some SPR hacks for oneMKL

add another matrix
2023-12-11 10:08:43 -07:00 · 2023-12-11 10:08:43 -07:00 · 2023-12-11 10:08:43 -07:00 · 2023-12-11 10:08:43 -07:00 · 2023-12-07 10:00:05 -08:00 · 2023-12-07 10:00:05 -08:00
17 changed files with 157 additions and 30 deletions
--- a/blake-h100-cuda12/download.sh
+++ b/blake-h100-cuda12/download.sh
@@ -0,0 +1,57 @@
+#! /bin/bash
+
+set -eou pipefail
+shopt -s globstar
+
+source "$( cd -- "$(dirname "$0")" >/dev/null 2>&1 ; pwd -P )"/env.sh
+
+LOG_DIR="$(date +"%Y%m%d_%H%M%S")_build"
+export LOG_DIR
+mkdir -p "$LOG_DIR"
+
+# intel blows up SSH for some reason?
+git clone git@github.com:kokkos/kokkos.git "$KOKKOS_SRC" || true
+(cd "$KOKKOS_SRC" && git checkout $KOKKOS_SHA) || true
+git clone git@github.com:kokkos/kokkos-kernels.git "$KERNELS_SRC" || true
+(cd "$KERNELS_SRC" && git checkout $KERNELS_SHA) || true
+
+module list |& tee "$LOG_DIR/module-list.log"
+lscpu |& tee "$LOG_DIR/lscpu.log"
+hostname |& tee "$LOG_DIR/hostname.log"
+env |& tee "$LOG_DIR/env.log" || true
+
+## Configure Kokkos
+cmake -S "$KOKKOS_SRC" -B "$KOKKOS_BUILD" \
+-DCMAKE_INSTALL_PREFIX="$KOKKOS_INSTALL" \
+-DCMAKE_CXX_STANDARD=17 \
+-DCMAKE_BUILD_TYPE=Release \
+-DCMAKE_CXX_COMPILER="$KOKKOS_SRC"/bin/nvcc_wrapper \
+-DKokkos_ENABLE_CUDA=ON \
+-DKokkos_ARCH_NATIVE=ON \
+-DKokkos_ARCH_HOPPER90=ON \
+|& tee "$LOG_DIR/kokkos-config.log"
+
+## Build & Install Kokkos
+cmake --build "$KOKKOS_BUILD" -j "$(nproc)" -t install \
+|& tee "$LOG_DIR/kokkos-build.log"
+
+## Configure Kernels
+cmake -S "$KERNELS_SRC" -B "$KERNELS_BUILD" \
+-DKokkos_DIR="$KOKKOS_INSTALL/lib64/cmake/Kokkos" \
+-DCMAKE_BUILD_TYPE=Release \
+-DCMAKE_CXX_COMPILER="$KOKKOS_SRC"/bin/nvcc_wrapper \
+-DKokkosKernels_ENABLE_TPL_CUSPARSE=ON \
+-DKokkosKernels_ENABLE_TPL_CUBLASE=ON \
+-DKokkosKernels_ENABLE_TESTS=ON \
+-DKokkosKernels_ENABLE_PERFTESTS=ON \
+-DKokkosKernels_ENABLE_BENCHMARK=ON \
+|& tee "$LOG_DIR/kernels-config.log"
+
+## Build Kernels
+for unit_test_dir in "$KERNELS_BUILD"/**/unit_test; do
+  VERBOSE=1 make -C "$unit_test_dir" -j "$(nproc)" |& tee -a "$LOG_DIR/kernels-build.log"
+done
+VERBOSE=1 make -C "$KERNELS_BUILD" -j "$(nproc)" \
+KokkosKernels_Blas3_gemm_benchmark \
+KokkosKernels_sparse_spmv_benchmark \
+|& tee -a "$LOG_DIR/kernels-build.log"
--- a/blake-h100-cuda12/env.sh
