upgraded to 1.4.1

3 files changed, 64 insertions, 66 deletions

diff --git a/.SRCINFO b/.SRCINFO
index 9523bce85eb9..6dfca7717917 100644
--- a/.SRCINFO
+++ b/.SRCINFO
@@ -1,7 +1,7 @@
 pkgbase = mxnet
 	pkgdesc = Flexible and Efficient Library for Deep Learning
-	pkgver = 1.4.0
-	pkgrel = 2
+	pkgver = 1.4.1
+	pkgrel = 1
 	url = http://mxnet.io/
 	arch = x86_64
 	license = Apache
@@ -10,6 +10,7 @@ pkgbase = mxnet
 	makedepends = cudnn
 	makedepends = gtk3
 	makedepends = gtkglext
+	makedepends = nccl
 	makedepends = python-graphviz
 	makedepends = opencv
 	makedepends = vtk
@@ -28,13 +29,34 @@ pkgbase = mxnet
 	optdepends = cudnn
 	optdepends = gtk3
 	optdepends = gtkglext
+	optdepends = nccl
 	optdepends = python-graphviz
 	optdepends = opencv
 	optdepends = vtk
 	optdepends = glew
-	source = git+https://github.com/apache/incubator-mxnet#tag=1.4.0
+	source = git+https://github.com/apache/incubator-mxnet#tag=1.4.1
+	source = git+https://github.com/dmlc/cub
+	source = git+https://github.com/dmlc/dlpack
+	source = git+https://github.com/dmlc/dmlc-core
+	source = git+https://github.com/google/googletest
+	source = git+https://github.com/intel/mkl-dnn
+	source = git+https://github.com/dmlc/mshadow
+	source = git+https://github.com/onnx/onnx-tensorrt
+	source = git+https://github.com/llvm-mirror/openmp
+	source = git+https://github.com/dmlc/ps-lite
+	source = git+https://github.com/dmlc/tvm
 	source = 13559.patch
 	md5sums = SKIP
+	md5sums = SKIP
+	md5sums = SKIP
+	md5sums = SKIP
+	md5sums = SKIP
+	md5sums = SKIP
+	md5sums = SKIP
+	md5sums = SKIP
+	md5sums = SKIP
+	md5sums = SKIP
+	md5sums = SKIP
 	md5sums = fd1c56e58357cd8ba82fcaf8ebfcc1fa
 
 pkgname = mxnet
diff --git a/PKGBUILD b/PKGBUILD
index 92e546047bd4..07ebb87e3154 100644
--- a/PKGBUILD
+++ b/PKGBUILD
@@ -2,25 +2,51 @@
 # Contributor: Carl Åkerlindh <carl.akerlindh at gmail dot com>
 pkgname=mxnet
 _gitname=incubator-mxnet
-pkgver=1.4.0
-pkgrel=2
+pkgver=1.4.1
+pkgrel=1
 pkgdesc="Flexible and Efficient Library for Deep Learning"
 arch=('x86_64')
 url="http://mxnet.io/"
 license=('Apache')
 depends=('hdf5' 'cblas' 'lapack' 'python-numpy' 'python-requests' 'intel-tbb' 'double-conversion')
-optdepends=('cairo' 'cuda' 'cudnn' 'gtk3' 'gtkglext' 'python-graphviz' 'opencv' 'vtk' 'glew')
+optdepends=('cairo' 'cuda' 'cudnn' 'gtk3' 'gtkglext' 'nccl' 'python-graphviz' 'opencv' 'vtk' 'glew')
 makedepends=(${optdepends[@]} 'git' 'cython')
 source=("git+https://github.com/apache/$_gitname#tag=$pkgver"
+	'git+https://github.com/dmlc/cub'
+	'git+https://github.com/dmlc/dlpack'
+	'git+https://github.com/dmlc/dmlc-core'
+	'git+https://github.com/google/googletest'
+	'git+https://github.com/intel/mkl-dnn'
+	'git+https://github.com/dmlc/mshadow'
+	'git+https://github.com/onnx/onnx-tensorrt'
+	'git+https://github.com/llvm-mirror/openmp'
+	'git+https://github.com/dmlc/ps-lite'
+	'git+https://github.com/dmlc/tvm'
 	'13559.patch')
 md5sums=('SKIP'
+         'SKIP'
+         'SKIP'
+         'SKIP'
+         'SKIP'
+         'SKIP'
+         'SKIP'
+         'SKIP'
+         'SKIP'
+         'SKIP'
+         'SKIP'
          'fd1c56e58357cd8ba82fcaf8ebfcc1fa')
 
 prepare() {
 	cd "$srcdir/$_gitname"
+
+	for i in cub dlpack dmlc-core googletest mshadow onnx-tensorrt openmp ps-lite tvm
+	do
+		git config submodule.3rdpary/$i.url "$srcdir/$i"
+	done
+	git config submodule.3rdpary/mkldnn.url "$srcdir/mkl-dnn"
+
 	git submodule update --init --recursive
 	(
-		#echo "CXX=g++"
 		echo "USE_BLAS=blas"
 		echo "USE_LAPACK=1"
 		echo "ADD_LDFLAGS+=-lcblas"
@@ -29,22 +55,13 @@ prepare() {
 		echo "USE_GPERFTOOLS=0"
 		echo "USE_JEMALLOC=0"
 
