Fix runTest segfault (remove cudaDeviceReset) and simplify googletest template usage#909
Merged
valassi merged 84 commits intomadgraph5:masterfrom valassi:gtest2Jul 16, 2024
+171,442-2,808
Commits
Commits on Jul 11, 2024
[gtest/june24] in gg_tt.mad runTest.cc, include two identical tests on both CPU and GPU (prepare for madgraph5#896) - the C++ tests succeed but the CUDA tests segfaults madgraph5#903
committed[gtest/june24] in CODEGEN cudacpp_test.mk, try to upgrade googletest from release-1.11.0 to v1.14.0 to solve madgraph5#903, but the segfault remains - will revert
committed- committed
[gtest/june24] in gg_tt.mad cudacpp.mk, TEMPORARELY disable testmisc.cc build in CUDA while debugging madgraph5#903
committed[gtest/june24] in gg_tt.mad cudacpp.mk, TEMPORARELY disable testmisc.cc build while debugging madgraph5#903 also for C++
committed[clang/june24] in gg_tt.mad cudacpp.mk, fix clang16 build of fcheck_cpp.exe by adding -no-pie madgraph5#904
committed[clang/june24] in gg_tt.mad cudacpp.mk, fix clang16 builds by disabling OMP only for clang16 madgraph5#904
committed[clang/june24] in CODEGEN (backport gg_tt.mad) cudacpp.mk, fix clang16 builds madgraph5#904 (disabling OMP only for clang16; add -no-pie for fcheck_cpp.exe)
committed- committed
[clang/june24] in gg_tt.mad cudacpp.mk, fix clang16 build of fcheck_cpp.exe by adding -no-pie madgraph5#904
committed[clang/june24] in gg_tt.mad cudacpp.mk, fix clang16 builds by disabling OMP only for clang16 madgraph5#904
committed[clang/june24] in CODEGEN (backport gg_tt.mad) cudacpp.mk, fix clang16 builds madgraph5#904 (disabling OMP only for clang16; add -no-pie for fcheck_cpp.exe)
committed- committed
- committed
- committed
- committed
[clang/june24] in gg_tt.mad cudacpp.mk, fix clang17 builds by disabling OMP only for clang17 madgraph5#904
committed[clang/june24] in CODEGEN (backport gg_tt.mad) cudacpp.mk, fix clang17 builds madgraph5#904 (disable OMP also for clang17)
committed- committed
[clang/june24] in gg_tt.mad cudacpp.mk, fix clang17 builds by disabling OMP only for clang17 madgraph5#904
committed[clang/june24] in CODEGEN (backport gg_tt.mad) cudacpp.mk, fix clang17 builds madgraph5#904 (disable OMP also for clang17)
committed- committed
[clang/june24] in gg_tt.mad cudacpp.mk, improve clang16/17 fix madgraph5#904: remove link-time -no-pie, add compiler-time -fPIC to fortran
committed[clang/june24] in CODEGEN cudacpp.mk, improve clang16/17 fix madgraph5#904: remove link-time -no-pie, add compiler-time -fPIC to fortran
committed[clang/june24] regenerate all processes with cudacpp.mk new fixes for madgraph5#904, adding -fPIC to fortran compilation
committed- committed
- committed
- committed
[clang/june24] in gg_tt.mad cudacpp.mk, improve clang16/17 fix madgraph5#904: remove link-time -no-pie, add compiler-time -fPIC to fortran
committed[clang/june24] in CODEGEN cudacpp.mk, improve clang16/17 fix madgraph5#904: remove link-time -no-pie, add compiler-time -fPIC to fortran
committed- committed
[gtest/june24] Remove the duplicate test... the segfault in c++ through valgrind is still there! madgraph5#903
committed- committed
[gtest/june24] in gg_tt.mad cudacpp.mk, reenable testmisc.cc build in CUDA while debugging madgraph5#903
committed[gtest/june24] in gg_tt.mad constexpr_math.h, temporarely remove constexpr and add runtime debug printouts for madgraph5#903
committed- committed
- committed