+++ b/blake-h100-cuda12/env.sh
@@ -0,0 +1,14 @@
+export ROOT_DIR=/projects/cwpears/kug-2023/blake-h100-cuda12
+
+export KOKKOS_SHA=f8788ef2ae1940b627cc6ebc6abeef2c34e7e8dc # 2023 11 30
+export KOKKOS_SRC="$ROOT_DIR/kokkos-${KOKKOS_SHA:0:8}"
+export KOKKOS_BUILD="$ROOT_DIR/kokkos-build-${KOKKOS_SHA:0:8}"
+export KOKKOS_INSTALL="$ROOT_DIR/kokkos-install-${KOKKOS_SHA:0:8}"
+
+export KERNELS_SHA=a80eb9114ddda2d9454e4f3cc8a3dd5143ecdfc8 # 2023 11 30
+export KERNELS_SRC="$ROOT_DIR/kernels-${KERNELS_SHA:0:8}"
+export KERNELS_BUILD="$ROOT_DIR/kernels-build-${KERNELS_SHA:0:8}"
+
+source /projects/x86-64-icelake-rocky8/spack-config/blake-setup-user-module-env.sh
+module load gcc/11.3.0 cuda/12.0.0
+module load cmake
--- a/blake-h100-cuda12/run.sh
+++ b/blake-h100-cuda12/run.sh
@@ -0,0 +1,23 @@
+#! /bin/bash
+#SBATCH -N 1
+#SBATCH -p H100
+
+set -eou pipefail
+
+source "$( cd -- "$(dirname "$0")" >/dev/null 2>&1 ; pwd -P )"/env.sh
+
+LOG_DIR="$(date +"%Y%m%d_%H%M%S")_run"
+export LOG_DIR
+mkdir -p "$LOG_DIR"
+
+
+srun -n1 -t 1 lscpu |& tee "$LOG_DIR/lscpu.log" || true
+srun -n1 -t 1 hostname |& tee "$LOG_DIR/hostname.log" || true
+srun -n1 -t 1 cat /proc/cpuinfo |& tee "$LOG_DIR/cpuinfo.log" || true
+srun -n1 -t 1 env |& tee "$LOG_DIR/env.log" || true
+
+
+srun -N 1 -p H100 -n 1 -t 60 "$KERNELS_BUILD"/perf_test/sparse/KokkosKernels_sparse_spmv_benchmark -f /projects/cwpears/kug-2023/dielFilterV3real/dielFilterV3real.mtx |& tee "$LOG_DIR/spmv2.log"
+srun -N 1 -p H100 -n 1 -t 60 "$KERNELS_BUILD"/perf_test/blas/blas3/KokkosKernels_Blas3_gemm_benchmark --cuda 0 |& tee "$LOG_DIR/gemm.log"
+srun -N 1 -p H100 -n 1 -t 60 "$KERNELS_BUILD"/perf_test/sparse/KokkosKernels_sparse_spmv_benchmark -f /projects/cwpears/sparc_gpu_problems/single_gpu/matrix.mm |& tee "$LOG_DIR/spmv1.log"
+srun -N 1 -p H100 -n 1 -t 60 ctest --test-dir "$KERNELS_BUILD" |& tee "$LOG_DIR/ctest.log"
--- a/blake-h100/download.sh
+++ b/blake-h100/download.sh
@@ -11,7 +11,7 @@ mkdir -p "$LOG_DIR"
 # intel blows up SSH for some reason?
 git clone git@github.com:kokkos/kokkos.git "$KOKKOS_SRC" || true
 (cd "$KOKKOS_SRC" && git checkout $KOKKOS_SHA) || true
-git clone git@github.com:kokkos/kokkos-kernels.git "$KERNELS_SRC" || true
+git clone "$KERNELS_REMOTE" "$KERNELS_SRC" || true
 (cd "$KERNELS_SRC" && git checkout $KERNELS_SHA) || true

