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.

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

Fix the work resource for checking diag #1440

Merged
merged 2 commits into from
Oct 25, 2023
Merged

Fix the work resource for checking diag #1440

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
e07dcd1
failed: missing diag in tail and ensure all diag
yhmtsai Oct 24, 2023
d6f5926
enough work for check and initial non-full-diag ex
yhmtsai Oct 24, 2023
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
48 changes: 28 additions & 20 deletions core/test/utils/matrix_utils_test.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -355,29 +355,37 @@ TEST(MatrixUtils, ModifyToEnsureAllDiagonalEntries)
using T = float;
using Csr = gko::matrix::Csr<T, int>;
auto exec = gko::ReferenceExecutor::create();
auto b = gko::initialize<Csr>(
{I<T>{2.0, 0.0, 1.1, 0.0}, I<T>{1.0, 2.4, 0.0, -1.0},
I<T>{0.0, -4.0, 2.2, -2.0}, I<T>{0.0, -3.0, 1.5, 1.0}},
exec);
auto check_all_diag = [](const Csr* csr) {
const auto rowptrs = csr->get_const_row_ptrs();
const auto colidxs = csr->get_const_col_idxs();
const auto ndiag =
static_cast<int>(std::min(csr->get_size()[0], csr->get_size()[1]));
bool all_diags = true;
for (int i = 0; i < ndiag; i++) {
bool has_diag = false;
for (int j = rowptrs[i]; j < rowptrs[i + 1]; j++) {
if (colidxs[j] == i) {
has_diag = true;
break;
}
}
if (!has_diag) {
all_diags = false;
break;
}
}
return all_diags;
};
auto b = gko::initialize<Csr>({I<T>{2.0, 0.0, 1.1}, I<T>{1.0, 0.0, 0.0},
I<T>{0.0, -4.0, 2.2}, I<T>{0.0, -3.0, 1.5}},
exec);
// ensure it misses some diag
bool prev_check = check_all_diag(b.get());

gko::utils::ensure_all_diagonal_entries(b.get());

const auto rowptrs = b->get_const_row_ptrs();
const auto colidxs = b->get_const_col_idxs();
bool all_diags = true;
for (int i = 0; i < 3; i++) {
bool has_diag = false;
for (int j = rowptrs[i]; j < rowptrs[i + 1]; j++) {
if (colidxs[j] == i) {
has_diag = true;
}
}
if (!has_diag) {
all_diags = false;
break;
}
}
ASSERT_TRUE(all_diags);
ASSERT_FALSE(prev_check);
ASSERT_TRUE(check_all_diag(b.get()));
}


