64-BitBrainstorm_2026/AMSS-NCKU
8
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Cache GSL in SyncPlan and apply async Sync to Z4c_class

Browse Source 
Major optimization: Pre-build grid segment lists (GSLs) once per Step() call
via SyncPreparePlan(), then reuse them across all 4 RK4 substep SyncBegin calls
via SyncBeginWithPlan(). This eliminates the O(cpusize * blocks^2) GSL rebuild
cost that was incurred on every ghost zone exchange.

Applied async SyncBegin/SyncEnd overlap pattern to Z4c_class.C (ABEtype==2,
the default configuration), which was still using blocking Parallel::Sync.
Both the regular and CPBC variants of Z4c Step() are now optimized.

Co-authored-by: copilot-swe-agent[bot] <198982749+copilot@users.noreply.github.com>
This commit is contained in:
ianchb
2026-02-08 08:36:21 +00:00
parent a918dc103e
commit afd4006da2
4 changed files with 225 additions and 25 deletions
Download Patch File Download Diff File Expand all files Collapse all files

21
AMSS_NCKU_source/Parallel.h
View File

@@ -98,6 +98,7 @@ namespace Parallel
MyList<var> *VarList1;
MyList<var> *VarList2;
int Symmetry;
bool owns_gsl; // true if this state owns and should free the GSLs
};
struct SyncHandle
{
@@ -107,6 +108,26 @@ namespace Parallel
SyncHandle *SyncBegin(Patch *Pat, MyList<var> *VarList, int Symmetry);
SyncHandle *SyncBegin(MyList<Patch> *PatL, MyList<var> *VarList, int Symmetry);
void SyncEnd(SyncHandle *handle);
// Cached GSL plan: pre-build grid segment lists once, reuse across multiple Sync calls
struct SyncPlanEntry
{
int cpusize;
MyList<gridseg> **transfer_src;
MyList<gridseg> **transfer_dst;
MyList<gridseg> **src;
MyList<gridseg> *dst;
};
struct SyncPlan
{
SyncPlanEntry *entries;
int num_entries;
int Symmetry;
};
SyncPlan *SyncPreparePlan(MyList<Patch> *PatL, int Symmetry);
void SyncFreePlan(SyncPlan *plan);
SyncHandle *SyncBeginWithPlan(SyncPlan *plan, MyList<var> *VarList);
SyncHandle *SyncBeginWithPlan(SyncPlan *plan, MyList<var> *VarList1, MyList<var> *VarList2);
void OutBdLow2Hi(Patch *Patc, Patch *Patf,
MyList<var> *VarList1 /* source */, MyList<var> *VarList2 /* target */,
int Symmetry);