frontier scripts

scripts/run_1.sh

@@ -20,49 +20,27 @@ export HSA_XNACK=1
 mkdir -p $SCOPE_RESULTS
 module list > $SCOPE_RESULTS/modules.r1.$SLURM_JOBID.txt 2>&1
 env > $SCOPE_RESULTS/env.r1.$SLURM_JOBID.txt
+rocm-smi > $SCOPE_RESULTS/rocm-smi.r1.$SLURM_JOBID.txt 2>&1
+lscpu > $SCOPE_RESULTS/lscpu.r1.$SLURM_JOBID.txt 2>&1
 
 date
 
 srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
 --benchmark_repetitions=5 \
---benchmark_filter='.*hipMemcpyAsync_GPUToPinned/0/.*' \
+--benchmark_filter='.*hipMemcpyAsync_GPUToPinned/0/0/.*' \
 --benchmark_out_format=json \
 --benchmark_out="$SCOPE_RESULTS/hipMemcpyAsync_GPUToPinned.json"
 
 srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
 --benchmark_repetitions=5 \
---benchmark_filter='.*hipMemcpyAsync_PinnedToGPU/0/.*' \
+--benchmark_filter='.*hipMemcpyAsync_PinnedToGPU/0/0/.*' \
 --benchmark_out_format=json \
 --benchmark_out="$SCOPE_RESULTS/hipMemcpyAsync_PinnedToGPU"
 
 srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
 --benchmark_repetitions=5 \
---benchmark_filter='.*hipMemcpyAsync_GPUToGPU/0/.*' \
---benchmark_out_format=json \
---benchmark_out="$SCOPE_RESULTS/hipMemcpyAsync_GPUToGPU.json"
-
-srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
---benchmark_repetitions=5 \
---benchmark_filter='.*hipMemcpyAsync_GPUToPageable/0/.*' \
+--benchmark_filter='.*hipMemcpyAsync_GPUToPageable/0/0/.*' \
 --benchmark_out_format=json \
 --benchmark_out="$SCOPE_RESULTS/hipMemcpyAsync_GPUToPageable.json"
 
-srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
---benchmark_repetitions=5 \
---benchmark_filter='.*hipMemcpyAsync_PageableToGPU/0/.*' \
---benchmark_out_format=json \
---benchmark_out="$SCOPE_RESULTS/hipMemcpyAsync_PageableToGPU.json"
-
-srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
---benchmark_repetitions=5 \
---benchmark_filter='.*prefetch_managed_GPUToHost/0/.*' \
---benchmark_out_format=json \
---benchmark_out="$SCOPE_RESULTS/prefetch_managed_GPUToHost.json"
-
-srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
---benchmark_repetitions=5 \
---benchmark_filter='.*prefetch_managed_HostToGPU/0/.*' \
---benchmark_out_format=json \
---benchmark_out="$SCOPE_RESULTS/prefetch_managed_HostToGPU.json"
-
 date

scripts/run_10.sh

@@ -0,0 +1,34 @@
+#!/bin/bash
+
+#SBATCH -A CSC465
+#SBATCH -J r10
+#SBATCH -o %x-%j.out
+#SBATCH -t 2:00:00
+#SBATCH -p batch
+#SBATCH -N 1
+
+# salloc -A CSC465 -N 1 -p batch -t 2:00:00
+
+export ROOT=/lustre/orion/csc465/scratch/cpearson/frontier-gpu-bandwidth
+export SCOPE_SRC=${ROOT}/comm_scope
+export SCOPE_BUILD=${ROOT}/build
+export SCOPE_RESULTS=${ROOT}/run
+
+module load PrgEnv-amd/8.3.3
+export HSA_XNACK=1
+
+mkdir -p $SCOPE_RESULTS
+module list > $SCOPE_RESULTS/modules.r10.$SLURM_JOBID.txt 2>&1
+env > $SCOPE_RESULTS/env.r10.$SLURM_JOBID.txt
+rocm-smi > $SCOPE_RESULTS/rocm-smi.r10.$SLURM_JOBID.txt 2>&1
+lscpu > $SCOPE_RESULTS/lscpu.r10.$SLURM_JOBID.txt 2>&1
+
+date
+
+srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
+--benchmark_repetitions=5 \
+--benchmark_filter='.*hipMemcpyAsync_GPUToGPU/0/.*' \
+--benchmark_out_format=json \
+--benchmark_out="$SCOPE_RESULTS/hipMemcpyAsync_GPUToGPU.json"
+
+date