 module list |& tee "$LOG_DIR/module-list.log"
--- a/blake-h100/env.sh
+++ b/blake-h100/env.sh
@@ -5,7 +5,8 @@ export KOKKOS_SRC="$ROOT_DIR/kokkos-${KOKKOS_SHA:0:8}"
 export KOKKOS_BUILD="$ROOT_DIR/kokkos-build-${KOKKOS_SHA:0:8}"
 export KOKKOS_INSTALL="$ROOT_DIR/kokkos-install-${KOKKOS_SHA:0:8}"

-export KERNELS_SHA=a80eb9114ddda2d9454e4f3cc8a3dd5143ecdfc8 # 2023 11 30
+export KERNELS_REMOTE="git@github.com:cwpearson/kokkos-kernels.git"
+export KERNELS_SHA=a91a1f24560692d6a294e764f82f497dc10aeb7d # 2023 12 08
 export KERNELS_SRC="$ROOT_DIR/kernels-${KERNELS_SHA:0:8}"
 export KERNELS_BUILD="$ROOT_DIR/kernels-build-${KERNELS_SHA:0:8}"

--- a/blake-h100/run.sh
+++ b/blake-h100/run.sh
@@ -16,6 +16,6 @@ srun -n1 -t 1 hostname |& tee "$LOG_DIR/hostname.log" || true
 srun -n1 -t 1 cat /proc/cpuinfo |& tee "$LOG_DIR/cpuinfo.log" || true
 srun -n1 -t 1 env |& tee "$LOG_DIR/env.log" || true

-
+srun -N 1 -p H100 -n 1 -t 60 "$KERNELS_BUILD"/perf_test/sparse/KokkosKernels_sparse_spmv_benchmark -f /projects/cwpears/kug-2023/dielFilterV3real/dielFilterV3real.mtx |& tee "$LOG_DIR/spmv2.log"
 srun -N 1 -p H100 -n 1 -t 60 "$KERNELS_BUILD"/perf_test/blas/blas3/KokkosKernels_Blas3_gemm_benchmark --cuda 0 |& tee "$LOG_DIR/gemm.log"
-srun -N 1 -p H100 -n 1 -t 60 "$KERNELS_BUILD"/perf_test/sparse/KokkosKernels_sparse_spmv_benchmark -f /projects/cwpears/sparc_gpu_problems/single_gpu/matrix.mm |& tee "$LOG_DIR/spmv.log"
+srun -N 1 -p H100 -n 1 -t 60 "$KERNELS_BUILD"/perf_test/sparse/KokkosKernels_sparse_spmv_benchmark -f /projects/cwpears/sparc_gpu_problems/single_gpu/matrix.mm |& tee "$LOG_DIR/spmv1.log"
--- a/blake-pvc/download-worker.sh
+++ b/blake-pvc/download-worker.sh
@@ -9,19 +9,31 @@ export LOG_DIR
 mkdir -p "$LOG_DIR"

 # intel blows up SSH for some reason?
-module del intel/oneAPI/hpc-toolkit/2022.1.2
+#module del intel/oneAPI/hpc-toolkit/2022.1.2
 git clone git@github.com:kokkos/kokkos.git "$KOKKOS_SRC" || true
 (cd "$KOKKOS_SRC" && git checkout $KOKKOS_SHA) || true
-git clone git@github.com:kokkos/kokkos-kernels.git "$KERNELS_SRC" || true
+git clone "$KERNELS_REMOTE" "$KERNELS_SRC" || true
 (cd "$KERNELS_SRC" && git checkout $KERNELS_SHA) || true
+git clone "$PTI_REMOTE" "$PTI_SRC" || true
+(cd "$PTI_SRC" && git checkout $PTI_SHA) || true

 # re-set up our environment
-source "$( cd -- "$(dirname "$0")" >/dev/null 2>&1 ; pwd -P )"/env.sh
+#source "$( cd -- "$(dirname "$0")" >/dev/null 2>&1 ; pwd -P )"/env.sh

 module list |& tee "$LOG_DIR/module-list.log"
 lscpu |& tee "$LOG_DIR/lscpu.log"
 hostname |& tee "$LOG_DIR/hostname.log"
-env |& tee "$LOG_DIR/env.log"
+env 2>&1 >> "$LOG_DIR/env.log"
+
+## Configure onetrace
+cmake -S "$PTI_SRC/tools/onetrace" -B "$PTI_BUILD" \
+-DCMAKE_BUILD_TYPE=Release \
+-DCMAKE_CXX_COMPILER=icpx \
+|& tee "$LOG_DIR"/onetrace-config.log
+
+## build onetrace
+make -C "$PTI_BUILD" \
+|& tee "$LOG_DIR/onetrace-build.log"

