Build a proxy NVRTC command line utility that can act as if it were a "real" compiler.

[jrhemstad]

No description provided.

[wmaxey]

My current thoughts on this utility:

NVRTCC

How to use:

Configure libcudacxx to test with NVRTC and cmake --build the project before executing lit.

cmake ... -DLIBCUDACXX_TEST_WITH_NVRTC=ON
cmake --build $BUILD_DIR
lit ... $TEST_DIR

How it works

nvrtcc processes incoming arguments matching for flags that modify its behavior, and passing the rest to NVRTC.
It will hopefully filter any that don't apply (gcc warnings and such).

The input file is processed to be compatible with NVRTC similarly to the nvrtc.sh scripts during compilation.
The resulting file is then compiled with NVRTC and stored as a fatbin. This is, in effect, a compilation pass for NVRTC.
.fail.cpp tests will be analyzable by lit and should work properly.

For execution, the fatbin file is provided to nvrtcc again, but will this time instead launch on GPU the precompiled
test unit.

flowchart TD;
    compile{nvrtcc};
    runfat{nvrtcc};

    start[Execute lit]
    build(Execute nvrtcc)
    pass(Pass: Continue)
    fail(Fail: Compile time failure)
    save[Save fatbin]

    load[Load fatbin]
    testpass[Pass: No failures]
    testfail[Fail: Report driver error]

    save --> load;

    subgraph Build Pass
    start   --  lit -sv -Denable_nvrtc=true    --> build;
    build   --  nvrtcc -x cu test.pass.cpp ... --> compile;
    compile --> fail;
    compile --> pass;
    pass    --> save;
    end


    subgraph Test Pass
    load -- nvrtcc test.pass.cpp.fatbin --> runfat;
    runfat --> testpass
    runfat --> testfail
    end

Loading

[benbarsdell]

In case it's useful, we made a similar thing for the purpose of adding NVRTC support to godbolt.org: https://github.com/NVIDIA/jitify/blob/master/nvrtc_cli.cpp