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

Eliminate hot-path heap allocations in TwoPunctures spectral solver

Browse Source 
Pre-allocate workspace buffers as class members to remove ~8M malloc/free
pairs per Newton iteration from LineRelax, ThomasAlgorithm, JFD_times_dv,
J_times_dv, chebft_Zeros, fourft, Derivatives_AB3, and F_of_v.
Rewrite ThomasAlgorithm to operate in-place on input arrays.
Results are bit-identical; no algorithmic changes.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
CGH0S7
2026-02-06 21:20:35 +08:00
parent 699e443c7a
commit 09ffdb553d
2 changed files with 167 additions and 136 deletions
Download Patch File Download Diff File Expand all files Collapse all files

27
AMSS_NCKU_source/TwoPunctures.h
View File

@@ -42,6 +42,33 @@ private:
int ntotal;
// Pre-allocated workspace buffers for hot-path allocation elimination
// LineRelax_be workspace (sized for n2)
double *ws_diag_be, *ws_e_be, *ws_f_be, *ws_b_be, *ws_x_be;
// LineRelax_al workspace (sized for n1)
double *ws_diag_al, *ws_e_al, *ws_f_al, *ws_b_al, *ws_x_al;
// ThomasAlgorithm workspace (sized for max(n1,n2))
double *ws_thomas_y;
// JFD_times_dv workspace (sized for nvar)
double *ws_jfd_values;
derivs ws_jfd_dU, ws_jfd_U;
// chebft_Zeros workspace (sized for max(n1,n2,n3)+1)
double *ws_cheb_c;
// fourft workspace (sized for max(n1,n2,n3)/2+1 each)
double *ws_four_a, *ws_four_b;
// Derivatives_AB3 workspace
double *ws_deriv_p, *ws_deriv_dp, *ws_deriv_d2p;
double *ws_deriv_q, *ws_deriv_dq;
double *ws_deriv_r, *ws_deriv_dr;
int *ws_deriv_indx;
// F_of_v workspace
double *ws_fov_sources;
double *ws_fov_values;
derivs ws_fov_U;
// J_times_dv workspace
double *ws_jtdv_values;
derivs ws_jtdv_dU, ws_jtdv_U;
struct parameters
{
int nvar, n1, n2, n3;