add failed example of loop unrolling

makefile

@@ -4,6 +4,9 @@ MAKEFLAGS += --no-print-directory
 STYLE_BLUEPRINT=webkit
 FORMATTER=clang-format -i -style=$(STYLE_BLUEPRINT)
 
+## Time to botec
+TTB=./examples/more/12_time_to_botec_parallel/example
+
 build-examples:
 	cd examples/core && make all
 	cd examples/more && make all
@@ -25,3 +28,6 @@ profile:
 	sudo perf report
 	rm perf.data
 
+time-linux: 
+	@echo "Running 100x and taking avg time: $(TTB)"
+	@t=$$(/usr/bin/time -f "%e" -p bash -c 'for i in {1..100}; do $(TTB); done' 2>&1 >/dev/null | grep real | awk '{print $$2}' ); echo "scale=2; 1000 * $$t / 100" | bc | sed "s|^|Time using 16 threads: |" | sed 's|$$|ms|' && echo

squiggle_more.c

@@ -64,6 +64,8 @@ void sampler_parallel(double (*sampler)(uint64_t* seed), double* results, int n_
     {
 #pragma omp for
         for (i = 0; i < n_threads; i++) {
+            // Simple version 
+            /*
             int quotient = n_samples / n_threads;
             int lower_bound_inclusive = i * quotient;
             int upper_bound_not_inclusive = ((i + 1) * quotient); // note the < in the for loop below,
@@ -71,6 +73,34 @@ void sampler_parallel(double (*sampler)(uint64_t* seed), double* results, int n_
                 results[j] = sampler(&(cache_box[i].seed));
                 // Could also result in inefficient cache stuff, but hopefully not too often
             }
+            */
+            
+            // Version with loop unrolling
+            int quotient = n_samples / n_threads;
+            int lower_bound_inclusive = i * quotient;
+            int upper_bound_not_inclusive = ((i + 1) * quotient); // note the < in the for loop below,
+            int delta = quotient;
+            int eighth_of_deltas = delta/8; // why 8? a double in 8 bytes, 8 doubles is the size of a cache line
+            int k;
+            // to do: simplify these variables. Maybe divide by n_threads * 8 directly
+            for(int j=0; j<eighth_of_deltas; j++){
+                k = lower_bound_inclusive + j*8;
+                results[k+0] = sampler(&(cache_box[i].seed));
+                results[k+1] = sampler(&(cache_box[i].seed));
+                results[k+2] = sampler(&(cache_box[i].seed));
+                results[k+3] = sampler(&(cache_box[i].seed));
+                results[k+4] = sampler(&(cache_box[i].seed));
+                results[k+5] = sampler(&(cache_box[i].seed));
+                results[k+6] = sampler(&(cache_box[i].seed));
+                results[k+7] = sampler(&(cache_box[i].seed));
+                // these all fit one single cache line
+                // name of the techique: loop unrolling.
+            }
+            for(int k=eighth_of_deltas*8; k<upper_bound_not_inclusive; k++){
+                results[k] = sampler(&(cache_box[i].seed));
+
+            }
+            
         }
     }
     for (int j = divisor_multiple; j < n_samples; j++) {