// gather_dependent.br// Tests that gather operations produce correct// results when the index is non-constant.// This utilizes the dependent texture read// capability of the GPU.#include <stdio.h>kernel void gather1( float index<>, float a[100], out float b<> ) {	b = a[ index.x ];}kernel void gather2( float2 index<>, float a[10][10], out float b<> ) {	b = a[ index.xy ];}int main() {	int i, j;	float index1< 100 >;	float a1< 100 >;	float b1< 100 >;	float2 index2< 10, 10 >;	float a2< 10, 10 >;	float b2< 10, 10 >;	float input_index1[100];	float input_a1[100];	float output_b1[100];	float2 input_index2[10][10];	float input_a2[10][10];	float output_b2[10][10];		// initialize data	for( i = 0; i < 10; i++ )	{		for( j = 0; j < 10; j++ )		{			input_index1[10*i+j] = (float)(10*i+j);			input_a1[10*i+j] = (float)(99 - (10*i+j));			input_index2[i][j].x = (float)j;			input_index2[i][j].y = (float)i;			input_a2[i][j] = (float)(99 - (10*i+j));		}	}	streamRead( index1, input_index1 );	streamRead( a1, input_a1 );	streamRead( index2, input_index2 );	streamRead( a2, input_a2 );		gather1( index1, a1, b1 );	gather2( index2, a2, b2 );		streamWrite( b1, output_b1 );	streamWrite( b2, output_b2 );		for( i = 0; i < 10; i++ )	{		for( j = 0; j < 10; j++ )			printf( "%3.2f ", output_b1[10*i+j] );		printf("\n");	}		for( i = 0; i < 10; i++ )	{		for( j = 0; j < 10; j++ )			printf( "%3.2f ", output_b2[i][j] );		printf("\n");	}		return 0;}