[INFRA] Import NVIDIA/CCCL upstream as optimization reference library

CCCL (CUDA C++ Core Libraries) provides:
- CUB: device/block/warp-level GPU primitives (reduce, scan, sort, topk)
- Thrust: high-level parallel algorithms (transform_reduce, sort, scan)
- libcudacxx: CUDA C++ standard library (atomics, barriers, memory)
- cudax: experimental features (memory resources, allocators)
- Tuning policies: per-SM hardware-specific algorithm parameters

Competition optimization vectors mapped to CCCL:
- Output TPS (83% weight): warp_reduce, block_reduce, device_topk
- Input TPS (14% weight): device_scan, block_load, prefetch
- Cache TPS (3% weight): prefix caching strategy patterns
- Memory (0.9 util): pooled/cached/buddy allocators

Source: https://github.com/NVIDIA/cccl (shallow clone, HEAD only)
License: Apache-2.0

cccl_upstream/thrust/examples/word_count.cu

@@ -0,0 +1,79 @@
+#include <thrust/device_vector.h>
+#include <thrust/functional.h>
+#include <thrust/host_vector.h>
+#include <thrust/inner_product.h>
+#include <thrust/reduce.h>
+
+#include <cuda/std/iterator> // Required for std::begin/std::end
+
+#include <iostream>
+
+// This example computes the number of words in a text sample
+// with a single call to thrust::inner_product.  The algorithm
+// counts the number of characters which start a new word, i.e.
+// the number of characters where input[i] is an alphabetical
+// character and input[i-1] is not an alphabetical character.
+
+// determines whether the character is alphabetical
+__host__ __device__ bool is_alpha(const char c)
+{
+  return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z');
+}
+
+int word_count(const thrust::device_vector<char>& input)
+{
+  // check for empty string
+  if (input.empty())
+  {
+    return 0;
+  }
+
+  // determines whether the right character begins a new word
+  auto is_word_start = [] __host__ __device__(const char& left, const char& right) {
+    return is_alpha(right) && !is_alpha(left);
+  };
+
+  // compute the number characters that start a new word
+  int wc = thrust::inner_product(
+    input.begin(),
+    input.end() - 1, // sequence of left characters
+    input.begin() + 1, // sequence of right characters
+    0, // initialize sum to 0
+    cuda::std::plus<int>{}, // sum values together
+    is_word_start // how to compare the left and right characters
+  );
+
+  // if the first character is alphabetical, then it also begins a word
+  if (is_alpha(input.front()))
+  {
+    wc++;
+  }
+
+  return wc;
+}
+
+int main()
+{
+  // Paragraph from 'The Raven' by Edgar Allan Poe
+  // http://en.wikipedia.org/wiki/The_Raven
+  const char raw_input[] =
+    "  But the raven, sitting lonely on the placid bust, spoke only,\n"
+    "  That one word, as if his soul in that one word he did outpour.\n"
+    "  Nothing further then he uttered - not a feather then he fluttered -\n"
+    "  Till I scarcely more than muttered `Other friends have flown before -\n"
+    "  On the morrow he will leave me, as my hopes have flown before.'\n"
+    "  Then the bird said, `Nevermore.'\n";
+
+  std::cout << "Text sample:\n";
+  std::cout << raw_input << "\n";
+
+  // transfer to device
+  thrust::device_vector<char> input(cuda::std::begin(raw_input), cuda::std::end(raw_input));
+
+  // count words
+  int wc = word_count(input);
+
+  std::cout << "Text sample contains " << wc << " words\n";
+
+  return 0;
+}