Value-View Interface #289
Value-View Interface #289
Conversation
…not be used in prod.
…nt. This might be useful again.
…; Moving SphArray.hh to Utilities/ManagedVector.hh
…e test for the new interface pattern.
…d organization for VVI macros; Differentiating between Ptr and Ref syntax metaclasses with vvi pattern.
…o provide a clearer API for using VVI to users.
… defines VIEW Ctor, Copy and Assign.
…Better default behavior for Value Interfaces, def copy-ctor, assign op and eq op behavior is baked in for free in the ValueInterface class.
…crtp; Cleaning up VVI_VALUE_ macros.
…or defining default Ctor, Copy, Ass and Eq operations.
…ngs in field_test
| endif() | ||
| message("-----------------------------------------------------------------------------") | ||
| find_package(umpire REQUIRED NO_DEFAULT_PATH PATHS ${umpire_DIR}) | ||
| if (umpire_FOUND) | ||
| message("Found umpire External Package.") | ||
| blt_convert_to_system_includes(TARGET umpire) |
There was a problem hiding this comment.
Should this blt_convert_to_system_includes() stuff also be in the block under "Found chai External Package" (and any other packages which could be found externally)?
There was a problem hiding this comment.
I`ll add it to CHAI. We already do this for all other external packages on line 127.
There was a problem hiding this comment.
We don't do it for the packages we use find_package on. We only do it on packages we manually bring in. Was this necessary for your stuff?
There was a problem hiding this comment.
It was, pedantic warnings from RAJA and umpire were causing our WARNINGS_AS_ERRORS builds to fail.
| @@ -85,7 +90,10 @@ if(ENABLE_CUDA) | |||
| #set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -arch=${CUDA_ARCH} --expt-relaxed-constexpr --extended-lambda -Xcudafe --display_error_number") | |||
| set(CMAKE_CUDA_STANDARD 17) | |||
| list(APPEND SPHERAL_CXX_DEPENDS cuda) | |||
| set(SPHERAL_ENABLE_CUDA On) | |||
| set(SPHERAL_ENABLE_VVI On) | |||
There was a problem hiding this comment.
Should the Value-View interface be enabled for CPU builds as well?
There was a problem hiding this comment.
It can be used for CPU builds but it is required for any GPU builds where we will be executing on the device. This just ensures it's enabled if you are building spheral w/ cuda.
Summary
ManagedVector(MV) implementation.MVunit testing.VVI) that is based around the experimentalchai::ManagedSharedPtr.VVIin Spheral RTD. Preview here.MVmember.MVmember.VVIforSpheral::QuadraticInterpolator.ctestc++ unit testing for ensuring semantic behavior of classes is working as expected withVVI.Spheral::QuadraticInterpolatorw/ and w/o VVI.Spheral::Field. Incompletemake testat the end of the build stage.SPHERAL_ENABLE_VVI=Onfor+cudabuilds.ToDo :
RELEASE_NOTES.mdwith notable changes.