BS: dos2unix; indent -linux

6 files changed, 632 insertions, 539 deletions

diff --git a/BS/baselines/cpu/bs_omp.c b/BS/baselines/cpu/bs_omp.c
index 871792d..2e4c300 100644
--- a/BS/baselines/cpu/bs_omp.c
+++ b/BS/baselines/cpu/bs_omp.c
@@ -19,275 +19,274 @@
 #include <numaif.h>
 #include <numa.h>
 
-void* mp_pages[1];
+void *mp_pages[1];
 int mp_status[1];
 int mp_nodes[1];
-struct bitmask* bitmask_in;
+struct bitmask *bitmask_in;
 int numa_node_in = -1;
 int numa_node_cpu = -1;
 #endif
 
-
 #if NUMA_MEMCPY
-struct bitmask* bitmask_cpu;
+struct bitmask *bitmask_cpu;
 int numa_node_cpu_memcpy = -1;
 int numa_node_local = -1;
 int numa_node_in_is_local = 0;
 #endif
 
-
 #define DTYPE uint64_t
 /*
 * @brief creates a "test file" by filling a bufferwith values
 */
-void create_test_file(DTYPE * input, uint64_t  nr_elements, DTYPE * querys, uint64_t n_querys) {
+void create_test_file(DTYPE *input, uint64_t nr_elements, DTYPE *querys,
+		      uint64_t n_querys)
+{
 
-  srand(time(NULL));
+	srand(time(NULL));
 
-  input[0] = 1;
-  for (uint64_t i = 1; i < nr_elements; i++) {
-        input[i] = input[i - 1] + (rand() % 10) + 1;
-  }
+	input[0] = 1;
+	for (uint64_t i = 1; i < nr_elements; i++) {
+		input[i] = input[i - 1] + (rand() % 10) + 1;
+	}
 
-  for(uint64_t i = 0; i < n_querys; i++)
-  {
-	querys[i] = input[rand() % nr_elements];
-  }
+	for (uint64_t i = 0; i < n_querys; i++) {
+		querys[i] = input[rand() % nr_elements];
+	}
 }
 
 /**
 * @brief compute output in the host
 */
-uint64_t binarySearch(DTYPE * input, uint64_t input_size, DTYPE* querys, unsigned n_querys)
+uint64_t binarySearch(DTYPE *input, uint64_t input_size, DTYPE *querys,
+		      unsigned n_querys)
 {
 
 	uint64_t found = -1;
 	uint64_t q, r, l, m;
-	
-       #pragma omp parallel for private(q,r,l,m)
-     	for(q = 0; q < n_querys; q++)
-      	{
+
+#pragma omp parallel for private(q,r,l,m)
+	for (q = 0; q < n_querys; q++) {
 		l = 0;
 		r = input_size;
-		while (l <= r) 
-		{
-	    		m = l + (r - l) / 2;
-
-	    		// Check if x is present at mid
-	     		if (input[m] == querys[q])
-			{	
-		    		found += m;
+		while (l <= r) {
+			m = l + (r - l) / 2;
+
+			// Check if x is present at mid
+			if (input[m] == querys[q]) {
+				found += m;
 				break;
 			}
-	    		// If x greater, ignore left half
-	    		if (input[m] < querys[q])
-			    	l = m + 1;
+			// If x greater, ignore left half
+			if (input[m] < querys[q])
+				l = m + 1;
 
-	    		// If x is smaller, ignore right half
+			// If x is smaller, ignore right half
 			else
-		    		r = m - 1;
-		
+				r = m - 1;
+
 		}
-       	}
+	}
 
-      	return found;
+	return found;
 }
 
   /**
   * @brief Main of the Host Application.
   */
-  int main(int argc, char **argv) {
-    (void)argc;
+int main(int argc, char **argv)
+{
+	(void)argc;
 #if WITH_BENCHMARK
-    Timer timer;
+	Timer timer;
 #endif
-    uint64_t input_size = atol(argv[1]);
-    uint64_t n_querys = atol(argv[2]);
+	uint64_t input_size = atol(argv[1]);
+	uint64_t n_querys = atol(argv[2]);
 #if NUMA
-    bitmask_in = numa_parse_nodestring(argv[3]);
-    numa_node_cpu = atoi(argv[4]);
+	bitmask_in = numa_parse_nodestring(argv[3]);
+	numa_node_cpu = atoi(argv[4]);
 #endif
 #if NUMA_MEMCPY
-    bitmask_cpu = numa_parse_nodestring(argv[5]);
-    numa_node_cpu_memcpy = atoi(argv[6]);
+	bitmask_cpu = numa_parse_nodestring(argv[5]);
+	numa_node_cpu_memcpy = atoi(argv[6]);
 #endif
 
-    printf("Vector size: %lu, num searches: %lu\n", input_size, n_querys);
+	printf("Vector size: %lu, num searches: %lu\n", input_size, n_querys);
 
 #if NUMA
-    if (bitmask_in) {
-        numa_set_membind(bitmask_in);
-        numa_free_nodemask(bitmask_in);
-    }
-    DTYPE * input = numa_alloc((input_size) * sizeof(DTYPE));
-    DTYPE * querys = numa_alloc((n_querys) * sizeof(DTYPE));
+	if (bitmask_in) {
+		numa_set_membind(bitmask_in);
+		numa_free_nodemask(bitmask_in);
+	}
+	DTYPE *input = numa_alloc((input_size) * sizeof(DTYPE));
+	DTYPE *querys = numa_alloc((n_querys) * sizeof(DTYPE));
 #else
-    DTYPE * input = malloc((input_size) * sizeof(DTYPE));
-    DTYPE * querys = malloc((n_querys) * sizeof(DTYPE));
+	DTYPE *input = malloc((input_size) * sizeof(DTYPE));
+	DTYPE *querys = malloc((n_querys) * sizeof(DTYPE));
 #endif
 
 #if NUMA
 #if NUMA_MEMCPY
-    if (bitmask_cpu) {
-        numa_set_membind(bitmask_cpu);
-        numa_free_nodemask(bitmask_cpu);
-    }
+	if (bitmask_cpu) {
+		numa_set_membind(bitmask_cpu);
+		numa_free_nodemask(bitmask_cpu);
+	}
 #else
-    struct bitmask *bitmask_all = numa_allocate_nodemask();
-    numa_bitmask_setall(bitmask_all);
-    numa_set_membind(bitmask_all);
-    numa_free_nodemask(bitmask_all);
-#endif // NUMA_MEMCPY
+	struct bitmask *bitmask_all = numa_allocate_nodemask();
+	numa_bitmask_setall(bitmask_all);
+	numa_set_membind(bitmask_all);
+	numa_free_nodemask(bitmask_all);
+#endif				// NUMA_MEMCPY
 #endif
 
-    DTYPE result_host = -1;
+	DTYPE result_host = -1;
 
-    // Create an input file with arbitrary data.
-    create_test_file(input, input_size, querys, n_querys);
+	// Create an input file with arbitrary data.
+	create_test_file(input, input_size, querys, n_querys);
 
 #if NUMA
-    mp_pages[0] = input;
-    if (move_pages(0, 1, mp_pages, NULL, mp_status, 0) == -1) {
-        perror("move_pages(A)");
-    }
-    else if (mp_status[0] < 0) {
-        printf("move_pages error: %d", mp_status[0]);
-    }
-    else {
-        numa_node_in = mp_status[0];
-    }
-
-    if (numa_node_cpu != -1) {
-        if (numa_run_on_node(numa_node_cpu) == -1) {
-            perror("numa_run_on_node");
-            numa_node_cpu = -1;
-        }
-    }
+	mp_pages[0] = input;
+	if (move_pages(0, 1, mp_pages, NULL, mp_status, 0) == -1) {
+		perror("move_pages(A)");
+	} else if (mp_status[0] < 0) {
+		printf("move_pages error: %d", mp_status[0]);
+	} else {
+		numa_node_in = mp_status[0];
+	}
+
+	if (numa_node_cpu != -1) {
+		if (numa_run_on_node(numa_node_cpu) == -1) {
+			perror("numa_run_on_node");
+			numa_node_cpu = -1;
+		}
+	}
 #endif
 
