fix(nccl-tests): use cuda-nvcc-12-6 to support Ampere/Volta (sm_70..sm_90)

This commit is contained in:
2026-03-26 23:50:36 +03:00
parent 5407c26e25
commit 896bdb6ee8
3 changed files with 9 additions and 5 deletions

View File

@@ -36,7 +36,7 @@ RUN wget -qO /tmp/cuda-keyring.gpg \
https://developer.download.nvidia.com/compute/cuda/repos/debian12/x86_64/ /" \
> /etc/apt/sources.list.d/cuda.list \
&& apt-get update -qq \
&& apt-get install -y cuda-nvcc-13-0 \
&& apt-get install -y cuda-nvcc-12-6 \
&& rm -rf /var/lib/apt/lists/*
RUN arch="$(dpkg --print-architecture)" \

View File

@@ -5,6 +5,7 @@ NCCL_VERSION=2.28.9-1
NCCL_CUDA_VERSION=13.0
NCCL_SHA256=2e6faafd2c19cffc7738d9283976a3200ea9db9895907f337f0c7e5a25563186
NCCL_TESTS_VERSION=2.13.10
NVCC_VERSION=12.6
CUBLAS_VERSION=13.0.2.14-1
CUDA_USERSPACE_VERSION=13.0.96-1
GO_VERSION=1.24.0

View File

@@ -34,9 +34,9 @@ if [ -f "${CACHE_DIR}/bin/all_reduce_perf" ]; then
exit 0
fi
# Resolve nvcc path (cuda-nvcc-13-0 installs to /usr/local/cuda-13.0/bin/nvcc)
# Resolve nvcc path (cuda-nvcc-12-6 installs to /usr/local/cuda-12.6/bin/nvcc)
NVCC=""
for candidate in nvcc /usr/local/cuda-13.0/bin/nvcc /usr/local/cuda/bin/nvcc; do
for candidate in nvcc /usr/local/cuda-12.6/bin/nvcc /usr/local/cuda-12/bin/nvcc /usr/local/cuda/bin/nvcc; do
if command -v "$candidate" >/dev/null 2>&1 || [ -x "$candidate" ]; then
NVCC="$candidate"
break
@@ -96,8 +96,11 @@ SRC_DIR=$(ls -d nccl-tests-* 2>/dev/null | head -1)
cd "$SRC_DIR"
echo "=== building all_reduce_perf ==="
# CUDA 13.0 supports only Hopper (sm_90) and newer.
GENCODE="-gencode=arch=compute_90,code=sm_90"
# CUDA 12.6 supports Volta through Hopper (sm_70..sm_90).
GENCODE="-gencode=arch=compute_70,code=sm_70 \
-gencode=arch=compute_80,code=sm_80 \
-gencode=arch=compute_86,code=sm_86 \
-gencode=arch=compute_90,code=sm_90"
make MPI=0 \
NVCC="$NVCC" \
CUDA_HOME="$CUDA_HOME" \