nvptx-tools: a collection of tools for use with nvptx-none GCC toolchains.

  * nvptx-none-as: "assembler" for PTX.
  * nvptx-none-ld: "linker" for PTX.
  * nvptx-none-run: run PTX binaries compiled with -mmainkernel.
  * nvptx-none-run-single: as nvptx-none-run, locked, such that system-wide
    only one instance of it is running at a time.

Additionally, the following symlinks are installed:

  * nvptx-none-ar: link to the GNU/Linux host system's ar.
  * nvptx-none-ranlib: link to the GNU/Linux host system's ranlib.


Testsuite
---------

    $ make check

In addition to the testsuite of the included libiberty support library
('check-libiberty'), this tests the nvptx-tools using a lit-based testsuite
('check-lit'); see 'test/README'.


DejaGnu Board File
------------------

There is a DejaGnu board file included, nvptx-none-run.exp, which can be used
as follows with the GCC testsuite:

    $ make check DEJAGNU=[build-nvptx-tools]/dejagnu.exp RUNTESTFLAGS='--target_board=nvptx-none-run' [...]

By default, this will use nvptx-none-run found in $PATH, but that can be
overridden by setting the NVPTX_NONE_RUN environment variable, for example to
[...]/nvptx-none-run-single.


Nvidia CUDA/PTX Dependency
--------------------------

The nvptx-tools are used for interfacing nvptx-none GCC toolchains with Nvidia
tools and GPUs via Nvidia CUDA/PTX.

Individual nvptx-tools depend on CUDA/PTX as follows:

  * nvptx-none-as

    In addition to its own minimalistic verification, it verifies the output
    file produced by feeding it into 'ptxas'.  If (at run time) that executable
    isn't available in the standard search path, this verification step is
    skipped (output file assumed compliant).

    Unless overridden with the '-m' option, the target architecture used for
    'ptxas' verification ('--gpu-name' option to 'ptxas') is deduced from the
    '.target' directive in the input's preamble.  This means that it's the
    responsibility of the producer (nvptx-none GCC toolchain) that the PTX code
    works with the CUDA version used.

  * nvptx-none-run

    Uses the CUDA Driver library ('cuda.h', 'libcuda') to run PTX binaries.  If
    (at configure time) that library isn't available, this tool isn't built.


For the testsuite (only), the baseline dependency is sm_35 functionality (or
higher).

Support for sm_35 has been introduced with PTX ISA 3.1, CUDA 5.0, driver r302,
has been deprecated in CUDA 10.2 (see
<https://docs.nvidia.com/cuda/archive/10.2/cuda-toolkit-release-notes/#deprecated-features>),
but has not yet been removed.

Testcases for the nvptx-tools depend on CUDA/PTX as follows:

  * nvptx-none-as

    Testcases depending on 'ptxas' are UNSUPPORTED if 'ptxas' isn't available.

  * nvptx-none-run

    All execution testcases are UNSUPPORTED if this tool isn't available,
    because the CUDA Driver library isn't available.


With reduced functionality as indicated, it's possible to build, test, and use
the nvptx-tools without a CUDA installation.


Report Issues
-------------

Report issues to <https://github.com/MentorEmbedded/nvptx-tools/issues>.