scripts/run_2.sh

@@ -20,49 +20,15 @@ export HSA_XNACK=1
 mkdir -p $SCOPE_RESULTS
 module list > $SCOPE_RESULTS/modules.r2.$SLURM_JOBID.txt 2>&1
 env > $SCOPE_RESULTS/env.r2.$SLURM_JOBID.txt
+rocm-smi > $SCOPE_RESULTS/rocm-smi.r2.$SLURM_JOBID.txt 2>&1
+lscpu > $SCOPE_RESULTS/lscpu.r2.$SLURM_JOBID.txt 2>&1
 
 date
 
 srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
 --benchmark_repetitions=5 \
---benchmark_filter='.*implicit_mapped_GPURdHost/0/.*' \
+--benchmark_filter='.*implicit_mapped_GPURdHost/0/0/.*' \
 --benchmark_out_format=json \
 --benchmark_out="$SCOPE_RESULTS/implicit_mapped_GPURdHost.json"
 
-srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
---benchmark_repetitions=5 \
---benchmark_filter='.*implicit_managed_HostWrGPU_fine/0/.*' \
---benchmark_out_format=json \
---benchmark_out="$SCOPE_RESULTS/implicit_managed_HostWrGPU_fine.json"
-
-srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
---benchmark_repetitions=5 \
---benchmark_filter='.*implicit_managed_HostWrGPU_coarse/0/.*' \
---benchmark_out_format=json \
---benchmark_out="$SCOPE_RESULTS/implicit_managed_HostWrGPU_coarse.json"
-
-srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
---benchmark_repetitions=5 \
---benchmark_filter='.*implicit_managed_GPUWrHost_fine/0/.*' \
---benchmark_out_format=json \
---benchmark_out="$SCOPE_RESULTS/implicit_managed_GPUWrHost_fine.json"
-
-srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
---benchmark_repetitions=5 \
---benchmark_filter='.*implicit_managed_GPUWrHost_coarse/0/.*' \
---benchmark_out_format=json \
---benchmark_out="$SCOPE_RESULTS/implicit_managed_GPUWrHost_coarse.json"
-
-srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
---benchmark_repetitions=5 \
---benchmark_filter='.*implicit_mapped_GPUWrHost/0/.*' \
---benchmark_out_format=json \
---benchmark_out="$SCOPE_RESULTS/implicit_mapped_GPUWrHost.json"
-
-srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
---benchmark_repetitions=5 \
---benchmark_filter='.*implicit_mapped_GPUWrGPU/0/.*' \
---benchmark_out_format=json \
---benchmark_out="$SCOPE_RESULTS/implicit_mapped_GPUWrGPU.json"
-
 date

scripts/run_3.sh

@@ -20,6 +20,8 @@ export HSA_XNACK=1
 mkdir -p $SCOPE_RESULTS
 module list > $SCOPE_RESULTS/modules.r3.$SLURM_JOBID.txt 2>&1
 env > $SCOPE_RESULTS/env.r3.$SLURM_JOBID.txt
+rocm-smi > $SCOPE_RESULTS/rocm-smi.r3.$SLURM_JOBID.txt 2>&1
+lscpu > $SCOPE_RESULTS/lscpu.r3.$SLURM_JOBID.txt 2>&1
 
 date
 
@@ -37,32 +39,20 @@ srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
 
 srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
 --benchmark_repetitions=5 \
---benchmark_filter='.*implicit_managed_GPURdHost_fine/0/.*' \
+--benchmark_filter='.*implicit_managed_GPURdHost_fine/0/0/.*' \
 --benchmark_out_format=json \
 --benchmark_out="$SCOPE_RESULTS/implicit_managed_GPURdHost_fine.json"
 
 srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
 --benchmark_repetitions=5 \
---benchmark_filter='.*implicit_managed_GPURdHost_coarse/0/.*' \
+--benchmark_filter='.*implicit_managed_GPURdHost_coarse/0/0/.*' \
 --benchmark_out_format=json \
 --benchmark_out="$SCOPE_RESULTS/implicit_managed_GPURdHost_coarse.json"
 
 srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
 --benchmark_repetitions=5 \
---benchmark_filter='.*hipManaged_HostToGPUWriteDst/0/.*' \
+--benchmark_filter='.*hipManaged_HostToGPUWriteDst/0/0/.*' \
 --benchmark_out_format=json \
 --benchmark_out="$SCOPE_RESULTS/hipManaged_HostToGPUWriteDst.json"
 
-srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
---benchmark_repetitions=5 \
---benchmark_filter='.*prefetch_managed_GPUToGPU/0/.*' \
---benchmark_out_format=json \
---benchmark_out="$SCOPE_RESULTS/prefetch_managed_GPUToGPU.json"
-
-srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
---benchmark_repetitions=5 \
---benchmark_filter='.*implicit_mapped_HostWrGPU/0/.*' \
---benchmark_out_format=json \
---benchmark_out="$SCOPE_RESULTS/implicit_mapped_HostWrGPU.json"
-
 date

scripts/run_4.sh

@@ -0,0 +1,34 @@
+#!/bin/bash
+
+#SBATCH -A CSC465
+#SBATCH -J r4
+#SBATCH -o %x-%j.out
+#SBATCH -t 2:00:00
+#SBATCH -p batch
+#SBATCH -N 1
+
+# salloc -A CSC465 -N 1 -p batch -t 2:00:00
+
+export ROOT=/lustre/orion/csc465/scratch/cpearson/frontier-gpu-bandwidth
+export SCOPE_SRC=${ROOT}/comm_scope
+export SCOPE_BUILD=${ROOT}/build
+export SCOPE_RESULTS=${ROOT}/run
+
+module load PrgEnv-amd/8.3.3
+export HSA_XNACK=1
+
+mkdir -p $SCOPE_RESULTS
+module list > $SCOPE_RESULTS/modules.r4.$SLURM_JOBID.txt 2>&1
+env > $SCOPE_RESULTS/env.r4.$SLURM_JOBID.txt
+rocm-smi > $SCOPE_RESULTS/rocm-smi.r4.$SLURM_JOBID.txt 2>&1
+lscpu > $SCOPE_RESULTS/lscpu.r4.$SLURM_JOBID.txt 2>&1
+
+date
+
+srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
+--benchmark_repetitions=5 \
+--benchmark_filter='.*hipMemcpyAsync_PageableToGPU/0/0/.*' \
+--benchmark_out_format=json \
+--benchmark_out="$SCOPE_RESULTS/hipMemcpyAsync_PageableToGPU.json"
+
+date

