sl@0: /*
sl@0:  * LIBOIL - Library of Optimized Inner Loops
sl@0:  * Copyright (c) 2003,2004 David A. Schleef <ds@schleef.org>
sl@0:  * All rights reserved.
sl@0:  *
sl@0:  * Redistribution and use in source and binary forms, with or without
sl@0:  * modification, are permitted provided that the following conditions
sl@0:  * are met:
sl@0:  * 1. Redistributions of source code must retain the above copyright
sl@0:  *    notice, this list of conditions and the following disclaimer.
sl@0:  * 2. Redistributions in binary form must reproduce the above copyright
sl@0:  *    notice, this list of conditions and the following disclaimer in the
sl@0:  *    documentation and/or other materials provided with the distribution.
sl@0:  * 
sl@0:  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
sl@0:  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
sl@0:  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
sl@0:  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
sl@0:  * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
sl@0:  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
sl@0:  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
sl@0:  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
sl@0:  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
sl@0:  * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
sl@0:  * POSSIBILITY OF SUCH DAMAGE.
sl@0:  */
sl@0: //Portions Copyright (c)  2008-2009 Nokia Corporation and/or its subsidiary(-ies). All rights reserved. 
sl@0: 
sl@0: #ifdef HAVE_CONFIG_H
sl@0: #include "config.h"
sl@0: #endif
sl@0: 
sl@0: #include <liboil/liboilfunction.h>
sl@0: 
sl@0: 
sl@0: /**
sl@0:  * oil_err_intra8x8_u8:
sl@0:  * @d_1:
sl@0:  * @s1_8x8:
sl@0:  * @ss1:
sl@0:  *
sl@0:  * Calculates the sum of squared differences from the mean over
sl@0:  * @s1_8x8 and places the result in @d_1.  This result is 64 times
sl@0:  * the variance of the mean of @s1_8x8.
sl@0:  */
sl@0: OIL_DEFINE_CLASS (err_intra8x8_u8,
sl@0:     "uint32_t *d_1, uint8_t *s1_8x8, int ss1");
sl@0: /**
sl@0:  * oil_err_inter8x8_u8:
sl@0:  * @d_1:
sl@0:  * @s1_8x8:
sl@0:  * @ss1:
sl@0:  * @s2_8x8:
sl@0:  * @ss2:
sl@0:  *
sl@0:  * Calculates an intermediate 8x8 block where each element is the
sl@0:  * difference between @s1_8x8 and @s2_8x8.
sl@0:  * The sum of squares of the difference of each element in the 
sl@0:  * intermediate block and the mean of the intermediate block is
sl@0:  * placed into @d_1.  This result is
sl@0:  * equal to 64 times the variance of the mean of the intermediate block.
sl@0:  */
sl@0: OIL_DEFINE_CLASS (err_inter8x8_u8,
sl@0:     "uint32_t *d_1, uint8_t *s1_8x8, int ss1, uint8_t *s2_8x8, int ss2");
sl@0: /**
sl@0:  * oil_err_inter8x8_u8_avg:
sl@0:  * @d_1:
sl@0:  * @s1_8x8:
sl@0:  * @ss1:
sl@0:  * @s2_8x8:
sl@0:  * @s3_8x8:
sl@0:  * @ss2:
sl@0:  *
sl@0:  * Calculates an intermediate 8x8 block where each element is the
sl@0:  * difference between @s1_8x8 and the average of @s2_8x8 and @s3_8x8.
sl@0:  * The sum of squares of the difference of each element in the 
sl@0:  * intermediate block and the mean of the intermediate block is
sl@0:  * placed into @d_1.
sl@0:  * This result is 64 times the variance of the mean of the intermediate
sl@0:  * block.
sl@0:  *
sl@0:  * FIXME: This function is broken, since the reference function uses
sl@0:  * @ss2 as the stride for both @s2_8x8 and @s3_8x8.