 ## Configure Kokkos
 cmake -S "$KOKKOS_SRC" -B "$KOKKOS_BUILD" \
@@ -44,6 +56,7 @@ cmake -S "$KERNELS_SRC" -B "$KERNELS_BUILD" \
 -DKokkos_DIR="$KOKKOS_INSTALL/lib64/cmake/Kokkos" \
 -DCMAKE_BUILD_TYPE=Release \
 -DCMAKE_CXX_COMPILER=icpx \
+-DKokkosKernels_ENABLE_TPL_MKL=ON \
 -DKokkosKernels_ENABLE_TESTS=ON \
 -DKokkosKernels_ENABLE_PERFTESTS=ON \
 -DKokkosKernels_ENABLE_BENCHMARK=ON \
--- a/blake-pvc/env.sh
+++ b/blake-pvc/env.sh
@@ -1,19 +1,23 @@
 export ROOT_DIR=/projects/cwpears/kug-2023/blake-pvc

 export KOKKOS_SHA=f8788ef2ae1940b627cc6ebc6abeef2c34e7e8dc # 2023 11 30
-# export KOKKOS_SHA="4.2.00"
 export KOKKOS_SRC="$ROOT_DIR/kokkos-${KOKKOS_SHA:0:8}"
 export KOKKOS_BUILD="$ROOT_DIR/kokkos-build-${KOKKOS_SHA:0:8}"
 export KOKKOS_INSTALL="$ROOT_DIR/kokkos-install-${KOKKOS_SHA:0:8}"

-export KERNELS_SHA=a80eb9114ddda2d9454e4f3cc8a3dd5143ecdfc8 # 2023 11 30
-# export KERNELS_SHA="4.2.00"
+export KERNELS_REMOTE="git@github.com:cwpearson/kokkos-kernels.git"
+export KERNELS_SHA=a91a1f24560692d6a294e764f82f497dc10aeb7d # 2023 12 08
 export KERNELS_SRC="$ROOT_DIR/kernels-${KERNELS_SHA:0:8}"
 export KERNELS_BUILD="$ROOT_DIR/kernels-build-${KERNELS_SHA:0:8}"

+export PTI_REMOTE="git@github.com:intel/pti-gpu.git"
+export PTI_SHA="f27829883defd69c18e7dc2e4cdd727a2483c40c" # 2023 12 11
+export PTI_SRC="$ROOT_DIR/pti-gpu-${PTI_SHA:0:8}"
+export PTI_BUILD="$ROOT_DIR/pti-gpu-build-${PTI_SHA:0:8}"
+
 source /projects/x86-64-icelake-rocky8/spack-config/blake-setup-user-module-env.sh
 module load cmake
 module load intel-oneapi-compilers/2023.1.0 intel-oneapi-dpl/2022.1.0 intel-oneapi-mkl/2023.1.0

 # Required for the hashmap accumulator
-export ZES_ENABLE_SYSMAN=1
+export ZES_ENABLE_SYSMAN=1
--- a/blake-pvc/run.sh
+++ b/blake-pvc/run.sh
@@ -11,8 +11,10 @@ mkdir -p "$LOG_DIR"

 srun -N 1 -p PV --exclude=blake15 -n1 -t 1 lscpu |& tee "$LOG_DIR/lscpu.log" || true
 srun -N 1 -p PV --exclude=blake15 -n1 -t 1 hostname |& tee "$LOG_DIR/hostname.log" || true
-srun -N 1 -p PV --exclude=blake15 -n1 -t 1 cat /proc/cpuinfo |& tee "$LOG_DIR/cpuinfo.log" || true
-srun -N 1 -p PV --exclude=blake15 -n1 -t 1 env |& tee "$LOG_DIR/env.log" || true
+srun -N 1 -p PV --exclude=blake15 -n1 -t 1 cat /proc/cpuinfo 2>&1 > "$LOG_DIR/cpuinfo.log" || true
+srun -N 1 -p PV --exclude=blake15 -n1 -t 1 env 2>&1 > "$LOG_DIR/env.log" || true

