1 /*

     2  * Copyright (c) 2005

     3  *	Eric Anholt.  All rights reserved.

     4  *

     5  * Redistribution and use in source and binary forms, with or without

     6  * modification, are permitted provided that the following conditions

     7  * are met:

     8  * 1. Redistributions of source code must retain the above copyright

     9  *    notice, this list of conditions and the following disclaimer.

    10  * 2. Redistributions in binary form must reproduce the above copyright

    11  *    notice, this list of conditions and the following disclaimer in the

    12  *    documentation and/or other materials provided with the distribution.

    13  *

    14  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND

    15  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE

    16  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE

    17  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR BE LIABLE

    18  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL

    19  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS

    20  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)

    21  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT

    22  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY

    23  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF

    24  * SUCH DAMAGE.

    25  */

    26 //Portions Copyright (c)  2008-2009 Nokia Corporation and/or its subsidiary(-ies). All rights reserved. 

    27 

    28 #ifdef HAVE_CONFIG_H

    29 #include "config.h"

    30 #endif

    31 #include <liboil/liboilclasses.h>

    32 #include <liboil/liboilfunction.h>

    33 #include <emmintrin.h>

    34 

    35 #define SSE_FUNCTION __attribute__((force_align_arg_pointer))

    36 

    37 SSE_FUNCTION static void

    38 splat_u32_ns_sse (uint32_t *dest, const uint32_t *param, int n)

    39 {

    40   __m128i v;

    41 

    42   v = _mm_set1_epi32(*param);

    43 

    44   for (; ((long)dest & 15) && (n > 0); n--) {

    45     *dest++ = *param;

    46   }

    47   for (; n >= 4; n -= 4) {

    48     _mm_store_si128((__m128i *)dest, v);

    49     dest += 4;

    50   }

    51   for (; n > 0; n--) {

    52     *dest++ = *param;

    53   }

    54 }

    55 OIL_DEFINE_IMPL_FULL (splat_u32_ns_sse, splat_u32_ns, OIL_IMPL_FLAG_SSE2);

    56 

    57 SSE_FUNCTION static void

    58 splat_u32_ns_sse_unroll2 (uint32_t *dest, const uint32_t *param, int n)

    59 {

    60   __m128i v;

    61 

    62   v = _mm_set1_epi32(*param);

    63 

    64   for (; ((long)dest & 15) && (n > 0); n--) {

    65     *dest++ = *param;

    66   }

    67   for (; n >= 8; n -= 8) {

    68     _mm_store_si128((__m128i *)dest, v);

    69     _mm_store_si128((__m128i *)(dest + 4), v);

    70     dest += 8;

    71   }

    72   for (; n >= 4; n -= 4) {

    73     _mm_store_si128((__m128i *)dest, v);

    74     dest += 4;

    75   }

    76   for (; n > 0; n--) {

    77     *dest++ = *param;

    78   }

    79 }

    80 OIL_DEFINE_IMPL_FULL (splat_u32_ns_sse_unroll2, splat_u32_ns, OIL_IMPL_FLAG_SSE2);

    81 

    82 SSE_FUNCTION static void

    83 splat_u8_ns_sse (uint8_t *dest, const uint8_t *param, int n)

    84 {

    85   __m128i v;

    86 

    87   v = _mm_set1_epi8(*param);

    88 

    89   for (; ((long)dest & 15) && (n > 0); n--) {

    90     *dest++ = *param;

    91   }

    92   for (; n >= 16; n -= 16) {

    93     _mm_store_si128((__m128i *)dest, v);

    94     dest += 16;

    95   }

    96   for (; n > 0; n--) {

    97     *dest++ = *param;

    98   }

    99 }

   100 OIL_DEFINE_IMPL_FULL (splat_u8_ns_sse, splat_u8_ns, OIL_IMPL_FLAG_SSE2);

   101 

   102 SSE_FUNCTION static void

   103 splat_u8_ns_sse_unroll2 (uint8_t *dest, const uint8_t *param, int n)

   104 {

   105   __m128i v;

   106 

   107   v = _mm_set1_epi8(*param);

   108 

   109   for (; ((long)dest & 15) && (n > 0); n--) {

   110     *dest++ = *param;

   111   }

   112   for (; n >= 32; n -= 32) {

   113     _mm_store_si128((__m128i *)dest, v);

   114     _mm_store_si128((__m128i *)(dest + 16), v);

   115     dest += 32;

   116   }

   117   for (; n >= 16; n -= 16) {

   118     _mm_store_si128((__m128i *)dest, v);

   119     dest += 16;

   120   }

   121   for (; n > 0; n--) {

   122     *dest++ = *param;

   123   }

   124 }

   125 OIL_DEFINE_IMPL_FULL (splat_u8_ns_sse_unroll2, splat_u8_ns, OIL_IMPL_FLAG_SSE2);

   126 

   127 

   128 #ifdef	__SYMBIAN32__

   129  

   130 OilFunctionImpl* __oil_function_impl_splat_u32_ns_sse, splat_u32_ns() {

   131 		return &_oil_function_impl_splat_u32_ns_sse, splat_u32_ns;

   132 }

   133 #endif

   134 

   135 #ifdef	__SYMBIAN32__

   136  

   137 OilFunctionImpl* __oil_function_impl_splat_u32_ns_sse_unroll2, splat_u32_ns() {

   138 		return &_oil_function_impl_splat_u32_ns_sse_unroll2, splat_u32_ns;

   139 }

   140 #endif

   141 

   142 #ifdef	__SYMBIAN32__

   143  

   144 OilFunctionImpl* __oil_function_impl_splat_u8_ns_sse, splat_u8_ns() {

   145 		return &_oil_function_impl_splat_u8_ns_sse, splat_u8_ns;

   146 }

   147 #endif

   148 

   149 #ifdef	__SYMBIAN32__

   150  

   151 OilFunctionImpl* __oil_function_impl_splat_u8_ns_sse_unroll2, splat_u8_ns() {

   152 		return &_oil_function_impl_splat_u8_ns_sse_unroll2, splat_u8_ns;

   153 }

   154 #endif

   155