Shell C functions must export Fortran-compatible symbols with trailing
underscore so bssn_rhs_ss.f90 and getnp4.f90 can link when WithShell is
active and USE_CXX_SHELL_KERNELS=1 replaces Fortran diff_new_sh.o.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
New bssn_em_rhs_c.C computes EM field RHS (E,B,Kpsi,Kphi) and stress-energy
tensor, then calls the C BSSN RHS kernel with source terms. Replaces empart.f90
when USE_CXX_EM_KERNEL=1. Supports all ghost_width orders via existing derivative
kernels. Controlled by USE_CXX_EM_KERNEL switch (default 0, experimental).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
New files provide C equivalents of Fortran diff_new_sh.f90 and kodiss_sh.f90:
- fderivs_sh_c.C: first derivatives in shell (rho, sigma, R) coords
- fdderivs_sh_c.C: second derivatives in shell coords
- fderivs_shc_c.C: shell first derivs + chain rule to Cartesian
- fdderivs_shc_c.C: shell second derivs + chain rule to Cartesian
- kodiss_sh_c.C: Kreiss-Oliger dissipation on shell patches
Also add symmetry_stbd() C implementation and shell fh indexing to share_func.h.
Controlled by USE_CXX_SHELL_KERNELS switch (default 0, experimental).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Wrap each C kernel in #if (ghost_width == N) blocks matching Fortran stencil
coefficients from diff_new.f90, kodiss.f90, and lopsidediff.f90. Add fast-path
indexing for ord=1,4,5 in share_func.h.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Split prolongpointstru into search-only (prolongpointstru_search) and
append-only (prolongpointstru_append) functions. Parallelize shell-point
interpolation table construction with #pragma omp parallel for collapse(3)
and per-thread linked lists (merged after the loop to avoid data races).
Add OMP_FLAG = -fopenmp in makefile.inc and ShellPatch.o override rule
in makefile for AOCC OpenMP runtime (-lomp already linked).
Speedup: setupintintstuff ~2.2x faster on multi-core.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Replace Intel compilers (ifx/icpx/icx) with AOCC (flang/clang++/clang),
Intel MPI (mpiicpx) with AOCC-built OpenMPI (mpicxx), and Intel MKL
with AOCL BLIS/libFLAME. Replace -xHost with -march=znver4, -ipo with
-flto, -fp-model fast=2 with -ffast-math, -qopenmp with -fopenmp.
Remove PGO, TBB allocator, and Intel-specific runtime libraries.
Fix MKL-specific includes in TwoPunctures.C and gaussj.C to use
standard CBLAS/LAPACKE headers from AOCL.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
bssnEScalar_class::Initialize() already calls setup_transfer_caches(),
but bssnEM_class::Initialize() did not. When USE_TRANSFER_CACHE=1,
the sync_cache pointers remain NULL, causing SIGSEGV in wrapper
methods that dereference sync_cache_*[lev].
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Seven Parallel::*_cached() calls in RestrictProlong and
RestrictProlong_aux were missed during the transfer-cache refactoring
(commits 9cd3741..8d28c29). When BSSN_USE_TRANSFER_CACHE=0, all
sync_cache pointers are NULL, so dereferencing sync_cache_*[lev]
triggers SIGSEGV.
Replace them with the equivalent wrapper methods (sync_evolution,
restrict_evolution, outbdlow2hi_evolution) that check
use_transfer_cache() and fall back to uncached direct calls.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- compute fi_min/fi_max from output i-range and derive ii_lo/ii_hi
- replace full ii sweep (-1:extf(1)) with windowed sweep in Z/Y precompute passes
- keep stencil math unchanged; add bounds sanity check for ii window
