[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/testing/mr_pool_options.cu

@@ -0,0 +1,64 @@
+#include <thrust/mr/pool_options.h>
+
+#include <unittest/unittest.h>
+
+void TestPoolOptionsBasicValidity()
+{
+  thrust::mr::pool_options options = thrust::mr::pool_options();
+  ASSERT_EQUAL(options.validate(), false);
+
+  options.max_blocks_per_chunk = 1024;
+  options.max_bytes_per_chunk  = 1024 * 1024;
+  options.smallest_block_size  = 8;
+  options.largest_block_size   = 1024;
+  ASSERT_EQUAL(options.validate(), true);
+
+  // the minimum number of blocks per chunk is bigger than the max
+  options.min_blocks_per_chunk = 1025;
+  ASSERT_EQUAL(options.validate(), false);
+  options.min_blocks_per_chunk = 128;
+  ASSERT_EQUAL(options.validate(), true);
+
+  // the minimum number of bytes per chunk is bigger than the max
+  options.min_bytes_per_chunk = 1025 * 1024;
+  ASSERT_EQUAL(options.validate(), false);
+  options.min_bytes_per_chunk = 1024;
+  ASSERT_EQUAL(options.validate(), true);
+
+  // smallest block size is bigger than the largest block size
+  options.smallest_block_size = 2048;
+  ASSERT_EQUAL(options.validate(), false);
+  options.smallest_block_size = 8;
+  ASSERT_EQUAL(options.validate(), true);
+}
+DECLARE_UNITTEST(TestPoolOptionsBasicValidity);
+
+void TestPoolOptionsComplexValidity()
+{
+  thrust::mr::pool_options options = thrust::mr::pool_options();
+  ASSERT_EQUAL(options.validate(), false);
+
+  options.max_blocks_per_chunk = 1024;
+  options.max_bytes_per_chunk  = 1024 * 1024;
+  options.smallest_block_size  = 8;
+  options.largest_block_size   = 1024;
+  ASSERT_EQUAL(options.validate(), true);
+
+  options.min_bytes_per_chunk = 2 * 1024;
+  options.max_bytes_per_chunk = 256 * 1024;
+
+  // the biggest allowed allocation (deduced from blocks in chunks)
+  // is smaller than the minimal allowed one (defined in bytes)
+  options.max_blocks_per_chunk = 1;
+  ASSERT_EQUAL(options.validate(), false);
+  options.max_blocks_per_chunk = 1024;
+  ASSERT_EQUAL(options.validate(), true);
+
+  // the smallest allowed allocation (deduced from blocks in chunks)
+  // is bigger than the maximum allowed one (defined in bytes)
+  options.min_blocks_per_chunk = 1024 * 1024;
+  ASSERT_EQUAL(options.validate(), false);
+  options.min_blocks_per_chunk = 128;
+  ASSERT_EQUAL(options.validate(), true);
+}
+DECLARE_UNITTEST(TestPoolOptionsComplexValidity);