/*

 * Copyright (c) 2005

 *	Eric Anholt.  All rights reserved.

 *

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

 * modification, are permitted provided that the following conditions

 * are met:

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

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

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

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

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

 *

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

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

 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE

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

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

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

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

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

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

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

 * SUCH DAMAGE.

 */

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

#ifdef HAVE_CONFIG_H

#include "config.h"

#endif

#include <liboil/liboilclasses.h>

#include <liboil/liboilfunction.h>

#include <emmintrin.h>

#include <xmmintrin.h>

#define SSE_FUNCTION __attribute__((force_align_arg_pointer))

SSE_FUNCTION static void

add_f32_sse_unroll2 (float *dest, float *src1, float *src2, int n)

{

  /* Initial operations to align the destination pointer */

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

    *dest++ = *src1++ + *src2++;

  }

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

    __m128 xmm0, xmm1;

    xmm0 = _mm_loadu_ps(src1);

    xmm1 = _mm_loadu_ps(src2);

    xmm0 = _mm_add_ps(xmm0, xmm1);

    _mm_store_ps(dest, xmm0);

    xmm0 = _mm_loadu_ps(src1 + 4);

    xmm1 = _mm_loadu_ps(src2 + 4);

    xmm0 = _mm_add_ps(xmm0, xmm1);

    _mm_store_ps(dest + 4, xmm0);

    dest += 8;

    src1 += 8;

    src2 += 8;

  }

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

    *dest++ = *src1++ + *src2++;

  }

}

OIL_DEFINE_IMPL_FULL (add_f32_sse_unroll2, add_f32, OIL_IMPL_FLAG_SSE);

SSE_FUNCTION static void

subtract_f32_sse_unroll2 (float *dest, float *src1, float *src2, int n)

{

  /* Initial operations to align the destination pointer */

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

    *dest++ = *src1++ - *src2++;

  }

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

    __m128 xmm0, xmm1;

    xmm0 = _mm_loadu_ps(src1);

    xmm1 = _mm_loadu_ps(src2);

    xmm0 = _mm_sub_ps(xmm0, xmm1);

    _mm_store_ps(dest, xmm0);

    xmm0 = _mm_loadu_ps(src1 + 4);

    xmm1 = _mm_loadu_ps(src2 + 4);

    xmm0 = _mm_sub_ps(xmm0, xmm1);

    _mm_store_ps(dest + 4, xmm0);

    dest += 8;

    src1 += 8;

    src2 += 8;

  }

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

    *dest++ = *src1++ - *src2++;

  }

}

OIL_DEFINE_IMPL_FULL (subtract_f32_sse_unroll2, subtract_f32, OIL_IMPL_FLAG_SSE);

SSE_FUNCTION static void

multiply_f32_sse_unroll2 (float *dest, float *src1, float *src2, int n)

{

  /* Initial operations to align the destination pointer */

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

    *dest++ = *src1++ * *src2++;

  }

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

    __m128 xmm0, xmm1;

    xmm0 = _mm_loadu_ps(src1);

    xmm1 = _mm_loadu_ps(src2);

    xmm0 = _mm_mul_ps(xmm0, xmm1);

    _mm_store_ps(dest, xmm0);

    xmm0 = _mm_loadu_ps(src1 + 4);

    xmm1 = _mm_loadu_ps(src2 + 4);

    xmm0 = _mm_mul_ps(xmm0, xmm1);

    _mm_store_ps(dest + 4, xmm0);

    dest += 8;

    src1 += 8;

    src2 += 8;

  }

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

    *dest++ = *src1++ * *src2++;

  }

}

OIL_DEFINE_IMPL_FULL (multiply_f32_sse_unroll2, multiply_f32, OIL_IMPL_FLAG_SSE);

SSE_FUNCTION static void

divide_f32_sse_unroll2 (float *dest, float *src1, float *src2, int n)

{

  /* Initial operations to align the destination pointer */

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

    *dest++ = *src1++ / *src2++;

  }

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

    __m128 xmm0, xmm1;

    xmm0 = _mm_loadu_ps(src1);

    xmm1 = _mm_loadu_ps(src2);

    xmm0 = _mm_div_ps(xmm0, xmm1);

    _mm_store_ps(dest, xmm0);

    xmm0 = _mm_loadu_ps(src1 + 4);

    xmm1 = _mm_loadu_ps(src2 + 4);

    xmm0 = _mm_div_ps(xmm0, xmm1);

    _mm_store_ps(dest + 4, xmm0);

    dest += 8;

    src1 += 8;

    src2 += 8;

  }

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

    *dest++ = *src1++ / *src2++;

  }

}

OIL_DEFINE_IMPL_FULL (divide_f32_sse_unroll2, divide_f32, OIL_IMPL_FLAG_SSE);

SSE_FUNCTION static void

minimum_f32_sse_unroll2 (float *dest, float *src1, float *src2, int n)

{

  /* Initial operations to align the destination pointer */

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

    *dest++ = *src1 < *src2 ? *src1 : *src2;

    src1++;

    src2++;

  }

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

    __m128 xmm0, xmm1;

    xmm0 = _mm_loadu_ps(src1);

    xmm1 = _mm_loadu_ps(src2);

    xmm0 = _mm_min_ps(xmm0, xmm1);

    _mm_store_ps(dest, xmm0);

    xmm0 = _mm_loadu_ps(src1 + 4);

    xmm1 = _mm_loadu_ps(src2 + 4);

    xmm0 = _mm_min_ps(xmm0, xmm1);

    _mm_store_ps(dest + 4, xmm0);

    dest += 8;

    src1 += 8;

    src2 += 8;

  }

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

    *dest++ = *src1 < *src2 ? *src1 : *src2;

    src1++;

    src2++;

  }

}

OIL_DEFINE_IMPL_FULL (minimum_f32_sse_unroll2, minimum_f32, OIL_IMPL_FLAG_SSE);

SSE_FUNCTION static void

maximum_f32_sse_unroll2 (float *dest, float *src1, float *src2, int n)

{

  /* Initial operations to align the destination pointer */

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

    *dest++ = *src1 > *src2 ? *src1 : *src2;

    src1++;

    src2++;

  }

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

    __m128 xmm0, xmm1;

    xmm0 = _mm_loadu_ps(src1);

    xmm1 = _mm_loadu_ps(src2);

    xmm0 = _mm_max_ps(xmm0, xmm1);

    _mm_store_ps(dest, xmm0);

    xmm0 = _mm_loadu_ps(src1 + 4);

    xmm1 = _mm_loadu_ps(src2 + 4);

    xmm0 = _mm_max_ps(xmm0, xmm1);

    _mm_store_ps(dest + 4, xmm0);

    dest += 8;

    src1 += 8;

    src2 += 8;

  }

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

    *dest++ = *src1 > *src2 ? *src1 : *src2;

    src1++;

    src2++;

  }

}

OIL_DEFINE_IMPL_FULL (maximum_f32_sse_unroll2, maximum_f32, OIL_IMPL_FLAG_SSE);

SSE_FUNCTION static void

inverse_f32_sse_unroll2 (float *dest, float *src1, int n)

{

  /* Initial operations to align the destination pointer */

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

    *dest++ = 1.0 / *src1++;

  }

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

    __m128 xmm0, xmm1;

    /* While _mm_rcp_ps sounds promising, the results it gives are rather

     * different from the 1.0 / src1 reference implementation, so do that.

     */

    xmm0 = _mm_set_ps1(1.0);

    xmm1 = _mm_loadu_ps(src1);

    xmm0 = _mm_div_ps(xmm0, xmm1);

    _mm_store_ps(dest, xmm0);

    xmm0 = _mm_set_ps1(1.0);

    xmm1 = _mm_loadu_ps(src1 + 4);

    xmm0 = _mm_div_ps(xmm0, xmm1);

    _mm_store_ps(dest + 4, xmm0);

    dest += 8;

    src1 += 8;

  }

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

    *dest++ = 1.0 / *src1++;

  }

}

OIL_DEFINE_IMPL_FULL (inverse_f32_sse_unroll2, inverse_f32, OIL_IMPL_FLAG_SSE);

SSE_FUNCTION static void

negative_f32_sse_unroll2 (float *dest, float *src1, int n)

{

  /* Initial operations to align the destination pointer */

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

    *dest++ = -(*src1++);

  }

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

    __m128 xmm0, xmm1;

    xmm0 = _mm_setzero_ps();

    xmm1 = _mm_loadu_ps(src1);

    xmm0 = _mm_sub_ps(xmm0, xmm1);

    _mm_store_ps(dest, xmm0);

    xmm0 = _mm_setzero_ps();

    xmm1 = _mm_loadu_ps(src1 + 4);

    xmm0 = _mm_sub_ps(xmm0, xmm1);

    _mm_store_ps(dest + 4, xmm0);

    dest += 8;

    src1 += 8;

  }

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

    *dest++ = -(*src1++);

  }

}

OIL_DEFINE_IMPL_FULL (negative_f32_sse_unroll2, negative_f32, OIL_IMPL_FLAG_SSE);

SSE_FUNCTION static void

scalaradd_f32_ns_sse_unroll2 (float *dest, float *src1, float *val, int n)

{

  __m128 xmm1;

  /* Initial operations to align the destination pointer */

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

    *dest++ = *src1++ + *val;

  }

  xmm1 = _mm_load_ps1(val);

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

    __m128 xmm0;

    xmm0 = _mm_loadu_ps(src1);

    xmm0 = _mm_add_ps(xmm0, xmm1);

    _mm_store_ps(dest, xmm0);

    xmm0 = _mm_loadu_ps(src1 + 4);

    xmm0 = _mm_add_ps(xmm0, xmm1);

    _mm_store_ps(dest + 4, xmm0);

    dest += 8;

    src1 += 8;

  }

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

    *dest++ = *src1++ + *val;

  }

}

OIL_DEFINE_IMPL_FULL (scalaradd_f32_ns_sse_unroll2, scalaradd_f32_ns, OIL_IMPL_FLAG_SSE);

SSE_FUNCTION static void

scalarmultiply_f32_ns_sse_unroll2 (float *dest, float *src1, float *val, int n)

{

  __m128 xmm1;

  /* Initial operations to align the destination pointer */

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

    *dest++ = *src1++ * *val;

  }

  xmm1 = _mm_load_ps1(val);

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

    __m128 xmm0;

    xmm0 = _mm_loadu_ps(src1);

    xmm0 = _mm_mul_ps(xmm0, xmm1);

    _mm_store_ps(dest, xmm0);

    xmm0 = _mm_loadu_ps(src1 + 4);

    xmm0 = _mm_mul_ps(xmm0, xmm1);

    _mm_store_ps(dest + 4, xmm0);

    dest += 8;

    src1 += 8;

  }

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

    *dest++ = *src1++ * *val;

  }

}

OIL_DEFINE_IMPL_FULL (scalarmultiply_f32_ns_sse_unroll2, scalarmultiply_f32_ns, OIL_IMPL_FLAG_SSE);

SSE_FUNCTION static void

scalarmultiply_f64_ns_sse2_unroll2 (double *dest, double *src1, double *val, int n)

{

  __m128d xmm1;

  /* Initial operations to align the destination pointer */

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

    *dest++ = *src1++ * *val;

  }

  xmm1 = _mm_load_pd1(val);

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

    __m128d xmm0;

    xmm0 = _mm_loadu_pd(src1);

    xmm0 = _mm_mul_pd(xmm0, xmm1);

    _mm_store_pd(dest, xmm0);

    xmm0 = _mm_loadu_pd(src1 + 2);

    xmm0 = _mm_mul_pd(xmm0, xmm1);

    _mm_store_pd(dest + 2, xmm0);

    dest += 4;

    src1 += 4;

  }

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

    *dest++ = *src1++ * *val;

  }

}

OIL_DEFINE_IMPL_FULL (scalarmultiply_f64_ns_sse2_unroll2, scalarmultiply_f64_ns, OIL_IMPL_FLAG_SSE2);

#ifdef	__SYMBIAN32__

OilFunctionImpl* __oil_function_impl_add_f32_sse_unroll2, add_f32() {

		return &_oil_function_impl_add_f32_sse_unroll2, add_f32;

}

#endif

#ifdef	__SYMBIAN32__

OilFunctionImpl* __oil_function_impl_subtract_f32_sse_unroll2, subtract_f32() {

		return &_oil_function_impl_subtract_f32_sse_unroll2, subtract_f32;

}

#endif

#ifdef	__SYMBIAN32__

OilFunctionImpl* __oil_function_impl_multiply_f32_sse_unroll2, multiply_f32() {

		return &_oil_function_impl_multiply_f32_sse_unroll2, multiply_f32;

}

#endif

#ifdef	__SYMBIAN32__

OilFunctionImpl* __oil_function_impl_divide_f32_sse_unroll2, divide_f32() {

		return &_oil_function_impl_divide_f32_sse_unroll2, divide_f32;

}

#endif

#ifdef	__SYMBIAN32__

OilFunctionImpl* __oil_function_impl_minimum_f32_sse_unroll2, minimum_f32() {

		return &_oil_function_impl_minimum_f32_sse_unroll2, minimum_f32;

}

#endif

#ifdef	__SYMBIAN32__

OilFunctionImpl* __oil_function_impl_maximum_f32_sse_unroll2, maximum_f32() {

		return &_oil_function_impl_maximum_f32_sse_unroll2, maximum_f32;

}

#endif

#ifdef	__SYMBIAN32__

OilFunctionImpl* __oil_function_impl_inverse_f32_sse_unroll2, inverse_f32() {

		return &_oil_function_impl_inverse_f32_sse_unroll2, inverse_f32;

}

#endif

#ifdef	__SYMBIAN32__

OilFunctionImpl* __oil_function_impl_negative_f32_sse_unroll2, negative_f32() {

		return &_oil_function_impl_negative_f32_sse_unroll2, negative_f32;

}

#endif

#ifdef	__SYMBIAN32__

OilFunctionImpl* __oil_function_impl_scalaradd_f32_ns_sse_unroll2, scalaradd_f32_ns() {

		return &_oil_function_impl_scalaradd_f32_ns_sse_unroll2, scalaradd_f32_ns;

}

#endif

#ifdef	__SYMBIAN32__

OilFunctionImpl* __oil_function_impl_scalarmultiply_f32_ns_sse_unroll2, scalarmultiply_f32_ns() {

		return &_oil_function_impl_scalarmultiply_f32_ns_sse_unroll2, scalarmultiply_f32_ns;

}

#endif

#ifdef	__SYMBIAN32__

OilFunctionImpl* __oil_function_impl_scalarmultiply_f64_ns_sse2_unroll2, scalarmultiply_f64_ns() {

		return &_oil_function_impl_scalarmultiply_f64_ns_sse2_unroll2, scalarmultiply_f64_ns;

}

#endif