yading@10: /*
yading@10:  * simple math operations
yading@10:  * Copyright (c) 2006 Michael Niedermayer <michaelni@gmx.at> et al
yading@10:  *
yading@10:  * This file is part of FFmpeg.
yading@10:  *
yading@10:  * FFmpeg is free software; you can redistribute it and/or
yading@10:  * modify it under the terms of the GNU Lesser General Public
yading@10:  * License as published by the Free Software Foundation; either
yading@10:  * version 2.1 of the License, or (at your option) any later version.
yading@10:  *
yading@10:  * FFmpeg is distributed in the hope that it will be useful,
yading@10:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
yading@10:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
yading@10:  * Lesser General Public License for more details.
yading@10:  *
yading@10:  * You should have received a copy of the GNU Lesser General Public
yading@10:  * License along with FFmpeg; if not, write to the Free Software
yading@10:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
yading@10:  */
yading@10: 
yading@10: #ifndef AVCODEC_X86_MATHOPS_H
yading@10: #define AVCODEC_X86_MATHOPS_H
yading@10: 
yading@10: #include "config.h"
yading@10: #include "libavutil/common.h"
yading@10: 
yading@10: #if HAVE_INLINE_ASM
yading@10: 
yading@10: #if ARCH_X86_32
yading@10: 
yading@10: #define MULL MULL
yading@10: static av_always_inline av_const int MULL(int a, int b, unsigned shift)
yading@10: {
yading@10:     int rt, dummy;
yading@10:     __asm__ (
yading@10:         "imull %3               \n\t"
yading@10:         "shrdl %4, %%edx, %%eax \n\t"
yading@10:         :"=a"(rt), "=d"(dummy)
yading@10:         :"a"(a), "rm"(b), "ci"((uint8_t)shift)
yading@10:     );
yading@10:     return rt;
yading@10: }
yading@10: 
yading@10: #define MULH MULH
yading@10: static av_always_inline av_const int MULH(int a, int b)
yading@10: {
yading@10:     int rt, dummy;
yading@10:     __asm__ (
yading@10:         "imull %3"
yading@10:         :"=d"(rt), "=a"(dummy)
yading@10:         :"a"(a), "rm"(b)
yading@10:     );
yading@10:     return rt;
yading@10: }
yading@10: 
yading@10: #define MUL64 MUL64
yading@10: static av_always_inline av_const int64_t MUL64(int a, int b)
yading@10: {
yading@10:     int64_t rt;
yading@10:     __asm__ (
yading@10:         "imull %2"
yading@10:         :"=A"(rt)
yading@10:         :"a"(a), "rm"(b)
yading@10:     );
yading@10:     return rt;
yading@10: }
yading@10: 
yading@10: #endif /* ARCH_X86_32 */
yading@10: 
yading@10: #if HAVE_CMOV
yading@10: /* median of 3 */
yading@10: #define mid_pred mid_pred
yading@10: static inline av_const int mid_pred(int a, int b, int c)
yading@10: {
yading@10:     int i=b;
yading@10:     __asm__ volatile(
yading@10:         "cmp    %2, %1 \n\t"
yading@10:         "cmovg  %1, %0 \n\t"
yading@10:         "cmovg  %2, %1 \n\t"
yading@10:         "cmp    %3, %1 \n\t"
yading@10:         "cmovl  %3, %1 \n\t"
yading@10:         "cmp    %1, %0 \n\t"
yading@10:         "cmovg  %1, %0 \n\t"
yading@10:         :"+&r"(i), "+&r"(a)
yading@10:         :"r"(b), "r"(c)
yading@10:     );
yading@10:     return i;
yading@10: }
yading@10: #endif
yading@10: 
yading@10: #if HAVE_CMOV
yading@10: #define COPY3_IF_LT(x, y, a, b, c, d)\
yading@10: __asm__ volatile(\
yading@10:     "cmpl  %0, %3       \n\t"\
yading@10:     "cmovl %3, %0       \n\t"\
yading@10:     "cmovl %4, %1       \n\t"\
yading@10:     "cmovl %5, %2       \n\t"\
yading@10:     : "+&r" (x), "+&r" (a), "+r" (c)\
yading@10:     : "r" (y), "r" (b), "r" (d)\
yading@10: );
yading@10: #endif
yading@10: 
yading@10: #define MASK_ABS(mask, level)                   \
yading@10:     __asm__ ("cltd                   \n\t"      \
yading@10:              "xorl %1, %0            \n\t"      \
yading@10:              "subl %1, %0            \n\t"      \
yading@10:              : "+a"(level), "=&d"(mask))
yading@10: 
yading@10: // avoid +32 for shift optimization (gcc should do that ...)
yading@10: #define NEG_SSR32 NEG_SSR32
yading@10: static inline  int32_t NEG_SSR32( int32_t a, int8_t s){
yading@10:     __asm__ ("sarl %1, %0\n\t"
yading@10:          : "+r" (a)
yading@10:          : "ic" ((uint8_t)(-s))
yading@10:     );
yading@10:     return a;
yading@10: }
yading@10: 
yading@10: #define NEG_USR32 NEG_USR32
yading@10: static inline uint32_t NEG_USR32(uint32_t a, int8_t s){
yading@10:     __asm__ ("shrl %1, %0\n\t"
yading@10:          : "+r" (a)
yading@10:          : "ic" ((uint8_t)(-s))
yading@10:     );
yading@10:     return a;
yading@10: }
yading@10: 
yading@10: #endif /* HAVE_INLINE_ASM */
yading@10: #endif /* AVCODEC_X86_MATHOPS_H */