sl@0:  */
sl@0: OIL_DEFINE_CLASS (err_inter8x8_u8_avg,
sl@0:     "uint32_t *d_1, uint8_t *s1_8x8, int ss1, uint8_t *s2_8x8, uint8_t *s3_8x8, int ss2");
sl@0: 
sl@0: #define DSP_OP_AVG(a,b) ((((int)(a)) + ((int)(b)))/2)
sl@0: #define DSP_OP_DIFF(a,b) (((int)(a)) - ((int)(b)))
sl@0: 
sl@0: static void
sl@0: err_intra8x8_u8_ref (uint32_t *dest, uint8_t *src1, int ss1)
sl@0: {
sl@0:   uint32_t  i;
sl@0:   uint32_t  xsum=0;
sl@0:   uint32_t  xxsum=0;
sl@0: 
sl@0:   for (i=8; i; i--) {
sl@0:     /* Examine alternate pixel locations. */
sl@0:     xsum += src1[0];
sl@0:     xxsum += src1[0]*src1[0];
sl@0:     xsum += src1[1];
sl@0:     xxsum += src1[1]*src1[1];
sl@0:     xsum += src1[2];
sl@0:     xxsum += src1[2]*src1[2];
sl@0:     xsum += src1[3];
sl@0:     xxsum += src1[3]*src1[3];
sl@0:     xsum += src1[4];
sl@0:     xxsum += src1[4]*src1[4];
sl@0:     xsum += src1[5];
sl@0:     xxsum += src1[5]*src1[5];
sl@0:     xsum += src1[6];
sl@0:     xxsum += src1[6]*src1[6];
sl@0:     xsum += src1[7];
sl@0:     xxsum += src1[7]*src1[7];
sl@0: 
sl@0:     /* Step to next row of block. */
sl@0:     src1 += ss1;
sl@0:   }
sl@0:   /* Compute population variance as mis-match metric. */
sl@0:   *dest = (((xxsum<<6) - xsum*xsum ));
sl@0: }
sl@0: OIL_DEFINE_IMPL_REF (err_intra8x8_u8_ref, err_intra8x8_u8);
sl@0: 
sl@0: static void
sl@0: err_inter8x8_u8_ref (uint32_t *dest, uint8_t *src1, int ss1, uint8_t *src2, int ss2)
sl@0: {
sl@0:   uint32_t  i;
sl@0:   uint32_t  xsum=0;
sl@0:   uint32_t  xxsum=0;
sl@0:   int32_t   diff;
sl@0: 
sl@0:   for (i=8; i; i--) {
sl@0:     diff = DSP_OP_DIFF (src1[0], src2[0]);
sl@0:     xsum += diff;
sl@0:     xxsum += diff*diff;
sl@0: 
sl@0:     diff = DSP_OP_DIFF (src1[1], src2[1]);
sl@0:     xsum += diff;
sl@0:     xxsum += diff*diff;
sl@0: 
sl@0:     diff = DSP_OP_DIFF (src1[2], src2[2]);
sl@0:     xsum += diff;
sl@0:     xxsum += diff*diff;
sl@0: 
sl@0:     diff = DSP_OP_DIFF (src1[3], src2[3]);
sl@0:     xsum += diff;
sl@0:     xxsum += diff*diff;
sl@0:         
sl@0:     diff = DSP_OP_DIFF (src1[4], src2[4]);
sl@0:     xsum += diff;
sl@0:     xxsum += diff*diff;
sl@0:         
sl@0:     diff = DSP_OP_DIFF (src1[5], src2[5]);
sl@0:     xsum += diff;
sl@0:     xxsum += diff*diff;
sl@0:         
sl@0:     diff = DSP_OP_DIFF (src1[6], src2[6]);
sl@0:     xsum += diff;
sl@0:     xxsum += diff*diff;
sl@0:         
sl@0:     diff = DSP_OP_DIFF (src1[7], src2[7]);
sl@0:     xsum += diff;
sl@0:     xxsum += diff*diff;
sl@0:         
sl@0:     /* Step to next row of block. */
sl@0:     src1 += ss1;
sl@0:     src2 += ss2;
sl@0:   }
sl@0: 
sl@0:   /* Compute and return population variance as mis-match metric. */
sl@0:   *dest = (((xxsum<<6) - xsum*xsum));
sl@0: }
sl@0: OIL_DEFINE_IMPL_REF (err_inter8x8_u8_ref, err_inter8x8_u8);
sl@0: 
sl@0: static void
sl@0: err_inter8x8_u8_avg_ref (uint32_t *dest, uint8_t *src1, int ss1, uint8_t *src2, uint8_t *src3, int ss2)
sl@0: {
sl@0:   uint32_t  i;
sl@0:   uint32_t  xsum=0;
sl@0:   uint32_t  xxsum=0;
sl@0:   int32_t   diff;
sl@0: 
sl@0:   for (i=8; i; i--) {
sl@0:     diff = DSP_OP_DIFF(src1[0], DSP_OP_AVG (src2[0], src3[0]));
sl@0:     xsum += diff;
sl@0:     xxsum += diff*diff;
sl@0: 
sl@0:     diff = DSP_OP_DIFF(src1[1], DSP_OP_AVG (src2[1], src3[1]));
