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

KokkosSparse spgemm: handling some compile errors and warnings #140

Merged
merged 1 commit into from
Jan 5, 2018
Merged

KokkosSparse spgemm: handling some compile errors and warnings #140

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
66 changes: 32 additions & 34 deletions src/sparse/impl/KokkosSparse_spgemm_impl_kkmem.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -97,10 +97,8 @@ struct KokkosSPGEMM
nnz_lno_t team_shared_memory_hash_func;
nnz_lno_t team_shmem_hash_size;

const nnz_lno_t first_nnz_lno_t_bit;
const nnz_lno_t rest_nnz_lno_t_bit;
nnz_lno_t team_cuckoo_key_size, team_cuckoo_hash_func;
const int MAX_FIRST_LEVEL_RATIO;

nnz_lno_t max_first_level_hash_size;
row_lno_persistent_work_view_t flops_per_row;
PortableNumericCHASH(
Expand Down Expand Up @@ -157,10 +155,8 @@ struct KokkosSPGEMM
team_shmem_key_size(),
team_shared_memory_hash_func(),
team_shmem_hash_size(1),
first_nnz_lno_t_bit (1 << (sizeof(nnz_lno_t) * 8 - 1)),
rest_nnz_lno_t_bit (~first_nnz_lno_t_bit),
team_cuckoo_key_size (1),
team_cuckoo_hash_func(1), MAX_FIRST_LEVEL_RATIO(0.90), max_first_level_hash_size( 1), flops_per_row(flops_per_row_)
team_cuckoo_hash_func(1), max_first_level_hash_size( 1), flops_per_row(flops_per_row_)


{
Expand Down Expand Up @@ -191,7 +187,6 @@ struct KokkosSPGEMM
if (KOKKOSKERNELS_VERBOSE_){
std::cout << "\t\tNumericCMEM -- adjusted hashsize:" << thread_shmem_hash_size << " shmem_key_size:" << thread_shmem_key_size << std::endl;
std::cout << "\t\tNumericCMEM -- adjusted team hashsize:" << team_shmem_hash_size << " team_shmem_key_size:" << team_shmem_key_size << std::endl;
std::cout << "\t\tfirst_nnz_lno_t_bit:" << first_nnz_lno_t_bit << " rest_nnz_lno_t_bit:" << rest_nnz_lno_t_bit << std::endl;
std::cout << "\t\tteam_cuckoo_key_size:" << team_cuckoo_key_size << " team_cuckoo_hash_func:" << team_cuckoo_hash_func << " max_first_level_hash_size:" << max_first_level_hash_size << std::endl;
std::cout << "\t\tpow2_hash_size:" << pow2_hash_size << " pow2_hash_func:" << pow2_hash_func << std::endl;
}
Expand Down Expand Up @@ -613,6 +608,7 @@ struct KokkosSPGEMM
const nnz_lno_t team_row_end = KOKKOSKERNELS_MACRO_MIN(team_row_begin + team_work_size, numrows);
char *all_shared_memory = (char *) (teamMember.team_shmem().get_shmem(shared_memory_size));

const nnz_lno_t init_value = -1;
volatile nnz_lno_t *used_hash_sizes = (volatile nnz_lno_t *) (all_shared_memory);
all_shared_memory += sizeof(nnz_lno_t) * 2;
//holds the keys
Expand Down Expand Up @@ -676,7 +672,7 @@ struct KokkosSPGEMM
// not needed as team_cuckoo_key_size is always pow2. + (team_cuckoo_key_size & (vector_size - 1)) * 1;
Kokkos::parallel_for( Kokkos::TeamThreadRange(teamMember, num_threads), [&] (nnz_lno_t teamind) {
Kokkos::parallel_for( Kokkos::ThreadVectorRange(teamMember, vector_size ), [&] (nnz_lno_t i) {
keys[teamind * vector_size + i] = -1; vals[teamind * vector_size + i] = 0;
keys[teamind * vector_size + i] = init_value; vals[teamind * vector_size + i] = 0;
});
});
}
Expand All @@ -689,7 +685,7 @@ struct KokkosSPGEMM

bool insert_is_on = true;
const size_type a_col_begin_offset = row_mapA[row_index];
const nnz_lno_t left_work = row_mapA[row_index + 1] - a_col_begin_offset;


nnz_lno_t a_col_ind = entriesA[a_col_begin_offset];
scalar_t a_col_val = valuesA[a_col_begin_offset];
Expand All @@ -707,7 +703,7 @@ struct KokkosSPGEMM
for (nnz_lno_t vector_read_shift = vector_shift; vector_read_shift< row_flops; vector_read_shift += bs){
{
nnz_lno_t my_b_col_shift = vector_read_shift - flops_on_the_left_of_offsett;
nnz_lno_t my_b_col = -1; scalar_t my_b_val = 0; nnz_lno_t hash = -1;
nnz_lno_t my_b_col = init_value; scalar_t my_b_val = 0; nnz_lno_t hash = init_value;
int fail = 0;

