Files
project_6/cccl_upstream/cudax/test/copy/copy_vectorize.cu
EngineX CI 56fd68e7dd [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
2026-07-30 09:35:51 +00:00

92 lines
3.0 KiB
Plaintext

//===----------------------------------------------------------------------===//
//
// Part of CUDA Experimental in CUDA C++ Core Libraries,
// under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
// SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES.
//
//===----------------------------------------------------------------------===//
#include "copy_common.cuh"
// All layouts in this file have 48 contiguous elements that should coalesce into a (48):(1) layout
// and generate 16B vectorized copies.
static constexpr int N = 48;
/***********************************************************************************************************************
* 1D Vectorization Test
**********************************************************************************************************************/
// src: (48):(1)
// dst: (48):(1)
TEST_CASE("copy d2d vectorize 48:1", "[copy][d2d][vectorize][1d]")
{
test_copy<layout_right>(make_iota<int>(N), N);
}
/***********************************************************************************************************************
* 2D Vectorization Tests
**********************************************************************************************************************/
// src: (6,8):(8,1)
// dst: (6,8):(8,1)
TEST_CASE("copy d2d vectorize (6,8):(8,1)", "[copy][d2d][vectorize][2d]")
{
test_copy<layout_right>(make_iota<int>(N), 6, 8);
}
// src: (8,6):(1,8)
// dst: (8,6):(1,8)
TEST_CASE("copy d2d vectorize (8,6):(1,8)", "[copy][d2d][vectorize][2d]")
{
test_copy<layout_left>(make_iota<int>(N), 8, 6);
}
// src: (6,8):(1,6)
// dst: (6,8):(1,6)
TEST_CASE("copy d2d vectorize (6,8):(1,6)", "[copy][d2d][vectorize][2d]")
{
test_copy<layout_left>(make_iota<int>(N), 6, 8);
}
// src: (8,6):(6,1)
// dst: (8,6):(6,1)
TEST_CASE("copy d2d vectorize (8,6):(6,1)", "[copy][d2d][vectorize][2d]")
{
test_copy<layout_right>(make_iota<int>(N), 8, 6);
}
/***********************************************************************************************************************
* 2D Non-Square Vectorization Tests
**********************************************************************************************************************/
// src: (3,16):(1,3)
// dst: (3,16):(1,3)
TEST_CASE("copy d2d vectorize (3,16):(1,3)", "[copy][d2d][vectorize][2d]")
{
test_copy<layout_left>(make_iota<int>(N), 3, 16);
}
// src: (3,16):(16,1)
// dst: (3,16):(16,1)
TEST_CASE("copy d2d vectorize (3,16):(16,1)", "[copy][d2d][vectorize][2d]")
{
test_copy<layout_right>(make_iota<int>(N), 3, 16);
}
// src: (16,3):(1,16)
// dst: (16,3):(1,16)
TEST_CASE("copy d2d vectorize (16,3):(1,16)", "[copy][d2d][vectorize][2d]")
{
test_copy<layout_left>(make_iota<int>(N), 16, 3);
}
// src: (16,3):(3,1)
// dst: (16,3):(3,1)
TEST_CASE("copy d2d vectorize (16,3):(3,1)", "[copy][d2d][vectorize][2d]")
{
test_copy<layout_right>(make_iota<int>(N), 16, 3);
}