sl@0:     xsum += diff;
sl@0:     xxsum += diff*diff;
sl@0: 
sl@0:     diff = DSP_OP_DIFF(src1[2], DSP_OP_AVG (src2[2], src3[2]));
sl@0:     xsum += diff;
sl@0:     xxsum += diff*diff;
sl@0: 
sl@0:     diff = DSP_OP_DIFF(src1[3], DSP_OP_AVG (src2[3], src3[3]));
sl@0:     xsum += diff;
sl@0:     xxsum += diff*diff;
sl@0: 
sl@0:     diff = DSP_OP_DIFF(src1[4], DSP_OP_AVG (src2[4], src3[4]));
sl@0:     xsum += diff;
sl@0:     xxsum += diff*diff;
sl@0: 
sl@0:     diff = DSP_OP_DIFF(src1[5], DSP_OP_AVG (src2[5], src3[5]));
sl@0:     xsum += diff;
sl@0:     xxsum += diff*diff;
sl@0: 
sl@0:     diff = DSP_OP_DIFF(src1[6], DSP_OP_AVG (src2[6], src3[6]));
sl@0:     xsum += diff;
sl@0:     xxsum += diff*diff;
sl@0: 
sl@0:     diff = DSP_OP_DIFF(src1[7], DSP_OP_AVG (src2[7], src3[7]));
sl@0:     xsum += diff;
sl@0:     xxsum += diff*diff;
sl@0: 
sl@0:     /* Step to next row of block. */
sl@0:     src1 += ss1;
sl@0:     src2 += ss2;
sl@0:     src3 += ss2;
sl@0:   }
sl@0: 
sl@0:   /* Compute and return population variance as mis-match metric. */
sl@0:   *dest = (((xxsum<<6) - xsum*xsum));
sl@0: }
sl@0: 
sl@0: OIL_DEFINE_IMPL_REF (err_inter8x8_u8_avg_ref, err_inter8x8_u8_avg);
sl@0: 
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: OilFunctionClass* __oil_function_class_err_intra8x8_u8() {
sl@0: 		return &_oil_function_class_err_intra8x8_u8;
sl@0: }
sl@0: #endif
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: OilFunctionClass* __oil_function_class_err_inter8x8_u8() {
sl@0: 		return &_oil_function_class_err_inter8x8_u8;
sl@0: }
sl@0: #endif
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: OilFunctionClass* __oil_function_class_err_inter8x8_u8_avg() {
sl@0: 		return &_oil_function_class_err_inter8x8_u8_avg;
sl@0: }
sl@0: #endif
sl@0: 
sl@0: 
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: OilFunctionImpl* __oil_function_impl_err_intra8x8_u8_ref() {
sl@0: 		return &_oil_function_impl_err_intra8x8_u8_ref;
sl@0: }
sl@0: #endif
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: OilFunctionImpl* __oil_function_impl_err_inter8x8_u8_ref() {
sl@0: 		return &_oil_function_impl_err_inter8x8_u8_ref;
sl@0: }
sl@0: #endif
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: OilFunctionImpl* __oil_function_impl_err_inter8x8_u8_avg_ref() {
sl@0: 		return &_oil_function_impl_err_inter8x8_u8_avg_ref;
sl@0: }
sl@0: #endif
sl@0: 
sl@0: 
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: EXPORT_C void** _oil_function_class_ptr_err_intra8x8_u8 ()	{
sl@0: 	oil_function_class_ptr_err_intra8x8_u8 = __oil_function_class_err_intra8x8_u8();
sl@0: 	return &oil_function_class_ptr_err_intra8x8_u8->func;
sl@0: 	}
sl@0: #endif
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: EXPORT_C void** _oil_function_class_ptr_err_inter8x8_u8 ()	{
sl@0: 	oil_function_class_ptr_err_inter8x8_u8 = __oil_function_class_err_inter8x8_u8();
sl@0: 	return &oil_function_class_ptr_err_inter8x8_u8->func;
sl@0: 	}
sl@0: #endif
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: EXPORT_C void** _oil_function_class_ptr_err_inter8x8_u8_avg ()	{
sl@0: 	oil_function_class_ptr_err_inter8x8_u8_avg = __oil_function_class_err_inter8x8_u8_avg();
sl@0: 	return &oil_function_class_ptr_err_inter8x8_u8_avg->func;
sl@0: 	}
sl@0: #endif
sl@0: