERT_Y_PIXEL(2);
}
SCALE_WIDTH;
SCALE_HEIGHT(420, 2);
}
#else
for ( i_x = i_width / 8; i_x--; )
{
__asm__ (
"movd (%1), %%mm0 # Load 4 Cb 00 00 00 00 u3 u2 u1 u0\n\t"
"movd (%2), %%mm1 # Load 4 Cr 00 00 00 00 v3 v2 v1 v0\n\t"
"pxor %%mm4, %%mm4 # zero mm4\n\t"
"movq (%0), %%mm6 # Load 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0\n\t"
//"movl $0, (%3) # cache preload for image\n\t"
: : "r" (p_y), "r" (p_u), "r" (p_v), "r" (p_buffer));
__asm__ (
".align 8 \n\t"
/* Do the multiply part of the conversion for even and odd pixels,
* register usage:
* mm0 -> Cblue, mm1 -> Cred, mm2 -> Cgreen even pixels,
* mm3 -> Cblue, mm4 -> Cred, mm5 -> Cgreen odd pixels,
* mm6 -> Y even, mm7 -> Y odd */
/* convert the chroma part */
"punpcklbw %%mm4, %%mm0 # scatter 4 Cb 00 u3 00 u2 00 u1 00 u0\n\t"
"punpcklbw %%mm4, %%mm1 # scatter 4 Cr 00 v3 00 v2 00 v1 00 v0\n\t"
"psubsw mmx_80w, %%mm0 # Cb -= 128\n\t"
"psubsw mmx_80w, %%mm1 # Cr -= 128\n\t"
"psllw $3, %%mm0 # Promote precision\n\t"
"psllw $3, %%mm1 # Promote precision\n\t"
"movq %%mm0, %%mm2 # Copy 4 Cb 00 u3 00 u2 00 u1 00 u0\n\t"
"movq %%mm1, %%mm3 # Copy 4 Cr 00 v3 00 v2 00 v1 00 v0\n\t"
"pmulhw mmx_U_green, %%mm2# Mul Cb with green coeff -> Cb green\n\t"
"pmulhw mmx_V_green, %%mm3# Mul Cr with green coeff -> Cr green\n\t"
"pmulhw mmx_U_blue, %%mm0 # Mul Cb -> Cblue 00 b3 00 b2 00 b1 00 b0\n\t"
"pmulhw mmx_V_red, %%mm1 # Mul Cr -> Cred 00 r3 00 r2 00 r1 00 r0\n\t"
"paddsw %%mm3, %%mm2 # Cb green + Cr green -> Cgreen\n\t"
/* convert the luma part */
"psubusb mmx_10w, %%mm6 # Y -= 16\n\t"
"movq %%mm6, %%mm7 # Copy 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0\n\t"
"pand mmx_00ffw, %%mm6 # get Y even 00 Y6 00 Y4 00 Y2 00 Y0\n\t"
"psrlw $8, %%mm7 # get Y odd 00 Y7 00 Y5 00 Y3 00 Y1\n\t"
"psllw $3, %%mm6 # Promote precision\n\t"
"psllw $3, %%mm7 # Promote precision\n\t"
"pmulhw mmx_Y_coeff, %%mm6# Mul 4 Y even 00 y6 00 y4 00 y2 00 y0\n\t"
"pmulhw mmx_Y_coeff, %%mm7# Mul 4 Y odd 00 y7 00 y5 00 y3 00 y1\n\t"
/* Do the addition part of the conversion for even and odd pixels,
* register usage:
* mm0 -> Cblue, mm1 -> Cred, mm2 -> Cgreen even pixels,
* mm3 -> Cblue, mm4 -> Cred, mm5 -> Cgreen odd pixels,
* mm6 -> Y even, mm7 -> Y odd */ /* Do horizontal and vertical scaling */
"movq %%mm0, %%mm3 # Copy Cblue\n\t"
"movq %%mm1, %%mm4 # Copy Cred\n\t"
"movq %%mm2, %%mm5 # Copy Cgreen\n\t"
"paddsw %%mm6, %%mm0 # Y even + Cblue 00 B6 00 B4 00 B2 00 B0\n\t"
"paddsw %%mm7, %%mm3 # Y odd + Cblue 00 B7 00 B5 00 B3 00 B1\n\t"
