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

GPU-accelerate Shell-Patch BSSN evolution

Browse Source 
Phase 1: Enable GPU resident state for Cartesian patches in Shell mode.
- Remove WithShell guard from bssn_cuda_use_resident_sync().
- Add GPU-to-CPU state sync before shell CPU consumers (SHStep,
  CS_Inter, inline shell RHS blocks).

Phase 2: GPU-accelerate BSSN Shell Patch RHS.
- Create bssn_gpu.h with RHS_SS_PARA macro and gpu_rhs_ss declaration.
- Fix compilation bugs in legacy bssn_gpu_rhs_ss.cu (deprecated
  cudaThreadSynchronize, tmp_con2 redeclaration, ijkmin3_h typo,
  CUDA_SAFE_CALL, missing compare_result guard).
- Add bssn_gpu_rhs_ss.o to CFILES_CUDA_BSSN with build rule.
- Write cuda_compute_rhs_bssn_ss() wrapper bridging Fortran and GPU
  parameter conventions, redirect all shell RHS call sites via #define.

Verified: 30-step Shell-Patch GPU run completes without errors/NaN.
Step wall time ~4.4s (step_fn ~2.0s + RP ~0.68s + constraint ~0.70s).

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
This commit is contained in:
CGH0S7
2026-05-09 18:50:10 +08:00
parent 5eb49949d9
commit bd4ce3fbf3
4 changed files with 234 additions and 46 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
AMSS_NCKU_source/makefile
View File

@@ -59,15 +59,19 @@ endif
.for.o:
$(f77) -c $< -o $@
.cu.o:
$(Cu) $(CUDA_APP_FLAGS) -c $< -o $@ $(CUDA_LIB_PATH)
# CUDA rewrite of BSSN RHS (drop-in replacement for bssn_rhs_c + stencil helpers)
bssn_rhs_cuda.o: bssn_rhs_cuda.cu bssn_rhs.h macrodef.h fd_cuda_helpers.cuh
$(Cu) $(CUDA_APP_FLAGS) -c $< -o $@ $(CUDA_LIB_PATH)
.cu.o:
$(Cu) $(CUDA_APP_FLAGS) -c $< -o $@ $(CUDA_LIB_PATH)
# CUDA rewrite of BSSN RHS (drop-in replacement for bssn_rhs_c + stencil helpers)
bssn_rhs_cuda.o: bssn_rhs_cuda.cu bssn_rhs.h macrodef.h fd_cuda_helpers.cuh
$(Cu) $(CUDA_APP_FLAGS) -c $< -o $@ $(CUDA_LIB_PATH)
# CUDA rewrite of BSSN Shell-Patch RHS (drop-in replacement for bssn_rhs_ss)
bssn_gpu_rhs_ss.o: bssn_gpu_rhs_ss.cu bssn_gpu.h gpu_rhsSS_mem.h bssn_macro.h macrodef.fh
$(Cu) $(CUDA_APP_FLAGS) -c $< -o $@ $(CUDA_LIB_PATH)
# CUDA rewrite of Z4C Cartesian RHS
z4c_rhs_cuda.o: z4c_rhs_cuda.cu z4c_rhs_cuda.h bssn_rhs.h macrodef.h ricci_gamma.h fd_cuda_helpers.cuh
z4c_rhs_cuda.o: z4c_rhs_cuda.cu z4c_rhs_cuda.h bssn_rhs.h macrodef.h ricci_gamma.h fd_cuda_helpers.cuh
$(Cu) $(CUDA_APP_FLAGS) -c $< -o $@ $(CUDA_LIB_PATH)
# C rewrite of BSSN RHS kernel and helpers
@@ -123,7 +127,7 @@ else
CFILES_CPU = bssn_rhs_c.o fderivs_c.o fdderivs_c.o kodiss_c.o lopsided_c.o lopsided_kodis_c.o
endif
CFILES_CUDA_BSSN = bssn_rhs_cuda.o
CFILES_CUDA_BSSN = bssn_rhs_cuda.o bssn_gpu_rhs_ss.o
ifeq ($(USE_CUDA_BSSN),1)
CFILES = $(CFILES_CUDA_BSSN)