Expand Down
7 changes: 4 additions & 3 deletions core/utils/matrix_utils.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -301,9 +301,10 @@ void ensure_all_diagonal_entries(MtxType* mtx)
using index_type = typename MtxType::index_type;
matrix_data<value_type, index_type> mdata;
mtx->write(mdata);
const auto nrows = static_cast<index_type>(mtx->get_size()[0]);
mdata.nonzeros.reserve(mtx->get_num_stored_elements() + nrows);
for (index_type i = 0; i < nrows; i++) {
const auto ndiag = static_cast<index_type>(
std::min(mtx->get_size()[0], mtx->get_size()[1]));
mdata.nonzeros.reserve(mtx->get_num_stored_elements() + ndiag);
for (index_type i = 0; i < ndiag; i++) {
mdata.nonzeros.push_back({i, i, zero<value_type>()});
}
mdata.sum_duplicates();
Expand Down
13 changes: 6 additions & 7 deletions cuda/matrix/csr_kernels.template.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1322,16 +1322,15 @@ void check_diagonal_entries_exist(
std::shared_ptr<const CudaExecutor> exec,
const matrix::Csr<ValueType, IndexType>* const mtx, bool& has_all_diags)
{
const size_type num_warps = mtx->get_size()[0];
if (num_warps > 0) {
const size_type num_blocks =
num_warps / (default_block_size / config::warp_size);
const auto num_diag = static_cast<IndexType>(
std::min(mtx->get_size()[0], mtx->get_size()[1]));
if (num_diag > 0) {
const IndexType num_blocks =
ceildiv(num_diag, default_block_size / config::warp_size);
array<bool> has_diags(exec, {true});
kernel::check_diagonal_entries<<<num_blocks, default_block_size, 0,
exec->get_stream()>>>(
static_cast<IndexType>(
std::min(mtx->get_size()[0], mtx->get_size()[1])),
mtx->get_const_row_ptrs(), mtx->get_const_col_idxs(),
num_diag, mtx->get_const_row_ptrs(), mtx->get_const_col_idxs(),
has_diags.get_data());
has_all_diags = exec->copy_val_to_host(has_diags.get_const_data());
} else {
Expand Down
13 changes: 6 additions & 7 deletions dpcpp/matrix/csr_kernels.dp.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2436,15 +2436,14 @@ void check_diagonal_entries_exist(
std::shared_ptr<const DpcppExecutor> exec,
const matrix::Csr<ValueType, IndexType>* const mtx, bool& has_all_diags)
{
const size_type num_subgroup = mtx->get_size()[0];
if (num_subgroup > 0) {
const size_type num_blocks =
num_subgroup / (default_block_size / config::warp_size);
const auto num_diag = static_cast<IndexType>(
std::min(mtx->get_size()[0], mtx->get_size()[1]));
if (num_diag > 0) {
const IndexType num_blocks =
ceildiv(num_diag, default_block_size / config::warp_size);
array<bool> has_diags(exec, {true});
kernel::check_diagonal_entries(
num_blocks, default_block_size, 0, exec->get_queue(),
static_cast<IndexType>(
std::min(mtx->get_size()[0], mtx->get_size()[1])),
num_blocks, default_block_size, 0, exec->get_queue(), num_diag,
mtx->get_const_row_ptrs(), mtx->get_const_col_idxs(),
has_diags.get_data());
has_all_diags = exec->copy_val_to_host(has_diags.get_const_data());
Expand Down
13 changes: 6 additions & 7 deletions hip/matrix/csr_kernels.template.hip.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1119,16 +1119,15 @@ void check_diagonal_entries_exist(
std::shared_ptr<const HipExecutor> exec,
const matrix::Csr<ValueType, IndexType>* const mtx, bool& has_all_diags)
{
const size_type num_warps = mtx->get_size()[0];
if (num_warps > 0) {
const size_type num_blocks =
num_warps / (default_block_size / config::warp_size);
const auto num_diag = static_cast<IndexType>(
std::min(mtx->get_size()[0], mtx->get_size()[1]));
if (num_diag > 0) {
const IndexType num_blocks =
ceildiv(num_diag, default_block_size / config::warp_size);
array<bool> has_diags(exec, {true});
kernel::check_diagonal_entries<<<num_blocks, default_block_size, 0,
exec->get_stream()>>>(
static_cast<IndexType>(
std::min(mtx->get_size()[0], mtx->get_size()[1])),
mtx->get_const_row_ptrs(), mtx->get_const_col_idxs(),
num_diag, mtx->get_const_row_ptrs(), mtx->get_const_col_idxs(),
has_diags.get_data());
has_all_diags = exec->copy_val_to_host(has_diags.get_const_data());
} else {
Expand Down
6 changes: 4 additions & 2 deletions test/matrix/csr_kernels2.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1315,10 +1315,12 @@ TEST_F(Csr, CanDetectMissingDiagonalEntry)
{
using T = double;
using Csr = Mtx;
auto ref_mtx = gen_mtx<Csr>(103, 98, 10);
auto ref_mtx = gen_mtx<Csr>(103, 104, 10);
const auto rowptrs = ref_mtx->get_row_ptrs();
const auto colidxs = ref_mtx->get_col_idxs();
const int testrow = 15;
gko::utils::ensure_all_diagonal_entries(ref_mtx.get());
// Choose the last row to ensure that kernel assign enough work
const int testrow = 102;
gko::utils::remove_diagonal_entry_from_row(ref_mtx.get(), testrow);
auto mtx = gko::clone(exec, ref_mtx);
bool has_diags = true;
Expand Down