int main(void)
{
#ifdef ALOCACAO_NORMAL
printf(">>>>>>>>>Versao Offload<<<<<<<<<<<\n");
#else
printf(">>>>>>>>>Versao Memoria Compartilhada<<<<<<<<<<<\n");
#endif
double t_start, t_end;
double t_start_init, t_end_init;
double t_offload_start, t_offload_end, t_start_init_off, t_end_init_off;
double total_kernel;
#ifdef ALOCACAO_NORMAL
A = (DATA_TYPE*)malloc(NI*NK*sizeof(DATA_TYPE));
B = (DATA_TYPE*)malloc(NK*NJ*sizeof(DATA_TYPE));
D = (DATA_TYPE*)malloc(NJ*NL*sizeof(DATA_TYPE));
E_outputFromGpu = (DATA_TYPE*)malloc(NI*NL*sizeof(DATA_TYPE));
#endif
C = (DATA_TYPE*)malloc(NI*NJ*sizeof(DATA_TYPE));
E = (DATA_TYPE*)malloc(NI*NL*sizeof(DATA_TYPE));
t_start_init = rtclock();
read_cl_file();
t_end_init = rtclock();
tmp_read_cl_file = t_end_init - t_start_init;
total_kernel = t_end_init - t_start_init;
t_start_init = rtclock();
#ifndef MALI
cl_initialization();
#else
cl_initialization_Mali();
#endif
t_end_init = rtclock();
tmp_cl_initialization = t_end_init - t_start_init;
total_kernel = t_end_init - t_start_init;
t_start_init = rtclock();
cl_mem_init();
t_end_init = rtclock();
tmp_cl_mem_init= t_end_init - t_start_init;
total_kernel += t_end_init - t_start_init;
//Está dentro da função a contagem
init();
//------------GPU---------------
//Inicia tempo GPU
#ifdef ALOCACAO_NORMAL
t_offload_start = rtclock();
errcode = clEnqueueWriteBuffer(clCommandQue, a_mem_obj, CL_TRUE, 0, sizeof(DATA_TYPE) * NI * NK, A, 0, NULL, NULL);
errcode = clEnqueueWriteBuffer(clCommandQue, b_mem_obj, CL_TRUE, 0, sizeof(DATA_TYPE) * NK * NJ, B, 0, NULL, NULL);
errcode = clEnqueueWriteBuffer(clCommandQue, c_mem_obj, CL_TRUE, 0, sizeof(DATA_TYPE) * NI * NJ, C, 0, NULL, NULL);
errcode = clEnqueueWriteBuffer(clCommandQue, d_mem_obj, CL_TRUE, 0, sizeof(DATA_TYPE) * NJ * NL, D, 0, NULL, NULL);
errcode = clEnqueueWriteBuffer(clCommandQue, e_mem_obj, CL_TRUE, 0, sizeof(DATA_TYPE) * NI * NL, E, 0, NULL, NULL);
if(errcode != CL_SUCCESS)printf("Error in writing buffers\n");
t_offload_end = rtclock();
tmp_clEnqueueWriteBuffer += t_offload_end - t_offload_start;
total_kernel+=t_offload_end - t_offload_start;
#endif
t_start_init = rtclock();
cl_load_prog();
t_end_init = rtclock();
tmp_cl_load_prog= t_end_init - t_start_init;
total_kernel += t_end_init - t_start_init;
t_start_init = rtclock();
cl_launch_kernel();
t_end_init = rtclock();
tmp_cl_launch_kernel += t_end_init - t_start_init;
total_kernel += t_end_init - t_start_init;
#ifdef ALOCACAO_NORMAL
t_offload_start = rtclock();
errcode = clEnqueueReadBuffer(clCommandQue, e_mem_obj, CL_TRUE, 0, sizeof(DATA_TYPE) * NI * NL, E_outputFromGpu, 0, NULL, NULL);
if(errcode != CL_SUCCESS) printf("Error in reading GPU mem\n");
t_offload_end = rtclock();
tmp_clEnqueueReadBuffer += t_offload_end - t_offload_start;
total_kernel+=t_offload_end - t_offload_start;
#endif
//-------------CPU---------------
t_start = rtclock();
mm2_cpu(A, B, C, D, E);
t_end = rtclock();
tmp_serial = t_end - t_start;
compareResults();
t_start_init_off = rtclock();