scripts/run_5.sh

@@ -0,0 +1,58 @@
+#!/bin/bash
+
+#SBATCH -A CSC465
+#SBATCH -J r5
+#SBATCH -o %x-%j.out
+#SBATCH -t 2:00:00
+#SBATCH -p batch
+#SBATCH -N 1
+
+# salloc -A CSC465 -N 1 -p batch -t 2:00:00
+
+export ROOT=/lustre/orion/csc465/scratch/cpearson/frontier-gpu-bandwidth
+export SCOPE_SRC=${ROOT}/comm_scope
+export SCOPE_BUILD=${ROOT}/build
+export SCOPE_RESULTS=${ROOT}/run
+
+module load PrgEnv-amd/8.3.3
+export HSA_XNACK=1
+
+mkdir -p $SCOPE_RESULTS
+module list > $SCOPE_RESULTS/modules.r5.$SLURM_JOBID.txt 2>&1
+env > $SCOPE_RESULTS/env.r5.$SLURM_JOBID.txt
+rocm-smi > $SCOPE_RESULTS/rocm-smi.r5.$SLURM_JOBID.txt 2>&1
+lscpu > $SCOPE_RESULTS/lscpu.r5.$SLURM_JOBID.txt 2>&1
+
+date
+
+srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
+--benchmark_repetitions=5 \
+--benchmark_filter='.*implicit_managed_HostWrGPU_fine/0/0/.*' \
+--benchmark_out_format=json \
+--benchmark_out="$SCOPE_RESULTS/implicit_managed_HostWrGPU_fine.json"
+
+srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
+--benchmark_repetitions=5 \
+--benchmark_filter='.*implicit_managed_HostWrGPU_coarse/0/0/.*' \
+--benchmark_out_format=json \
+--benchmark_out="$SCOPE_RESULTS/implicit_managed_HostWrGPU_coarse.json"
+
+srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
+--benchmark_repetitions=5 \
+--benchmark_filter='.*implicit_managed_GPUWrHost_fine/0/0/.*' \
+--benchmark_out_format=json \
+--benchmark_out="$SCOPE_RESULTS/implicit_managed_GPUWrHost_fine.json"
+
+srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
+--benchmark_repetitions=5 \
+--benchmark_filter='.*implicit_managed_GPUWrHost_coarse/0/0/.*' \
+--benchmark_out_format=json \
+--benchmark_out="$SCOPE_RESULTS/implicit_managed_GPUWrHost_coarse.json"
+
+srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
+--benchmark_repetitions=5 \
+--benchmark_filter='.*implicit_mapped_GPUWrHost/0/0/.*' \
+--benchmark_out_format=json \
+--benchmark_out="$SCOPE_RESULTS/implicit_mapped_GPUWrHost.json"
+
+date

scripts/run_6.sh

@@ -0,0 +1,34 @@
+#!/bin/bash
+
+#SBATCH -A CSC465
+#SBATCH -J r6
+#SBATCH -o %x-%j.out
+#SBATCH -t 2:00:00
+#SBATCH -p batch
+#SBATCH -N 1
+
+# salloc -A CSC465 -N 1 -p batch -t 2:00:00
+
+export ROOT=/lustre/orion/csc465/scratch/cpearson/frontier-gpu-bandwidth
+export SCOPE_SRC=${ROOT}/comm_scope
+export SCOPE_BUILD=${ROOT}/build
+export SCOPE_RESULTS=${ROOT}/run
+
+module load PrgEnv-amd/8.3.3
+export HSA_XNACK=1
+
+mkdir -p $SCOPE_RESULTS
+module list > $SCOPE_RESULTS/modules.r6.$SLURM_JOBID.txt 2>&1
+env > $SCOPE_RESULTS/env.r6.$SLURM_JOBID.txt
+rocm-smi > $SCOPE_RESULTS/rocm-smi.r6.$SLURM_JOBID.txt 2>&1
+lscpu > $SCOPE_RESULTS/lscpu.r6.$SLURM_JOBID.txt 2>&1
+
+date
+
+srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
+--benchmark_repetitions=5 \
+--benchmark_filter='.*prefetch_managed_GPUToGPU/0/.*' \
+--benchmark_out_format=json \
+--benchmark_out="$SCOPE_RESULTS/prefetch_managed_GPUToGPU.json"
+
+date

scripts/run_7.sh

@@ -0,0 +1,40 @@
+#!/bin/bash
+
+#SBATCH -A CSC465
+#SBATCH -J r7
+#SBATCH -o %x-%j.out
+#SBATCH -t 2:00:00
+#SBATCH -p batch
+#SBATCH -N 1
+
+# salloc -A CSC465 -N 1 -p batch -t 2:00:00
+
+export ROOT=/lustre/orion/csc465/scratch/cpearson/frontier-gpu-bandwidth
+export SCOPE_SRC=${ROOT}/comm_scope
+export SCOPE_BUILD=${ROOT}/build
+export SCOPE_RESULTS=${ROOT}/run
+
+module load PrgEnv-amd/8.3.3
+export HSA_XNACK=1
+
+mkdir -p $SCOPE_RESULTS
+module list > $SCOPE_RESULTS/modules.r7.$SLURM_JOBID.txt 2>&1
+env > $SCOPE_RESULTS/env.r7.$SLURM_JOBID.txt
+rocm-smi > $SCOPE_RESULTS/rocm-smi.r7.$SLURM_JOBID.txt 2>&1
+lscpu > $SCOPE_RESULTS/lscpu.r7.$SLURM_JOBID.txt 2>&1
+
+date
+
+srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
+--benchmark_repetitions=5 \
+--benchmark_filter='.*prefetch_managed_GPUToHost/0/0/.*' \
+--benchmark_out_format=json \
+--benchmark_out="$SCOPE_RESULTS/prefetch_managed_GPUToHost.json"
+
+srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
+--benchmark_repetitions=5 \
+--benchmark_filter='.*prefetch_managed_HostToGPU/0/0/.*' \
+--benchmark_out_format=json \
+--benchmark_out="$SCOPE_RESULTS/prefetch_managed_HostToGPU.json"
+
+date

