Updating benchmark data

Author: audictheprogrammer

src/data/benchmark_blocks_BLAS.txt

@@ -1,3 +1,5 @@
-256 0.104650 
-512 0.873165 
-1024 7.320097 
+256 0.060490 
+512 0.512852 
+1024 4.203045 
+2048 34.561330 
+4096 291.356990 

src/data/benchmark_blocks_BLAS_MP.txt

@@ -1,3 +1,5 @@
-256 0.133424 
-512 0.738715 
-1024 4.116122 
+256 0.020453 
+512 0.236462 
+1024 1.788613 
+2048 15.762803 
+4096 130.438630 

src/data/benchmark_blocks_NoBLAS.txt

@@ -1,3 +1,5 @@
-256 0.601513 
-512 2.010825 
-1024 14.243098 
+256 0.125000 
+512 0.953125 
+1024 7.640625 
+2048 60.562500 
+4096 485.718750 

src/data/benchmark_float_18.txt

@@ -1,3 +1,5 @@
-256 0.004334 
-512 0.019533 
-1024 0.147355 
+256 0.001228 
+512 0.007793 
+1024 0.052957 
+2048 0.389164 
+4096 3.035921 

src/data/benchmark_float_20.txt

@@ -1,3 +1,5 @@
-256 0.002996 
-512 0.019488 
-1024 0.141876 
+256 0.001223 
+512 0.007750 
+1024 0.053096 
+2048 0.387539 
+4096 3.042942 

src/data/benchmark_float_22.txt

@@ -1,3 +1,5 @@
-256 0.004537 
-512 0.019813 
-1024 0.161059 
+256 0.001764 
+512 0.008584 
+1024 0.061115 
+2048 0.470926 
+4096 3.803679 

src/data/benchmark_float_24.txt

@@ -1,3 +1,5 @@
-256 0.010669 
-512 0.074771 
-1024 0.699697 
+256 0.004415 
+512 0.035783 
+1024 0.286922 
+2048 2.353680 
+4096 19.387153 

src/data/benchmark_float_26.txt

@@ -1,3 +1,5 @@
-256 0.131005 
-512 1.299020 
-1024 8.850115 
+256 0.056815 
+512 0.506669 
+1024 4.147431 
+2048 34.695722 
+4096 290.457733 

src/data/benchmark_integer_18.txt

@@ -1,3 +1,5 @@
-256 0.092843 
-512 0.759280 
-1024 5.759017 
+256 0.060725 
+512 0.480752 
+1024 3.813639 
+2048 30.555814 
+4096 245.128241 

src/data/benchmark_integer_20.txt

@@ -1,3 +1,5 @@
-256 0.093085 
-512 0.780829 
-1024 5.916821 
+256 0.060548 
+512 0.476654 
+1024 3.803149 
+2048 30.443846 
+4096 245.004002 

src/data/benchmark_integer_22.txt

@@ -1,3 +1,5 @@
-256 0.090962 
-512 0.711311 
-1024 5.932683 
+256 0.060411 
+512 0.478618 
+1024 3.801294 
+2048 30.433349 
+4096 245.126253 

src/data/benchmark_integer_24.txt

@@ -1,3 +1,5 @@
-256 0.089326 
-512 0.716516 
-1024 5.863130 
+256 0.068369 
+512 0.478048 
+1024 3.829007 
+2048 30.668390 
+4096 245.311900 

src/data/benchmark_integer_26.txt

@@ -1,3 +1,5 @@
-256 0.090416 
-512 0.737959 
-1024 5.879954 
+256 0.060966 
+512 0.481131 
+1024 3.843690 
+2048 30.489705 
+4096 245.108636 

src/data/benchmark_modulo_Barrett.txt

@@ -1,3 +1,4 @@
-256 0.349905
-512 2.884986
-1024 21.453475
+256 0.218750 
+512 1.687500 
+1024 13.531250 
+2048 107.671875 

src/data/benchmark_modulo_MP_Barrett.txt

@@ -1,3 +1,4 @@
-256 0.060192 
-512 0.509851 
-1024 3.235947 
+256 0.277262 
+512 1.320997 
+1024 6.976223 
+2048 40.347728 

src/data/benchmark_modulo_MP_SIMD1.txt

@@ -1,3 +1,4 @@
-256 0.055740 
-512 0.544654 
-1024 4.393463 
+256 0.300370 
+512 1.281565 
+1024 6.697673 
+2048 37.871269 