cl_clean_up();
t_end_init_off = rtclock();
tmp_cl_clean_up+=t_end_init_off - t_start_init_off;
total_kernel += t_end_init_off - t_start_init_off;
#ifdef ALOCACAO_NORMAL
free(C);
free(A);
free(B);
free(D);
free(E_outputFromGpu);
#endif
free(E);
printf("\n-------RESULTS-------\n");
printf("Sizes NI=%d, NJ=%d, NK=%d e NL=%d\n\n", NI, NJ, NK, NL);
printf("read_cl_file -------------> %lf\n", tmp_read_cl_file);
printf("cl_initialization --------> %lf\n", tmp_cl_initialization);
printf("cl_mem_init --------------> %lf\n", tmp_cl_mem_init);
printf("init ---------------------> %lf\n", tmp_init);
printf("cl_load_prog -------------> %lf\n", tmp_cl_load_prog);
printf("cl_launch_kernel ---------> %lf\n", tmp_cl_launch_kernel);
printf("serialExecution ----------> %lf\n", tmp_serial);
printf("cl_clean_up --------------> %lf\n", tmp_cl_clean_up);
printf("clEnqueueWriteBuffer -----> %lf\n", tmp_clEnqueueWriteBuffer);
printf("clEnqueueReadBuffer-------> %lf\n", tmp_clEnqueueReadBuffer);
printf("clEnqueueMapBuffer -------> %lf\n", tmp_clEnqueueMapBuffer);
printf("clEnqueueUnmapMemObject --> %lf\n", tmp_clEnqueueUnmapMemObject);
return 0;
}
int main(int argc, char *argv[])
{
/* Retrieve problem size. */
int ni = NI;
int nj = NJ;
int nk = NK;
int nl = NL;
/* Variable declaration/allocation. */
DATA_TYPE alpha;
DATA_TYPE beta;
POLYBENCH_2D_ARRAY_DECL(tmp,DATA_TYPE,NI,NJ,ni,nj);
POLYBENCH_2D_ARRAY_DECL(A,DATA_TYPE,NI,NK,ni,nk);
POLYBENCH_2D_ARRAY_DECL(B,DATA_TYPE,NK,NJ,nk,nj);
POLYBENCH_2D_ARRAY_DECL(C,DATA_TYPE,NL,NJ,nl,nj);
POLYBENCH_2D_ARRAY_DECL(D,DATA_TYPE,NI,NL,ni,nl);
POLYBENCH_2D_ARRAY_DECL(D_outputFromGpu,DATA_TYPE,NI,NL,ni,nl);
/* Initialize array(s). */
init_array(ni, nj, nk, nl, &alpha, &beta, POLYBENCH_ARRAY(A), POLYBENCH_ARRAY(B), POLYBENCH_ARRAY(C), POLYBENCH_ARRAY(D));
read_cl_file();
cl_initialization();
cl_mem_init(POLYBENCH_ARRAY(tmp), POLYBENCH_ARRAY(A), POLYBENCH_ARRAY(B), POLYBENCH_ARRAY(C), POLYBENCH_ARRAY(D_outputFromGpu));
cl_load_prog();
cl_launch_kernel(ni, nj, nk, nl, alpha, beta);
errcode = clEnqueueReadBuffer(clCommandQue, dOutputFromGpu_mem_obj, CL_TRUE, 0, sizeof(DATA_TYPE) * NI * NL, POLYBENCH_ARRAY(D_outputFromGpu), 0, NULL, NULL);
if(errcode != CL_SUCCESS) printf("Error in reading GPU mem\n");
#ifdef RUN_ON_CPU
/* Start timer. */
polybench_start_instruments;
mm2_cpu(ni, nj, nk, nl, alpha, beta, POLYBENCH_ARRAY(tmp), POLYBENCH_ARRAY(A), POLYBENCH_ARRAY(B), POLYBENCH_ARRAY(C), POLYBENCH_ARRAY(D));
/* Stop and print timer. */
printf("CPU Time in seconds:\n");
polybench_stop_instruments;
polybench_print_instruments;
compareResults(ni, nl, POLYBENCH_ARRAY(D), POLYBENCH_ARRAY(D_outputFromGpu));
#else //prevent dead code elimination
polybench_prevent_dce(print_array(ni, nl, POLYBENCH_ARRAY(D_outputFromGpu)));
#endif //RUN_ON_CPU
cl_clean_up();
POLYBENCH_FREE_ARRAY(tmp);
POLYBENCH_FREE_ARRAY(A);
POLYBENCH_FREE_ARRAY(B);
POLYBENCH_FREE_ARRAY(C);
POLYBENCH_FREE_ARRAY(D);
POLYBENCH_FREE_ARRAY(D_outputFromGpu);
return 0;
}