 #if NUMA_MEMCPY
-    numa_node_in_is_local = ((numa_node_cpu == numa_node_in) || (numa_node_cpu + 8 == numa_node_in)) * 1;
+	numa_node_in_is_local = ((numa_node_cpu == numa_node_in)
+				 || (numa_node_cpu + 8 == numa_node_in)) * 1;
 #endif
 
 #if NUMA_MEMCPY
-    DTYPE *input_local = input;
-    DTYPE *querys_local = querys;
-    start(&timer, 1, 0);
-    if (!numa_node_in_is_local) {
-        input_local = numa_alloc((input_size) * sizeof(DTYPE));
-        querys_local = numa_alloc((n_querys) * sizeof(DTYPE));
-    }
-    stop(&timer, 1);
-    if (!numa_node_in_is_local) {
-        if (numa_node_cpu_memcpy != -1) {
-            if (numa_run_on_node(numa_node_cpu_memcpy) == -1) {
-                perror("numa_run_on_node");
-                numa_node_cpu_memcpy = -1;
-            }
-        }
-    }
-    start(&timer, 2, 0);
-    if (!numa_node_in_is_local) {
-        memcpy(input_local, input, input_size * sizeof(DTYPE));
-        memcpy(querys_local, querys, n_querys * sizeof(DTYPE));
-    } else {
-        input_local = input;
-        querys_local = querys;
-    }
-    stop(&timer, 2);
-    if (numa_node_cpu != -1) {
-        if (numa_run_on_node(numa_node_cpu) == -1) {
-            perror("numa_run_on_node");
-            numa_node_cpu = -1;
-        }
-    }
-    mp_pages[0] = input_local;
-    if (move_pages(0, 1, mp_pages, NULL, mp_status, 0) == -1) {
-        perror("move_pages(input_local)");
-    }
-    else if (mp_status[0] < 0) {
-        printf("move_pages error: %d", mp_status[0]);
-    }
-    else {
-        numa_node_local = mp_status[0];
-    }
+	DTYPE *input_local = input;
+	DTYPE *querys_local = querys;
+	start(&timer, 1, 0);
+	if (!numa_node_in_is_local) {
+		input_local = numa_alloc((input_size) * sizeof(DTYPE));
+		querys_local = numa_alloc((n_querys) * sizeof(DTYPE));
+	}
+	stop(&timer, 1);
+	if (!numa_node_in_is_local) {
+		if (numa_node_cpu_memcpy != -1) {
+			if (numa_run_on_node(numa_node_cpu_memcpy) == -1) {
+				perror("numa_run_on_node");
+				numa_node_cpu_memcpy = -1;
+			}
+		}
+	}
+	start(&timer, 2, 0);
+	if (!numa_node_in_is_local) {
+		memcpy(input_local, input, input_size * sizeof(DTYPE));
+		memcpy(querys_local, querys, n_querys * sizeof(DTYPE));
+	} else {
+		input_local = input;
+		querys_local = querys;
+	}
+	stop(&timer, 2);
+	if (numa_node_cpu != -1) {
+		if (numa_run_on_node(numa_node_cpu) == -1) {
+			perror("numa_run_on_node");
+			numa_node_cpu = -1;
+		}
+	}
+	mp_pages[0] = input_local;
+	if (move_pages(0, 1, mp_pages, NULL, mp_status, 0) == -1) {
+		perror("move_pages(input_local)");
+	} else if (mp_status[0] < 0) {
+		printf("move_pages error: %d", mp_status[0]);
+	} else {
+		numa_node_local = mp_status[0];
+	}
 #endif
 
 #if NOP_SYNC
-    for(int rep = 0; rep < 200000; rep++) {
-        asm volatile("nop" ::);
-    }
+	for (int rep = 0; rep < 200000; rep++) {
+		asm volatile ("nop"::);
+	}
 #endif
 
-    start(&timer, 0, 0);
+	start(&timer, 0, 0);
 #if NUMA_MEMCPY
-    result_host = binarySearch(input_local, input_size - 1, querys_local, n_querys);
+	result_host =
+	    binarySearch(input_local, input_size - 1, querys_local, n_querys);
 #else
-    result_host = binarySearch(input, input_size - 1, querys, n_querys);
+	result_host = binarySearch(input, input_size - 1, querys, n_querys);
 #endif
-    stop(&timer, 0);
+	stop(&timer, 0);
 
 #if NOP_SYNC
-    for(int rep = 0; rep < 200000; rep++) {
-        asm volatile("nop" ::);
-    }
+	for (int rep = 0; rep < 200000; rep++) {
+		asm volatile ("nop"::);
+	}
 #endif
 
 #if NUMA_MEMCPY
-    start(&timer, 3, 0);
-    if (!numa_node_in_is_local) {
-        numa_free(input_local, input_size * sizeof(DTYPE));
-        numa_free(querys_local, n_querys * sizeof(DTYPE));
-    }
-    stop(&timer, 3);
+	start(&timer, 3, 0);
+	if (!numa_node_in_is_local) {
+		numa_free(input_local, input_size * sizeof(DTYPE));
+		numa_free(querys_local, n_querys * sizeof(DTYPE));
+	}
+	stop(&timer, 3);
 #endif
 
-    int status = (result_host);
+	int status = (result_host);
 #if WITH_BENCHMARK
-    unsigned int nr_threads = 0;
+	unsigned int nr_threads = 0;
 #pragma omp parallel
 #pragma omp atomic
-    nr_threads++;
+	nr_threads++;
 
-    if (status) {
+	if (status) {
 #if NUMA_MEMCPY
-        printf("[::] BS-CPU-MEMCPY | n_threads=%d e_type=%s n_elements=%lu"
-            " numa_node_in=%d numa_node_local=%d numa_node_cpu=%d numa_node_cpu_memcpy=%d numa_distance_in_cpu=%d"
-            " | throughput_MBps=%f throughput_MOpps=%f"
-            " latency_kernel_us=%f latency_alloc_us=%f latency_memcpy_us=%f latency_free_us=%f latency_total_us=%f\n",
-            nr_threads, "uint64_t", input_size,
-            numa_node_in, numa_node_local, numa_node_cpu, numa_node_cpu_memcpy, numa_distance(numa_node_in, numa_node_cpu),
-            n_querys * sizeof(DTYPE) / timer.time[0], n_querys / timer.time[0],
-            timer.time[0], timer.time[1], timer.time[2], timer.time[3],
-            timer.time[0] + timer.time[1] + timer.time[2] + timer.time[3]);
+		printf
+		    ("[::] BS-CPU-MEMCPY | n_threads=%d e_type=%s n_elements=%lu"
+		     " numa_node_in=%d numa_node_local=%d numa_node_cpu=%d numa_node_cpu_memcpy=%d numa_distance_in_cpu=%d"
+		     " | throughput_MBps=%f throughput_MOpps=%f"
+		     " latency_kernel_us=%f latency_alloc_us=%f latency_memcpy_us=%f latency_free_us=%f latency_total_us=%f\n",
+		     nr_threads, "uint64_t", input_size, numa_node_in,
+		     numa_node_local, numa_node_cpu, numa_node_cpu_memcpy,
+		     numa_distance(numa_node_in, numa_node_cpu),
+		     n_querys * sizeof(DTYPE) / timer.time[0],
+		     n_querys / timer.time[0], timer.time[0], timer.time[1],
+		     timer.time[2], timer.time[3],
+		     timer.time[0] + timer.time[1] + timer.time[2] +
+		     timer.time[3]);
 #else
-        printf("[::] BS-CPU | n_threads=%d e_type=%s n_elements=%lu"
+		printf("[::] BS-CPU | n_threads=%d e_type=%s n_elements=%lu"
 #if NUMA
-            " numa_node_in=%d numa_node_cpu=%d numa_distance_in_cpu=%d"
+		       " numa_node_in=%d numa_node_cpu=%d numa_distance_in_cpu=%d"
 #endif
-            " | throughput_MBps=%f",
-            nr_threads, "uint64_t", input_size,
+		       " | throughput_MBps=%f",
+		       nr_threads, "uint64_t", input_size,
 #if NUMA
-            numa_node_in, numa_node_cpu, numa_distance(numa_node_in, numa_node_cpu),
+		       numa_node_in, numa_node_cpu, numa_distance(numa_node_in,
+								  numa_node_cpu),
 #endif
-            n_querys * sizeof(DTYPE) / timer.time[0]);
-        printf(" throughput_MOpps=%f latency_us=%f\n",
-            n_querys / timer.time[0], timer.time[0]);
+		       n_querys * sizeof(DTYPE) / timer.time[0]);
+		printf(" throughput_MOpps=%f latency_us=%f\n",
+		       n_querys / timer.time[0], timer.time[0]);
 #endif
-    } else {
-        printf("[ERROR]\n");
-    }
-#endif // WITH_BENCHMARK
+	} else {
+		printf("[ERROR]\n");
+	}
+#endif				// WITH_BENCHMARK
 
