Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add sanity checks for predicates on vectors #702

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 3 additions & 0 deletions cmake/traccc-compiler-options-cuda.cmake
Original file line number Diff line number Diff line change
Expand Up @@ -27,6 +27,9 @@ set( CMAKE_CUDA_ARCHITECTURES "52" CACHE STRING
# not marked with __device__.
traccc_add_flag( CMAKE_CUDA_FLAGS "--expt-relaxed-constexpr" )

# Allow the use of lambdas with __device__ specifiers.
traccc_add_flag( CMAKE_CUDA_FLAGS "--extended-lambda" )

# Make CUDA generate debug symbols for the device code as well in a debug
# build.
traccc_add_flag( CMAKE_CUDA_FLAGS_DEBUG "-G --keep -src-in-ptx" )
Expand Down
129 changes: 129 additions & 0 deletions device/cuda/src/sanity/predicate.cuh
Original file line number Diff line number Diff line change
@@ -0,0 +1,129 @@
/**
* traccc library, part of the ACTS project (R&D line)
*
* (c) 2024 CERN for the benefit of the ACTS project
*
* Mozilla Public License Version 2.0
*/

#pragma once

// Project include(s).
#include "../utils/cuda_error_handling.hpp"
#include "traccc/cuda/utils/stream.hpp"

// VecMem include(s).
#include <vecmem/containers/data/vector_view.hpp>
#include <vecmem/containers/device_vector.hpp>
#include <vecmem/memory/memory_resource.hpp>
#include <vecmem/memory/unique_ptr.hpp>
#include <vecmem/utils/copy.hpp>

// CUDA include
#include <cuda_runtime.h>

// System include
#include <concepts>

namespace traccc::cuda {
namespace kernels {
template <typename P, typename T>
requires std::predicate<P, T> __global__ void true_for_all_kernel(
P projection, vecmem::data::vector_view<T> _in, bool* out) {
int tid = threadIdx.x + blockIdx.x * blockDim.x;

vecmem::device_vector<T> in(_in);

if (tid < in.size()) {
if (!projection(in.at(tid))) {
*out = false;
}
}
}
} // namespace kernels

/**
* @brief Sanity check that a predicate is true for all elements of a vector.
*
* @note This function runs in O(n) time.
*
* @tparam P The type of the predicate.
* @tparam T The type of the vector.
* @param predicate A projection object of type `P`.
* @param mr A memory resource used for allocating intermediate memory.
* @param copy A copy object.
* @param stream A wrapped CUDA stream.
* @param vector The vector which to check for contiguity.
* @return true If `predicate` is true for all elements of `vector`.
* @return false Otherwise.
*/
template <typename P, typename T>
requires std::predicate<P, T> bool true_for_all(
P&& predicate, vecmem::memory_resource& mr, vecmem::copy& copy,
stream& stream, vecmem::data::vector_view<T> vector) {
// This should never be a performance-critical step, so we can keep the
// block size fixed.
constexpr int block_size = 512;

cudaStream_t cuda_stream =
reinterpret_cast<cudaStream_t>(stream.cudaStream());

// Grab the number of elements in our vector.
const std::uint32_t n = copy.get_size(vector);

// Allocate memory for outputs, then set them up.
vecmem::unique_alloc_ptr<bool> device_out =
vecmem::make_unique_alloc<bool>(mr);

bool initial_out = true;

TRACCC_CUDA_ERROR_CHECK(
cudaMemcpyAsync(device_out.get(), &initial_out, sizeof(bool),
cudaMemcpyHostToDevice, cuda_stream));

// Launch the main kernel.
kernels::true_for_all_kernel<P, T>
<<<(n + block_size - 1) / block_size, block_size, 0, cuda_stream>>>(
predicate, vector, device_out.get());

TRACCC_CUDA_ERROR_CHECK(cudaGetLastError());

// Copy the total number of squashed elements, e.g. the size of the
// resulting vector.
bool host_out;

TRACCC_CUDA_ERROR_CHECK(
cudaMemcpyAsync(&host_out, device_out.get(), sizeof(bool),
cudaMemcpyDeviceToHost, cuda_stream));

stream.synchronize();

return host_out;
}

template <typename P, typename T>
requires std::predicate<P, T> bool false_for_all(
P&& projection, vecmem::memory_resource& mr, vecmem::copy& copy,
stream& stream, vecmem::data::vector_view<T> vector) {
return true_for_all(
[projection] __device__<typename... Args>(Args && ... args) {
return !projection(std::forward<Args>(args)...);
niermann999 marked this conversation as resolved.
Show resolved Hide resolved
},
mr, copy, stream, vector);
}

template <typename P, typename T>
requires std::predicate<P, T> bool true_for_any(
P&& projection, vecmem::memory_resource& mr, vecmem::copy& copy,
stream& stream, vecmem::data::vector_view<T> vector) {
return !false_for_all(std::forward<P>(projection), mr, copy, stream,
vector);
}

template <typename P, typename T>
requires std::predicate<P, T> bool false_for_any(
P&& projection, vecmem::memory_resource& mr, vecmem::copy& copy,
stream& stream, vecmem::data::vector_view<T> vector) {
return !true_for_all(std::forward<P>(projection), mr, copy, stream, vector);
}
} // namespace traccc::cuda
130 changes: 130 additions & 0 deletions device/sycl/src/sanity/predicate.hpp
Original file line number Diff line number Diff line change
@@ -0,0 +1,130 @@
/**
* traccc library, part of the ACTS project (R&D line)
*
* (c) 2024 CERN for the benefit of the ACTS project
*
* Mozilla Public License Version 2.0
*/