-		if (pacman -Q cuda &>/dev/null && pacman -Q cudnn &>/dev/null); then
-			msg2 "CUDA support enabled"
-			echo "USE_CUDA=1"
-			echo "USE_CUDA_PATH=/opt/cuda"
-			echo "USE_CUDNN=1"
-			echo "ADD_LDFLAGS+=-L/opt/cuda/lib64/stubs"
-		else
-			msg2 "CUDA support disabled"
-		fi
-		if (pacman -Q opencv &>/dev/null); then
-			msg2 "OpenCV support enabled"
-			echo "USE_OPENCV=1"
-		else
-			msg2 "OpenCV support disabled"
-			echo "USE_OPENCV=0"
-		fi
+		echo "USE_CUDA=1"
+		echo "USE_CUDA_PATH=/opt/cuda"
+		echo "USE_CUDNN=1"
+		echo "ADD_LDFLAGS+=-L/opt/cuda/lib64/stubs"
+		echo "USE_NCCL=1"
+		echo "USE_NCCL_PATH=/usr"
+		echo "USE_OPENCV=1"
 	) >> make/config.mk
 
 	# https://github.com/archlinuxcn/repo/issues/684
@@ -75,5 +92,7 @@ package() {
 	mv $pkgdir/usr/$pkgname/* $pkgdir/usr/lib/python3.7/site-packages/$pkgname/
 	rmdir $pkgdir/usr/$pkgname
 
+	cp -r $srcdir/$_gitname/include $pkgdir/usr/lib/python3.7/site-packages/$pkgname/
+
 	install -Dm644 "$srcdir/$_gitname/LICENSE" "${pkgdir}/usr/share/licenses/${pkgname}/LICENSE"
 }
diff --git a/cuda_call.patch b/cuda_call.patch
deleted file mode 100644
index 7f1dae135808..000000000000
--- a/cuda_call.patch
+++ /dev/null
@@ -1,43 +0,0 @@
-diff --git a/src/common/random_generator.cu b/src/common/random_generator.cu
-index 930e5e07b..e116b8c5f 100644
---- a/src/common/random_generator.cu
-+++ b/src/common/random_generator.cu
-@@ -59,6 +59,17 @@ void RandGenerator<gpu, float>::Seed(mshadow::Stream<gpu> *s, uint32_t seed) {
-   s->Wait();
- }
- 
-+template<>
-+void RandGenerator<gpu, float>::AllocState(RandGenerator<gpu> *inst) {
-+  CUDA_CALL(cudaMalloc(&inst->states_,
-+                       kNumRandomStates * sizeof(curandStatePhilox4_32_10_t)));
-+}
-+
-+template<>
-+void RandGenerator<gpu, float>::FreeState(RandGenerator<gpu> *inst) {
-+  CUDA_CALL(cudaFree(inst->states_));
-+}
-+
- }  // namespace random
- }  // namespace common
- }  // namespace mxnet
-diff --git a/src/common/random_generator.h b/src/common/random_generator.h
-index 5d78b616e..1c8ae01de 100644
---- a/src/common/random_generator.h
-+++ b/src/common/random_generator.h
-@@ -150,14 +150,9 @@ class RandGenerator<gpu, DType> {
-     curandStatePhilox4_32_10_t state_;
-   };  // class RandGenerator<gpu, DType>::Impl
- 
--  static void AllocState(RandGenerator<gpu, DType> *inst) {
--    CUDA_CALL(cudaMalloc(&inst->states_,
--                         kNumRandomStates * sizeof(curandStatePhilox4_32_10_t)));
--  }
-+  static void AllocState(RandGenerator<gpu, DType> *inst);
- 
--  static void FreeState(RandGenerator<gpu, DType> *inst) {
--    CUDA_CALL(cudaFree(inst->states_));
--  }
-+  static void FreeState(RandGenerator<gpu, DType> *inst);
- 
-   void Seed(mshadow::Stream<gpu> *s, uint32_t seed);
-