Move CUDA builds from C++14 to C++17#354

Merged

valassi merged 24 commits intomadgraph5:masterfrom

valassi:cpp17

Jan 26, 2022

Commits on Jan 26, 2022

[cpp17] move CUDA builds from c++14 to c++17 (#333)
valassi
committed
[cpp17] require cuda >=11.2 to be able to use c++17 (#333 - see also #96)
valassi
committed
[cpp17] improve a comment about nvcc builds of c++ code with c++17 (#318, #319, #333)
valassi
committed
[cpp17] use "if constexpr" in MemoryAccessHelpers also in cuda, thanks to c++17 (#333)
valassi
committed
[cpp17] enable if constexpr in Bridge.h (needs c++17 and cuda>=11.2)
valassi
committed
[cpp17] bug fix in OMPFLAGS - effectively reenable OpenMP (code was not built with OpenMP for a long time)
valassi
committed
[cpp17] remove OMPFLAGS support for gcc<9.3 in Makefile
valassi
committed
[cpp17] require gcc>=9.3 (#96) - it was observed in the past that OpenMP required this (#269)
valassi
committed
[cpp17] fix OpenMP MT builds after effectively reenabling OpenMP
valassi
committed
[cpp17] remove CUFLAGS CXXFLAGS CPPFLAGS from link instructions (keep for compilation)
valassi
committed
[cpp17] add back OMPFAGS to linker after removing CXXFLAGS
valassi
committed
[cpp17] Use __CUDACC_VER_MAJOR__ instead of __CUDACC__ in getCompiler to detect if nvcc was used
valassi
committed
[cpp17] improve the previous patch: use __NVCC__ instead of __CUDA_MAJOR_VERSION__ to detect if nvcc was used as compiler
valassi
committed
[cpp17] TEMPORARELY DISABLE OMP (need to reassess MT)
valassi
committed
[cpp17] further use __NVCC__ instead of __CUDACC__ throughout getCompiler
valassi
committed
[cpp17] complete backport to codegen and regenerate eemumu auto
valassi
committed
[cpp17] BUG FIX IN LOGS! recover gcc102 baselines from c2e67b4a425319339d902b58b1bf931b0a846968
valassi
committed
[cpp17] rerun eemumu basic log, check all is ok
valassi
committed
[cpp17] add ggtg and ggttggg to sync script
valassi
committed
[cpp17] regenerate ggtt* auto and resync manu
valassi
committed
[cpp17] ** COMPLETE CPP17 ** rerun ggtt/ggttg/ggttgg, all ok
valassi
committed
[cpp17] allow gcc<=9.3 if gcc is used below clang (#96 and #355)
valassi
committed
[cpp17] backport to codegen and regenerate eemumu auto
valassi
committed
[cpp17] ** COMPLETE CPP17 ** regenerate gg_tt* with gcc check bug fix (#355)
valassi
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Move CUDA builds from C++14 to C++17#354

Move CUDA builds from C++14 to C++17#354
valassi merged 24 commits intomadgraph5:masterfrom
valassi:cpp17

Commits on Jan 26, 2022

[cpp17] move CUDA builds from c++14 to c++17 (#333)

[cpp17] require cuda >=11.2 to be able to use c++17 (#333 - see also #96)

[cpp17] improve a comment about nvcc builds of c++ code with c++17 (#318, #319, #333)

[cpp17] use "if constexpr" in MemoryAccessHelpers also in cuda, thanks to c++17 (#333)

[cpp17] enable if constexpr in Bridge.h (needs c++17 and cuda>=11.2)

[cpp17] bug fix in OMPFLAGS - effectively reenable OpenMP (code was not built with OpenMP for a long time)

[cpp17] remove OMPFLAGS support for gcc<9.3 in Makefile

[cpp17] require gcc>=9.3 (#96) - it was observed in the past that OpenMP required this (#269)

[cpp17] fix OpenMP MT builds after effectively reenabling OpenMP

[cpp17] remove CUFLAGS CXXFLAGS CPPFLAGS from link instructions (keep for compilation)

[cpp17] add back OMPFAGS to linker after removing CXXFLAGS

[cpp17] Use __CUDACC_VER_MAJOR instead of CUDACC__ in getCompiler to detect if nvcc was used

[cpp17] improve the previous patch: use NVCC instead of __CUDA_MAJOR_VERSION__ to detect if nvcc was used as compiler

[cpp17] TEMPORARELY DISABLE OMP (need to reassess MT)

[cpp17] further use NVCC instead of CUDACC throughout getCompiler

[cpp17] complete backport to codegen and regenerate eemumu auto

[cpp17] BUG FIX IN LOGS! recover gcc102 baselines from c2e67b4a425319339d902b58b1bf931b0a846968

[cpp17] rerun eemumu basic log, check all is ok

[cpp17] add ggtg and ggttggg to sync script

[cpp17] regenerate ggtt* auto and resync manu

[cpp17] COMPLETE CPP17 rerun ggtt/ggttg/ggttgg, all ok

[cpp17] allow gcc<=9.3 if gcc is used below clang (#96 and #355)

[cpp17] backport to codegen and regenerate eemumu auto

[cpp17] COMPLETE CPP17 regenerate gg_tt* with gcc check bug fix (#355)