//************************************************************************** // Multi-threaded Matrix Multiply benchmark //-------------------------------------------------------------------------- // TA : Christopher Celio // Student: // // // This benchmark multiplies two 2-D arrays together and writes the results to // a third vector. The input data (and reference data) should be generated // using the matmul_gendata.pl perl script and dumped to a file named // dataset.h. // print out arrays, etc. //#define DEBUG //-------------------------------------------------------------------------- // Includes #include #include #include //-------------------------------------------------------------------------- // Input/Reference Data typedef float data_t; #include "dataset.h" //-------------------------------------------------------------------------- // Basic Utilities and Multi-thread Support __thread unsigned long coreid; unsigned long ncores; #include "util.h" #define stringify_1(s) #s #define stringify(s) stringify_1(s) #define stats(code) do { \ unsigned long _c = -rdcycle(), _i = -rdinstret(); \ code; \ _c += rdcycle(), _i += rdinstret(); \ if (coreid == 0) \ printf("%s: %ld cycles, %ld.%ld cycles/iter, %ld.%ld CPI\n", \ stringify(code), _c, _c/DIM_SIZE/DIM_SIZE/DIM_SIZE, 10*_c/DIM_SIZE/DIM_SIZE/DIM_SIZE%10, _c/_i, 10*_c/_i%10); \ } while(0) //-------------------------------------------------------------------------- // Helper functions void printArray( char name[], int n, data_t arr[] ) { int i; if (coreid != 0) return; printf( " %10s :", name ); for ( i = 0; i < n; i++ ) printf( " %3ld ", (long) arr[i] ); printf( "\n" ); } void __attribute__((noinline)) verify(size_t n, const data_t* test, const data_t* correct) { if (coreid != 0) return; size_t i; for (i = 0; i < n; i++) { if (test[i] != correct[i]) { printf("FAILED test[%d]= %3ld, correct[%d]= %3ld\n", i, (long)test[i], i, (long)correct[i]); exit(-1); } } return; } //-------------------------------------------------------------------------- // matmul function // single-thread, naive version void __attribute__((noinline)) matmul_naive(const int lda, const data_t A[], const data_t B[], data_t C[] ) { int i, j, k; if (coreid > 0) return; for ( i = 0; i < lda; i++ ) for ( j = 0; j < lda; j++ ) { for ( k = 0; k < lda; k++ ) { C[i + j*lda] += A[j*lda + k] * B[k*lda + i]; } } } void __attribute__((noinline)) matmul(const int lda, const data_t A[], const data_t B[], data_t C[] ) { int i, j, k, x; data_t temp0, temp1, temp2, temp3, temp4, temp5, temp6, temp7; data_t temp8, temp9, temp10, temp11, temp12, temp13, temp14, temp15; //complete Q1 if(coreid == 0) { for(j = 0; j < 32; j++) { temp0 = C[j*lda]; temp1 = C[1 + j*lda]; temp2 = C[2 + j*lda]; temp3 = C[3 + j*lda]; temp4 = C[4 + j*lda]; temp5 = C[5 + j*lda]; temp6 = C[6 + j*lda]; temp7 = C[7 + j*lda]; temp8 = C[8 + j*lda]; temp9 = C[9 + j*lda]; temp10 = C[10 + j*lda]; temp11 = C[11 + j*lda]; temp12 = C[12 + j*lda]; temp13 = C[13 + j*lda]; temp14 = C[14 + j*lda]; temp15 = C[15 + j*lda]; for(k = 0; k < 32; k++) { temp0 += A[j*lda + k] * B[k*lda]; temp1 += A[j*lda + k] * B[1+k*lda]; temp2 += A[j*lda + k] * B[2+k*lda]; temp3 += A[j*lda + k] * B[3+k*lda]; temp4 += A[j*lda + k] * B[4+k*lda]; temp5 += A[j*lda + k] * B[5+k*lda]; temp6 += A[j*lda + k] * B[6+k*lda]; temp7 += A[j*lda + k] * B[7+k*lda]; temp8 += A[j*lda + k] * B[8+k*lda]; temp9 += A[j*lda + k] * B[9+k*lda]; temp10 += A[j*lda + k] * B[10+k*lda]; temp11 += A[j*lda + k] * B[11+k*lda]; temp12 += A[j*lda + k] * B[12+k*lda]; temp13 += A[j*lda + k] * B[13+k*lda]; temp14 += A[j*lda + k] * B[14+k*lda]; temp15 += A[j*lda + k] * B[15+k*lda]; } C[j*lda] = temp0; C[1 + j*lda] = temp1; C[2 + j*lda] = temp2; C[3 + j*lda] = temp3; C[4 + j*lda] = temp4; C[5 + j*lda] = temp5; C[6 + j*lda] = temp6; C[7 + j*lda] = temp7; C[8 + j*lda] = temp8; C[9 + j*lda] = temp9; C[10 + j*lda] = temp10; C[11 + j*lda] = temp11; C[12 + j*lda] = temp12; C[13 + j*lda] = temp13; C[14 + j*lda] = temp14; C[15 + j*lda] = temp15; } } else { for(j = 0; j < 32; j++) { temp0 = C[16 + j*lda]; temp1 = C[17 + j*lda]; temp2 = C[18 + j*lda]; temp3 = C[19 + j*lda]; temp4 = C[20 + j*lda]; temp5 = C[21 + j*lda]; temp6 = C[22 + j*lda]; temp7 = C[23 + j*lda]; temp8 = C[24 + j*lda]; temp9 = C[25 + j*lda]; temp10 = C[26 + j*lda]; temp11 = C[27 + j*lda]; temp12 = C[28 + j*lda]; temp13 = C[29 + j*lda]; temp14 = C[30 + j*lda]; temp15 = C[31 + j*lda]; for(k = 0; k < 32; k++) { temp0 += A[j*lda + k] * B[16 + k*lda]; temp1 += A[j*lda + k] * B[17 + k*lda]; temp2 += A[j*lda + k] * B[18 + k*lda]; temp3 += A[j*lda + k] * B[19 + k*lda]; temp4 += A[j*lda + k] * B[20 + k*lda]; temp5 += A[j*lda + k] * B[21 + k*lda]; temp6 += A[j*lda + k] * B[22 + k*lda]; temp7 += A[j*lda + k] * B[23 + k*lda]; temp8 += A[j*lda + k] * B[24 + k*lda]; temp9 += A[j*lda + k] * B[25 + k*lda]; temp10 += A[j*lda + k] * B[26 + k*lda]; temp11 += A[j*lda + k] * B[27 + k*lda]; temp12 += A[j*lda + k] * B[28 + k*lda]; temp13 += A[j*lda + k] * B[29 + k*lda]; temp14 += A[j*lda + k] * B[30 + k*lda]; temp15 += A[j*lda + k] * B[31 + k*lda]; } C[16 + j*lda] = temp0; C[17 + j*lda] = temp1; C[18 + j*lda] = temp2; C[19 + j*lda] = temp3; C[20 + j*lda] = temp4; C[21 + j*lda] = temp5; C[22 + j*lda] = temp6; C[23 + j*lda] = temp7; C[24 + j*lda] = temp8; C[25 + j*lda] = temp9; C[26 + j*lda] = temp10; C[27 + j*lda] = temp11; C[28 + j*lda] = temp12; C[29 + j*lda] = temp13; C[30 + j*lda] = temp14; C[31 + j*lda] = temp15; } } } //-------------------------------------------------------------------------- // Main // // all threads start executing thread_entry(). Use their "coreid" to // differentiate between threads (each thread is running on a separate core). void thread_entry(int cid, int nc) { coreid = cid; ncores = nc; // static allocates data in the binary, which is visible to both threads static data_t results_data[ARRAY_SIZE]; // // Execute the provided, naive matmul // barrier(); // stats(matmul_naive(DIM_SIZE, input1_data, input2_data, results_data); barrier()); // // // // verify // verify(ARRAY_SIZE, results_data, verify_data); // // // clear results from the first trial // size_t i; // if (coreid == 0) // for (i=0; i < ARRAY_SIZE; i++) // results_data[i] = 0; // barrier(); // Execute your faster matmul barrier(); stats(matmul(DIM_SIZE, input1_data, input2_data, results_data); barrier()); #ifdef DEBUG printArray("results:", ARRAY_SIZE, results_data); printArray("verify :", ARRAY_SIZE, verify_data); #endif // verify verify(ARRAY_SIZE, results_data, verify_data); barrier(); exit(0); }