Skip to content

Replace std::vector() with sycl::span() by prompt from OneMKL #1235

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.

Sign up for GitHub

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

Merged
merged 2 commits into from
Nov 15, 2022
Merged

Replace std::vector() with sycl::span() by prompt from OneMKL #1235

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
6c4cf11
Replace std::vector() with sycl::span() by prompt from OneMKL
antonwolfy Nov 15, 2022
85d38b1
Use std::vector() in constructors with old OneMKL
antonwolfy Nov 15, 2022
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
52 changes: 36 additions & 16 deletions dpnp/backend/kernels/dpnp_krnl_random.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -973,33 +973,43 @@ DPCTLSyclEventRef dpnp_rng_multinomial_c(DPCTLSyclQueueRef q_ref,
else
{
DPNPC_ptr_adapter<double> p_ptr(q_ref, p_in, p_size, true);
const double* p = p_ptr.get_ptr();
std::vector<double> p_vec(p, p + p_size);
double* p_data = p_ptr.get_ptr();

// size = size
// `result` is a array for random numbers
// `size` is a `result`'s len. `size = n * p.size()`
// `size` is a `result`'s len. `size = n * p_size`
// `n` is a number of random values to be generated.
size_t n = size / p_vec.size();
size_t n = size / p_size;

size_t is_cpu_queue = dpnp_queue_is_cpu_c();

// math library supports the distribution generation on GPU device with input parameters
// which follow the condition
if (is_cpu_queue || (!is_cpu_queue && (p_size >= ((size_t)ntrial * 16)) && (ntrial <= 16)))
{
DPNPC_ptr_adapter<std::int32_t> result_ptr(q_ref, result, size, false, true);
std::int32_t* result1 = result_ptr.get_ptr();
mkl_rng::multinomial<std::int32_t> distribution(ntrial, p_vec);
DPNPC_ptr_adapter<_DataType> result_ptr(q_ref, result, size, true, true);
_DataType* result1 = result_ptr.get_ptr();

#if (INTEL_MKL_VERSION < __INTEL_MKL_2023_SWITCHOVER)
std::vector<double> p(p_data, p_data + p_size);
#else
auto p = sycl::span<double>{p_data, p_size};
#endif
mkl_rng::multinomial<_DataType> distribution(ntrial, p);

// perform generation
event_out = mkl_rng::generate(distribution, DPNP_RNG_ENGINE, n, result1);
event_ref = reinterpret_cast<DPCTLSyclEventRef>(&event_out);

p_ptr.depends_on(event_out);
result_ptr.depends_on(event_out);
}
else
{
DPNPC_ptr_adapter<std::int32_t> result_ptr(q_ref, result, size, true, true);
std::int32_t* result1 = result_ptr.get_ptr();
DPNPC_ptr_adapter<_DataType> result_ptr(q_ref, result, size, true, true);
_DataType* result1 = result_ptr.get_ptr();
int errcode = viRngMultinomial(
VSL_RNG_METHOD_MULTINOMIAL_MULTPOISSON, get_rng_stream(), n, result1, ntrial, p_size, p);
VSL_RNG_METHOD_MULTINOMIAL_MULTPOISSON, get_rng_stream(), n, result1, ntrial, p_size, p_data);
if (errcode != VSL_STATUS_OK)
{
throw std::runtime_error("DPNP RNG Error: dpnp_rng_multinomial_c() failed.");
Expand All @@ -1023,6 +1033,7 @@ void dpnp_rng_multinomial_c(
size,
dep_event_vec_ref);
DPCTLEvent_WaitAndThrow(event_ref);
DPCTLEvent_Delete(event_ref);
}

template <typename _DataType>
Expand Down Expand Up @@ -1065,24 +1076,32 @@ DPCTLSyclEventRef dpnp_rng_multivariate_normal_c(DPCTLSyclQueueRef q_ref,
sycl::queue q = *(reinterpret_cast<sycl::queue*>(q_ref));

DPNPC_ptr_adapter<double> mean_ptr(q_ref, mean_in, mean_size, true);
const double* mean = mean_ptr.get_ptr();
double* mean_data = mean_ptr.get_ptr();
DPNPC_ptr_adapter<double> cov_ptr(q_ref, cov_in, cov_size, true);
const double* cov = cov_ptr.get_ptr();
double* cov_data = cov_ptr.get_ptr();

_DataType* result1 = reinterpret_cast<_DataType*>(result);
_DataType* result1 = static_cast<_DataType *>(result);

std::vector<double> mean_vec(mean, mean + mean_size);
std::vector<double> cov_vec(cov, cov + cov_size);
#if (INTEL_MKL_VERSION < __INTEL_MKL_2023_SWITCHOVER)
std::vector<double> mean(mean_data, mean_data + mean_size);
std::vector<double> cov(cov_data, cov_data + cov_size);
#else
auto mean = sycl::span<double>{mean_data, mean_size};
auto cov = sycl::span<double>{cov_data, cov_size};
#endif