 #if NUMA
-    numa_free(input, input_size * sizeof(DTYPE));
-    numa_free(querys, n_querys * sizeof(DTYPE));
+	numa_free(input, input_size * sizeof(DTYPE));
+	numa_free(querys, n_querys * sizeof(DTYPE));
 #else
-    free(input);
-    free(querys);
+	free(input);
+	free(querys);
 #endif
 
-
-    return status ? 0 : 1;
+	return status ? 0 : 1;
 }
-
diff --git a/BS/dpu/task.c b/BS/dpu/task.c
index acf66f2..5881dd1 100644
--- a/BS/dpu/task.c
+++ b/BS/dpu/task.c
@@ -17,140 +17,168 @@ __host dpu_arguments_t DPU_INPUT_ARGUMENTS;
 __host dpu_results_t DPU_RESULTS[NR_TASKLETS];
 
 // Search
-static DTYPE search(DTYPE *bufferA, DTYPE searching_for, size_t search_size) {
-  DTYPE found = -2;
-  if(bufferA[0] <= searching_for)
-  {
-    found = -1;
-    for (uint32_t i = 0; i < search_size / sizeof(DTYPE); i++){
-      if(bufferA[i] == searching_for)
-      {
-        found = i;
-        break;
-      }
-    }
-  }
-  return found;
+static DTYPE search(DTYPE *bufferA, DTYPE searching_for, size_t search_size)
+{
+	DTYPE found = -2;
+	if (bufferA[0] <= searching_for) {
+		found = -1;
+		for (uint32_t i = 0; i < search_size / sizeof(DTYPE); i++) {
+			if (bufferA[i] == searching_for) {
+				found = i;
+				break;
+			}
+		}
+	}
+	return found;
 }
 
 BARRIER_INIT(my_barrier, NR_TASKLETS);
 
 extern int main_kernel1(void);
 
-int(*kernels[nr_kernels])(void) = {main_kernel1};
+int (*kernels[nr_kernels])(void) = { main_kernel1 };
 
-int main(void){
-  // Kernel
-  return kernels[DPU_INPUT_ARGUMENTS.kernel]();
+int main(void)
+{
+	// Kernel
+	return kernels[DPU_INPUT_ARGUMENTS.kernel] ();
 }
 
 // main_kernel1
-int main_kernel1() {
-  unsigned int tasklet_id = me();
-  #if PRINT
-  printf("tasklet_id = %u\n", tasklet_id);
-  #endif
-  if(tasklet_id == 0){
-    mem_reset(); // Reset the heap
-  }
-  // Barrier
-  barrier_wait(&my_barrier);
-
-  DTYPE searching_for, found;
-  uint64_t input_size = DPU_INPUT_ARGUMENTS.input_size;
-
-  // Address of the current processing block in MRAM
-  uint32_t start_mram_block_addr_A       = (uint32_t) DPU_MRAM_HEAP_POINTER;
-  uint32_t start_mram_block_addr_aux     = start_mram_block_addr_A;
-  uint32_t end_mram_block_addr_A         = start_mram_block_addr_A + sizeof(DTYPE) * input_size;
-  uint32_t current_mram_block_addr_query = end_mram_block_addr_A + tasklet_id * (DPU_INPUT_ARGUMENTS.slice_per_dpu / NR_TASKLETS) * sizeof(DTYPE);
-
-  // Initialize a local cache to store the MRAM block
-  DTYPE *cache_A     = (DTYPE *) mem_alloc(BLOCK_SIZE);
-  DTYPE *cache_aux_A = (DTYPE *) mem_alloc(BLOCK_SIZE);
-  DTYPE *cache_aux_B = (DTYPE *) mem_alloc(BLOCK_SIZE);
-
-  dpu_results_t *result = &DPU_RESULTS[tasklet_id];
-
-  for(uint64_t targets = 0; targets < (DPU_INPUT_ARGUMENTS.slice_per_dpu / NR_TASKLETS); targets++)
-  {
-    found = -1;
-
-    mram_read((__mram_ptr void const *) current_mram_block_addr_query, &searching_for, 8);
-    current_mram_block_addr_query += 8;
-
-    // Initialize input vector boundaries
-    start_mram_block_addr_A    = (uint32_t) DPU_MRAM_HEAP_POINTER;
-    start_mram_block_addr_aux  = start_mram_block_addr_A;
-    end_mram_block_addr_A      = start_mram_block_addr_A + sizeof(DTYPE) * input_size;
-
-    uint32_t current_mram_block_addr_A = start_mram_block_addr_A;
-
-    // Bring first and last values to WRAM
-    mram_read((__mram_ptr void const *) current_mram_block_addr_A, cache_aux_A, BLOCK_SIZE);
-    mram_read((__mram_ptr void const *) (end_mram_block_addr_A - BLOCK_SIZE * sizeof(DTYPE)),   cache_aux_B, BLOCK_SIZE);
-
-    while(1)
-    {
-      // Locate the address of the mid mram block
-      current_mram_block_addr_A = (start_mram_block_addr_A + end_mram_block_addr_A) / 2;
-      current_mram_block_addr_A &= WORD_MASK;
-      
-      // Boundary check
-      if(current_mram_block_addr_A < (start_mram_block_addr_A + BLOCK_SIZE))
-      {
-	// Search inside (start_mram_block_addr_A, start_mram_block_addr_A + BLOCK_SIZE)
-        mram_read((__mram_ptr void const *) start_mram_block_addr_A, cache_A, BLOCK_SIZE);
-        found = search(cache_A, searching_for, BLOCK_SIZE);
-
-        if(found > -1)
-        {
-          result->found = found + (start_mram_block_addr_A - start_mram_block_addr_aux) / sizeof(DTYPE);
-        }
-	// Search inside (start_mram_block_addr_A + BLOCK_SIZE, end_mram_block_addr_A)
-	else
-	{
-	  size_t remain_bytes_to_search = end_mram_block_addr_A - (start_mram_block_addr_A + BLOCK_SIZE);
-          mram_read((__mram_ptr void const *) start_mram_block_addr_A + BLOCK_SIZE, cache_A, remain_bytes_to_search);
-          found = search(cache_A, searching_for, remain_bytes_to_search);
-	  
-	  if(found > -1)
-          {
-            result->found = found + (start_mram_block_addr_A + BLOCK_SIZE - start_mram_block_addr_aux) / sizeof(DTYPE);
-          }
-	  else
-	  {
-	    printf("%lld NOT found\n", searching_for);
-	  }
+int main_kernel1()
+{
+	unsigned int tasklet_id = me();
+#if PRINT
+	printf("tasklet_id = %u\n", tasklet_id);
+#endif
+	if (tasklet_id == 0) {
+		mem_reset();	// Reset the heap
+	}
+	// Barrier
+	barrier_wait(&my_barrier);
+
+	DTYPE searching_for, found;
+	uint64_t input_size = DPU_INPUT_ARGUMENTS.input_size;
+
+	// Address of the current processing block in MRAM
+	uint32_t start_mram_block_addr_A = (uint32_t) DPU_MRAM_HEAP_POINTER;
+	uint32_t start_mram_block_addr_aux = start_mram_block_addr_A;
+	uint32_t end_mram_block_addr_A =
+	    start_mram_block_addr_A + sizeof(DTYPE) * input_size;
+	uint32_t current_mram_block_addr_query =
+	    end_mram_block_addr_A +
+	    tasklet_id * (DPU_INPUT_ARGUMENTS.slice_per_dpu / NR_TASKLETS) *
+	    sizeof(DTYPE);
+
+	// Initialize a local cache to store the MRAM block
+	DTYPE *cache_A = (DTYPE *) mem_alloc(BLOCK_SIZE);
+	DTYPE *cache_aux_A = (DTYPE *) mem_alloc(BLOCK_SIZE);
+	DTYPE *cache_aux_B = (DTYPE *) mem_alloc(BLOCK_SIZE);
+
+	dpu_results_t *result = &DPU_RESULTS[tasklet_id];
+
+	for (uint64_t targets = 0;
+	     targets < (DPU_INPUT_ARGUMENTS.slice_per_dpu / NR_TASKLETS);
+	     targets++) {
+		found = -1;
+
+		mram_read((__mram_ptr void const *)
+			  current_mram_block_addr_query, &searching_for, 8);
+		current_mram_block_addr_query += 8;
+
+		// Initialize input vector boundaries
+		start_mram_block_addr_A = (uint32_t) DPU_MRAM_HEAP_POINTER;
+		start_mram_block_addr_aux = start_mram_block_addr_A;
+		end_mram_block_addr_A =
+		    start_mram_block_addr_A + sizeof(DTYPE) * input_size;
+
+		uint32_t current_mram_block_addr_A = start_mram_block_addr_A;
+
+		// Bring first and last values to WRAM
+		mram_read((__mram_ptr void const *)current_mram_block_addr_A,
+			  cache_aux_A, BLOCK_SIZE);
+		mram_read((__mram_ptr void const *)(end_mram_block_addr_A -
+						    BLOCK_SIZE * sizeof(DTYPE)),
+			  cache_aux_B, BLOCK_SIZE);
+
+		while (1) {
+			// Locate the address of the mid mram block
+			current_mram_block_addr_A =
+			    (start_mram_block_addr_A +
+			     end_mram_block_addr_A) / 2;
+			current_mram_block_addr_A &= WORD_MASK;
+
+			// Boundary check
+			if (current_mram_block_addr_A <
+			    (start_mram_block_addr_A + BLOCK_SIZE)) {
+				// Search inside (start_mram_block_addr_A, start_mram_block_addr_A + BLOCK_SIZE)
+				mram_read((__mram_ptr void const *)
+					  start_mram_block_addr_A, cache_A,
+					  BLOCK_SIZE);
+				found =
+				    search(cache_A, searching_for, BLOCK_SIZE);
+
+				if (found > -1) {
+					result->found =
+					    found + (start_mram_block_addr_A -
+						     start_mram_block_addr_aux)
+					    / sizeof(DTYPE);
+				}
+				// Search inside (start_mram_block_addr_A + BLOCK_SIZE, end_mram_block_addr_A)
+				else {
+					size_t remain_bytes_to_search =
+					    end_mram_block_addr_A -
+					    (start_mram_block_addr_A +
+					     BLOCK_SIZE);
+					mram_read((__mram_ptr void const *)
+						  start_mram_block_addr_A +
+						  BLOCK_SIZE, cache_A,
+						  remain_bytes_to_search);
+					found =
+					    search(cache_A, searching_for,
+						   remain_bytes_to_search);
+
+					if (found > -1) {
+						result->found =
+						    found +
+						    (start_mram_block_addr_A +
+						     BLOCK_SIZE -
+						     start_mram_block_addr_aux)
+						    / sizeof(DTYPE);
+					} else {
+						printf("%lld NOT found\n",
+						       searching_for);
+					}
+				}
+				break;
+			}
+			// Load cache with current MRAM block
+			mram_read((__mram_ptr void const *)
+				  current_mram_block_addr_A, cache_A,
+				  BLOCK_SIZE);
+
+			// Search inside block
+			found = search(cache_A, searching_for, BLOCK_SIZE);
+
+			// If found > -1, we found the searching_for query
+			if (found > -1) {
+				result->found =
+				    found + (current_mram_block_addr_A -
+					     start_mram_block_addr_aux) /
+				    sizeof(DTYPE);
+				break;
+			}
+			// If found == -2, we need to discard right part of the input vector
+			if (found == -2) {
+				end_mram_block_addr_A =
+				    current_mram_block_addr_A;
+			}
+			// If found == -1, we need to discard left part of the input vector
+			else if (found == -1) {
+				start_mram_block_addr_A =
+				    current_mram_block_addr_A;
+			}
+		}
 	}
-	break;
-      }
-      
-      // Load cache with current MRAM block
-      mram_read((__mram_ptr void const *) current_mram_block_addr_A, cache_A, BLOCK_SIZE);
-
-      // Search inside block
-      found = search(cache_A, searching_for, BLOCK_SIZE);
-
-      // If found > -1, we found the searching_for query
-      if(found > -1)
-      {
-        result->found = found + (current_mram_block_addr_A - start_mram_block_addr_aux) / sizeof(DTYPE);
-        break;
-      }
-
-      // If found == -2, we need to discard right part of the input vector
-      if(found == -2)
-      {
-        end_mram_block_addr_A     = current_mram_block_addr_A;
-      }
-
-      // If found == -1, we need to discard left part of the input vector
-      else if (found == -1)
-      {
-        start_mram_block_addr_A   = current_mram_block_addr_A;
-      }
-    }
-  }
-  return 0;
+	return 0;
 }
diff --git a/BS/host/app.c b/BS/host/app.c
index f267645..217ea99 100644
--- a/BS/host/app.c
+++ b/BS/host/app.c
@@ -31,7 +31,9 @@
 #define DPU_BINARY "./bin/bs_dpu"
 
 // Create input arrays
-void create_test_file(DTYPE * input, DTYPE * querys, uint64_t  nr_elements, uint64_t nr_querys) {
+void create_test_file(DTYPE *input, DTYPE *querys, uint64_t nr_elements,
+		      uint64_t nr_querys)
+{
 
 	srand(time(NULL));
 
@@ -45,12 +47,12 @@ void create_test_file(DTYPE * input, DTYPE * querys, uint64_t  nr_elements, uint
 }
 
 // Compute output in the host
-int64_t binarySearch(DTYPE * input, DTYPE * querys, DTYPE input_size, uint64_t num_querys)
+int64_t binarySearch(DTYPE *input, DTYPE *querys, DTYPE input_size,
+		     uint64_t num_querys)
 {
 	uint64_t result = -1;
 	DTYPE r;
-	for(uint64_t q = 0; q < num_querys; q++)
-	{
+	for (uint64_t q = 0; q < num_querys; q++) {
 		DTYPE l = 0;
 		r = input_size;
 		while (l <= r) {
@@ -59,92 +61,96 @@ int64_t binarySearch(DTYPE * input, DTYPE * querys, DTYPE input_size, uint64_t n
 			// XXX shouldn't this short-circuit?
 			// Check if x is present at mid
 			if (input[m] == querys[q])
-			result = m;
+				result = m;
 
 			// If x greater, ignore left half
 			if (input[m] < querys[q])
-			l = m + 1;
+				l = m + 1;
 
 			// If x is smaller, ignore right half
 			else
-			r = m - 1;
+				r = m - 1;
 		}
 	}
 	return result;
 }
 
-
 // Main of the Host Application
-int main(int argc, char **argv) {
+int main(int argc, char **argv)
+{
 
 	struct Params p = input_params(argc, argv);
 	struct dpu_set_t dpu_set, dpu;
 	uint32_t nr_of_dpus;
-    uint32_t nr_of_ranks;
+	uint32_t nr_of_ranks;
 	uint64_t input_size = INPUT_SIZE;
 	uint64_t num_querys = p.num_querys;
 	DTYPE result_host = -1;
-	DTYPE result_dpu  = -1;
+	DTYPE result_dpu = -1;
 
-    // Timer declaration
-    Timer timer;
+	// Timer declaration
+	Timer timer;
 
-    int numa_node_rank = -2;
+	int numa_node_rank = -2;
 
-    // Allocate DPUs and load binary
+	// Allocate DPUs and load binary
 #if !WITH_ALLOC_OVERHEAD
-    DPU_ASSERT(dpu_alloc(NR_DPUS, NULL, &dpu_set));
-    timer.time[0] = 0; // alloc
+	DPU_ASSERT(dpu_alloc(NR_DPUS, NULL, &dpu_set));
+	timer.time[0] = 0;	// alloc
 #endif
 #if !WITH_LOAD_OVERHEAD
-    DPU_ASSERT(dpu_load(dpu_set, DPU_BINARY, NULL));
-    DPU_ASSERT(dpu_get_nr_dpus(dpu_set, &nr_of_dpus));
-    DPU_ASSERT(dpu_get_nr_ranks(dpu_set, &nr_of_ranks));
-    assert(nr_of_dpus == NR_DPUS);
-    timer.time[1] = 0; // load
+	DPU_ASSERT(dpu_load(dpu_set, DPU_BINARY, NULL));
+	DPU_ASSERT(dpu_get_nr_dpus(dpu_set, &nr_of_dpus));
+	DPU_ASSERT(dpu_get_nr_ranks(dpu_set, &nr_of_ranks));
+	assert(nr_of_dpus == NR_DPUS);
+	timer.time[1] = 0;	// load
 #endif
 #if !WITH_FREE_OVERHEAD
-    timer.time[6] = 0; // free
+	timer.time[6] = 0;	// free
 #endif
 
-	#if ENERGY
+#if ENERGY
 	struct dpu_probe_t probe;
 	DPU_ASSERT(dpu_probe_init("energy_probe", &probe));
-	#endif
+#endif
 
 	// Query number adjustement for proper partitioning
-	if(num_querys % (NR_DPUS * NR_TASKLETS))
-	num_querys = num_querys + (NR_DPUS * NR_TASKLETS - num_querys % (NR_DPUS * NR_TASKLETS));
+	if (num_querys % (NR_DPUS * NR_TASKLETS))
+		num_querys =
+		    num_querys + (NR_DPUS * NR_TASKLETS -
+				  num_querys % (NR_DPUS * NR_TASKLETS));
 
-	assert(num_querys % (NR_DPUS * NR_TASKLETS) == 0 && "Input dimension");    // Allocate input and querys vectors
+	assert(num_querys % (NR_DPUS * NR_TASKLETS) == 0 && "Input dimension");	// Allocate input and querys vectors
 
-	DTYPE * input  = malloc((input_size) * sizeof(DTYPE));
-	DTYPE * querys = malloc((num_querys) * sizeof(DTYPE));
+	DTYPE *input = malloc((input_size) * sizeof(DTYPE));
+	DTYPE *querys = malloc((num_querys) * sizeof(DTYPE));
 
 	// Create an input file with arbitrary data
 	create_test_file(input, querys, input_size, num_querys);
 
 	// Create kernel arguments
-	uint64_t slice_per_dpu          = num_querys / NR_DPUS;
-	dpu_arguments_t input_arguments = {input_size, slice_per_dpu, 0};
+	uint64_t slice_per_dpu = num_querys / NR_DPUS;
+	dpu_arguments_t input_arguments = { input_size, slice_per_dpu, 0 };
 
 	for (unsigned int rep = 0; rep < p.n_warmup + p.n_reps; rep++) {
 		// Perform input transfers
 		uint64_t i = 0;
 
 #if WITH_ALLOC_OVERHEAD
-		if(rep >= p.n_warmup) {
+		if (rep >= p.n_warmup) {
 			start(&timer, 0, 0);
 		}
 		DPU_ASSERT(dpu_alloc(NR_DPUS, NULL, &dpu_set));
-		if(rep >= p.n_warmup) {
+		if (rep >= p.n_warmup) {
 			stop(&timer, 0);
 		}
 #endif
 #if WITH_DPUINFO
 		printf("DPUs:");
-		DPU_FOREACH (dpu_set, dpu) {
-			int rank = dpu_get_rank_id(dpu_get_rank(dpu_from_set(dpu))) & DPU_TARGET_MASK;
+		DPU_FOREACH(dpu_set, dpu) {
+			int rank =
+			    dpu_get_rank_id(dpu_get_rank(dpu_from_set(dpu))) &
+			    DPU_TARGET_MASK;
 			int slice = dpu_get_slice_id(dpu_from_set(dpu));
 			int member = dpu_get_member_id(dpu_from_set(dpu));
 			printf(" %d(%d.%d)", rank, slice, member);
@@ -152,11 +158,11 @@ int main(int argc, char **argv) {
 		printf("\n");
 #endif
 #if WITH_LOAD_OVERHEAD
-		if(rep >= p.n_warmup) {
+		if (rep >= p.n_warmup) {
 			start(&timer, 1, 0);
 		}
 		DPU_ASSERT(dpu_load(dpu_set, DPU_BINARY, NULL));
-		if(rep >= p.n_warmup) {
+		if (rep >= p.n_warmup) {
 			stop(&timer, 1);
 		}
 		DPU_ASSERT(dpu_get_nr_dpus(dpu_set, &nr_of_dpus));
@@ -166,27 +172,35 @@ int main(int argc, char **argv) {
 
 		// int prev_rank_id = -1;
 		int rank_id = -1;
-		DPU_FOREACH (dpu_set, dpu) {
-			rank_id = dpu_get_rank_id(dpu_get_rank(dpu_from_set(dpu))) & DPU_TARGET_MASK;
-			if ((numa_node_rank != -2) && numa_node_rank != dpu_get_rank_numa_node(dpu_get_rank(dpu_from_set(dpu)))) {
+		DPU_FOREACH(dpu_set, dpu) {
+			rank_id =
+			    dpu_get_rank_id(dpu_get_rank(dpu_from_set(dpu))) &
+			    DPU_TARGET_MASK;
+			if ((numa_node_rank != -2)
+			    && numa_node_rank !=
+			    dpu_get_rank_numa_node(dpu_get_rank
+						   (dpu_from_set(dpu)))) {
 				numa_node_rank = -1;
 			} else {
-				numa_node_rank = dpu_get_rank_numa_node(dpu_get_rank(dpu_from_set(dpu)));
+				numa_node_rank =
+				    dpu_get_rank_numa_node(dpu_get_rank
+							   (dpu_from_set(dpu)));
 			}
 			/*
-			if (rank_id != prev_rank_id) {
-				printf("/dev/dpu_rank%d @ NUMA node %d\n", rank_id, numa_node_rank);
-				prev_rank_id = rank_id;
-			}
-			*/
+			   if (rank_id != prev_rank_id) {
+			   printf("/dev/dpu_rank%d @ NUMA node %d\n", rank_id, numa_node_rank);
+			   prev_rank_id = rank_id;
+			   }
+			 */
 		}
 
 		// Compute host solution
-		if(rep >= p.n_warmup) {
+		if (rep >= p.n_warmup) {
 			start(&timer, 2, 0);
 		}
-		result_host = binarySearch(input, querys, input_size - 1, num_querys);
-		if(rep >= p.n_warmup) {
+		result_host =
+		    binarySearch(input, querys, input_size - 1, num_querys);
+		if (rep >= p.n_warmup) {
 			stop(&timer, 2);
 		}
 
@@ -194,103 +208,110 @@ int main(int argc, char **argv) {
 			start(&timer, 3, 0);
 		}
 
-		DPU_FOREACH(dpu_set, dpu, i)
-		{
+		DPU_FOREACH(dpu_set, dpu, i) {
 			DPU_ASSERT(dpu_prepare_xfer(dpu, &input_arguments));
 		}
 
-		DPU_ASSERT(dpu_push_xfer(dpu_set, DPU_XFER_TO_DPU, "DPU_INPUT_ARGUMENTS", 0, sizeof(input_arguments), DPU_XFER_DEFAULT));
+		DPU_ASSERT(dpu_push_xfer
+			   (dpu_set, DPU_XFER_TO_DPU, "DPU_INPUT_ARGUMENTS", 0,
+			    sizeof(input_arguments), DPU_XFER_DEFAULT));
 
 		i = 0;
 
-		DPU_FOREACH(dpu_set, dpu, i)
-		{
+		DPU_FOREACH(dpu_set, dpu, i) {
 			DPU_ASSERT(dpu_prepare_xfer(dpu, input));
 		}
 
-		DPU_ASSERT(dpu_push_xfer(dpu_set, DPU_XFER_TO_DPU, DPU_MRAM_HEAP_POINTER_NAME, 0, input_size * sizeof(DTYPE), DPU_XFER_DEFAULT));
+		DPU_ASSERT(dpu_push_xfer
+			   (dpu_set, DPU_XFER_TO_DPU,
+			    DPU_MRAM_HEAP_POINTER_NAME, 0,
+			    input_size * sizeof(DTYPE), DPU_XFER_DEFAULT));
 
 		i = 0;
 
-		DPU_FOREACH(dpu_set, dpu, i)
-		{
-			DPU_ASSERT(dpu_prepare_xfer(dpu, querys + slice_per_dpu * i));
+		DPU_FOREACH(dpu_set, dpu, i) {
+			DPU_ASSERT(dpu_prepare_xfer
+				   (dpu, querys + slice_per_dpu * i));
 		}
 
-		DPU_ASSERT(dpu_push_xfer(dpu_set, DPU_XFER_TO_DPU, DPU_MRAM_HEAP_POINTER_NAME, input_size * sizeof(DTYPE), slice_per_dpu * sizeof(DTYPE), DPU_XFER_DEFAULT));
+		DPU_ASSERT(dpu_push_xfer
+			   (dpu_set, DPU_XFER_TO_DPU,
+			    DPU_MRAM_HEAP_POINTER_NAME,
+			    input_size * sizeof(DTYPE),
+			    slice_per_dpu * sizeof(DTYPE), DPU_XFER_DEFAULT));
 
 		if (rep >= p.n_warmup) {
 			stop(&timer, 3);
 		}
-
 		// Run kernel on DPUs
-		if (rep >= p.n_warmup)
-		{
+		if (rep >= p.n_warmup) {
 			start(&timer, 4, 0);
-			#if ENERGY
+#if ENERGY
 			DPU_ASSERT(dpu_probe_start(&probe));
-			#endif
+#endif
 		}
 
 		DPU_ASSERT(dpu_launch(dpu_set, DPU_SYNCHRONOUS));
 
-		if (rep >= p.n_warmup)
-		{
+		if (rep >= p.n_warmup) {
 			stop(&timer, 4);
-			#if ENERGY
+#if ENERGY
 			DPU_ASSERT(dpu_probe_stop(&probe));
-			#endif
+#endif
 		}
 		// Print logs if required
-		#if PRINT
+#if PRINT
 		unsigned int each_dpu = 0;
 		printf("Display DPU Logs\n");
-		DPU_FOREACH(dpu_set, dpu)
-		{
+		DPU_FOREACH(dpu_set, dpu) {
 			printf("DPU#%d:\n", each_dpu);
 			DPU_ASSERT(dpulog_read_for_dpu(dpu.dpu, stdout));
 			each_dpu++;
 		}
-		#endif
+#endif
 
 		// Retrieve results
-		dpu_results_t* results_retrieve[NR_DPUS];
+		dpu_results_t *results_retrieve[NR_DPUS];
 		if (rep >= p.n_warmup) {
 			start(&timer, 5, 0);
 		}
 		i = 0;
-		DPU_FOREACH(dpu_set, dpu, i)
-		{
-			results_retrieve[i] = (dpu_results_t*)malloc(NR_TASKLETS * sizeof(dpu_results_t));
+		DPU_FOREACH(dpu_set, dpu, i) {
+			results_retrieve[i] =
+			    (dpu_results_t *) malloc(NR_TASKLETS *
+						     sizeof(dpu_results_t));
 			DPU_ASSERT(dpu_prepare_xfer(dpu, results_retrieve[i]));
 		}
 
-		DPU_ASSERT(dpu_push_xfer(dpu_set, DPU_XFER_FROM_DPU, "DPU_RESULTS", 0, NR_TASKLETS * sizeof(dpu_results_t), DPU_XFER_DEFAULT));
-
-		DPU_FOREACH(dpu_set, dpu, i)
-		{
-			for(unsigned int each_tasklet = 0; each_tasklet < NR_TASKLETS; each_tasklet++)
-			{
-				if(results_retrieve[i][each_tasklet].found > result_dpu)
-				{
-					result_dpu = results_retrieve[i][each_tasklet].found;
+		DPU_ASSERT(dpu_push_xfer
+			   (dpu_set, DPU_XFER_FROM_DPU, "DPU_RESULTS", 0,
+			    NR_TASKLETS * sizeof(dpu_results_t),
+			    DPU_XFER_DEFAULT));
+
+		DPU_FOREACH(dpu_set, dpu, i) {
+			for (unsigned int each_tasklet = 0;
+			     each_tasklet < NR_TASKLETS; each_tasklet++) {
+				if (results_retrieve[i][each_tasklet].found >
+				    result_dpu) {
+					result_dpu =
+					    results_retrieve[i][each_tasklet].
+					    found;
 				}
 			}
 			free(results_retrieve[i]);
 		}
-		if(rep >= p.n_warmup) {
+		if (rep >= p.n_warmup) {
 			stop(&timer, 5);
 		}
-
 #if WITH_ALLOC_OVERHEAD
 #if WITH_FREE_OVERHEAD
-		if(rep >= p.n_warmup) {
+		if (rep >= p.n_warmup) {
 			start(&timer, 6, 0);
 		}
 #endif
 		DPU_ASSERT(dpu_free(dpu_set));
 #if WITH_FREE_OVERHEAD
-		if(rep >= p.n_warmup) {
+		if (rep >= p.n_warmup) {
 			stop(&timer, 6);
 		}
 #endif
@@ -298,58 +319,91 @@ int main(int argc, char **argv) {
 
 		int status = (result_dpu == result_host);
 		if (status) {
-			printf("[" ANSI_COLOR_GREEN "OK" ANSI_COLOR_RESET "] results are equal\n");
+			printf("[" ANSI_COLOR_GREEN "OK" ANSI_COLOR_RESET
+			       "] results are equal\n");
 			if (rep >= p.n_warmup) {
-				printf("[::] BS-UPMEM | n_dpus=%d n_ranks=%d n_tasklets=%d e_type=%s block_size_B=%d n_elements=%lu",
-					NR_DPUS, nr_of_ranks, NR_TASKLETS, XSTR(DTYPE), BLOCK_SIZE, input_size);
-				printf(" b_with_alloc_overhead=%d b_with_load_overhead=%d b_with_free_overhead=%d numa_node_rank=%d ",
-					WITH_ALLOC_OVERHEAD, WITH_LOAD_OVERHEAD, WITH_FREE_OVERHEAD, numa_node_rank);
-				printf("| latency_alloc_us=%f latency_load_us=%f latency_cpu_us=%f latency_write_us=%f latency_kernel_us=%f latency_read_us=%f latency_free_us=%f",
-					timer.time[0],
-					timer.time[1],
-					timer.time[2],
-					timer.time[3],
-					timer.time[4],
-					timer.time[5],
-					timer.time[6]);
-				printf(" throughput_cpu_MBps=%f throughput_upmem_kernel_MBps=%f throughput_upmem_total_MBps=%f",
-					num_querys * sizeof(DTYPE) / timer.time[2],
-					num_querys * sizeof(DTYPE) / (timer.time[4]),
-					num_querys * sizeof(DTYPE) / (timer.time[0] + timer.time[1] + timer.time[3] + timer.time[4] + timer.time[5] + timer.time[6]));
-				printf(" throughput_upmem_wxr_MBps=%f throughput_upmem_lwxr_MBps=%f throughput_upmem_alwxr_MBps=%f",
-					num_querys * sizeof(DTYPE) / (timer.time[3] + timer.time[4] + timer.time[5]),
-					num_querys * sizeof(DTYPE) / (timer.time[1] + timer.time[3] + timer.time[4] + timer.time[5]),
-					num_querys * sizeof(DTYPE) / (timer.time[0] + timer.time[1] + timer.time[3] + timer.time[4] + timer.time[5]));
-				printf(" throughput_cpu_MOpps=%f throughput_upmem_kernel_MOpps=%f throughput_upmem_total_MOpps=%f",
-					num_querys / timer.time[2],
-					num_querys / (timer.time[4]),
-					num_querys / (timer.time[0] + timer.time[1] + timer.time[3] + timer.time[4] + timer.time[5] + timer.time[6]));
-				printf(" throughput_upmem_wxr_MOpps=%f throughput_upmem_lwxr_MOpps=%f throughput_upmem_alwxr_MOpps=%f\n",
-					num_querys / (timer.time[3] + timer.time[4] + timer.time[5]),
-					num_querys / (timer.time[1] + timer.time[3] + timer.time[4] + timer.time[5]),
-					num_querys / (timer.time[0] + timer.time[1] + timer.time[3] + timer.time[4] + timer.time[5]));
+				printf
+				    ("[::] BS-UPMEM | n_dpus=%d n_ranks=%d n_tasklets=%d e_type=%s block_size_B=%d n_elements=%lu",
+				     NR_DPUS, nr_of_ranks, NR_TASKLETS,
+				     XSTR(DTYPE), BLOCK_SIZE, input_size);
+				printf
+				    (" b_with_alloc_overhead=%d b_with_load_overhead=%d b_with_free_overhead=%d numa_node_rank=%d ",
+				     WITH_ALLOC_OVERHEAD, WITH_LOAD_OVERHEAD,
+				     WITH_FREE_OVERHEAD, numa_node_rank);
+				printf
+				    ("| latency_alloc_us=%f latency_load_us=%f latency_cpu_us=%f latency_write_us=%f latency_kernel_us=%f latency_read_us=%f latency_free_us=%f",
+				     timer.time[0], timer.time[1],
+				     timer.time[2], timer.time[3],
+				     timer.time[4], timer.time[5],
+				     timer.time[6]);
+				printf
+				    (" throughput_cpu_MBps=%f throughput_upmem_kernel_MBps=%f throughput_upmem_total_MBps=%f",
+				     num_querys * sizeof(DTYPE) / timer.time[2],
+				     num_querys * sizeof(DTYPE) /
+				     (timer.time[4]),
+				     num_querys * sizeof(DTYPE) /
+				     (timer.time[0] + timer.time[1] +
+				      timer.time[3] + timer.time[4] +
+				      timer.time[5] + timer.time[6]));
+				printf
+				    (" throughput_upmem_wxr_MBps=%f throughput_upmem_lwxr_MBps=%f throughput_upmem_alwxr_MBps=%f",
+				     num_querys * sizeof(DTYPE) /
+				     (timer.time[3] + timer.time[4] +
+				      timer.time[5]),
+				     num_querys * sizeof(DTYPE) /
+				     (timer.time[1] + timer.time[3] +
+				      timer.time[4] + timer.time[5]),
+				     num_querys * sizeof(DTYPE) /
+				     (timer.time[0] + timer.time[1] +
+				      timer.time[3] + timer.time[4] +
+				      timer.time[5]));
+				printf
+				    (" throughput_cpu_MOpps=%f throughput_upmem_kernel_MOpps=%f throughput_upmem_total_MOpps=%f",
+				     num_querys / timer.time[2],
+				     num_querys / (timer.time[4]),
+				     num_querys / (timer.time[0] +
+						   timer.time[1] +
+						   timer.time[3] +
+						   timer.time[4] +
+						   timer.time[5] +
+						   timer.time[6]));
+				printf
+				    (" throughput_upmem_wxr_MOpps=%f throughput_upmem_lwxr_MOpps=%f throughput_upmem_alwxr_MOpps=%f\n",
+				     num_querys / (timer.time[3] +
+						   timer.time[4] +
+						   timer.time[5]),
+				     num_querys / (timer.time[1] +
+						   timer.time[3] +
+						   timer.time[4] +
+						   timer.time[5]),
+				     num_querys / (timer.time[0] +
+						   timer.time[1] +
+						   timer.time[3] +
+						   timer.time[4] +
+						   timer.time[5]));
 			}
 		} else {
-			printf("[" ANSI_COLOR_RED "ERROR" ANSI_COLOR_RESET "] results differ!\n");
+			printf("[" ANSI_COLOR_RED "ERROR" ANSI_COLOR_RESET
+			       "] results differ!\n");
 		}
 	}
 	// Print timing results
 	/*
-	printf("CPU Version Time (ms): ");
-	print(&timer, 0, p.n_reps);
-	printf("CPU-DPU Time (ms): ");
-	print(&timer, 1, p.n_reps);
-	printf("DPU Kernel Time (ms): ");
-	print(&timer, 2, p.n_reps);
-	printf("DPU-CPU Time (ms): ");
-	print(&timer, 3, p.n_reps);
-	*/
-
-	#if ENERGY
+	   printf("CPU Version Time (ms): ");
+	   print(&timer, 0, p.n_reps);
+	   printf("CPU-DPU Time (ms): ");
+	   print(&timer, 1, p.n_reps);
+	   printf("DPU Kernel Time (ms): ");
+	   print(&timer, 2, p.n_reps);
+	   printf("DPU-CPU Time (ms): ");
+	   print(&timer, 3, p.n_reps);
+	 */
+
+#if ENERGY
 	double energy;
 	DPU_ASSERT(dpu_probe_get(&probe, DPU_ENERGY, DPU_AVERAGE, &energy));
 	printf("DPU Energy (J): %f\t", energy * num_iterations);
-	#endif
+#endif
 
 	free(input);
 #if !WITH_ALLOC_OVERHEAD
diff --git a/BS/support/common.h b/BS/support/common.h
index dbd050c..54adc39 100755
--- a/BS/support/common.h
+++ b/BS/support/common.h
@@ -38,7 +38,7 @@ typedef struct {
 