if (my_b_col_shift >= current_a_column_flops ){
Expand Down Expand Up @@ -738,12 +734,12 @@ struct KokkosSPGEMM
fail = 0;
break;
}
else if (keys[trial] == -1){
else if (keys[trial] == init_value){
if (!insert_is_on) {
try_to_insert = false;
break;
}
else if (Kokkos::atomic_compare_exchange_strong(keys + trial, -1, my_b_col)){
else if (Kokkos::atomic_compare_exchange_strong(keys + trial, init_value, my_b_col)){
Kokkos::atomic_add(vals + trial, my_b_val);
Kokkos::atomic_increment(used_hash_sizes);
if (used_hash_sizes[0] > max_first_level_hash_size) insert_is_on = false;
Expand All @@ -764,11 +760,11 @@ struct KokkosSPGEMM
fail = 0;
break;
}
else if (keys[trial] == -1){
else if (keys[trial] == init_value){
if (!insert_is_on) {
break;
}
else if (Kokkos::atomic_compare_exchange_strong(keys + trial, -1, my_b_col)){
else if (Kokkos::atomic_compare_exchange_strong(keys + trial, init_value, my_b_col)){
Kokkos::atomic_add(vals + trial, my_b_val);
Kokkos::atomic_increment(used_hash_sizes);
if (used_hash_sizes[0] > max_first_level_hash_size) insert_is_on = false;
Expand All @@ -793,8 +789,8 @@ struct KokkosSPGEMM
fail = 0;
break;
}
else if (global_acc_row_keys[trial ] == -1){
if (Kokkos::atomic_compare_exchange_strong(global_acc_row_keys + trial , -1, my_b_col)){
else if (global_acc_row_keys[trial ] == init_value){
if (Kokkos::atomic_compare_exchange_strong(global_acc_row_keys + trial , init_value, my_b_col)){
Kokkos::atomic_add(global_acc_row_vals + trial , my_b_val);
//Kokkos::atomic_increment(used_hash_sizes + 1);
//c_row_vals[trial] = my_b_val;
Expand All @@ -814,8 +810,8 @@ struct KokkosSPGEMM

break;
}
else if (global_acc_row_keys[trial ] == -1){
if (Kokkos::atomic_compare_exchange_strong(global_acc_row_keys + trial , -1, my_b_col)){
else if (global_acc_row_keys[trial ] == init_value){
if (Kokkos::atomic_compare_exchange_strong(global_acc_row_keys + trial , init_value, my_b_col)){
//Kokkos::atomic_increment(used_hash_sizes + 1);
Kokkos::atomic_add(global_acc_row_vals + trial , my_b_val);
//c_row_vals[trial] = my_b_val;
Expand All @@ -840,7 +836,7 @@ struct KokkosSPGEMM

for (nnz_lno_t my_index = vector_shift; my_index < pow2_hash_size;my_index += bs){
nnz_lno_t my_b_col = global_acc_row_keys[my_index];
if (my_b_col != -1){
if (my_b_col != init_value){
scalar_t my_b_val = global_acc_row_vals[my_index];
int fail = 1;
{
Expand All @@ -854,7 +850,7 @@ struct KokkosSPGEMM
fail = 0;
break;
}
else if (keys[trial] == -1){
else if (keys[trial] == init_value){
break;
}
}
Expand All @@ -866,7 +862,7 @@ struct KokkosSPGEMM
fail = 0;
break;
}
else if (keys[trial] == -1){
else if (keys[trial] == init_value){
break;
}
}
Expand All @@ -878,7 +874,7 @@ struct KokkosSPGEMM
c_row[write_index] = my_b_col;
c_row_vals[write_index] = my_b_val;
}
global_acc_row_keys[my_index] = -1;
global_acc_row_keys[my_index] = init_value;
}
}

Expand All @@ -891,7 +887,7 @@ struct KokkosSPGEMM

for (nnz_lno_t my_index = vector_shift; my_index < team_cuckoo_key_size;my_index += bs){
nnz_lno_t my_key = keys[my_index];
if (my_key != -1){
if (my_key != init_value){
scalar_t my_val = vals[my_index];
nnz_lno_t write_index = 0;
write_index = Kokkos::atomic_fetch_add(used_hash_sizes + 1, 1);
Expand All @@ -909,6 +905,8 @@ struct KokkosSPGEMM
KOKKOS_INLINE_FUNCTION
void operator()(const GPUTag4&, const team_member_t & teamMember) const {


const nnz_lno_t init_value = -1;
nnz_lno_t team_row_begin = teamMember.league_rank() * team_work_size;
const nnz_lno_t team_row_end = KOKKOSKERNELS_MACRO_MIN(team_row_begin + team_work_size, numrows);

