llvm dev - Nov 2014 - [LLVMdev] About inlining the modulo function in ARM architecture

Hi all,

Sorry for bothering those not interested. I found that ARM backend will
insert a modulo function
(like __modsi3) instead of the modulo instruction. I wonder how we can
inline the modulo function
into the program. Is there any OPTION we can use in the opt or llc?

Thanks,
Ray
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20141110/8ded710d/attachment.html>

Hi Ray,

On 10 November 2014 10:13, Alex S <alexmountain13 at gmail.com>
wrote:
> Sorry for bothering those not interested. I found that ARM backend will
> insert a modulo function
> (like __modsi3) instead of the modulo instruction. I wonder how we can
> inline the modulo function
> into the program. Is there any OPTION we can use in the opt or llc?
It depends on whether the CPU actually supports integer division.
Surprisingly modern CPUs don't (the Cortex-A9 for example), and LLVM
doesn't synthesise an entire divide loop.

Oddly, it also seems to call the library at -O0, because that's what
the FastISel decides to do. That strikes me as a bug, or at least an
oddity.

Cheers.

Tim.

Hi Tim,

Thanks for your reply.

On Mon, Nov 10, 2014 at 1:26 PM, Tim Northover <t.p.northover at
gmail.com>
wrote:
> Hi Ray,
>
> On 10 November 2014 10:13, Alex S <alexmountain13 at gmail.com>
wrote:
> > Sorry for bothering those not interested. I found that ARM backend
will
> > insert a modulo function
> > (like __modsi3) instead of the modulo instruction. I wonder how we can
> > inline the modulo function
> > into the program. Is there any OPTION we can use in the opt or llc?
>
> It depends on whether the CPU actually supports integer division.
> Surprisingly modern CPUs don't (the Cortex-A9 for example), and LLVM
> doesn't synthesise an entire divide loop.
>

I know this may degrade the performance. I wonder whether there is any easy
way to inline these functions into the callsite? Thanks again!



>
> Oddly, it also seems to call the library at -O0, because that's what
> the FastISel decides to do. That strikes me as a bug, or at least an
> oddity.
>
> Cheers.
>
> Tim.
>
Ray
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20141110/9b1a02a4/attachment.html>