diff options
Diffstat (limited to 'mt/az_matmul/matmul_mi.c')
-rwxr-xr-x | mt/az_matmul/matmul_mi.c | 416 |
1 files changed, 416 insertions, 0 deletions
diff --git a/mt/az_matmul/matmul_mi.c b/mt/az_matmul/matmul_mi.c new file mode 100755 index 0000000..56f02d3 --- /dev/null +++ b/mt/az_matmul/matmul_mi.c @@ -0,0 +1,416 @@ +//************************************************************************** +// 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 <string.h> +#include <stdlib.h> +#include <stdio.h> + + +//-------------------------------------------------------------------------- +// 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]; + } + } + +} + + +data_t ffmul(data_t a, data_t b) { + data_t result = 0; + + for (int i=0; i < b; i++) { + result += a; + } + + return result; +} + + +//void __attribute__((noinline)) matmul(const int lda, const data_t A[], const data_t B[], data_t C[] ) +//{ +// +// // ***************************** // +// // **** ADD YOUR CODE HERE ***** // +// // ***************************** // +// // +// // feel free to make a separate function for MI and MSI versions. +// +// static __thread int i, j, k; +// static __thread int jlda, ilda; +// static __thread data_t tempA1, tempA2, tempA3, tempA4, tempA5, tempA6, tempA7, tempA8; +// static __thread int start, end; +// +// start = coreid*(lda>>1); +// end = (coreid+1)*(lda>>1); +// +// for (j=start; j < end; j+=1) { +// jlda = j * lda; +// for ( i=0; i < lda; i+=1 ) { +// ilda = i*lda; +// tempA1 = A[i + jlda]; +// //tempA2 = A[i+1 + jlda]; +// //tempA3 = A[i+2 + jlda]; +// //tempA4 = A[i+3 + jlda]; +// //tempA5 = A[i+4 + jlda]; +// //tempA6 = A[i+5 + jlda]; +// //tempA7 = A[i+6 + jlda]; +// //tempA8 = A[i+7 + jlda]; +// //tempC1 = C[i + j*lda]; +// //tempC2 = C[i+1 + j*lda]; +// for(k=0; k < lda; k+=1) { +// //C[k + jlda] += tempA1 * B[k + i*lda] + tempA2 * B[k + (i+1)*lda] + tempA3 * B[k + (i+2)*lda] + tempA4 * B[k + (i+3)*lda] + +// // tempA5 * B[k + (i+4)*lda] + tempA6 * B[k + (i+5)*lda] + tempA7 * B[k + (i+6)*lda] + tempA8 * B[k + (i+7)*lda]; +// +// C[k + jlda] += tempA1* B[k + i*lda];// + ffmul(tempA2,B[k + (i+1)*lda]) + tempA3 * B[k + (i+2)*lda] + tempA4 * B[k + (i+3)*lda] + +// // tempA5 * B[k + (i+4)*lda] + tempA6 * B[k + (i+5)*lda] + tempA7 * B[k + (i+6)*lda] + tempA8 * B[k + (i+7)*lda]; +// // +// //C[k+1 + jlda] += tempA1 * B[k+1 + i*lda] + tempA2 * B[k+1 + (i+1)*lda] + tempA3 * B[k+1 + (i+2)*lda] + tempA4 * B[k+1 + (i+3)*lda] + +// // tempA5 * B[k+1 + (i+4)*lda] + tempA6 * B[k+1 + (i+5)*lda] + tempA7 * B[k+1 + (i+6)*lda] + tempA8 * B[k+1 + (i+7)*lda]; +// // +// //C[k+2 + jlda] += tempA1 * B[k+2 + i*lda] + tempA2 * B[k+2 + (i+1)*lda] + tempA3 * B[k+2 + (i+2)*lda] + tempA4 * B[k+2 + (i+3)*lda] + +// // tempA5 * B[k+2 + (i+4)*lda] + tempA6 * B[k+2 + (i+5)*lda] + tempA7 * B[k+2 + (i+6)*lda] + tempA8 * B[k+2 + (i+7)*lda]; +// // +// //C[k+3 + jlda] += tempA1 * B[k+3 + i*lda] + tempA2 * B[k+3 + (i+1)*lda] + tempA3 * B[k+3 + (i+2)*lda] + tempA4 * B[k+3 + (i+3)*lda] + +// // tempA5 * B[k+3 + (i+4)*lda] + tempA6 * B[k+3 + (i+5)*lda] + tempA7 * B[k+3 + (i+6)*lda] + tempA8 * B[k+3 + (i+7)*lda]; +// // +// //C[k+4 + jlda] += tempA1 * B[k+4 + i*lda] + tempA2 * B[k+4 + (i+1)*lda] + tempA3 * B[k+4 + (i+2)*lda] + tempA4 * B[k+4 + (i+3)*lda] + +// // tempA5 * B[k+4 + (i+4)*lda] + tempA6 * B[k+4 + (i+5)*lda] + tempA7 * B[k+4 + (i+6)*lda] + tempA8 * B[k+4 + (i+7)*lda]; +// // +// //C[k+5 + jlda] += tempA1 * B[k+5 + i*lda] + tempA2 * B[k+5 + (i+1)*lda] + tempA3 * B[k+5 + (i+2)*lda] + tempA4 * B[k+5 + (i+3)*lda] + +// // tempA5 * B[k+5 + (i+4)*lda] + tempA6 * B[k+5 + (i+5)*lda] + tempA7 * B[k+5 + (i+6)*lda] + tempA8 * B[k+5 + (i+7)*lda]; +// // +// //C[k+6 + jlda] += tempA1 * B[k+6 + i*lda] + tempA2 * B[k+6 + (i+1)*lda] + tempA3 * B[k+6 + (i+2)*lda] + tempA4 * B[k+6 + (i+3)*lda] + +// // tempA5 * B[k+6 + (i+4)*lda] + tempA6 * B[k+6 + (i+5)*lda] + tempA7 * B[k+6 + (i+6)*lda] + tempA8 * B[k+6 + (i+7)*lda]; +// // +// //C[k+7 + jlda] += tempA1 * B[k+7 + i*lda] + tempA2 * B[k+7 + (i+1)*lda] + tempA3 * B[k+7 + (i+2)*lda] + tempA4 * B[k+7 + (i+3)*lda] + +// // tempA5 * B[k+7 + (i+4)*lda] + tempA6 * B[k+7 + (i+5)*lda] + tempA7 * B[k+7 + (i+6)*lda] + tempA8 * B[k+7 + (i+7)*lda]; +// +// +// } +// } +// } +//} + + +void __attribute__((noinline)) matmul(const int lda, const data_t A[], const data_t B[], data_t C[] ) +{ + + // ***************************** // + // **** ADD YOUR CODE HERE ***** // + // ***************************** // + // + // feel free to make a separate function for MI and MSI versions. + + static __thread int i, j, k; + static __thread data_t tempA0, tempA1, tempA2, tempA3, tempA4, tempA5, tempA6, tempA7; + static __thread data_t tempC0, tempC1, tempC2, tempC3, tempC4, tempC5, tempC6, tempC7; //tempC8, tempC9, tempC10, tempC11, tempC12, tempC13, tempC14, tempC15; + + static __thread int start, end, jStride, jToRow, jToCol, iToRow; + + start = coreid << 9; + end = (coreid+1) << 9; + jStride = 8; + + for (j=start; j < end; j+=jStride) { + jToRow = (j>>5)<<5; + jToCol = j%32; + tempC0 = 0; + tempC1 = 0; + tempC2 = 0; + tempC3 = 0; + tempC4 = 0; + tempC5 = 0; + tempC6 = 0; + tempC7 = 0; + //tempC8 = 0; + //tempC9 = 0; + //tempC10 = 0; + //tempC11 = 0; + //tempC12 = 0; + //tempC13 = 0; + //tempC14 = 0; + //tempC15 = 0; + + for ( i=0; i < lda; i+=2 ) { + iToRow = i << 5; + + tempA0 = A[i + jToRow]; + tempA1 = A[i+1 + jToRow]; + //tempA2 = A[i+2 + jToRow]; + //tempA3 = A[i+3 + jToRow]; + //tempA4 = A[i+4 + jToRow]; + //tempA5 = A[i+5 + jToRow]; + //tempA6 = A[i+6 + jToRow]; + //tempA7 = A[i+7 + jToRow]; + + tempC0 += tempA0 * B[(jToCol ) + (iToRow)]; + tempC1 += tempA0 * B[(jToCol+1 ) + (iToRow)]; + tempC2 += tempA0 * B[(jToCol+2 ) + (iToRow)]; + tempC3 += tempA0 * B[(jToCol+3 ) + (iToRow)]; + tempC4 += tempA0 * B[(jToCol+4 ) + (iToRow)]; + tempC5 += tempA0 * B[(jToCol+5 ) + (iToRow)]; + tempC6 += tempA0 * B[(jToCol+6 ) + (iToRow)]; + tempC7 += tempA0 * B[(jToCol+7 ) + (iToRow)]; + //tempC8 += tempA0 * B[(jToCol+8 ) + (iToRow)]; + //tempC9 += tempA0 * B[(jToCol+9 ) + (iToRow)]; + //tempC10 += tempA0 * B[(jToCol+10) + (iToRow)]; + //tempC11 += tempA0 * B[(jToCol+11) + (iToRow)]; + //tempC12 += tempA0 * B[(jToCol+12) + (iToRow)]; + //tempC13 += tempA0 * B[(jToCol+13) + (iToRow)]; + //tempC14 += tempA0 * B[(jToCol+14) + (iToRow)]; + //tempC15 += tempA0 * B[(jToCol+15) + (iToRow)]; + + iToRow += 32; + tempC0 += tempA1 * B[(jToCol ) + (iToRow)]; + tempC1 += tempA1 * B[(jToCol+1 ) + (iToRow)]; + tempC2 += tempA1 * B[(jToCol+2 ) + (iToRow)]; + tempC3 += tempA1 * B[(jToCol+3 ) + (iToRow)]; + tempC4 += tempA1 * B[(jToCol+4 ) + (iToRow)]; + tempC5 += tempA1 * B[(jToCol+5 ) + (iToRow)]; + tempC6 += tempA1 * B[(jToCol+6 ) + (iToRow)]; + tempC7 += tempA1 * B[(jToCol+7 ) + (iToRow)]; + //tempC8 += tempA1 * B[(jToCol+8 ) + (iToRow+32)]; + //tempC9 += tempA1 * B[(jToCol+9 ) + (iToRow+32)]; + //tempC10 += tempA1 * B[(jToCol+10) + (iToRow+32)]; + //tempC11 += tempA1 * B[(jToCol+11) + (iToRow+32)]; + //tempC12 += tempA1 * B[(jToCol+12) + (iToRow+32)]; + //tempC13 += tempA1 * B[(jToCol+13) + (iToRow+32)]; + //tempC14 += tempA1 * B[(jToCol+14) + (iToRow+32)]; + //tempC15 += tempA1 * B[(jToCol+15) + (iToRow+32)]; + + //iToRow += 32; + //tempC0 += tempA2 * B[(jToCol ) + (iToRow)]; + //tempC1 += tempA2 * B[(jToCol+1 ) + (iToRow)]; + //tempC2 += tempA2 * B[(jToCol+2 ) + (iToRow)]; + //tempC3 += tempA2 * B[(jToCol+3 ) + (iToRow)]; + //tempC4 += tempA2 * B[(jToCol+4 ) + (iToRow)]; + //tempC5 += tempA2 * B[(jToCol+5 ) + (iToRow)]; + //tempC6 += tempA2 * B[(jToCol+6 ) + (iToRow)]; + //tempC7 += tempA2 * B[(jToCol+7 ) + (iToRow)]; + //tempC8 += tempA2 * B[(jToCol+8 ) + (iToRow)]; + //tempC9 += tempA2 * B[(jToCol+9 ) + (iToRow)]; + //tempC10 += tempA2 * B[(jToCol+10) + (iToRow)]; + //tempC11 += tempA2 * B[(jToCol+11) + (iToRow)]; + //tempC12 += tempA2 * B[(jToCol+12) + (iToRow)]; + //tempC13 += tempA2 * B[(jToCol+13) + (iToRow)]; + //tempC14 += tempA2 * B[(jToCol+14) + (iToRow)]; + //tempC15 += tempA2 * B[(jToCol+15) + (iToRow)]; + + //iToRow += 32; + //tempC0 += tempA3 * B[(jToCol ) + (iToRow)]; + //tempC1 += tempA3 * B[(jToCol+1 ) + (iToRow)]; + //tempC2 += tempA3 * B[(jToCol+2 ) + (iToRow)]; + //tempC3 += tempA3 * B[(jToCol+3 ) + (iToRow)]; + //tempC4 += tempA3 * B[(jToCol+4 ) + (iToRow)]; + //tempC5 += tempA3 * B[(jToCol+5 ) + (iToRow)]; + //tempC6 += tempA3 * B[(jToCol+6 ) + (iToRow)]; + //tempC7 += tempA3 * B[(jToCol+7 ) + (iToRow)]; + //tempC8 += tempA3 * B[(jToCol+8 ) + (iToRow)]; + //tempC9 += tempA3 * B[(jToCol+9 ) + (iToRow)]; + //tempC10 += tempA3 * B[(jToCol+10) + (iToRow)]; + //tempC11 += tempA3 * B[(jToCol+11) + (iToRow)]; + //tempC12 += tempA3 * B[(jToCol+12) + (iToRow)]; + //tempC13 += tempA3 * B[(jToCol+13) + (iToRow)]; + //tempC14 += tempA3 * B[(jToCol+14) + (iToRow)]; + //tempC15 += tempA3 * B[(jToCol+15) + (iToRow)]; + + //iToRow += 32; + //tempC0 += tempA4 * B[(jToCol ) + (iToRow)]; + //tempC1 += tempA4 * B[(jToCol+1 ) + (iToRow)]; + //tempC2 += tempA4 * B[(jToCol+2 ) + (iToRow)]; + //tempC3 += tempA4 * B[(jToCol+3 ) + (iToRow)]; + //tempC4 += tempA4 * B[(jToCol+4 ) + (iToRow)]; + //tempC5 += tempA4 * B[(jToCol+5 ) + (iToRow)]; + //tempC6 += tempA4 * B[(jToCol+6 ) + (iToRow)]; + //tempC7 += tempA4 * B[(jToCol+7 ) + (iToRow)]; + // + //iToRow += 32; + //tempC0 += tempA5 * B[(jToCol ) + (iToRow)]; + //tempC1 += tempA5 * B[(jToCol+1 ) + (iToRow)]; + //tempC2 += tempA5 * B[(jToCol+2 ) + (iToRow)]; + //tempC3 += tempA5 * B[(jToCol+3 ) + (iToRow)]; + //tempC4 += tempA5 * B[(jToCol+4 ) + (iToRow)]; + //tempC5 += tempA5 * B[(jToCol+5 ) + (iToRow)]; + //tempC6 += tempA5 * B[(jToCol+6 ) + (iToRow)]; + //tempC7 += tempA5 * B[(jToCol+7 ) + (iToRow)]; + // + //iToRow += 32; + //tempC0 += tempA6 * B[(jToCol ) + (iToRow)]; + //tempC1 += tempA6 * B[(jToCol+1 ) + (iToRow)]; + //tempC2 += tempA6 * B[(jToCol+2 ) + (iToRow)]; + //tempC3 += tempA6 * B[(jToCol+3 ) + (iToRow)]; + //tempC4 += tempA6 * B[(jToCol+4 ) + (iToRow)]; + //tempC5 += tempA6 * B[(jToCol+5 ) + (iToRow)]; + //tempC6 += tempA6 * B[(jToCol+6 ) + (iToRow)]; + //tempC7 += tempA6 * B[(jToCol+7 ) + (iToRow)]; + // + //iToRow += 32; + //tempC0 += tempA7 * B[(jToCol ) + (iToRow)]; + //tempC1 += tempA7 * B[(jToCol+1 ) + (iToRow)]; + //tempC2 += tempA7 * B[(jToCol+2 ) + (iToRow)]; + //tempC3 += tempA7 * B[(jToCol+3 ) + (iToRow)]; + //tempC4 += tempA7 * B[(jToCol+4 ) + (iToRow)]; + //tempC5 += tempA7 * B[(jToCol+5 ) + (iToRow)]; + //tempC6 += tempA7 * B[(jToCol+6 ) + (iToRow)]; + //tempC7 += tempA7 * B[(jToCol+7 ) + (iToRow)]; + + } + C[j ] = tempC0; + C[j + 1 ] = tempC1; + C[j + 2 ] = tempC2; + C[j + 3 ] = tempC3; + C[j + 4 ] = tempC4; + C[j + 5 ] = tempC5; + C[j + 6 ] = tempC6; + C[j + 7 ] = tempC7; + //C[j + 8 ] = tempC8 ; + //C[j + 9 ] = tempC9 ; + //C[j + 10] = tempC10; + //C[j + 11] = tempC11; + //C[j + 12] = tempC12; + //C[j + 13] = tempC13; + //C[j + 14] = tempC14; + //C[j + 15] = tempC15; + } +} + + + +//-------------------------------------------------------------------------- +// 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); +} + |