"paddsw %%mm6, %%mm1 # Y even + Cred 00 R6 00 R4 00 R2 00 R0\n\t"
"paddsw %%mm7, %%mm4 # Y odd + Cred 00 R7 00 R5 00 R3 00 R1\n\t"
"paddsw %%mm6, %%mm2 # Y even + Cgreen 00 G6 00 G4 00 G2 00 G0\n\t"
"paddsw %%mm7, %%mm5 # Y odd + Cgreen 00 G7 00 G5 00 G3 00 G1\n\t"
/* Limit RGB even to 0..255 */
"packuswb %%mm0, %%mm0 # B6 B4 B2 B0 | B6 B4 B2 B0\n\t"
"packuswb %%mm1, %%mm1 # R6 R4 R2 R0 | R6 R4 R2 R0\n\t"
"packuswb %%mm2, %%mm2 # G6 G4 G2 G0 | G6 G4 G2 G0\n\t"
/* Limit RGB odd to 0..255 */
"packuswb %%mm3, %%mm3 # B7 B5 B3 B1 | B7 B5 B3 B1\n\t"
"packuswb %%mm4, %%mm4 # R7 R5 R3 R1 | R7 R5 R3 R1\n\t"
"packuswb %%mm5, %%mm5 # G7 G5 G3 G1 | G7 G5 G3 G1\n\t"
/* Interleave RGB even and odd */
"punpcklbw %%mm3, %%mm0 # B7 B6 B5 B4 B3 B2 B1 B0\n\t"
"punpcklbw %%mm4, %%mm1 # R7 R6 R5 R4 R3 R2 R1 R0\n\t"
"punpcklbw %%mm5, %%mm2 # G7 G6 G5 G4 G3 G2 G1 G0\n\t"
/* mask unneeded bits off */
"pand mmx_redmask, %%mm0# b7b6b5b4 b3_0_0_0 b7b6b5b4 b3_0_0_0\n\t"
"pand mmx_grnmask, %%mm2# g7g6g5g4 g3g2_0_0 g7g6g5g4 g3g2_0_0\n\t"
"pand mmx_redmask, %%mm1# r7r6r5r4 r3_0_0_0 r7r6r5r4 r3_0_0_0\n\t"
"psrlw mmx_blueshift,%%mm0#0_0_0_b7 b6b5b4b3 0_0_0_b7 b6b5b4b3\n\t"
"pxor %%mm4, %%mm4 # zero mm4\n\t"
"movq %%mm0, %%mm5 # Copy B7-B0\n\t"
"movq %%mm2, %%mm7 # Copy G7-G0\n\t"
/* convert rgb24 plane to rgb16 pack for pixel 0-3 */
"punpcklbw %%mm4, %%mm2 # 0_0_0_0 0_0_0_0 g7g6g5g4 g3g2_0_0\n\t"
"punpcklbw %%mm1, %%mm0 # r7r6r5r4 r3_0_0_0 0_0_0_b7 b6b5b4b3\n\t"
"psllw mmx_blueshift,%%mm2# 0_0_0_0 0_g7g6g5 g4g3g2_0 0_0_0_0\n\t"
"por %%mm2, %%mm0 # r7r6r5r4 r3g7g6g5 g4g3g2b7 b6b5b4b3\n\t"
"movq 8(%0), %%mm6 # Load 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0\n\t"
"movq %%mm0, (%3) # store pixel 0-3\n\t"
/* convert rgb24 plane to rgb16 pack for pixel 0-3 */
"punpckhbw %%mm4, %%mm7 # 0_0_0_0 0_0_0_0 g7g6g5g4 g3g2_0_0\n\t"
"punpckhbw %%mm1, %%mm5 # r7r6r5r4 r3_0_0_0 0_0_0_b7 b6b5b4b3\n\t"
"psllw mmx_blueshift,%%mm7# 0_0_0_0 0_g7g6g5 g4g3g2_0 0_0_0_0\n\t"
"movd 4(%1), %%mm0 # Load 4 Cb 00 00 00 00 u3 u2 u1 u0\n\t"
"por %%mm7, %%mm5 # r7r6r5r4 r3g7g6g5 g4g3g2b7 b6b5b4b3\n\t"
"movd 4(%2), %%mm1 # Load 4 Cr 00 00 00 00 v3 v2 v1 v0\n\t"
"movq %%mm5, 8(%3) # store pixel 4-7\n\t"
: : "r" (p_y), "r" (p_u), "r" (p_v), "r" (p_buffer));
p_y += 8;
p_u += 4;
p_v += 4;
p_buffer += 8;
}
/* Do horizontal and vertical scaling */
SCALE_WIDTH;
SCALE_HEIGHT(420, 2);
}
__asm__ ("emms\n\t");
#endif
}
/*****************************************************************************