#pragma once

// Project include(s).
#include <traccc/sycl/utils/queue_wrapper.hpp>

#include "../utils/get_queue.hpp"

// VecMem include(s).
#include <vecmem/containers/data/vector_view.hpp>
#include <vecmem/containers/device_vector.hpp>
#include <vecmem/memory/memory_resource.hpp>
#include <vecmem/memory/unique_ptr.hpp>
#include <vecmem/utils/copy.hpp>

// SYCL include
#include <CL/sycl.hpp>

// System include
#include <concepts>

namespace traccc::sycl {
namespace kernels {
template <typename P, typename T>
class TrueForAllPredicate {};
} // namespace kernels

/**
* @brief Sanity check that a predicate is true for all elements of a vector.
*
* @note This function runs in O(n) time.
*
* @tparam P The type of the predicate.
* @tparam T The type of the vector.
* @param predicate A projection object of type `P`.
* @param mr A memory resource used for allocating intermediate memory.
* @param copy A copy object.
* @param stream A wrapped CUDA stream.
* @param vector The vector which to check for contiguity.
* @return true If `predicate` is true for all elements of `vector`.
* @return false Otherwise.
*/
template <typename P, typename T>
requires std::predicate<P, T> bool true_for_all(
P&& predicate, vecmem::memory_resource& mr, vecmem::copy& copy,
queue_wrapper& queue_wrapper, vecmem::data::vector_view<T> vector) {
// This should never be a performance-critical step, so we can keep the
// block size fixed.
constexpr int block_size = 512;

cl::sycl::queue& queue = details::get_queue(queue_wrapper);

// Grab the number of elements in our vector.
const std::uint32_t n = copy.get_size(vector);

// Allocate memory for outputs, then set them up.
vecmem::unique_alloc_ptr<bool> device_out =
vecmem::make_unique_alloc<bool>(mr);

bool initial_out = true;

cl::sycl::event kernel1_memcpy1 =
queue.memcpy(device_out.get(), &initial_out, sizeof(bool));

// Launch the main kernel.
cl::sycl::nd_range<1> kernel_range{
cl::sycl::range<1>(((n + block_size - 1) / block_size) * block_size),
cl::sycl::range<1>(block_size)};

cl::sycl::event kernel1 = queue.submit([&](cl::sycl::handler& h) {
h.depends_on(kernel1_memcpy1);
h.parallel_for<kernels::TrueForAllPredicate<P, T>>(
kernel_range, [projection, vector,
out = device_out.get()](cl::sycl::nd_item<1> item) {
std::size_t tid = item.get_global_linear_id();

vecmem::device_vector<T> in(vector);

if (tid < in.size()) {
if (!projection(in.at(tid))) {
*out = false;
}
}
});
});

// Copy the total number of squashed elements, e.g. the size of the
// resulting vector.
bool host_out;

queue.memcpy(&host_out, out.get(), sizeof(bool), {kernel1})
.wait_and_throw();

return host_out;
}

template <typename P, typename T>
requires std::predicate<P, T> bool false_for_all(
P&& projection, vecmem::memory_resource& mr, vecmem::copy& copy,
queue_wrapper& queue_wrapper, vecmem::data::vector_view<T> vector) {
return true_for_all(
[projection]<typename... Args>(Args && ... args) {
return !projection(std::forward<Args>(args)...);
niermann999 marked this conversation as resolved.
Show resolved Hide resolved
},
mr, copy, queue_wrapper, vector);
}

template <typename P, typename T>
requires std::predicate<P, T> bool true_for_any(
P&& projection, vecmem::memory_resource& mr, vecmem::copy& copy,
queue_wrapper& queue_wrapper, vecmem::data::vector_view<T> vector) {
return !false_for_all(std::forward<P>(projection), mr, copy, queue_wrapper,
vector);
}

template <typename P, typename T>
requires std::predicate<P, T> bool false_for_any(
P&& projection, vecmem::memory_resource& mr, vecmem::copy& copy,
queue_wrapper& queue_wrapper, vecmem::data::vector_view<T> vector) {
return !true_for_all(std::forward<P>(projection), mr, copy, queue_wrapper,
vector);
}
} // namespace traccc::sycl
1 change: 1 addition & 0 deletions tests/cuda/CMakeLists.txt
Original file line number Diff line number Diff line change
Expand Up @@ -44,6 +44,7 @@ traccc_add_test(
test_unique_lock.cu
test_sanity_contiguous_on.cu
test_sanity_ordered_on.cu
test_sanity_predicate.cu
test_sort.cu

LINK_LIBRARIES
Expand Down
Loading
Loading