From 06eb24cf4d7d54e56abfb37ea062a7cb0c887550 Mon Sep 17 00:00:00 2001
From: Gael Guennebaud <g.gael@free.fr>
Date: Fri, 13 Jul 2018 16:04:27 +0200
Subject: Introduce gpu_assert for assertion in device-code, and disable them
 with clang-cuda.

---
 unsupported/Eigen/CXX11/src/Tensor/TensorConvolution.h | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

(limited to 'unsupported/Eigen/CXX11/src/Tensor/TensorConvolution.h')

diff --git a/unsupported/Eigen/CXX11/src/Tensor/TensorConvolution.h b/unsupported/Eigen/CXX11/src/Tensor/TensorConvolution.h
index 3110887e1..25131600d 100644
--- a/unsupported/Eigen/CXX11/src/Tensor/TensorConvolution.h
+++ b/unsupported/Eigen/CXX11/src/Tensor/TensorConvolution.h
@@ -903,7 +903,7 @@ struct TensorEvaluator<const TensorConvolutionOp<Indices, InputArgType, KernelAr
         }
 
         const int shared_mem = block_size.y * (maxX + kernel_size - 1) * sizeof(Scalar);
-        assert(shared_mem <= maxSharedMem);
+        gpu_assert(shared_mem <= maxSharedMem);
 
         const int num_x_blocks = ceil(numX, maxX);
         const int blocksPerProcessor = numext::mini(maxBlocksPerProcessor, maxSharedMem / shared_mem);
@@ -960,7 +960,7 @@ struct TensorEvaluator<const TensorConvolutionOp<Indices, InputArgType, KernelAr
         block_size.z = numext::mini<int>(1024/(block_size.x*block_size.y), maxP);
 
         const int shared_mem = block_size.z * (maxX + kernel_size_x - 1) * (maxY + kernel_size_y - 1) * sizeof(Scalar);
-        assert(shared_mem <= maxSharedMem);
+        gpu_assert(shared_mem <= maxSharedMem);
 
         const int num_x_blocks = ceil(numX, maxX);
         const int num_y_blocks = ceil(numY, maxY);
@@ -1040,7 +1040,7 @@ struct TensorEvaluator<const TensorConvolutionOp<Indices, InputArgType, KernelAr
         dim3 num_blocks(ceil(numX, maxX), ceil(numY, maxY), ceil(numZ, maxZ));
 
         const int shared_mem = (maxX + kernel_size_x - 1) * (maxY + kernel_size_y - 1) * (maxZ + kernel_size_z - 1) * sizeof(Scalar);
-        assert(shared_mem <= maxSharedMem);
+        gpu_assert(shared_mem <= maxSharedMem);
 
         //cout << "launching 3D kernel with block_size.x: " << block_size.x << " block_size.y: " << block_size.y  << " block_size.z: " << block_size.z << " num_blocks.x: " << num_blocks.x << " num_blocks.y: " << num_blocks.y << " num_blocks.z: " << num_blocks.z  << " shared_mem: " << shared_mem << " in stream " << m_device.stream() << endl;
         const array<Index, 3> indices(m_indices[idxX], m_indices[idxY],
-- 
cgit v1.2.3