Hi, The compiler generates calls to special functions (like the one in question _multi3) if it doesn’t have a way how to solve it (via compiler backend rules). This function is implemented by the libgcc which will have a default method to address the functionality in question. Now, from the link which you gave me, I understand users are not very happy with the _multi3 implementation (128 bit multiplication) in MIPS cores, therefore, they come up with an patched implementation. In ARC backend we try to address these issues, but if there is a problem, I will be happy to fix it. Anyhow, if you see any issue with code gen (improvements/bugs), I would recommend you to send it over and I will spend some time in fixing them directly in the compiler.
Cheers, Claudiu > -----Original Message----- > From: Alexey Brodkin > Sent: Monday, January 29, 2018 2:38 PM > To: Claudiu Zissulescu <claz...@synopsys.com> > Cc: j...@phrozen.org; lede-dev@lists.infradead.org; ha...@hauke-m.de; > evgeniy.di...@synopsys.com; linux-snps-...@lists.infradead.org > Subject: Re: [LEDE-DEV] [PATCH] archs38: bump kernel to 4.14. > > Hi Claudiu, > > Could you please comment on the question from Hauke below? > > On Sat, 2018-01-27 at 16:36 +0100, Hauke Mehrtens wrote: > > Does the arc CPU also need some special handling like some MIPS CPUs: > > > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i > d=bbc25bee37d2b32cf3a1fab9195b6da3a185614a > > -Alexey _______________________________________________ Lede-dev mailing list Lede-dev@lists.infradead.org http://lists.infradead.org/mailman/listinfo/lede-dev