// `result` is a array for random numbers
// `size` is a `result`'s len.
// `size1` is a number of random values to be generated for each dimension.
size_t size1 = size / dimen;

mkl_rng::gaussian_mv<_DataType> distribution(dimen, mean_vec, cov_vec);
mkl_rng::gaussian_mv<_DataType> distribution(dimen, mean, cov);
auto event_out = mkl_rng::generate(distribution, DPNP_RNG_ENGINE, size1, result1);
event_ref = reinterpret_cast<DPCTLSyclEventRef>(&event_out);

mean_ptr.depends_on(event_out);
cov_ptr.depends_on(event_out);

return DPCTLEvent_Copy(event_ref);
}

Expand All @@ -1107,6 +1126,7 @@ void dpnp_rng_multivariate_normal_c(void* result,
size,
dep_event_vec_ref);
DPCTLEvent_WaitAndThrow(event_ref);
DPCTLEvent_Delete(event_ref);
}

template <typename _DataType>
Expand Down
7 changes: 7 additions & 0 deletions dpnp/backend/src/dpnp_utils.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -49,6 +49,13 @@
#define __SYCL_COMPILER_2023_SWITCHOVER 20221102L
#endif

/**
* Version of Intel MKL at which transition to OneMKL release 2023.0.0 occurs.
*/
#ifndef __INTEL_MKL_2023_SWITCHOVER
#define __INTEL_MKL_2023_SWITCHOVER 20230000
#endif

/**
* @defgroup BACKEND_UTILS Backend C++ library utilities
* @{
Expand Down
3 changes: 3 additions & 0 deletions tests/skipped_tests.tbl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -199,9 +199,12 @@ tests/test_linalg.py::test_svd[(5,3)-float64]
tests/test_linalg.py::test_svd[(16,16)-float64]

tests/test_mathematical.py::TestGradient::test_gradient_y1_dx[3.5-array1]

tests/test_random.py::TestPermutationsTestShuffle::test_shuffle1[lambda x: dpnp.asarray([[i, i] for i in x])]
tests/test_random.py::TestPermutationsTestShuffle::test_shuffle1[lambda x: (dpnp.asarray([(i, i) for i in x], [("a", int), ("b", int)]).view(dpnp.recarray))]
tests/test_random.py::TestPermutationsTestShuffle::test_shuffle1[lambda x: dpnp.asarray([(i, i) for i in x], [("a", object), ("b", dpnp.int32)])]]
tests/test_random.py::TestPermutationsTestShuffle::test_shuffle1[lambda x: dpnp.asarray(x).astype(dpnp.int8)]

tests/test_sort.py::test_partition[[[1, 0], [3, 0]]-float32-1]
tests/third_party/cupy/binary_tests/test_elementwise.py::TestElementwise::test_bitwise_and
tests/third_party/cupy/binary_tests/test_elementwise.py::TestElementwise::test_bitwise_or
Expand Down
6 changes: 1 addition & 5 deletions tests/skipped_tests_gpu.tbl
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,3 @@
tests/test_random.py::TestDistributionsMultinomial::test_extreme_value
tests/test_random.py::TestDistributionsMultinomial::test_seed1
tests/test_random.py::TestDistributionsMultivariateNormal::test_moments
tests/test_random.py::TestDistributionsMultivariateNormal::test_output_shape_check
tests/test_random.py::TestDistributionsMultivariateNormal::test_seed
Expand Down Expand Up @@ -219,9 +217,7 @@ tests/test_mathematical.py::TestGradient::test_gradient_y1[array2]
tests/test_mathematical.py::TestGradient::test_gradient_y1_dx[2-array0]
tests/test_mathematical.py::TestGradient::test_gradient_y1_dx[2-array1]
tests/test_mathematical.py::TestGradient::test_gradient_y1_dx[2-array2]
tests/test_random.py::TestDistributionsMultinomial::test_check_sum
tests/test_random.py::TestDistributionsMultinomial::test_moments
tests/test_random.py::TestDistributionsMultinomial::test_seed

tests/test_random.py::TestPermutationsTestShuffle::test_shuffle1[lambda x: dpnp.astype(dpnp.asarray(x), dpnp.int8)]
tests/test_random.py::TestPermutationsTestShuffle::test_shuffle1[lambda x: dpnp.astype(dpnp.asarray(x), object)]
tests/test_random.py::TestPermutationsTestShuffle::test_shuffle1[lambda x: dpnp.vstack([x, x]).T]
Expand Down