windows fix

Summary: Attempt to reduce nvcc trouble on windows by (1) avoiding flag for c++14 and (2) avoiding `torch/extension.h`, which introduces pybind11, in `.cu` files.

Reviewed By: patricklabatut

Differential Revision: D34969868

fbshipit-source-id: f3878d6a2ba9d644e87ae7b6377cb5008b4b6ce3

pytorch3d/csrc/ball_query/ball_query.cu

@@ -12,7 +12,6 @@
 #include <math.h>
 #include <stdio.h>
 #include <stdlib.h>
-#include "utils/pytorch3d_cutils.h"
 
 // A chunk of work is blocksize-many points of P1.
 // The number of potential chunks to do is N*(1+(P1-1)/blocksize)

pytorch3d/csrc/iou_box3d/iou_box3d.cu

@@ -15,7 +15,6 @@
 #include <thrust/device_vector.h>
 #include <thrust/tuple.h>
 #include "iou_box3d/iou_utils.cuh"
-#include "utils/pytorch3d_cutils.h"
 
 // Parallelize over N*M computations which can each be done
 // independently

pytorch3d/csrc/sample_farthest_points/sample_farthest_points.cu

@@ -12,7 +12,6 @@
 #include <math.h>
 #include <stdio.h>
 #include <stdlib.h>
-#include "utils/pytorch3d_cutils.h"
 #include "utils/warp_reduce.cuh"
 
 template <unsigned int block_size>
@@ -170,6 +169,9 @@ at::Tensor FarthestPointSamplingCuda(
   // This will ensure each thread processes the minimum necessary number of
   // points (P/threads).
   const int points_pow_2 = std::log(static_cast<double>(P)) / std::log(2.0);
+
+  // Max possible threads per block
+  const int MAX_THREADS_PER_BLOCK = 1024;
   const size_t threads = max(min(1 << points_pow_2, MAX_THREADS_PER_BLOCK), 1);
 
   // Create the accessors

pytorch3d/csrc/utils/pytorch3d_cutils.h

@@ -15,6 +15,3 @@
 #define CHECK_CONTIGUOUS_CUDA(x) \
   CHECK_CUDA(x);                 \
   CHECK_CONTIGUOUS(x)
-
-// Max possible threads per block
-const int MAX_THREADS_PER_BLOCK = 1024;