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

Optimize memory allocation in JFD_times_dv

Browse Source 
This should reduce the pressure on the memory allocator, indirectly improving caching behavior.

Co-authored-by: copilot-swe-agent[bot] <198982749+copilot@users.noreply.github.com>
This commit is contained in:
ianchb
2026-02-07 15:55:45 +08:00
committed by jaunatisblue
parent c5f8a18ba4
commit 32ed7ec5bd
1 changed files with 20 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files

33
AMSS_NCKU_source/TwoPunctures.C
View File

@@ -2111,10 +2111,19 @@ void TwoPunctures::JFD_times_dv(int i, int j, int k, int nvar, int n1, int n2,
double sin_be, sin_be_i1, sin_be_i2, sin_be_i3, cos_be;
double dV0, dV1, dV2, dV3, dV11, dV12, dV13, dV22, dV23, dV33,
ha, ga, ga2, hb, gb, gb2, hp, gp, gp2, gagb, gagp, gbgp;
derivs dU, U;
allocate_derivs(&dU, nvar);
allocate_derivs(&U, nvar);
// Stack-allocated derivs (nvar=1) — no malloc/free!
double dU_d0[1], dU_d1[1], dU_d2[1], dU_d3[1];
double dU_d11[1], dU_d12[1], dU_d13[1], dU_d22[1], dU_d23[1], dU_d33[1];
double U_d0[1], U_d1[1], U_d2[1], U_d3[1];
double U_d11[1], U_d12[1], U_d13[1], U_d22[1], U_d23[1], U_d33[1];
derivs dU, U;
dU.d0=dU_d0; dU.d1=dU_d1; dU.d2=dU_d2; dU.d3=dU_d3;
dU.d11=dU_d11; dU.d12=dU_d12; dU.d13=dU_d13;
dU.d22=dU_d22; dU.d23=dU_d23; dU.d33=dU_d33;
U.d0=U_d0; U.d1=U_d1; U.d2=U_d2; U.d3=U_d3;
U.d11=U_d11; U.d12=U_d12; U.d13=U_d13;
U.d22=U_d22; U.d23=U_d23; U.d33=U_d33;
if (k < 0)
k = k + n3;
@@ -2182,12 +2191,9 @@ void TwoPunctures::JFD_times_dv(int i, int j, int k, int nvar, int n1, int n2,
dV11 = ga2 * (dv.d0[ipcc] + dv.d0[imcc] - 2 * dv.d0[iccc]);
dV22 = gb2 * (dv.d0[icpc] + dv.d0[icmc] - 2 * dv.d0[iccc]);
dV33 = gp2 * (dv.d0[iccp] + dv.d0[iccm] - 2 * dv.d0[iccc]);
dV12 =
0.25 * gagb * (dv.d0[ippc] - dv.d0[ipmc] + dv.d0[immc] - dv.d0[impc]);
dV13 =
0.25 * gagp * (dv.d0[ipcp] - dv.d0[imcp] + dv.d0[imcm] - dv.d0[ipcm]);
dV23 =
0.25 * gbgp * (dv.d0[icpp] - dv.d0[icpm] + dv.d0[icmm] - dv.d0[icmp]);
dV12 = 0.25 * gagb * (dv.d0[ippc] - dv.d0[ipmc] + dv.d0[immc] - dv.d0[impc]);
dV13 = 0.25 * gagp * (dv.d0[ipcp] - dv.d0[imcp] + dv.d0[imcm] - dv.d0[ipcm]);
dV23 = 0.25 * gbgp * (dv.d0[icpp] - dv.d0[icpm] + dv.d0[icmm] - dv.d0[icmp]);
/* Derivatives of (dv) w.r.t. (A,B,phi):*/
dV11 = sin_al_i3 * (sin_al * dV11 - cos_al * dV1);
dV12 = sin_al_i1 * sin_be_i1 * dV12;
@@ -2230,11 +2236,12 @@ void TwoPunctures::JFD_times_dv(int i, int j, int k, int nvar, int n1, int n2,
/* (dU, dU_x, dU_y, dU_z, dU_xx, dU_xy, dU_xz, dU_yy, dU_yz, dU_zz)*/
rx3_To_xyz(nvar, x, r, phi, &y, &z, dU);
LinEquations(A, B, X, R, x, r, phi, y, z, dU, U, values);
for (ivar = 0; ivar < nvar; ivar++)
values[ivar] *= FAC;
free_derivs(&dU, nvar);
free_derivs(&U, nvar);
double FAC_val = sin_al * sin_be * sin_al * sin_be * sin_al * sin_be;
for (ivar = 0; ivar < nvar; ivar++)
values[ivar] *= FAC_val;
// No free_derivs needed — everything is on the stack
}
#undef FAC
/*-----------------------------------------------------------*/