avoid CPU/GPU sync in sample_farthest_points

Summary:
Optimizing sample_farthest_poinst by reducing CPU/GPU sync:
1. replacing iterative randint for starting indexes for 1 function call, if length is constant
2. Avoid sync in fetching maxumum of sample points, if we sample the same amount
3. Initializing 1 tensor for samples and indixes

compare
https://fburl.com/mlhub/7wk0xi98
Before
{F1980383703}
after
{F1980383707}

Histogram match pretty closely
{F1980464338}

Reviewed By: bottler

Differential Revision: D78731869

fbshipit-source-id: 060528ae7a1e0fbbd005d129c151eaf9405841de

pytorch3d/csrc/sample_farthest_points/sample_farthest_points.cu

@@ -107,7 +107,8 @@ at::Tensor FarthestPointSamplingCuda(
     const at::Tensor& points, // (N, P, 3)
     const at::Tensor& lengths, // (N,)
     const at::Tensor& K, // (N,)
-    const at::Tensor& start_idxs) {
+    const at::Tensor& start_idxs,
+    const int64_t max_K_known = -1) {
   // Check inputs are on the same device
   at::TensorArg p_t{points, "points", 1}, lengths_t{lengths, "lengths", 2},
       k_t{K, "K", 3}, start_idxs_t{start_idxs, "start_idxs", 4};
@@ -129,7 +130,12 @@ at::Tensor FarthestPointSamplingCuda(
 
   const int64_t N = points.size(0);
   const int64_t P = points.size(1);
-  const int64_t max_K = at::max(K).item<int64_t>();
+  int64_t max_K;
+  if (max_K_known > 0) {
+    max_K = max_K_known;
+  } else {
+    max_K = at::max(K).item<int64_t>();
+  }
 
   // Initialize the output tensor with the sampled indices
   auto idxs = at::full({N, max_K}, -1, lengths.options());

pytorch3d/csrc/sample_farthest_points/sample_farthest_points.h

@@ -43,7 +43,8 @@ at::Tensor FarthestPointSamplingCuda(
     const at::Tensor& points,
     const at::Tensor& lengths,
     const at::Tensor& K,
-    const at::Tensor& start_idxs);
+    const at::Tensor& start_idxs,
+    const int64_t max_K_known = -1);
 
 at::Tensor FarthestPointSamplingCpu(
     const at::Tensor& points,
@@ -56,14 +57,16 @@ at::Tensor FarthestPointSampling(
     const at::Tensor& points,
     const at::Tensor& lengths,
     const at::Tensor& K,
-    const at::Tensor& start_idxs) {
+    const at::Tensor& start_idxs,
+    const int64_t max_K_known = -1) {
   if (points.is_cuda() || lengths.is_cuda() || K.is_cuda()) {
 #ifdef WITH_CUDA
     CHECK_CUDA(points);
     CHECK_CUDA(lengths);
     CHECK_CUDA(K);
     CHECK_CUDA(start_idxs);
-    return FarthestPointSamplingCuda(points, lengths, K, start_idxs);
+    return FarthestPointSamplingCuda(
+        points, lengths, K, start_idxs, max_K_known);
 #else
     AT_ERROR("Not compiled with GPU support.");
 #endif

pytorch3d/ops/sample_farthest_points.py

@@ -55,6 +55,7 @@ def sample_farthest_points(
     N, P, D = points.shape
     device = points.device
 
+    constant_length = lengths is None
     # Validate inputs
     if lengths is None:
         lengths = torch.full((N,), P, dtype=torch.int64, device=device)
@@ -65,7 +66,9 @@ def sample_farthest_points(
             raise ValueError("A value in lengths was too large.")
 
     # TODO: support providing K as a ratio of the total number of points instead of as an int
+    max_K = -1
     if isinstance(K, int):
+        max_K = K
         K = torch.full((N,), K, dtype=torch.int64, device=device)
     elif isinstance(K, list):
         K = torch.tensor(K, dtype=torch.int64, device=device)
@@ -82,15 +85,17 @@ def sample_farthest_points(
         K = K.to(torch.int64)
 
     # Generate the starting indices for sampling
-    start_idxs = torch.zeros_like(lengths)
     if random_start_point:
-        for n in range(N):
-            # pyre-fixme[6]: For 1st param expected `int` but got `Tensor`.
-            start_idxs[n] = torch.randint(high=lengths[n], size=(1,)).item()
+        if constant_length:
+            start_idxs = torch.randint(high=P, size=(N,), device=device)
+        else:
+            start_idxs = (lengths * torch.rand(lengths.size())).to(torch.int64)
+    else:
+        start_idxs = torch.zeros_like(lengths)
 
     with torch.no_grad():
         # pyre-fixme[16]: `pytorch3d_._C` has no attribute `sample_farthest_points`.
-        idx = _C.sample_farthest_points(points, lengths, K, start_idxs)
+        idx = _C.sample_farthest_points(points, lengths, K, start_idxs, max_K)
     sampled_points = masked_gather(points, idx)
 
     return sampled_points, idx