Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add build configure option for cuda_arch #2807

Merged
merged 3 commits into from
Nov 1, 2018
Merged

Add build configure option for cuda_arch #2807

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
feccddf
[src] Add configure option for cuda_arch
luitjens Oct 31, 2018
89b2990
Fix documentation bug
ryanleary Oct 31, 2018
3464e90
Reduce duplicated code
ryanleary Oct 31, 2018
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
41 changes: 27 additions & 14 deletions src/configure
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,6 +22,8 @@
# ./configure --atlas-root=../tools/ATLAS/build
# ./configure --use-cuda=no # disable CUDA detection (will build cpu-only
# # version of kaldi even on CUDA-enabled machine
# ./configure --use-cuda --cudatk-dir=/usr/local/cuda/ --cuda-arch=-arch=sm_70
# # Use cuda in /usr/local/cuda and set the arch to sm_70
# ./configure --static --fst-root=/opt/cross/armv8hf \
# --atlas-root=/opt/cross/armv8hf --host=armv8-rpi3-linux-gnueabihf
# # Cross compile for armv8hf, this assumes that you have openfst built
Expand Down Expand Up @@ -65,6 +67,7 @@ Configuration options:
--shared Build and link against shared libraries [default=no]
--use-cuda Build with CUDA [default=yes]
--cudatk-dir=DIR CUDA toolkit directory
--cuda-arch=FLAGS Override the default CUDA_ARCH flags. See https://docs.nvidia.com/cuda/cuda-compiler-driver-nvcc/index.html#nvcc-examples.
--double-precision Build with BaseFloat set to double if yes [default=no],
mostly useful for testing purposes.
--static-fst Build with static OpenFst libraries [default=no]
Expand Down Expand Up @@ -114,8 +117,13 @@ function rel2abs {
fi
}

function read_value {
local val=`expr "X$1" : '[^=]*=\(.*\)'`;
echo $val
}

function read_dirname {
local dir_name=`expr "X$1" : '[^=]*=\(.*\)'`;
local dir_name=`read_value $1`
local retval=`rel2abs $dir_name`
[ -z $retval ] && echo "Bad option '$1': no such directory" && exit 1;
echo $retval
Expand Down Expand Up @@ -421,15 +429,17 @@ function configure_cuda {
fi
fi