[gtest/june24] in gg_tt.mad constexpr_math.h, FIX madgraph5#903 in constexpr_cos_quad (second fix: for cos, not sin)
committed[gtest/june24] in gg_tt.mad constexpr_math.h, FIX madgraph5#903 add special handling for pi/4 (avoid asserts in sinTaylor)
committed
Commits on Jul 12, 2024
[gtest/june24] in gg_tt.mad constexpr_math.h, comprehensive rewrite of fixes and debug printouts for bug madgraph5#903
committed[gtest/june24] in gg_tt.mad, include header valgrind.h (BSD license) to detect running on valgrind madgraph5#906
committed[gtest/june24] in gg_tt.mad testmisc.cc, use higher tolerances when running on valgrind madgraph5#906
committed[gtest/june24] in gg_tt.mad testmisc.cc, allow tan(x)=-inf if ctan(x)=+inf and viceversa when running on valgrind madgraph5#906
committed- committed
- committed
[gtest/june24] in CODEGEN (backport gg_tt.mad), include header valgrind.h (BSD license) to detect when running on valgrind madgraph5#906
committed[gtest/june24] in CODEGEN (backport gg_tt.mad) constexpr_math.h, comprehensive fixes and debug printouts for bug madgraph5#903 (recursive iteration, stack overflow, segfault etc)
committed[gtest/june24] in CODEGEN (from gg_tt.mad) cudacpp.mk, add an 'asan' target (address sanitizer madgraph5#207), but keep it commented out
committed[gtest/june24] in CODEGEN (from gg_tt.mad) testmisc.cc, use higher tolerances when running on valgrind madgraph5#906
committed- committed
[gtest/june24] regenerate all processes, with fixes for constexpr_math.h madgraph5#903 and test_misc.cc/valgrind.h madgraph5#906
committed- committed
- committed
- committed
- committed
- committed
- committed
- committed
- committed
- committed
[gtest2/june24] in gg_tt.mad, temporarely revert the simplication of tests and add back the duplication of tests - the segfault madgraph5#907 disappears if gpuDeviceReset is removed!
committed[gtest2/june24] in gg_tt.mad runTest.cc, add back gpuDeviceReset(), the segfault does reappear! THIS PROVES THAT GPUDEVICERESET CAUSES madgraph5#907
committed- committed
- committed
- committed
[gtest2/june24] in gg_tt.mad testxxx.cc, try to move gpuDeviceReset() to the atexit function, but this STILL crashes! madgraph5#907
committed[gtest2/june24] in gg_tt.mad testxxx.cc, COMMENT OUT gpuDeviceReset() to avoid all crashes madgraph5#907 (FIXME? avoid cuda api calls in dtors?)
committed- committed
[gtest2/june24] in gg_tt.mad runTest.cc, reorder test1 and test2, prepare to comment out test2 (preparatory work for madgraph5#896)
committed- committed
[gtest2/june24] in CODEGEN (backport gg_tt.mad) MadgraphTest.h, runTest.cc, testxxx.cc: simplify gtest templates, remove cudaDeviceReset to fix madgraph5#907, complete preparation of two-test infra…
committed- committed
- committed
- committed
Commits on Jul 16, 2024
Merge remote-tracking branch 'upstream/master' (including OMP madgraph5#900 and submod madgraph5#897) into clang
committed[clang/june24] in gg_tt.mad and CODEGEN cudacpp.mk, fail with an error if OpenMP builds are attempted on clang16/17 (as discussed with Olivier in madgraph5#905)
committed- committed
- committed
Merge branch 'clang' (madgraph5#905, also including latest upstream/master with OMP madgraph5#900 and submod madgraph5#897) into gtest
committed- committed
- committed
Merge branch 'gtest' (madgraph5#908, also including latest upstream/master with clang madgraph5#905, OMP madgraph5#900 and submod madgraph5#897) into gtest
committed- committed
- committed