fix bug to add large tensor support

Signed-off-by: kaixuan <[email protected]>

pyg_lib/csrc/ops/cpu/radix_sort.h

@@ -60,16 +60,16 @@ void radix_sort_kernel(K* input_keys,
                        V* input_values,
                        K* output_keys,
                        V* output_values,
-                       int elements_count,
-                       int* histogram,
-                       int* histogram_ps,
+                       int64_t elements_count,
+                       int64_t* histogram,
+                       int64_t* histogram_ps,
                        int pass) {
   int tid = omp_get_thread_num();
   int nthreads = omp_get_num_threads();
-  int elements_count_4 = elements_count / 4 * 4;
+  int64_t elements_count_4 = elements_count / 4 * 4;
 
-  int* local_histogram = &histogram[RDX_HIST_SIZE * tid];
-  int* local_histogram_ps = &histogram_ps[RDX_HIST_SIZE * tid];
+  int64_t* local_histogram = &histogram[RDX_HIST_SIZE * tid];
+  int64_t* local_histogram_ps = &histogram_ps[RDX_HIST_SIZE * tid];
 
   // Step 1: compute histogram
   for (int i = 0; i < RDX_HIST_SIZE; i++) {
@@ -97,7 +97,7 @@ void radix_sort_kernel(K* input_keys,
 #pragma omp barrier
   // Step 2: prefix sum
   if (tid == 0) {
-    int sum = 0, prev_sum = 0;
+    int64_t sum = 0, prev_sum = 0;
     for (int bins = 0; bins < RDX_HIST_SIZE; bins++) {
       for (int t = 0; t < nthreads; t++) {
         sum += histogram[t * RDX_HIST_SIZE + bins];
@@ -123,7 +123,7 @@ void radix_sort_kernel(K* input_keys,
     int bin_3 = (key_3 >> (pass * 8)) & 0xFF;
     int bin_4 = (key_4 >> (pass * 8)) & 0xFF;
 
-    int pos;
+    int64_t pos;
     pos = local_histogram_ps[bin_1]++;
     output_keys[pos] = key_1;
     output_values[pos] = input_values[i];
@@ -140,7 +140,7 @@ void radix_sort_kernel(K* input_keys,
   if (tid == (nthreads - 1)) {
     for (int64_t i = elements_count_4; i < elements_count; ++i) {
       K key = input_keys[i];
-      int pos = local_histogram_ps[(key >> (pass * 8)) & 0xFF]++;
+      int64_t pos = local_histogram_ps[(key >> (pass * 8)) & 0xFF]++;
       output_keys[pos] = key;
       output_values[pos] = input_values[i];
     }
@@ -161,11 +161,11 @@ std::pair<K*, V*> radix_sort_parallel(K* inp_key_buf,
                                       int64_t elements_count,
                                       int64_t max_value) {
   int maxthreads = omp_get_max_threads();
-  std::unique_ptr<int[]> histogram_tmp(new int[RDX_HIST_SIZE * maxthreads]);
-  std::unique_ptr<int[]> histogram_ps_tmp(
-      new int[RDX_HIST_SIZE * maxthreads + 1]);
-  int* histogram = histogram_tmp.get();
-  int* histogram_ps = histogram_ps_tmp.get();
+  std::unique_ptr<int64_t[]> histogram_tmp(new int64_t[RDX_HIST_SIZE * maxthreads]);
+  std::unique_ptr<int64_t[]> histogram_ps_tmp(
+      new int64_t[RDX_HIST_SIZE * maxthreads + 1]);
+  int64_t* histogram = histogram_tmp.get();
+  int64_t* histogram_ps = histogram_ps_tmp.get();
   if (max_value == 0) {
     return std::make_pair(inp_key_buf, inp_value_buf);
   }