I do not know if we have any "simpler bugs"; you have chosen a hard-mode
configuration to work on!
You can search gcc.gnu.org/bugzilla for "nvptx" and see if there's
anything you'd like to work on.
This one is an interesting minor feature to implement, perhaps?
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104926 . Naming the
registers would assist toolchain developers, but does not change the
behaviour of the running code at all, and yet you would get to look into
how things work in the internals, and what NVPTX assembler code looks
like, and you'll have to figure out the list of "special" registers for
yourself. (Hint: "print_operand", but beware rogue "printf".)
Hope that helps
Andrew
On 09/10/2025 11:34, Amruth Tetakali wrote:
Thank you very much for the link to the wiki page. I have now
successfully built the complete nvptx-none-gcc toolchain from the
latest git source on my Ubuntu system.
Now that I have a working build environment, I am eager to start contributing.
As a newcomer to the codebase, could you perhaps recommend any
simpler bugs that would be a good starting point?
Amruth
On Tue, 30 Sept 2025 at 13:07, Andrew Stubbs <[email protected]> wrote:
On 28/09/2025 08:51, Amruth Tetakali via Gcc wrote:
Hello,
I am a student exploring the GCC source code with an interest in GPU
programming and the NVPTX backend. I am in the process of setting up my
development environment.
I have successfully cloned the latest source from the git repository and
installed the required prerequisites (GMP, MPFR, MPC, etc.) on my Linux
system. I am now at the configure step.
Could you please guide me on the essential ./configure flags that are
required to build GCC with support for NVPTX offloading?
There's a Wiki page that should cover this. You should probably read
over the whole page, but especially this section should help you:
https://gcc.gnu.org/wiki/Offloading#For_Nvidia_PTX: (the colon is part
of the URL).
If the instructions aren't clear or are broken then please give feedback. :)
Andrew
