sl@0: /*
sl@0: * Copyright (c) 2009 Nokia Corporation and/or its subsidiary(-ies).
sl@0: * All rights reserved.
sl@0: * This component and the accompanying materials are made available
sl@0: * under the terms of "Eclipse Public License v1.0"
sl@0: * which accompanies this distribution, and is available
sl@0: * at the URL "http://www.eclipse.org/legal/epl-v10.html".
sl@0: *
sl@0: * Initial Contributors:
sl@0: * Nokia Corporation - initial contribution.
sl@0: *
sl@0: * Contributors:
sl@0: *
sl@0: * Description: 
sl@0: *
sl@0: */
sl@0: 
sl@0: #ifdef HAVE_CONFIG_H
sl@0: #include "config.h"
sl@0: #endif
sl@0: #include <liboil/liboilclasses.h>
sl@0: #include <liboil/liboilfunction.h>
sl@0: #include <emmintrin.h>
sl@0: 
sl@0: #define SSE_FUNCTION __attribute__((force_align_arg_pointer))
sl@0: 
sl@0: #define MULTSUM_SSE2_NSTRIDED(i) { \
sl@0:   t1 = _mm_load_pd(&OIL_GET(src1, i, double)); \
sl@0:   t2 = _mm_load_pd(&OIL_GET(src2, i, double)); \
sl@0:   t1 = _mm_mul_pd(t1,t2); \
sl@0:   sum.reg = _mm_add_pd(sum.reg,t1); \
sl@0: }
sl@0: #define MULTSUM_SSE2_NSTRIDEDP(i) { \
sl@0:   t1 = _mm_load_pd(&OIL_GET(src1, i*sstr1, double)); \
sl@0:   t2 = _mm_loadl_pd(t2, &OIL_GET(src2, i*sstr2, double)); \
sl@0:   t2 = _mm_loadh_pd(t2, &OIL_GET(src2, (i+1)*sstr2, double)); \
sl@0:   t1 = _mm_mul_pd(t1,t2); \
sl@0:   sum.reg = _mm_add_pd(sum.reg,t1); \
sl@0: }
sl@0: #define MULTSUM_SSE2_STRIDED(i) { \
sl@0:   t1 = _mm_loadl_pd(t1, &OIL_GET(src1, i*sstr1, double)); \
sl@0:   t1 = _mm_loadh_pd(t1, &OIL_GET(src1, (i+1)*sstr1, double)); \
sl@0:   t2 = _mm_loadl_pd(t2, &OIL_GET(src2, i*sstr2, double)); \
sl@0:   t2 = _mm_loadh_pd(t2, &OIL_GET(src2, (i+1)*sstr2, double)); \
sl@0:   t1 = _mm_mul_pd(t1,t2); \
sl@0:   sum.reg = _mm_add_pd(sum.reg,t1); \
sl@0: }
sl@0: 
sl@0: 
sl@0: #ifdef ENABLE_BROKEN_IMPLS
sl@0: SSE_FUNCTION static void
sl@0: multsum_f64_sse2_unroll4(double *dest,
sl@0:      const double *src1, int sstr1,
sl@0:      const double *src2, int sstr2,
sl@0:      int n)
sl@0: {
sl@0:   __m128d t1, t2;
sl@0:   union {
sl@0:     __m128d reg;
sl@0:     double vals[2];
sl@0:   } sum;
sl@0:   int i = 0;
sl@0: 
sl@0:   sum.reg = _mm_setzero_pd();
sl@0:   while (i < n-3) {
sl@0:     MULTSUM_SSE2_STRIDED(0);
sl@0:     MULTSUM_SSE2_STRIDED(2);
sl@0: 
sl@0:     OIL_INCREMENT(src1, 4*sstr1);
sl@0:     OIL_INCREMENT(src2, 4*sstr2);
sl@0:     i += 4;
sl@0:   }
sl@0:   while (i < n-1) {
sl@0:     MULTSUM_SSE2_STRIDED(0);
sl@0: 
sl@0:     OIL_INCREMENT(src1, 2*sstr1);
sl@0:     OIL_INCREMENT(src2, 2*sstr2);
sl@0:     i+=2;
sl@0:   }
sl@0:   *dest = sum.vals[0] + sum.vals[1];
sl@0:   if (i < n) {
sl@0:     *dest += (OIL_GET(src1,0,double)*OIL_GET(src2,0,double));
sl@0:   }
sl@0: }
sl@0: OIL_DEFINE_IMPL_FULL (multsum_f64_sse2_unroll4, multsum_f64, OIL_IMPL_FLAG_SSE2);
sl@0: #endif
sl@0: 
sl@0: 
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: OilFunctionImpl* __oil_function_impl_multsum_f64_sse2_unroll4, multsum_f64() {
sl@0: 		return &_oil_function_impl_multsum_f64_sse2_unroll4, multsum_f64;
sl@0: }
sl@0: #endif
sl@0: