sl@0: #!/usr/bin/perl
sl@0: #
sl@0: 
sl@0: 
sl@0: 
sl@0: print <<EOF
sl@0: /* This file is autogenerated.  Do not edit. */
sl@0: /*
sl@0:  * LIBOIL - Library of Optimized Inner Loops
sl@0:  * Copyright (c) 2005 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: 
sl@0: #ifdef HAVE_CONFIG_H
sl@0: #include "config.h"
sl@0: #endif
sl@0: 
sl@0: #include <math.h>
sl@0: 
sl@0: #include <liboil/liboil.h>
sl@0: #include <liboil/liboilclasses.h>
sl@0: 
sl@0: EOF
sl@0: ;
sl@0: 
sl@0: 
sl@0: sub binary_pointer
sl@0: {
sl@0: 	my $kernel = shift;
sl@0: 	my $precision = shift;
sl@0: 	my $type = "oil_type_$precision";
sl@0: 	my $operator = shift;
sl@0: 
sl@0: 	print <<EOF
sl@0: static void
sl@0: ${kernel}_${precision}_pointer (${type} *dest, ${type} *src1, ${type} *src2, int n)
sl@0: {
sl@0:   while (n) {
sl@0:     *dest = *src1 ${operator} *src2;
sl@0:     dest++;
sl@0:     src1++;
sl@0:     src2++;
sl@0:     n--;
sl@0:   }
sl@0: }
sl@0: OIL_DEFINE_IMPL (${kernel}_${precision}_pointer, ${kernel}_${precision});
sl@0: 
sl@0: EOF
sl@0: ;
sl@0: }
sl@0: 
sl@0: sub binary_unroll2
sl@0: {
sl@0: 	my $kernel = shift;
sl@0: 	my $precision = shift;
sl@0: 	my $type = "oil_type_$precision";
sl@0: 	my $operator = shift;
sl@0: 
sl@0: 	print <<EOF
sl@0: static void
sl@0: ${kernel}_${precision}_unroll2 (${type} *dest, ${type} *src1, ${type} *src2, int n)
sl@0: {
sl@0:   int i;
sl@0: 
sl@0:   if (n & 1) {
sl@0:     dest[0] = src1[0] ${operator} src2[0];
sl@0:     dest++;
sl@0:     src1++;
sl@0:     src2++;
sl@0:     n--;
sl@0:   }
sl@0:   for(i=0;i<n;i+=2){
sl@0:     dest[i] = src1[i] ${operator} src2[i];
sl@0:     dest[i+1] = src1[i+1] ${operator} src2[i+1];
sl@0:   }
sl@0: }
sl@0: OIL_DEFINE_IMPL (${kernel}_${precision}_unroll2, ${kernel}_${precision});
sl@0: 
sl@0: EOF
sl@0: ;
sl@0: }
sl@0: 
sl@0: sub binary_unroll4a
sl@0: {
sl@0: 	my $kernel = shift;
sl@0: 	my $precision = shift;
sl@0: 	my $type = "oil_type_$precision";
sl@0: 	my $operator = shift;
sl@0: 
sl@0: 	print <<EOF
sl@0: static void
sl@0: ${kernel}_${precision}_unroll4a (${type} *dest, ${type} *src1, ${type} *src2, int n)
sl@0: {
sl@0:   int i;
sl@0: 
sl@0:   while (n & 3) {
sl@0:     dest[0] = src1[0] ${operator} src2[0];
sl@0:     dest++;
sl@0:     src1++;
sl@0:     src2++;
sl@0:     n--;
sl@0:   }
sl@0:   for(i=0;i<n;i+=4){
sl@0:     dest[i] = src1[i] ${operator} src2[i];
sl@0:     dest[i+1] = src1[i+1] ${operator} src2[i+1];
sl@0:     dest[i+2] = src1[i+2] ${operator} src2[i+2];
sl@0:     dest[i+3] = src1[i+3] ${operator} src2[i+3];
sl@0:   }
sl@0: }
sl@0: OIL_DEFINE_IMPL (${kernel}_${precision}_unroll4a, ${kernel}_${precision});
sl@0: 
sl@0: EOF
sl@0: ;
sl@0: }
sl@0: 
sl@0: sub binary_unroll4b
sl@0: {
sl@0: 	my $kernel = shift;
sl@0: 	my $precision = shift;
sl@0: 	my $type = "oil_type_$precision";
sl@0: 	my $operator = shift;
sl@0: 
sl@0: 	print <<EOF
sl@0: static void
sl@0: ${kernel}_${precision}_unroll4b (${type} *dest, ${type} *src1, ${type} *src2, int n)
sl@0: {
sl@0:   int i;
sl@0: 
sl@0:   for(i=0;i<(n&(~0x3));i+=4){
sl@0:     dest[i+0] = src1[i+0] ${operator} src2[i+0];
sl@0:     dest[i+1] = src1[i+1] ${operator} src2[i+1];
sl@0:     dest[i+2] = src1[i+2] ${operator} src2[i+2];
sl@0:     dest[i+3] = src1[i+3] ${operator} src2[i+3];
sl@0:   }
sl@0:   for(;i<n;i++){
sl@0:     dest[i] = src1[i] ${operator} src2[i];
sl@0:   }
sl@0: }
sl@0: OIL_DEFINE_IMPL (${kernel}_${precision}_unroll4b, ${kernel}_${precision});
sl@0: 
sl@0: EOF
sl@0: ;
sl@0: }
sl@0: 
sl@0: sub binary_unroll4c
sl@0: {
sl@0: 	my $kernel = shift;
sl@0: 	my $precision = shift;
sl@0: 	my $type = "oil_type_$precision";
sl@0: 	my $operator = shift;
sl@0: 
sl@0: 	print <<EOF
sl@0: static void
sl@0: ${kernel}_${precision}_unroll4c (${type} *dest, ${type} *src1, ${type} *src2, int n)
sl@0: {
sl@0:   int i;
sl@0: 
sl@0:   for(i=0;i<(n&(~0x3));i+=4){
sl@0:     *dest++ = *src1++ ${operator} *src2++;
sl@0:     *dest++ = *src1++ ${operator} *src2++;
sl@0:     *dest++ = *src1++ ${operator} *src2++;
sl@0:     *dest++ = *src1++ ${operator} *src2++;
sl@0:   }
sl@0:   for(;i<n;i++){
sl@0:     *dest++ = *src1++ ${operator} *src2++;
sl@0:   }
sl@0: }
sl@0: OIL_DEFINE_IMPL (${kernel}_${precision}_unroll4c, ${kernel}_${precision});
sl@0: 
sl@0: EOF
sl@0: ;
sl@0: }
sl@0: 
sl@0: my %binary_operators = (
sl@0:  "add" => "+",
sl@0:  "subtract" => "-",
sl@0:  "multiply" => "*",
sl@0:  "divide" => "/"
sl@0: );
sl@0: 
sl@0: my @types = ( "f32", "f64" );
sl@0: 
sl@0: while ( ($name, $op) = each %binary_operators ) {
sl@0:   foreach $prec (@types) {
sl@0:     binary_pointer($name, $prec, $op);
sl@0:     binary_unroll2($name, $prec, $op);
sl@0:     binary_unroll4a($name, $prec, $op);
sl@0:     binary_unroll4b($name, $prec, $op);
sl@0:     binary_unroll4c($name, $prec, $op);
sl@0:   }
sl@0: }
sl@0: 
sl@0: exit 0;
sl@0: 
sl@0: binary_pointer("subtract", "f32", "-");
sl@0: binary_unroll2("subtract", "f32", "-");
sl@0: binary_unroll4a("subtract", "f32", "-");
sl@0: binary_unroll4b("subtract", "f32", "-");
sl@0: binary_unroll4c("subtract", "f32", "-");
sl@0: 
sl@0: binary_pointer("add", "f32", "+");
sl@0: binary_unroll2("add", "f32", "+");
sl@0: binary_unroll4a("add", "f32", "+");
sl@0: binary_unroll4b("add", "f32", "+");
sl@0: binary_unroll4c("add", "f32", "+");
sl@0: 
sl@0: binary_pointer("multiply", "f32", "*");
sl@0: binary_unroll2("multiply", "f32", "*");
sl@0: binary_unroll4a("multiply", "f32", "*");
sl@0: binary_unroll4b("multiply", "f32", "*");
sl@0: binary_unroll4c("multiply", "f32", "*");
sl@0: 
sl@0: binary_pointer("divide", "f32", "/");
sl@0: binary_unroll2("divide", "f32", "/");
sl@0: binary_unroll4a("divide", "f32", "/");
sl@0: binary_unroll4b("divide", "f32", "/");
sl@0: binary_unroll4c("divide", "f32", "/");
sl@0: 
sl@0: binary_pointer("subtract", "f64", "-");
sl@0: binary_unroll2("subtract", "f64", "-");
sl@0: binary_unroll4a("subtract", "f64", "-");
sl@0: binary_unroll4b("subtract", "f64", "-");
sl@0: binary_unroll4c("subtract", "f64", "-");
sl@0: 
sl@0: binary_pointer("add", "f64", "+");
sl@0: binary_unroll2("add", "f64", "+");
sl@0: binary_unroll4a("add", "f64", "+");
sl@0: binary_unroll4b("add", "f64", "+");
sl@0: binary_unroll4c("add", "f64", "+");
sl@0: 
sl@0: binary_pointer("multiply", "f64", "*");
sl@0: binary_unroll2("multiply", "f64", "*");
sl@0: binary_unroll4a("multiply", "f64", "*");
sl@0: binary_unroll4b("multiply", "f64", "*");
sl@0: binary_unroll4c("multiply", "f64", "*");
sl@0: 
sl@0: binary_pointer("divide", "f64", "/");
sl@0: binary_unroll2("divide", "f64", "/");
sl@0: binary_unroll4a("divide", "f64", "/");
sl@0: binary_unroll4b("divide", "f64", "/");
sl@0: binary_unroll4c("divide", "f64", "/");
sl@0: 
sl@0: $blah = "
sl@0: static void
sl@0: subtract_f32_ref (float *dest, float *src1, float *src2, int n)
sl@0: {
sl@0:   int i;
sl@0: 
sl@0:   for(i=0;i<n;i++){
sl@0:     dest[i] = src1[i] - src2[i];
sl@0:   }
sl@0: }
sl@0: OIL_DEFINE_IMPL (subtract_f32_ref, subtract_f32);
sl@0: 
sl@0: static void
sl@0: multiply_f32_ref (float *dest, float *src1, float *src2, int n)
sl@0: {
sl@0:   int i;
sl@0: 
sl@0:   for(i=0;i<n;i++){
sl@0:     dest[i] = src1[i] * src2[i];
sl@0:   }
sl@0: }
sl@0: OIL_DEFINE_IMPL (multiply_f32_ref, multiply_f32);
sl@0: 
sl@0: static void
sl@0: divide_f32_ref (float *dest, float *src1, float *src2, int n)
sl@0: {
sl@0:   int i;
sl@0: 
sl@0:   for(i=0;i<n;i++){
sl@0:     dest[i] = src1[i] / src2[i];
sl@0:   }
sl@0: }
sl@0: OIL_DEFINE_IMPL_REF (divide_f32_ref, divide_f32);
sl@0: 
sl@0: static void
sl@0: minimum_f32_ref (float *dest, float *src1, float *src2, int n)
sl@0: {
sl@0:   int i;
sl@0: 
sl@0:   for(i=0;i<n;i++){
sl@0:     dest[i] = (src1[i] < src2[i]) ? src1[i] : src2[i];
sl@0:   }
sl@0: }
sl@0: OIL_DEFINE_IMPL_REF (minimum_f32_ref, minimum_f32);
sl@0: 
sl@0: static void
sl@0: maximum_f32_ref (float *dest, float *src1, float *src2, int n)
sl@0: {
sl@0:   int i;
sl@0: 
sl@0:   for(i=0;i<n;i++){
sl@0:     dest[i] = (src1[i] > src2[i]) ? src1[i] : src2[i];
sl@0:   }
sl@0: }
sl@0: OIL_DEFINE_IMPL_REF (maximum_f32_ref, maximum_f32);
sl@0: 
sl@0: static void
sl@0: negative_f32_ref (float *dest, float *src1, int n)
sl@0: {
sl@0:   int i;
sl@0: 
sl@0:   for(i=0;i<n;i++){
sl@0:     dest[i] = -src1[i];
sl@0:   }
sl@0: }
sl@0: OIL_DEFINE_IMPL_REF (negative_f32_ref, negative_f32);
sl@0: 
sl@0: static void
sl@0: inverse_f32_ref (float *dest, float *src1, int n)
sl@0: {
sl@0:   int i;
sl@0: 
sl@0:   for(i=0;i<n;i++){
sl@0:     dest[i] = 1.0/src1[i];
sl@0:   }
sl@0: }
sl@0: OIL_DEFINE_IMPL_REF (inverse_f32_ref, inverse_f32);
sl@0: 
sl@0: static void
sl@0: sign_f32_ref (float *dest, float *src1, int n)
sl@0: {
sl@0:   int i;
sl@0: 
sl@0:   for(i=0;i<n;i++){
sl@0:     dest[i] = (src1[i] < 0) ? -src1[i] : src1[i];
sl@0:   }
sl@0: }
sl@0: OIL_DEFINE_IMPL_REF (sign_f32_ref, sign_f32);
sl@0: 
sl@0: static void
sl@0: floor_f32_ref (float *dest, float *src1, int n)
sl@0: {
sl@0:   int i;
sl@0: 
sl@0:   for(i=0;i<n;i++){
sl@0:     dest[i] = floor(src1[i]);
sl@0:   }
sl@0: }
sl@0: OIL_DEFINE_IMPL_REF (floor_f32_ref, floor_f32);
sl@0: 
sl@0: 
sl@0: 
sl@0: static void
sl@0: scalaradd_f32_ns_ref (float *dest, float *src1, float *src2, int n)
sl@0: {
sl@0:   int i;
sl@0: 
sl@0:   for(i=0;i<n;i++){
sl@0:     dest[i] = src1[i] + src2[0];
sl@0:   }
sl@0: }
sl@0: OIL_DEFINE_IMPL_REF (scalaradd_f32_ns_ref, scalaradd_f32_ns);
sl@0: 
sl@0: static void
sl@0: scalarmultiply_f32_ns_ref (float *dest, float *src1, float *src2, int n)
sl@0: {
sl@0:   int i;
sl@0: 
sl@0:   for(i=0;i<n;i++){
sl@0:     dest[i] = src1[i] * src2[0];
sl@0:   }
sl@0: }
sl@0: OIL_DEFINE_IMPL_REF (scalarmultiply_f32_ns_ref, scalarmultiply_f32_ns);
sl@0: 
sl@0: 
sl@0: ";