Skip to content

VirtAPI: reuse a per-process cache in the Uncached calls (performance) #367

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

Closed
ramic-k wants to merge 1 commit into from
Closed

VirtAPI: reuse a per-process cache in the Uncached calls (performance) #367

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
20 changes: 14 additions & 6 deletions ncrystal_core/src/virtualapi/NCVirtAPI_Type1_v1_impl.hh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,7 @@
#include "NCrystal/virtualapi/NCVirtAPI_Type1_v1.hh"
#include "NCrystal/factories/NCFactImpl.hh"
#include "NCVirtAPIUtils.hh"
#include <unordered_map>

namespace NCRYSTAL_NAMESPACE {

Expand Down Expand Up @@ -56,13 +57,21 @@ namespace NCRYSTAL_NAMESPACE {
delete reinterpret_cast<const ScatterProcess*>(sp);
}

// Reuse a persistent cache per (thread, process) instead of allocating a
// fresh one on every call. The cache is keyed by process pointer so it is
// never shared between processes: accessCache does not check process
// identity, and m_nHistory (the only reset trigger) is not unique per
// process, so a single shared cache could be applied to the wrong process.
// The cross-section and scatter paths use separate maps, matching the two
// original CachePtr objects. thread_local keeps it thread-safe without
// locks, and replaces the per-call allocation the original flagged as
// "Fully MT safe, fully inefficient".
double crossSectionUncached( const PubScatterProcess& pub_sp,
const double* n ) const override
{
auto sp = reinterpret_cast<const ScatterProcess*>(&pub_sp);
CachePtr dummycache;//<--- Fully MT safe, fully inefficient. To be
//revisited in a future api version!
return sp->procptr->crossSection( dummycache,
thread_local std::unordered_map<const ScatterProcess*, CachePtr> xs_caches;
return sp->procptr->crossSection( xs_caches[sp],
NeutronEnergy{ n[0] },
NeutronDirection( n[1], n[2], n[3] )
).dbl();
Expand All @@ -75,10 +84,9 @@ namespace NCRYSTAL_NAMESPACE {
double* n ) const override
{
auto sp = reinterpret_cast<const ScatterProcess*>(&pub_sp);
CachePtr dummycache;//<--- Fully MT safe, fully inefficient. To be
//revisited in a future api version!
thread_local std::unordered_map<const ScatterProcess*, CachePtr> scat_caches;
VirtAPIUtils::RNGWrapper rng( &rng_fct );
auto out = sp->procptr->sampleScatter( dummycache, rng,
auto out = sp->procptr->sampleScatter( scat_caches[sp], rng,
NeutronEnergy{ n[0] },
NeutronDirection( n[1],
n[2],
Expand Down
Loading