src/data/benchmark_modulo_MP_SIMD2.txt

@@ -1,3 +1,4 @@
-256 0.054218 
-512 0.442351 
-1024 3.072420 
+256 0.306009 
+512 1.242683 
+1024 7.175216 
+2048 38.146906 

src/data/benchmark_modulo_MP_SIMD3.txt

@@ -1,3 +1,4 @@
-256 0.090017 
-512 0.436566 
-1024 2.908950 
+256 0.348979 
+512 1.307635 
+1024 6.600367 
+2048 38.324097 

src/data/benchmark_modulo_MP_naive.txt

@@ -1,3 +1,4 @@
-256 0.090896 
-512 0.444606 
-1024 3.990551 
+256 0.285025 
+512 1.408056 
+1024 7.688562 
+2048 47.684510 

src/data/benchmark_modulo_SIMD1.txt

@@ -1,3 +1,4 @@
-256 0.284792
-512 2.405989
-1024 18.936896
+256 0.171875 
+512 1.515625 
+1024 12.140625 
+2048 91.812500 

src/data/benchmark_modulo_SIMD2.txt

@@ -1,3 +1,4 @@
-256 0.283364
-512 2.401201
-1024 17.964999
+256 0.187500 
+512 1.468750 
+1024 11.625000 
+2048 87.187500 

src/data/benchmark_modulo_SIMD3.txt

@@ -1,3 +1,4 @@
-256 0.331526
-512 2.526606
-1024 18.769032
+256 0.171875 
+512 1.500000 
+1024 12.046875 
+2048 90.187500 

src/data/benchmark_modulo_naive.txt

@@ -1,3 +1,4 @@
-256 0.222980
-512 1.881384
-1024 14.705631
+256 0.265625 
+512 2.078125 
+1024 16.687500 
+2048 133.937500 

src/data/benchmark_order_ijk.txt

@@ -1,3 +1,4 @@
-256 0.344155 
-512 1.089599 
-1024 37.425532 
+256 0.546875 
+512 0.937500 
+1024 14.156250 
+2048 194.296875 

src/data/benchmark_order_ikj.txt

@@ -1,3 +1,4 @@
-256 0.116009 
-512 0.877172 
-1024 7.293895 
+256 0.062500 
+512 0.500000 
+1024 4.015625 
+2048 32.359375 

src/data/benchmark_order_jik.txt

@@ -1,3 +1,4 @@
-256 0.123963 
-512 1.067085 
-1024 37.076965 
+256 0.109375 
+512 0.906250 
+1024 8.671875 
+2048 123.671875 

src/data/benchmark_order_jki.txt

@@ -1,3 +1,4 @@
-256 0.145821 
-512 1.343724 
-1024 61.732275 
+256 0.125000 
+512 1.062500 
+1024 15.375000 
+2048 220.890625 

src/data/benchmark_order_kij.txt

@@ -1,3 +1,4 @@
-256 0.107567 
-512 0.913795 
-1024 7.402726 
+256 0.062500 
+512 0.500000 
+1024 4.078125 
+2048 32.390625 

src/data/benchmark_order_kji.txt

@@ -1,3 +1,4 @@
-256 0.168071 
-512 1.326327 
-1024 62.548309 
+256 0.125000 
+512 1.187500 
+1024 14.171875 
+2048 240.890625 

src/main.c

@@ -305,7 +305,7 @@ void benchmark_loops_order(double p){
     The most efficient one is IKJ.
     */
     int m = 1;  // Executes m times each algo
-    for (int i=8; i<11; i++){
+    for (int i=8; i<12; i++){
         int n = (int) pow(2, i);
         double sum_ijk = 0;
         double sum_kij = 0;
@@ -340,10 +340,10 @@ void benchmark_loops_order(double p){
 
 void benchmark_modulos(double p, double u, double u_overline, double u_b, u_int32_t s, u_int32_t t){
     /* Benchmarking different modulos.
-    The most efficient one SIMD2.
+    The 3 SIMD are the most efficient.
     */
     int m = 1;  // Executes m times each algo
-    for (int i=8; i<11; i++){
+    for (int i=8; i<12; i++){
         int n = (int) pow(2, i);
         double sum_mod_naive = 0;
         double sum_mod_SIMD1 = 0;
@@ -376,10 +376,9 @@ void benchmark_modulos(double p, double u, double u_overline, double u_b, u_int3
 
