1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/os/ossrv/genericopenlibs/liboil/src/simdpack/clip_fast.c Fri Jun 15 03:10:57 2012 +0200
1.3 @@ -0,0 +1,136 @@
1.4 +/*
1.5 + * LIBOIL - Library of Optimized Inner Loops
1.6 + * Copyright (c) 2003,2004 David A. Schleef <ds@schleef.org>
1.7 + * All rights reserved.
1.8 + *
1.9 + * Redistribution and use in source and binary forms, with or without
1.10 + * modification, are permitted provided that the following conditions
1.11 + * are met:
1.12 + * 1. Redistributions of source code must retain the above copyright
1.13 + * notice, this list of conditions and the following disclaimer.
1.14 + * 2. Redistributions in binary form must reproduce the above copyright
1.15 + * notice, this list of conditions and the following disclaimer in the
1.16 + * documentation and/or other materials provided with the distribution.
1.17 + *
1.18 + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
1.19 + * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
1.20 + * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
1.21 + * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
1.22 + * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
1.23 + * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
1.24 + * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
1.25 + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
1.26 + * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
1.27 + * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
1.28 + * POSSIBILITY OF SUCH DAMAGE.
1.29 + */
1.30 +//Portions Copyright (c) 2008-2009 Nokia Corporation and/or its subsidiary(-ies). All rights reserved.
1.31 +
1.32 +#ifdef HAVE_CONFIG_H
1.33 +#include "config.h"
1.34 +#endif
1.35 +
1.36 +#include <liboil/liboilfunction.h>
1.37 +#include "liboil/simdpack/simdpack.h"
1.38 +
1.39 +/* This is a suprisingly fast implementation of clipping
1.40 + * in straight C. It would be difficult to do it faster in asm
1.41 + * without specialized opcodes. */
1.42 +
1.43 +static void
1.44 +clip_s16_fast (int16_t *dest, int dstr, int16_t *src, int sstr, int n,
1.45 + int16_t *low, int16_t *hi)
1.46 +{
1.47 + int i;
1.48 + int32_t x;
1.49 +
1.50 + for(i=0;i<n;i++){
1.51 + x = OIL_GET(src,i*sstr, int16_t);
1.52 + x = x - (((x-*low)>>31)&(x-*low)) + (((*hi-x)>>31)&(*hi-x));
1.53 + OIL_GET(dest,i*dstr, int16_t) = x;
1.54 + }
1.55 +}
1.56 +
1.57 +OIL_DEFINE_IMPL (clip_s16_fast, clip_s16);
1.58 +
1.59 +static void
1.60 +clip_s16_fast2 (int16_t *dest, int dstr, int16_t *src, int sstr, int n,
1.61 + int16_t *low, int16_t *hi)
1.62 +{
1.63 + int32_t x1, x2;
1.64 +
1.65 + if (n&1) {
1.66 + x1 = *src;
1.67 + OIL_INCREMENT(src,sstr);
1.68 + x1 -= (((x1-*low)>>31)&(x1-*low));
1.69 + *dest = x1 + (((*hi-x1)>>31)&(*hi-x1));
1.70 + OIL_INCREMENT(dest,dstr);
1.71 + n--;
1.72 + }
1.73 + while(n>0){
1.74 + x1 = *src;
1.75 + OIL_INCREMENT(src,sstr);
1.76 + x2 = *src;
1.77 + OIL_INCREMENT(src,sstr);
1.78 + x1 -= (((x1-*low)>>31)&(x1-*low));
1.79 + x2 -= (((x2-*low)>>31)&(x2-*low));
1.80 + *dest = x1 + (((*hi-x1)>>31)&(*hi-x1));
1.81 + OIL_INCREMENT(dest,dstr);
1.82 + *dest = x2 + (((*hi-x2)>>31)&(*hi-x2));
1.83 + OIL_INCREMENT(dest,dstr);
1.84 + n-=2;
1.85 + }
1.86 +}
1.87 +OIL_DEFINE_IMPL (clip_s16_fast2, clip_s16);
1.88 +
1.89 +
1.90 +
1.91 +
1.92 +#if 0
1.93 +/* broken */
1.94 +
1.95 +/* This trick clips
1.96 + * the range min^(1<<31) to max^(1<<31) incorrectly with int32_t.
1.97 + * Thus the use of int64_t. */
1.98 +
1.99 +static void
1.100 +clip_s32_fast (int32_t *dest, int dstr, int32_t *src, int sstr, int n,
1.101 + int32_t *low, int32_t *hi)
1.102 +{
1.103 + int i;
1.104 + int64_t x;
1.105 +
1.106 + for(i=0;i<n;i++){
1.107 + x = OIL_GET(src,i*sstr,int32_t);
1.108 + x = x - (((x-*low)>>63)&(x-*low)) + (((*hi-x)>>63)&(*hi-x));
1.109 + OIL_GET(dest,i*dstr,int32_t) = x;
1.110 + }
1.111 +}
1.112 +
1.113 +OIL_DEFINE_IMPL (clip_s32_fast, clip_s32);
1.114 +#endif
1.115 +
1.116 +
1.117 +
1.118 +
1.119 +#ifdef __SYMBIAN32__
1.120 +
1.121 +OilFunctionImpl* __oil_function_impl_clip_s16_fast() {
1.122 + return &_oil_function_impl_clip_s16_fast;
1.123 +}
1.124 +#endif
1.125 +
1.126 +#ifdef __SYMBIAN32__
1.127 +
1.128 +OilFunctionImpl* __oil_function_impl_clip_s16_fast2() {
1.129 + return &_oil_function_impl_clip_s16_fast2;
1.130 +}
1.131 +#endif
1.132 +#if 0
1.133 +#ifdef __SYMBIAN32__
1.134 +
1.135 +OilFunctionImpl* __oil_function_impl_clip_s32_fast() {
1.136 + return &_oil_function_impl_clip_s32_fast;
1.137 +}
1.138 +#endif
1.139 +#endif