+srun -N 1 -p PV --exclude=blake15 -n1 -t 60 "$KERNELS_BUILD"/perf_test/sparse/KokkosKernels_sparse_spmv_benchmark -f /projects/cwpears/kug-2023/dielFilterV3real/dielFilterV3real.mtx |& tee "$LOG_DIR/spmv2.log"
+srun -N 1 -p PV --exclude=blake15 -n1 -t 60 "$PTI_BUILD"/onetrace -d "$KERNELS_BUILD"/perf_test/sparse/KokkosKernels_sparse_spmv_benchmark -f /projects/cwpears/kug-2023/dielFilterV3real/dielFilterV3real.mtx |& tee "$LOG_DIR/spmv2-onetrace.log"
 srun -N 1 -p PV --exclude=blake15 -n 1 -t 60 "$KERNELS_BUILD"/perf_test/blas/blas3/KokkosKernels_Blas3_gemm_benchmark --sycl 0 |& tee "$LOG_DIR/gemm.log"
-srun -N 1 -p PV --exclude=blake15 -n 1 -t 60 "$KERNELS_BUILD"/perf_test/sparse/KokkosKernels_sparse_spmv_benchmark -f /projects/cwpears/sparc_gpu_problems/single_gpu/matrix.mm |& tee "$LOG_DIR/spmv.log"
+srun -N 1 -p PV --exclude=blake15 -n 1 -t 60 "$KERNELS_BUILD"/perf_test/sparse/KokkosKernels_sparse_spmv_benchmark -f /projects/cwpears/sparc_gpu_problems/single_gpu/matrix.mm |& tee "$LOG_DIR/spmv1.log"
--- a/blake-spr/download-worker.sh
+++ b/blake-spr/download-worker.sh
@@ -12,7 +12,7 @@ mkdir -p "$LOG_DIR"
 # module del intel/oneAPI/hpc-toolkit/2022.1.2
 git clone git@github.com:kokkos/kokkos.git "$KOKKOS_SRC" || true
 (cd "$KOKKOS_SRC" && git checkout $KOKKOS_SHA) || true
-git clone git@github.com:kokkos/kokkos-kernels.git "$KERNELS_SRC" || true
+git clone "$KOKKOS_REMOTE" "$KERNELS_SRC" || true
 (cd "$KERNELS_SRC" && git checkout $KERNELS_SHA) || true

 # re-set up our environment
--- a/blake-spr/env.sh
+++ b/blake-spr/env.sh
@@ -5,7 +5,8 @@ export KOKKOS_SRC="$ROOT_DIR/kokkos-${KOKKOS_SHA:0:8}"
 export KOKKOS_BUILD="$ROOT_DIR/kokkos-build-${KOKKOS_SHA:0:8}"
 export KOKKOS_INSTALL="$ROOT_DIR/kokkos-install-${KOKKOS_SHA:0:8}"

-export KERNELS_SHA=a80eb9114ddda2d9454e4f3cc8a3dd5143ecdfc8 # 2023 11 30
+export KERNELS_REMOTE="git@github.com:cwpearson/kokkos-kernels.git"
+export KERNELS_SHA=a91a1f24560692d6a294e764f82f497dc10aeb7d # 2023 12 08
 export KERNELS_SRC="$ROOT_DIR/kernels-${KERNELS_SHA:0:8}"
 export KERNELS_BUILD="$ROOT_DIR/kernels-build-${KERNELS_SHA:0:8}"

--- a/blake-spr/run.sh
+++ b/blake-spr/run.sh
@@ -14,5 +14,6 @@ srun -n1 -t 1 hostname |& tee "$LOG_DIR/hostname.log" || true
 srun -n1 -t 1 cat /proc/cpuinfo |& tee "$LOG_DIR/cpuinfo.log" || true
 srun -n1 -t 1 env |& tee "$LOG_DIR/env.log" || true

+srun -N 1 -p all -n1 -t 60 "$KERNELS_BUILD"/perf_test/sparse/KokkosKernels_sparse_spmv_benchmark -f /projects/cwpears/kug-2023/dielFilterV3real/dielFilterV3real.mtx |& tee "$LOG_DIR/spmv2.log"
 srun -N 1 -p all -n1 -t 60 "$KERNELS_BUILD"/perf_test/blas/blas3/KokkosKernels_Blas3_gemm_benchmark |& tee "$LOG_DIR/gemm.log"
-srun -N 1 -p all -n1 -t 60 "$KERNELS_BUILD"/perf_test/sparse/KokkosKernels_sparse_spmv_benchmark -f /projects/cwpears/sparc_gpu_problems/single_gpu/matrix.mm |& tee "$LOG_DIR/spmv.log"
+srun -N 1 -p all -n1 -t 60 "$KERNELS_BUILD"/perf_test/sparse/KokkosKernels_sparse_spmv_benchmark -f /projects/cwpears/sparc_gpu_problems/single_gpu/matrix.mm |& tee "$LOG_DIR/spmv1.log"
--- a/frontier/download.sh
+++ b/frontier/download.sh
@@ -43,6 +43,9 @@ cmake -S "$KERNELS_SRC" -B "$KERNELS_BUILD" \
 -DKokkosKernels_ENABLE_BENCHMARK=ON \
 -DKokkosKernels_ENABLE_TPL_ROCSPARSE=ON \
 -DKokkosKernels_ENABLE_TPL_ROCBLAS=ON \
+-DKokkosKernels_INST_OFFSET_INT=OFF \
+-DKokkosKernels_INST_OFFSET_INT64_T=ON \
+-DKokkosKernels_INST_OFFSET_SIZE_T=OFF \
 |& tee "$LOG_DIR/kernels-config.log"

 ## Build Kernels
--- a/frontier/env.sh
+++ b/frontier/env.sh
@@ -5,7 +5,7 @@ export KOKKOS_SRC="$ROOT_DIR/kokkos-${KOKKOS_SHA:0:8}"
 export KOKKOS_BUILD="$ROOT_DIR/kokkos-build-${KOKKOS_SHA:0:8}"
 export KOKKOS_INSTALL="$ROOT_DIR/kokkos-install-${KOKKOS_SHA:0:8}"

-export KERNELS_SHA=a80eb9114ddda2d9454e4f3cc8a3dd5143ecdfc8 # 2023 11 30
+export KERNELS_SHA="4ce619e161efd731b87df4160821328cff17a4ee" # 2023 12 06
 export KERNELS_SRC="$ROOT_DIR/kernels-${KERNELS_SHA:0:8}"
 export KERNELS_BUILD="$ROOT_DIR/kernels-build-${KERNELS_SHA:0:8}"

--- a/frontier/run-worker.sh
+++ b/frontier/run-worker.sh
@@ -1,4 +1,4 @@
-
+#!/bin/bash
 set -eou pipefail

 source "$( cd -- "$(dirname "$0")" >/dev/null 2>&1 ; pwd -P )"/env.sh
@@ -7,7 +7,7 @@ LOG_DIR="$(date +"%Y%m%d_%H%M%S")_run"
 export LOG_DIR
 mkdir -p "$LOG_DIR"

-srun -n 1 -t 1 /opt/rocm-5.7.0/bin/rocm-smi \
+/opt/rocm-5.7.0/bin/rocm-smi \
  --showdriverversion \
  --showproductname \
  --showmclkrange \
@@ -15,11 +15,11 @@ srun -n 1 -t 1 /opt/rocm-5.7.0/bin/rocm-smi \
  --showsclkrange \
  --showfwinfo \
  |& tee "$LOG_DIR/rocm-smi.log"  || true
-srun -n 1 -t 1 /opt/rocm-5.7.0/bin/rocminfo |& tee "$LOG_DIR/rocminfo.log" || true
-srun -n 1 -t 1 lscpu |& tee "$LOG_DIR/lscpu.log" || true
-srun -n 1 -t 1 hostname |& tee "$LOG_DIR/hostname.log" || true
-srun -n 1 -t 1 cat /proc/cpuinfo |& tee "$LOG_DIR/cpuinfo.log" || true
-srun -n 1 -t 1 env |& tee "$LOG_DIR/env.log" || true
+/opt/rocm-5.7.0/bin/rocminfo |& tee "$LOG_DIR/rocminfo.log" || true
+lscpu |& tee "$LOG_DIR/lscpu.log" || true
+hostname |& tee "$LOG_DIR/hostname.log" || true
+cat /proc/cpuinfo |& tee "$LOG_DIR/cpuinfo.log" || true
+env |& tee "$LOG_DIR/env.log" || true

-srun -n 1 -t 60 "$KERNELS_BUILD"/perf_test/blas/blas3/KokkosKernels_Blas3_gemm_benchmark --hip 0 |& tee "$LOG_DIR/gemm.log"
-srun -n 1 -t 60 "$KERNELS_BUILD"/perf_test/sparse/KokkosKernels_sparse_spmv_benchmark -f /projects/cwpears/sparc_gpu_problems/single_gpu/matrix.mm |& tee "$LOG_DIR/spmv.log"
+"$KERNELS_BUILD"/perf_test/blas/blas3/KokkosKernels_Blas3_gemm_benchmark --hip 0 |& tee "$LOG_DIR/gemm.log"
+"$KERNELS_BUILD"/perf_test/sparse/KokkosKernels_sparse_spmv_benchmark -f "$ROOT_DIR"/matrix.mm |& tee "$LOG_DIR/spmv.log"
--- a/perlmutter-a100/run-worker.sh
+++ b/perlmutter-a100/run-worker.sh
@@ -13,6 +13,8 @@ hostname |& tee "$LOG_DIR/hostname.log" || true
 cat /proc/cpuinfo |& tee "$LOG_DIR/cpuinfo.log" || true
 env |& tee "$LOG_DIR/env.log" || true

-
+date
 "$KERNELS_BUILD"/perf_test/blas/blas3/KokkosKernels_Blas3_gemm_benchmark --cuda 0 |& tee "$LOG_DIR/gemm.log"
+date
 "$KERNELS_BUILD"/perf_test/sparse/KokkosKernels_sparse_spmv_benchmark -f "$ROOT_DIR"/matrix.mm |& tee "$LOG_DIR/spmv.log"
+date
--- a/perlmutter-a100/run.sh
+++ b/perlmutter-a100/run.sh
@@ -1,5 +1,11 @@
 #!/bin/bash
+#SBATCH  -A m3953
+#SBATCH -C "gpu&hbm40g"
+#SBATCH  -G 1
+#SBATCH --qos regular
+#SBATCH -N 1
+#SBATCH -t 60

 set -eou pipefail

-srun -A m3953 -C "gpu&hbm40g" --qos regular -N 1 -t 120 ./run-worker.sh
+srun -A m3953 -C "gpu&hbm40g" -G 1 --qos regular -N 1 -t 60 ./run-worker.sh
Author	SHA1	Message	Date
Carl Pearson	280d9d21c9	some SPR hacks for oneMKL	2023-12-11 10:08:43 -07:00
Carl Pearson	682af43c18	add another matrix	2023-12-11 10:08:43 -07:00
Carl Pearson	63f79d7484	add another matrix	2023-12-11 10:08:43 -07:00
Carl Pearson	c2c8096280	add pti-gpu/onetrace build, remove irrelevant lmod stuff, add MKL	2023-12-11 10:08:43 -07:00
Carl Pearson	218c88bcb4	perlmutter fixes	2023-12-07 10:00:05 -08:00
Carl Pearson	f2b795c5ab	perlmutter fixes	2023-12-07 10:00:05 -08:00
Carl Pearson	aa25898ebd	add CUDA12 H100 build	2023-12-07 10:45:57 -07:00
Carl Pearson	60e2f6e819	newer kernels SHA	2023-12-07 10:45:56 -07:00
Carl Pearson	2f70fdc726	try int64_t on frontier	2023-12-06 18:10:39 -05:00
Carl Pearson	ecb4053a6d	fix frontier runs	2023-12-06 17:30:46 -05:00