 // Structures used by both the host and the dpu to communicate information
 typedef struct {
-    DTYPE found;
+	DTYPE found;
 } dpu_results_t;
 
 #ifndef ENERGY
diff --git a/BS/support/params.h b/BS/support/params.h
index 02bd750..c91202f 100644
--- a/BS/support/params.h
+++ b/BS/support/params.h
@@ -4,49 +4,56 @@
 #include "common.h"
 
 typedef struct Params {
-  long  num_querys;
-  unsigned   n_warmup;
-  unsigned   n_reps;
-}Params;
+	long num_querys;
+	unsigned n_warmup;
+	unsigned n_reps;
+} Params;
 
-void usage() {
-  fprintf(stderr,
-    "\nUsage:  ./program [options]"
-    "\n"
-    "\nGeneral options:"
-    "\n    -h        help"
-    "\n    -w <W>    # of untimed warmup iterations (default=1)"
-    "\n    -e <E>    # of timed repetition iterations (default=3)"
-    "\n"
-    "\nBenchmark-specific options:"
-    "\n    -i <I>    problem size (default=2 queries)"
-    "\n");
-  }
+void usage()
+{
+	fprintf(stderr,
+		"\nUsage:  ./program [options]"
+		"\n"
+		"\nGeneral options:"
+		"\n    -h        help"
+		"\n    -w <W>    # of untimed warmup iterations (default=1)"
+		"\n    -e <E>    # of timed repetition iterations (default=3)"
+		"\n"
+		"\nBenchmark-specific options:"
+		"\n    -i <I>    problem size (default=2 queries)" "\n");
+}
 
-  struct Params input_params(int argc, char **argv) {
-    struct Params p;
-    p.num_querys    = PROBLEM_SIZE;
-    p.n_warmup      = 1;
-    p.n_reps        = 3;
+struct Params input_params(int argc, char **argv)
+{
+	struct Params p;
+	p.num_querys = PROBLEM_SIZE;
+	p.n_warmup = 1;
+	p.n_reps = 3;
 
-    int opt;
-    while((opt = getopt(argc, argv, "h:i:w:e:")) >= 0) {
-      switch(opt) {
-        case 'h':
-        usage();
-        exit(0);
-        break;
-        case 'i': p.num_querys    = atol(optarg); break;
-        case 'w': p.n_warmup      = atoi(optarg); break;
-        case 'e': p.n_reps        = atoi(optarg); break; 
-	default:
-        	fprintf(stderr, "\nUnrecognized option!\n");
-        	usage();
-        	exit(0);
-      }
-    }
-    assert(NR_DPUS > 0 && "Invalid # of dpus!");
+	int opt;
+	while ((opt = getopt(argc, argv, "h:i:w:e:")) >= 0) {
+		switch (opt) {
+		case 'h':
+			usage();
+			exit(0);
+			break;
+		case 'i':
+			p.num_querys = atol(optarg);
+			break;
+		case 'w':
+			p.n_warmup = atoi(optarg);
+			break;
+		case 'e':
+			p.n_reps = atoi(optarg);
+			break;
+		default:
+			fprintf(stderr, "\nUnrecognized option!\n");
+			usage();
+			exit(0);
+		}
+	}
+	assert(NR_DPUS > 0 && "Invalid # of dpus!");
 
-    return p;
-  }
-  #endif
+	return p;
+}
+#endif
diff --git a/BS/support/timer.h b/BS/support/timer.h
index ff1ae1b..256447a 100755
--- a/BS/support/timer.h
+++ b/BS/support/timer.h
@@ -1,66 +1,71 @@
-/*

- * Copyright (c) 2016 University of Cordoba and University of Illinois

- * All rights reserved.

- *

- * Developed by:    IMPACT Research Group

- *                  University of Cordoba and University of Illinois

- *                  http://impact.crhc.illinois.edu/

- *

- * Permission is hereby granted, free of charge, to any person obtaining a copy

- * of this software and associated documentation files (the "Software"), to deal

- * with the Software without restriction, including without limitation the 

- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or

- * sell copies of the Software, and to permit persons to whom the Software is

- * furnished to do so, subject to the following conditions:

- *

- *      > Redistributions of source code must retain the above copyright notice,

- *        this list of conditions and the following disclaimers.

- *      > Redistributions in binary form must reproduce the above copyright

- *        notice, this list of conditions and the following disclaimers in the

- *        documentation and/or other materials provided with the distribution.

- *      > Neither the names of IMPACT Research Group, University of Cordoba, 

- *        University of Illinois nor the names of its contributors may be used 

- *        to endorse or promote products derived from this Software without 

- *        specific prior written permission.

- *

- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR

- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,

- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE 

- * CONTRIBUTORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER

- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,

- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS WITH

- * THE SOFTWARE.

- *

- */

-

-#include <sys/time.h>

-

-typedef struct Timer{

-

-    struct timeval startTime[7];

-    struct timeval stopTime[7];

-    double         time[7];

-

-}Timer;

-

-void start(Timer *timer, int i, int rep) {

-    if(rep == 0) {

-        timer->time[i] = 0.0;

-    }

-    gettimeofday(&timer->startTime[i], NULL);

-}

-

-void stop(Timer *timer, int i) {

-    gettimeofday(&timer->stopTime[i], NULL);

-    timer->time[i] += (timer->stopTime[i].tv_sec - timer->startTime[i].tv_sec) * 1000000.0 +

-                      (timer->stopTime[i].tv_usec - timer->startTime[i].tv_usec);

-}

-

-void print(Timer *timer, int i, int REP) { printf("%f\t", timer->time[i] / (1000 * REP)); }

-

-void printall(Timer *timer, int maxt) {

-    for (int i = 0; i <= maxt; i++) {

-        printf(" timer%d_us=%f", i, timer->time[i]);

-    }

-    printf("\n");

-}

+/*
+ * Copyright (c) 2016 University of Cordoba and University of Illinois
+ * All rights reserved.
+ *
+ * Developed by:    IMPACT Research Group
+ *                  University of Cordoba and University of Illinois
+ *                  http://impact.crhc.illinois.edu/
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * with the Software without restriction, including without limitation the 
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ *      > Redistributions of source code must retain the above copyright notice,
+ *        this list of conditions and the following disclaimers.
+ *      > Redistributions in binary form must reproduce the above copyright
+ *        notice, this list of conditions and the following disclaimers in the
+ *        documentation and/or other materials provided with the distribution.
+ *      > Neither the names of IMPACT Research Group, University of Cordoba, 
+ *        University of Illinois nor the names of its contributors may be used 
+ *        to endorse or promote products derived from this Software without 
+ *        specific prior written permission.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE 
+ * CONTRIBUTORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS WITH
+ * THE SOFTWARE.
+ *
+ */
+
+#include <sys/time.h>
+typedef struct Timer {
+	struct timeval startTime[7];
+	struct timeval stopTime[7];
+	double time[7];
+} Timer;
+
+void start(Timer *timer, int i, int rep)
+{
+	if (rep == 0) {
+		timer->time[i] = 0.0;
+	}
+	gettimeofday(&timer->startTime[i], NULL);
+}
+
+void stop(Timer *timer, int i)
+{
+	gettimeofday(&timer->stopTime[i], NULL);
+	timer->time[i] +=
+	    (timer->stopTime[i].tv_sec -
+	     timer->startTime[i].tv_sec) * 1000000.0 +
+	    (timer->stopTime[i].tv_usec - timer->startTime[i].tv_usec);
+}
+
+void print(Timer *timer, int i, int REP)
+{
+	printf("%f\t", timer->time[i] / (1000 * REP));
+}
+
+void printall(Timer *timer, int maxt)
+{
+	for (int i = 0; i <= maxt; i++) {
+		printf(" timer%d_us=%f", i, timer->time[i]);
+	}
+	printf("\n");
+}