scripts/run_8.sh

@@ -0,0 +1,34 @@
+#!/bin/bash
+
+#SBATCH -A CSC465
+#SBATCH -J r8
+#SBATCH -o %x-%j.out
+#SBATCH -t 2:00:00
+#SBATCH -p batch
+#SBATCH -N 1
+
+# salloc -A CSC465 -N 1 -p batch -t 2:00:00
+
+export ROOT=/lustre/orion/csc465/scratch/cpearson/frontier-gpu-bandwidth
+export SCOPE_SRC=${ROOT}/comm_scope
+export SCOPE_BUILD=${ROOT}/build
+export SCOPE_RESULTS=${ROOT}/run
+
+module load PrgEnv-amd/8.3.3
+export HSA_XNACK=1
+
+mkdir -p $SCOPE_RESULTS
+module list > $SCOPE_RESULTS/modules.r8.$SLURM_JOBID.txt 2>&1
+env > $SCOPE_RESULTS/env.r8.$SLURM_JOBID.txt
+rocm-smi > $SCOPE_RESULTS/rocm-smi.r8.$SLURM_JOBID.txt 2>&1
+lscpu > $SCOPE_RESULTS/lscpu.r8.$SLURM_JOBID.txt 2>&1
+
+date
+
+srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
+--benchmark_repetitions=5 \
+--benchmark_filter='.*implicit_mapped_HostWrGPU/0/.*' \
+--benchmark_out_format=json \
+--benchmark_out="$SCOPE_RESULTS/implicit_mapped_HostWrGPU.json"
+
+date

scripts/run_9.sh

@@ -0,0 +1,34 @@
+#!/bin/bash
+
+#SBATCH -A CSC465
+#SBATCH -J r9
+#SBATCH -o %x-%j.out
+#SBATCH -t 2:00:00
+#SBATCH -p batch
+#SBATCH -N 1
+
+# salloc -A CSC465 -N 1 -p batch -t 2:00:00
+
+export ROOT=/lustre/orion/csc465/scratch/cpearson/frontier-gpu-bandwidth
+export SCOPE_SRC=${ROOT}/comm_scope
+export SCOPE_BUILD=${ROOT}/build
+export SCOPE_RESULTS=${ROOT}/run
+
+module load PrgEnv-amd/8.3.3
+export HSA_XNACK=1
+
+mkdir -p $SCOPE_RESULTS
+module list > $SCOPE_RESULTS/modules.r9.$SLURM_JOBID.txt 2>&1
+env > $SCOPE_RESULTS/env.r9.$SLURM_JOBID.txt
+rocm-smi > $SCOPE_RESULTS/rocm-smi.r9.$SLURM_JOBID.txt 2>&1
+lscpu > $SCOPE_RESULTS/lscpu.r9.$SLURM_JOBID.txt 2>&1
+
+date
+
+srun -c 56 -n 1 --gpus 8 $SCOPE_BUILD/comm_scope \
+--benchmark_repetitions=5 \
+--benchmark_filter='.*implicit_mapped_GPUWrGPU/0/.*' \
+--benchmark_out_format=json \
+--benchmark_out="$SCOPE_RESULTS/implicit_mapped_GPUWrGPU.json"
+
+date