[llvm-dev] Clang for the PlayStation 2 (original) (raw)

Dan Ravensloft via llvm-dev llvm-dev at lists.llvm.org
Thu Sep 6 08:55:52 PDT 2018

On Thu, 6 Sep 2018, 16:35 Daniel Sanders, <daniel_l_sanders at apple.com> wrote:

On 6 Sep 2018, at 08:00, Dan Ravensloft via llvm-dev <_ _llvm-dev at lists.llvm.org> wrote: On Mon, 3 Sep 2018 at 13:31, Tim Northover <t.p.northover at gmail.com> wrote: So the next step is to debug where Mips is producing those TruncIntFP nodes. There'll be some constraint it's not checking or an unexpected node type, probably related to -msingle-float. I'm afraid I'm not sure what yet. FWIW, I'm not sure how well tested -msingle-float was on MIPS. I don't think we had any bots testing it.

Well, now we have hardware, if all of one machine, where it's worth testing. I'm happy to try it myself if needed, though rigging it up to happen automatically would be difficult.

I'm reasonably sure the function producing that node is lowerFPTOSINTSTORE in lib/Target/Mips/MipsISelLowering.cpp.

The node before that function executes has an fptosint node which seems to want to convert an i64 to an f32 which seems...rather odd to me, honestly. The PS2, for what it's worth, only has an i32 -> f32 instruction, so I think there's an impedance mismatch somewhere.

Did you mean those types to be the other way around? fptosint is supposed to take a floating point type and produce an integer type so if you're seeing them backwards like this then that would definitely be a bug.

Yes, my mistake; I got confused by the direction of the arrows on the graph, they're pointing to their parents, but I thought they were pointing to their children.

If you're referring to the size mismatch though, that's ok within the IR and DAG nodes. There's no relationship between the size of the input and output.

LLVM Developers mailing list llvm-dev at lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev

-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20180906/6e064ffc/attachment.html>

More information about the llvm-dev mailing list