Expand Down Expand Up @@ -941,7 +939,7 @@ struct KokkosSPGEMM
teamMember.team_barrier();
#endif
const size_type c_row_begin = rowmapC[row_index];
const size_type c_row_end = rowmapC[row_index + 1];
//const size_type c_row_end = rowmapC[row_index + 1];
//const nnz_lno_t c_row_size = c_row_end - c_row_begin;
nnz_lno_t *c_row = entriesC.data() + c_row_begin;
scalar_t *c_row_vals = valuesC.data() + c_row_begin;
Expand All @@ -953,7 +951,7 @@ struct KokkosSPGEMM
//nnz_lno_t team_shift = teamind * vector_size;
//nnz_lno_t work_to_handle = KOKKOSKERNELS_MACRO_MIN(vector_size, team_shmem_hash_size - team_shift);
Kokkos::parallel_for( Kokkos::ThreadVectorRange(teamMember, vector_size ), [&] (nnz_lno_t i) {
keys[teamind * vector_size + i] = -1; vals[teamind * vector_size + i] = 0;
keys[teamind * vector_size + i] = init_value; vals[teamind * vector_size + i] = 0;
});
});
}
Expand All @@ -965,8 +963,8 @@ struct KokkosSPGEMM
Kokkos::single(Kokkos::PerTeam(teamMember),[&] () {

for (int i = 0; i < team_shmem_hash_size; ++i){
if (begins[i] != -1){
std::cout << "row_index:" << row_index << " i:" << i << " team_shmem_hash_size:" << team_shmem_hash_size << " is not -1 begins[i]:" << begins[i] << std::endl;
if (begins[i] != init_value){
std::cout << "row_index:" << row_index << " i:" << i << " team_shmem_hash_size:" << team_shmem_hash_size << " is not init_value begins[i]:" << begins[i] << std::endl;
}
}
});
Expand All @@ -990,7 +988,7 @@ struct KokkosSPGEMM
nnz_lno_t *globally_used_hash_indices = NULL;
#endif
const size_type a_col_begin_offset = row_mapA[row_index];
const nnz_lno_t left_work = row_mapA[row_index + 1] - a_col_begin_offset;


nnz_lno_t a_col_ind = entriesA[a_col_begin_offset];
scalar_t a_col_val = valuesA[a_col_begin_offset];
Expand All @@ -1012,7 +1010,7 @@ struct KokkosSPGEMM
for (nnz_lno_t vector_read_shift = vector_shift; vector_read_shift< row_flops; vector_read_shift += bs){
{
nnz_lno_t my_b_col_shift = vector_read_shift - flops_on_the_left_of_offsett;
nnz_lno_t my_b_col = -1; scalar_t my_b_val = 0; nnz_lno_t hash = -1;
nnz_lno_t my_b_col = init_value; scalar_t my_b_val = 0; nnz_lno_t hash = init_value;
int fail = 0;


Expand Down Expand Up @@ -1046,8 +1044,8 @@ struct KokkosSPGEMM
fail = 0;
break;
}
else if (keys[trial] == -1){
if (Kokkos::atomic_compare_exchange_strong(keys + trial, -1, my_b_col)){
else if (keys[trial] == init_value){
if (Kokkos::atomic_compare_exchange_strong(keys + trial, init_value, my_b_col)){
Kokkos::atomic_add(vals + trial, my_b_val);
fail = 0;
break;
Expand All @@ -1065,8 +1063,8 @@ struct KokkosSPGEMM
fail = 0;
break;
}
else if (keys[trial] == -1){
if (Kokkos::atomic_compare_exchange_strong(keys + trial, -1, my_b_col)){
else if (keys[trial] == init_value){
if (Kokkos::atomic_compare_exchange_strong(keys + trial, init_value, my_b_col)){
Kokkos::atomic_add(vals + trial, my_b_val);
fail = 0;
break;
Expand All @@ -1084,7 +1082,7 @@ struct KokkosSPGEMM
teamMember.team_barrier();
for (nnz_lno_t my_index = vector_shift; my_index < team_cuckoo_key_size;my_index += bs){
nnz_lno_t my_key = keys[my_index];
if (my_key != -1){
if (my_key != init_value){
scalar_t my_val = vals[my_index];
nnz_lno_t write_index = Kokkos::atomic_fetch_add(used_hash_sizes, 1);
c_row[write_index] = my_key;
Expand Down
3 changes: 2 additions & 1 deletion src/sparse/impl/KokkosSparse_spgemm_impl_symbolic.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2068,9 +2068,10 @@ void KokkosSPGEMM
std::cout << "\tStructureC Kernel time:" << timer1.seconds() << std::endl<< std::endl;
}

int read_write_cost = this->handle->get_spgemm_handle()->get_read_write_cost_calc();

#if 0
int read_write_cost = this->handle->get_spgemm_handle()->get_read_write_cost_calc();

if (read_write_cost && KOKKOSKERNELS_VERBOSE)
{
std::cout << "\t\t!!!!DELETE THIS PART!!!! PRINTING STATS HERE!!!!!" << std::endl;
Expand Down