 void benchmark_modulos_MP(double p, double u, double u_overline, double u_b, u_int32_t s, u_int32_t t){
     /* Benchmarking different modulos.
-    The most efficient one SIMD2.
     */
     int m = 1;  // Executes m times each algo
-    for (int i=8; i<11; i++){
+    for (int i=8; i<12; i++){
         int n = (int) pow(2, i);
 
         double sum_mod_MP_naive = 0;
@@ -416,7 +415,8 @@ void benchmark_blocks(double p, double u_overline){
     /* Benchmarking different modulos.
     */
     int m = 1;  // Executes m times each algo
-    for (int i=8; i<11; i++){
+    openblas_set_num_threads(1);  // 8 is slower than 1.
+    for (int i=8; i<13; i++){
         int n = (int) pow(2, i);
         int b = get_blocksize(get_bitsize(p), n);
 
@@ -444,8 +444,9 @@ void benchmark_blocks(double p, double u_overline){
 void benchmark_blocks_MP(double p, double u_overline){
     /* Benchmarking different modulos.
     */
-    int m = 3;  // Executes m times each algo
-    for (int i=8; i<11; i++){
+    int m = 1;  // Executes m times each algo
+    openblas_set_num_threads(1);
+    for (int i=8; i<13; i++){
         int n = (int) pow(2, i);
         int b = get_blocksize(get_bitsize(p), n);
 
@@ -472,7 +473,7 @@ void benchmark_float_integer(double p, double u_overline, double u_b, u_int32_t
     */
     int m = 1;  // Executes m times each algo
     openblas_set_num_threads(1);
-    for (int i=8; i<11; i++){
+    for (int i=8; i<13; i++){
         int n = (int) pow(2, i);
         int b = get_blocksize(get_bitsize(p), n);
 
@@ -580,42 +581,45 @@ void clean_file_float_integer(){
 int main(){
     // Initialization
     srand(time(NULL));
-    // double p = pow(2, 26) - 5;
+    double p = pow(2, 26) - 5;
     // double p = pow(2, 24) - 3;
     // double p = pow(2, 22) - 3;
     // double p = pow(2, 20) - 3;
-    double p = pow(2, 18) - 5;
+    // double p = pow(2, 18) - 5;
 
     // Precomputed constants for Modular functions
     double u = 1.0 / p;  // Constant for SIMD
     fesetround(FE_UPWARD);
     double u_overline = 1.0 / p;  // Constant for SIMD2 and SIMD3
     fesetround(FE_TONEAREST);
-    // Constants for Barrett which is the integer version of SIMD
+    // Constants for Barrett reductionm, which is the integer version of SIMD
     u_int32_t t = 32;
     u_int32_t s = 30 + get_bitsize(p) - t;
     u_int32_t u_b = (int) (pow(2, s+t) / p);
 
 
-    // Benchmarking order of loop.
-    // clean_file_loops();
-    // benchmark_loops_order(p);
+    /* Benchmarking the order of loop. */
+    clean_file_loops();
+    benchmark_loops_order(p);
 
-    // Benchmarking different modulos.
-    // clean_file_modulos();
-    // benchmark_modulos(p, u, u_overline, u_b);
+    /* Benchmarking different modulo techniques. */
+    clean_file_modulos();
+    benchmark_modulos(p, u, u_overline, u_b, s, t);
 
-    // Benchmarking different modulos with OpenMP
-    // clean_file_modulos_MP();
-    // benchmark_modulos_MP(p, u, u_overline, u_b);
+    /* Benchmarking parallelism apporach with previous modulo techniques. */
+    clean_file_modulos_MP();
+    benchmark_modulos_MP(p, u, u_overline, u_b, s, t);
 
-    // Benchmarking blocks.
-    // clean_file_blocks();
-    // benchmark_blocks(p, u_overline);
+    /* Benchmarking block product approach. */
+    clean_file_blocks();
+    benchmark_blocks(p, u_overline);
 
-    // clean_file_blocks_MP();
-    // benchmark_blocks_MP(p, u_overline);
+    /* Benchmarking the final mixed implementation. */
+    clean_file_blocks_MP();
+    benchmark_blocks_MP(p, u_overline);
 
+    /* Benchmarking the progress made during this internship and work completed
+    last year, using the same environment: 1 thread only, same A, B and p. */
     clean_file_float_integer();
     double P[5];
     P[0] = pow(2, 26) - 5;