BS: nits

3 files changed, 13 insertions, 9 deletions

diff --git a/BS/dimes-hetsim-hbm.sh b/BS/dimes-hetsim-hbm.sh
index b4a1d6d..c5a3a85 100755
--- a/BS/dimes-hetsim-hbm.sh
+++ b/BS/dimes-hetsim-hbm.sh
@@ -1,5 +1,7 @@
 #!/bin/bash
 
+echo PrIM BS
+
 cd baselines/cpu
 make -B NUMA=1
 
@@ -20,7 +22,7 @@ num_queries_hbm=16777216
 
 run_benchmark() {
 	local "$@"
-	OMP_NUM_THREADS=${nr_threads} ./bs_omp ${input_size} ${num_queries} $ram $cpu
+	OMP_NUM_THREADS=${nr_threads} ./bs_omp ${input_size} ${num_queries} $ram $cpu 2>&1
 	return $?
 }
 
@@ -53,7 +55,7 @@ echo "multi-node execution, DPU ref (3/4)" >&2
 parallel -j1 --eta --joblog ${fn}.3.joblog --resume --header : \
 	run_benchmark i={i} nr_threads={nr_threads} ram={ram} cpu={cpu} \
 	input_size=${input_size_dpu} num_queries=${num_queries_dpu} \
-	::: i $(seq 1 5) \
+	::: i $(seq 1 40) \
 	::: nr_threads 32 48 64 96 128 \
 	::: cpu -1 \
 	::: ram $(seq 0 15)
@@ -63,7 +65,7 @@ echo "multi-node execution, HBM ref (4/4)" >&2
 parallel -j1 --eta --joblog ${fn}.4.joblog --resume --header : \
 	run_benchmark i={i} nr_threads={nr_threads} ram={ram} cpu={cpu} \
 	input_size=${input_size_hbm} num_queries=${num_queries_hbm} \
-	::: i $(seq 1 5) \
+	::: i $(seq 1 40) \
 	::: nr_threads 32 48 64 96 128 \
 	::: cpu -1 \
 	::: ram $(seq 0 15)
diff --git a/BS/dimes-hetsim-nmc.sh b/BS/dimes-hetsim-nmc.sh
index 7a2b12a..a14eded 100755
--- a/BS/dimes-hetsim-nmc.sh
+++ b/BS/dimes-hetsim-nmc.sh
@@ -1,5 +1,7 @@
 #!/bin/bash
 
+echo PrIM BS
+
 mkdir -p log/$(hostname) baselines/cpu/log/$(hostname)
 fn=log/$(hostname)/$(date +%Y%m%d)
 
@@ -15,7 +17,7 @@ run_benchmark_nmc() {
 	local "$@"
 	sudo limit_ranks_to_numa_node ${numa_rank}
 	if make -B NR_DPUS=${nr_dpus} NR_TASKLETS=${nr_tasklets} WITH_ALLOC_OVERHEAD=1 WITH_LOAD_OVERHEAD=1 WITH_FREE_OVERHEAD=1 INPUT_SIZE=${input_size} PROBLEM_SIZE=${num_queries}; then
-		bin/bs_host -w 0 -e 100
+		bin/bs_host -w 0 -e 100 2>&1
 	fi
 	return $?
 }
@@ -24,7 +26,7 @@ export -f run_benchmark_nmc
 
 run_benchmark_baseline() {
 	local "$@"
-	OMP_NUM_THREADS=${nr_threads} ./bs_omp ${input_size} ${num_queries} ${ram} ${cpu}
+	OMP_NUM_THREADS=${nr_threads} ./bs_omp ${input_size} ${num_queries} ${ram} ${cpu} 2>&1
 	return $?
 }
 
@@ -42,7 +44,7 @@ parallel -j1 --eta --joblog ${fn}.1.joblog --resume --header : \
 
 echo "NMC multi-node upstream-ref (2/4)" >&2
 
-parallel -j1 --eta --joblog ${fn}.1.joblog --resume --header : \
+parallel -j1 --eta --joblog ${fn}.2.joblog --resume --header : \
 	run_benchmark_nmc nr_dpus={nr_dpus} nr_tasklets=16 numa_rank={numa_rank} \
 	num_queries=${num_queries_upstream} input_size=${input_size_upstream} \
 	::: numa_rank -1 \
@@ -50,7 +52,7 @@ parallel -j1 --eta --joblog ${fn}.1.joblog --resume --header : \
 
 echo "NMC single-node DPU-ref (3/4)" >&2
 
-parallel -j1 --eta --joblog ${fn}.2.joblog --resume --header : \
+parallel -j1 --eta --joblog ${fn}.3.joblog --resume --header : \
 	run_benchmark_nmc nr_dpus={nr_dpus} nr_tasklets=16 numa_rank={numa_rank} \
 	num_queries=${num_queries_dpu} input_size=${input_size_dpu} \
 	::: numa_rank 0 1 \
@@ -58,7 +60,7 @@ parallel -j1 --eta --joblog ${fn}.2.joblog --resume --header : \
 
 echo "NMC multi-node DPU-ref (4/4)" >&2
 
-parallel -j1 --eta --joblog ${fn}.2.joblog --resume --header : \
+parallel -j1 --eta --joblog ${fn}.4.joblog --resume --header : \
 	run_benchmark_nmc nr_dpus={nr_dpus} nr_tasklets=16 numa_rank={numa_rank} \
 	num_queries=${num_queries_dpu} input_size=${input_size_dpu} \
 	::: numa_rank -1 \
diff --git a/BS/host/app.c b/BS/host/app.c
index 85eb883..10d76f1 100644
--- a/BS/host/app.c
+++ b/BS/host/app.c
@@ -300,7 +300,7 @@ int main(int argc, char **argv) {
 			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(" with_alloc_overhead=%d b_with_load_overhead=%d b_with_free_overhead=%d numa_node_rank=%d ",
+				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],