use less thrust, maybe help Windows

Summary: I think we include more thrust than needed, and maybe removing it will help things like  facebookresearch#1610 with DebugSyncStream errors on Windows.

Reviewed By: shapovalov

Differential Revision: D48949888

fbshipit-source-id: add889c0acf730a039dc9ffd6bbcc24ded20ef27

pytorch3d/csrc/iou_box3d/iou_box3d.cu

@@ -12,8 +12,6 @@
 #include <math.h>
 #include <stdio.h>
 #include <stdlib.h>
-#include <thrust/device_vector.h>
-#include <thrust/tuple.h>
 #include "iou_box3d/iou_utils.cuh"
 
 // Parallelize over N*M computations which can each be done

pytorch3d/csrc/iou_box3d/iou_utils.cuh

@@ -8,7 +8,6 @@
 
 #include <float.h>
 #include <math.h>
-#include <thrust/device_vector.h>
 #include <cstdio>
 #include "utils/float_math.cuh"
 

pytorch3d/csrc/marching_cubes/marching_cubes.cu

@@ -9,8 +9,6 @@
 #include <ATen/ATen.h>
 #include <ATen/cuda/CUDAContext.h>
 #include <c10/cuda/CUDAGuard.h>
-#include <thrust/device_vector.h>
-#include <thrust/scan.h>
 #include <cstdio>
 #include "marching_cubes/tables.h"
 
@@ -40,20 +38,6 @@ through" each cube in the grid.
 // EPS: Used to indicate if two float values are close
 __constant__ const float EPSILON = 1e-5;
 
-// Thrust wrapper for exclusive scan
-//
-// Args:
-//    output: pointer to on-device output array
-//    input: pointer to on-device input array, where scan is performed
-//    numElements: number of elements for the input array
-//
-void ThrustScanWrapper(int* output, int* input, int numElements) {
-  thrust::exclusive_scan(
-      thrust::device_ptr<int>(input),
-      thrust::device_ptr<int>(input + numElements),
-      thrust::device_ptr<int>(output));
-}
-
 // Linearly interpolate the position where an isosurface cuts an edge
 // between two vertices, based on their scalar values
 //
@@ -455,19 +439,24 @@ std::tuple<at::Tensor, at::Tensor, at::Tensor> MarchingCubesCuda(
     grid.x = 65535;
   }
 
+  using at::indexing::None;
+  using at::indexing::Slice;
+
   auto d_voxelVerts =
-      at::zeros({numVoxels}, at::TensorOptions().dtype(at::kInt))
+      at::zeros({numVoxels + 1}, at::TensorOptions().dtype(at::kInt))
           .to(vol.device());
+  auto d_voxelVerts_ = d_voxelVerts.index({Slice(1, None)});
   auto d_voxelOccupied =
-      at::zeros({numVoxels}, at::TensorOptions().dtype(at::kInt))
+      at::zeros({numVoxels + 1}, at::TensorOptions().dtype(at::kInt))
           .to(vol.device());
+  auto d_voxelOccupied_ = d_voxelOccupied.index({Slice(1, None)});
 