case $CUDA_VERSION in
5_5) CUDA_ARCH="-gencode arch=compute_30,code=sm_30 -gencode arch=compute_35,code=sm_35" ;;
6_*) CUDA_ARCH="-gencode arch=compute_30,code=sm_30 -gencode arch=compute_35,code=sm_35 -gencode arch=compute_50,code=sm_50" ;;
7_*) CUDA_ARCH="-gencode arch=compute_30,code=sm_30 -gencode arch=compute_35,code=sm_35 -gencode arch=compute_50,code=sm_50 -gencode arch=compute_53,code=sm_53" ;;
8_*) CUDA_ARCH="-gencode arch=compute_30,code=sm_30 -gencode arch=compute_35,code=sm_35 -gencode arch=compute_50,code=sm_50 -gencode arch=compute_53,code=sm_53 -gencode arch=compute_60,code=sm_60 -gencode arch=compute_61,code=sm_61 -gencode arch=compute_62,code=sm_62" ;;
9_*) CUDA_ARCH="-gencode arch=compute_30,code=sm_30 -gencode arch=compute_35,code=sm_35 -gencode arch=compute_50,code=sm_50 -gencode arch=compute_53,code=sm_53 -gencode arch=compute_60,code=sm_60 -gencode arch=compute_61,code=sm_61 -gencode arch=compute_62,code=sm_62 -gencode arch=compute_70,code=sm_70" ;;
10_*) CUDA_ARCH="-gencode arch=compute_30,code=sm_30 -gencode arch=compute_35,code=sm_35 -gencode arch=compute_50,code=sm_50 -gencode arch=compute_53,code=sm_53 -gencode arch=compute_60,code=sm_60 -gencode arch=compute_61,code=sm_61 -gencode arch=compute_62,code=sm_62 -gencode arch=compute_70,code=sm_70 -gencode arch=compute_72,code=sm_72 -gencode arch=compute_75,code=sm_75" ;;
*) echo "Unsupported CUDA_VERSION (CUDA_VERSION=$CUDA_VERSION), please report it to Kaldi mailing list, together with 'nvcc -h' or 'ptxas -h' which lists allowed -gencode values..."; exit 1 ;;
esac
if [ -z "$CUDA_ARCH" ]; then
case $CUDA_VERSION in
5_5) CUDA_ARCH="-gencode arch=compute_30,code=sm_30 -gencode arch=compute_35,code=sm_35" ;;
6_*) CUDA_ARCH="-gencode arch=compute_30,code=sm_30 -gencode arch=compute_35,code=sm_35 -gencode arch=compute_50,code=sm_50" ;;
7_*) CUDA_ARCH="-gencode arch=compute_30,code=sm_30 -gencode arch=compute_35,code=sm_35 -gencode arch=compute_50,code=sm_50 -gencode arch=compute_53,code=sm_53" ;;
8_*) CUDA_ARCH="-gencode arch=compute_30,code=sm_30 -gencode arch=compute_35,code=sm_35 -gencode arch=compute_50,code=sm_50 -gencode arch=compute_53,code=sm_53 -gencode arch=compute_60,code=sm_60 -gencode arch=compute_61,code=sm_61 -gencode arch=compute_62,code=sm_62" ;;
9_*) CUDA_ARCH="-gencode arch=compute_30,code=sm_30 -gencode arch=compute_35,code=sm_35 -gencode arch=compute_50,code=sm_50 -gencode arch=compute_53,code=sm_53 -gencode arch=compute_60,code=sm_60 -gencode arch=compute_61,code=sm_61 -gencode arch=compute_62,code=sm_62 -gencode arch=compute_70,code=sm_70" ;;
10_*) CUDA_ARCH="-gencode arch=compute_30,code=sm_30 -gencode arch=compute_35,code=sm_35 -gencode arch=compute_50,code=sm_50 -gencode arch=compute_53,code=sm_53 -gencode arch=compute_60,code=sm_60 -gencode arch=compute_61,code=sm_61 -gencode arch=compute_62,code=sm_62 -gencode arch=compute_70,code=sm_70 -gencode arch=compute_72,code=sm_72 -gencode arch=compute_75,code=sm_75" ;;
*) echo "Unsupported CUDA_VERSION (CUDA_VERSION=$CUDA_VERSION), please report it to Kaldi mailing list, together with 'nvcc -h' or 'ptxas -h' which lists allowed -gencode values..."; exit 1 ;;
esac
fi

echo "Using CUDA toolkit $CUDATKDIR (nvcc compiler and runtime libraries)"
echo >> kaldi.mk
Expand Down Expand Up @@ -939,7 +949,7 @@ do
mkl_threading=sequential;
shift ;;
--mkl-threading=*)
mkl_threading=`expr "X$1" : '[^=]*=\(.*\)'`;
mkl_threading=`read_value $1`;
threaded_atlas=true;
shift ;;
--fst-root=*)
Expand Down Expand Up @@ -970,19 +980,22 @@ do
OMPLIBDIR=`read_dirname $1`;
shift ;;
--mathlib=*)
MATHLIB=`expr "X$1" : '[^=]*=\(.*\)'`;
MATHLIB=`read_value $1`;
shift ;;
--cudatk-dir=*)
CUDATKDIR=`read_dirname $1`;
shift ;; #CUDA is used in src/cudamatrix and src/nnet{,bin} only
--cuda-arch=*)
CUDA_ARCH=`read_value $1`;
shift;;
--fst-version=*)
OPENFST_VER=`expr "X$1" : '[^=]*=\(.*\)'`;
OPENFST_VER=`read_value $1`;
shift;;
--host=*)
# The type of system where built programs and libraries will run.
# It should be in the format cpu-vendor-os. If specified, this script
# will infer the target architecture from the specified host triple.
HOST=`expr "X$1" : '[^=]*=\(.*\)'`;
HOST=`read_value $1`;
shift ;;
--android-incdir=*)
android=true;
Expand Down