Hi Uros, > -----Original Message----- > From: Uros Bizjak [mailto:ubiz...@gmail.com] > Sent: Tuesday, October 13, 2015 9:12 PM > To: Kumar, Venkataramanan > Cc: gcc-patches@gcc.gnu.org > Subject: Re: [Patch] [x86_64]: Add bdver4 for multi versioning and fix AMD > cpu model detection. > > On Tue, Oct 13, 2015 at 5:16 PM, Kumar, Venkataramanan > <venkataramanan.ku...@amd.com> wrote: > > Hi Uros, > > > > I realized both GCC 4.9 and GCC 5 branches includes processor subtype > AMDFAM15H_BDVER4. > > So I need to back port not only model selection fix but also the detection > > of > model for bdver4. > > > > Is that fine? > > OK, but to avoid ABI mismatches, please double check that enum values > passed between library and compiled code are always the same in all gcc > releases. >
I verified and same enum value is passed by gcc and libgcc in GCC 5 and GCC 4-9 branches. Also ran gcc bootstrap and gcc regression tests on a bdver4 machine . Completed the back ports. GCC 4_9 https://gcc.gnu.org/viewcvs/gcc?view=revision&revision=228953 GCC 5 https://gcc.gnu.org/viewcvs/gcc?view=revision&revision=228951 > Uros. > > > Regards, > > Venkat. > > > >> -----Original Message----- > >> From: Kumar, Venkataramanan > >> Sent: Friday, October 09, 2015 3:31 PM > >> To: 'Uros Bizjak' > >> Cc: gcc-patches@gcc.gnu.org > >> Subject: RE: [Patch] [x86_64]: Add bdver4 for multi versioning and > >> fix AMD cpu model detection. > >> > >> Thank you Uros, > >> > >> I will test and commit model selection change in all release branches as > well. > >> > >> Regards, > >> Venkat. > >> > >> > -----Original Message----- > >> > From: Uros Bizjak [mailto:ubiz...@gmail.com] > >> > Sent: Friday, October 09, 2015 3:25 PM > >> > To: Kumar, Venkataramanan > >> > Cc: gcc-patches@gcc.gnu.org > >> > Subject: Re: [Patch] [x86_64]: Add bdver4 for multi versioning and > >> > fix AMD cpu model detection. > >> > > >> > On Fri, Oct 9, 2015 at 11:50 AM, Kumar, Venkataramanan > >> > <venkataramanan.ku...@amd.com> wrote: > >> > > Hi Uros, > >> > > > >> > > Please find below patch that adds bdver4 target for multi versioning. > >> > > Also I while computing model, the extended_model is incorrectly > >> > > left > >> > shifted by 4. I have removed it now. > >> > > > >> > > Is below patch Ok for trunk ? > >> > > GCC bootstrap and regressions passed. > >> > > >> > OK for trunk and release branches, where applicable. IMO, model > >> > selection fix should be applied to all release branches. > >> > > >> > Thanks, > >> > Uros. > >> > > >> > > diff --git a/libgcc/ChangeLog b/libgcc/ChangeLog index > >> > > bb3a722..8676747 100644 > >> > > --- a/libgcc/ChangeLog > >> > > +++ b/libgcc/ChangeLog > >> > > @@ -1,3 +1,8 @@ > >> > > +2015-10-09 Venkataramanan Kumar > >> > <venkataramanan.ku...@amd.com> > >> > > + > >> > > + * config/i386/cpuinfo.c (get_amd_cpu): Detect bdver4. > >> > > + (__cpu_indicator_init): Fix model selection for AMD CPUs. > >> > > + > >> > > 2015-10-05 Kirill Yukhin <kirill.yuk...@intel.com> > >> > > > >> > > * config/i386/cpuinfo.c (get_intel_cpu): Detect > >> > > "skylake-avx512". > >> > > diff --git a/libgcc/config/i386/cpuinfo.c > >> > > b/libgcc/config/i386/cpuinfo.c index 0cbbc85..1313ca3 100644 > >> > > --- a/libgcc/config/i386/cpuinfo.c > >> > > +++ b/libgcc/config/i386/cpuinfo.c > >> > > @@ -169,6 +169,9 @@ get_amd_cpu (unsigned int family, unsigned > >> > > int > >> > model) > >> > > /* Bulldozer version 3 "Steamroller" */ > >> > > if (model >= 0x30 && model <= 0x4f) > >> > > __cpu_model.__cpu_subtype = AMDFAM15H_BDVER3; > >> > > + /* Bulldozer version 4 "Excavator" */ > >> > > + if (model >= 0x60 && model <= 0x7f) > >> > > + __cpu_model.__cpu_subtype = AMDFAM15H_BDVER4; > >> > > break; > >> > > /* AMD Family 16h "btver2" */ > >> > > case 0x16: > >> > > @@ -455,7 +458,7 @@ __cpu_indicator_init (void) > >> > > if (family == 0x0f) > >> > > { > >> > > family += extended_family; > >> > > - model += (extended_model << 4); > >> > > + model += extended_model; > >> > > } > >> > > > >> > > /* Get CPU type. */ > >> > > > >> > > Regards, > >> > > Venkat. > >> > > > >> > > > >> > > > >> > > > >> > > > >> > > Regards, Venkat.