   // Execute "ClassifyVoxelKernel" kernel to precompute
   // two arrays - d_voxelOccupied and d_voxelVertices to global memory,
   // which stores the occupancy state and number of voxel vertices per voxel.
   ClassifyVoxelKernel<<<grid, threads, 0, stream>>>(
-      d_voxelVerts.packed_accessor32<int, 1, at::RestrictPtrTraits>(),
-      d_voxelOccupied.packed_accessor32<int, 1, at::RestrictPtrTraits>(),
+      d_voxelVerts_.packed_accessor32<int, 1, at::RestrictPtrTraits>(),
+      d_voxelOccupied_.packed_accessor32<int, 1, at::RestrictPtrTraits>(),
       vol.packed_accessor32<float, 3, at::RestrictPtrTraits>(),
       isolevel);
   AT_CUDA_CHECK(cudaGetLastError());
@@ -477,18 +466,12 @@ std::tuple<at::Tensor, at::Tensor, at::Tensor> MarchingCubesCuda(
   // count for voxels in the grid and compute the number of active voxels.
   // If the number of active voxels is 0, return zero tensor for verts and
   // faces.
-  auto d_voxelOccupiedScan =
-      at::zeros({numVoxels}, at::TensorOptions().dtype(at::kInt))
-          .to(vol.device());
-  ThrustScanWrapper(
-      d_voxelOccupiedScan.data_ptr<int>(),
-      d_voxelOccupied.data_ptr<int>(),
-      numVoxels);
+
+  auto d_voxelOccupiedScan = at::cumsum(d_voxelOccupied, 0);
+  auto d_voxelOccupiedScan_ = d_voxelOccupiedScan.index({Slice(1, None)});
 
   // number of active voxels
-  int lastElement = d_voxelVerts[numVoxels - 1].cpu().item<int>();
-  int lastScan = d_voxelOccupiedScan[numVoxels - 1].cpu().item<int>();
-  int activeVoxels = lastElement + lastScan;
+  int activeVoxels = d_voxelOccupiedScan[numVoxels].cpu().item<int>();
 
   const int device_id = vol.device().index();
   auto opt = at::TensorOptions().dtype(at::kInt).device(at::kCUDA, device_id);
@@ -509,22 +492,17 @@ std::tuple<at::Tensor, at::Tensor, at::Tensor> MarchingCubesCuda(
   CompactVoxelsKernel<<<grid, threads, 0, stream>>>(
       d_compVoxelArray.packed_accessor32<int, 1, at::RestrictPtrTraits>(),
       d_voxelOccupied.packed_accessor32<int, 1, at::RestrictPtrTraits>(),
-      d_voxelOccupiedScan.packed_accessor32<int, 1, at::RestrictPtrTraits>(),
+      d_voxelOccupiedScan_.packed_accessor32<int, 1, at::RestrictPtrTraits>(),
       numVoxels);
   AT_CUDA_CHECK(cudaGetLastError());
   cudaDeviceSynchronize();
 
   // Scan d_voxelVerts array to generate offsets of vertices for each voxel
-  auto d_voxelVertsScan = at::zeros({numVoxels}, opt);
-  ThrustScanWrapper(
-      d_voxelVertsScan.data_ptr<int>(),
-      d_voxelVerts.data_ptr<int>(),
-      numVoxels);
+  auto d_voxelVertsScan = at::cumsum(d_voxelVerts, 0);
+  auto d_voxelVertsScan_ = d_voxelVertsScan.index({Slice(1, None)});
 
   // total number of vertices
-  lastElement = d_voxelVerts[numVoxels - 1].cpu().item<int>();
-  lastScan = d_voxelVertsScan[numVoxels - 1].cpu().item<int>();
-  int totalVerts = lastElement + lastScan;
+  int totalVerts = d_voxelVertsScan[numVoxels].cpu().item<int>();
 
   // Execute "GenerateFacesKernel" kernel
   // This runs only on the occupied voxels.
@@ -544,7 +522,7 @@ std::tuple<at::Tensor, at::Tensor, at::Tensor> MarchingCubesCuda(
       faces.packed_accessor<int64_t, 2, at::RestrictPtrTraits>(),
       ids.packed_accessor<int64_t, 1, at::RestrictPtrTraits>(),
       d_compVoxelArray.packed_accessor32<int, 1, at::RestrictPtrTraits>(),
-      d_voxelVertsScan.packed_accessor32<int, 1, at::RestrictPtrTraits>(),
+      d_voxelVertsScan_.packed_accessor32<int, 1, at::RestrictPtrTraits>(),
       activeVoxels,
       vol.packed_accessor32<float, 3, at::RestrictPtrTraits>(),
       faceTable.packed_accessor32<int, 2, at::RestrictPtrTraits>(),