Re: [cfarm-users] Interest for a RISC-V host with Vector extension? (BananaPi BPI-F3 / SpacemiT K1)

[Jeffrey Walton via cfarm-users]

On Wed, May 29, 2024 at 12:23 PM Baptiste Jonglez via cfarm-users
 wrote:
>
> We may have an opportunity to add the following RISC-V board to the farm:
>
> https://docs.banana-pi.org/en/BPI-F3/BananaPi_BPI-F3
> https://docs.banana-pi.org/en/BPI-F3/SpacemiT_K1
>
> The CPU core is advertised as having "RVA22 Profile and 256-bit RVV 1.0",
> that is, support for the Vector extension.
>
> To gauge interest for this board and be able to apply for a hardware
> donation, we need some feedback:
>
> 1) do you have an interest for the RISC-V Vector extension in general?

Yes.

> 2) do you have an interest for this specific CPU core?

Yes.

> 3) what can of work would you be able to do with this board, assuming it
>becomes available in the farm?

SIMD on RISC-V to speed up algorithms.

> 4) do you care about running an upstream Linux kernel on this board, or is
>a vendor kernel sufficient for your work?

No. I don't have a dog in this fight.

> We don't have any estimate about how long it would take to integrate such
> a board, but it could easily take 6 months to 1 year (possibly much more if
> we need to run an upstream kernel).

To add to (1), it would be nice to get my hands on a RISC-V machine
with security extensions too, like AES or SHA.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] New Solaris 11.4 hosts (x86 and SPARC)

[Jeffrey Walton via cfarm-users]

On Mon, Mar 25, 2024 at 8:22 PM Bruno Haible via cfarm-users
 wrote:
>
> > We are pleased to announce the availability of two new machines, cfarm215
> > and cfarm216. They run the latest Solaris 11.4 SRU (Support Repository
> > Update). The first one is based on a recent Intel x86 CPU, while the
> > second one is based on a SPARC T8-1 CPU.
>
> Thanks a lot for these machines!
>
> > Basic development packages are installed from the Solaris support
> > repository.
>
> Unfortunately, the /opt/developerstudio12.6/bin/cc compiler on these
> machines is a bit of a joke^H^H^H^Hjerk. It cannot compile packages that
> use GNU obstacks:
>
> In 32-bit mode:
> $ cc -c i32.c
> "i32.c", line 61: internal compiler error: ir_process_arrloc_tree: bad arrloc 
> tree
> cc: acomp failed for i32.c
>
> In 64-bit mode:
> $ cc -m64 -c i64.c
> "i64.c", line 61: internal compiler error: ir_process_arrloc_tree: bad arrloc 
> tree
> cc: acomp failed for i64.c
>
> Fortunately gcc is present (and clang as well).

Lol... Peter Gutmann calls SunCC is a collection of scripts that
pretend to be a compiler and linker.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] bootstrapping gcc 13.2.0 on power9/10

[Jeffrey Walton via cfarm-users]

On Fri, Mar 15, 2024 at 8:42 PM Sean McGovern via cfarm-users
 wrote:
> On Fri, Mar 15, 2024 at 6:40 PM Peter Bergner  wrote:
> > On 3/14/24 4:15 AM, Sean McGovern via cfarm-users wrote:
> > > I have not been sucessful doing a simple bootstrap of gcc 13.2.0 on
> > > cfarm29 or cfarm120. It worked fine on cfarm110.
> > >
> > > Are there any special tricks since they are little-endian?
> > >
> > > $ ~/Projects/gcc-13.2.0/configure --prefix=$HOME/tools
> > > --libdir=$HOME/tools/lib64 --enable-languages='c,c++,lto'
> > >
> > > with this configure line and running make I get a bunch of errors when
> > > compiling tree-vec-loop.cc, similar to the following:
> > >
> > > gcc-13.2.0/gcc/tree-vect-loop.cc:5564:12: error: expected unqualified-id
> > [...]
>
> Well, this is embarrassing! It was far simpler than that:
>
> I had CC and CXX environment variables pointing to clang/clang++ in my 
> .bashrc.
> They have been removed.

If Clang is (still?) defining __GNUC__, then it sounds like a Clang
bug. If Clang is going to pretend to be GCC, then it needs to compile
the source files GCC can.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] [cfarm-announces] A new /home for cfarm23 and cfarm91

[Jeffrey Walton via cfarm-users]

On Wed, Jan 10, 2024 at 9:14 PM Vincent Lefevre via cfarm-users
 wrote:
>
> On 2024-01-10 18:02:14 -0800, Paul H. Hargrove wrote:
> > It might just be you, Vincent.
> >
> > I am logged in to cfarm23 right now, copying files from /oldhome
> > So it *is* possible to log in.
>
> This is strange. I can log in on other machines, including the new
> cfarm29.cfarm.net, but not cfarm23.cfarm.net. With "ssh -v", I can
> see that the right key is picked up. But then I get
>
> debug1: send_pubkey_test: no mutual signature algorithm
>
> for cfarm23 instead of the usual
>
> debug1: Server accepts key: [...]
>
> This is a ssh-rsa key.
>
> For cfarm91.cfarm.net, it takes several dozens of seconds to get
> the "debug1: Server accepts key: [...]" line, and several dozens
> of seconds again until I get the prompt.

OpenSSH tightened up RSA key usage in both 9.2 and 9.1. RSA moduli
need to meet a minimum length, and use RSA/SHA256 nowadays. Small RSA
moduli will cause trouble, as will RSA/SHA1 usage. Also see
.

Use ecdsa and ed25519 keys nowadays.

My authorized_keys file includes ed25519, ecdsa and rsa. My config
prefers the newer algorithms:

$ cat ~/.ssh/config
...
Host *.fsffrance.org fsffrance.org
   User noloader
Host *.cfarm.net cfarm.net
   User noloader
...
Host *
   User jwalton

# Identities are tried in order
IdentityFile ~/.ssh/id_ed25519
IdentityFile ~/.ssh/id_ecdsa
IdentityFile ~/.ssh/id_rsa

Protocol 2
PreferredAuthentications publickey,password

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] [cfarm-announces] Transition to the new cfarm.net domain name

[Jeffrey Walton via cfarm-users]

On Fri, Oct 13, 2023 at 3:04 PM CFarm Annoucements via cfarm-announces
 wrote:
>
> The transition to our new cfarm.net domain is complete. All machines are
> now referred to as cfarmXX, or cfarmXX.cfarm.net. The old naming
> convention and domain (gccXX and gccXX.fsffrance.org) are obsolete and
> should not be used.
>
> The website formerly at https://cfarm.tetaneutral.net is now available
> (with redirects) at https://portal.cfarm.net, and a new landing page for
> the project, complete with documentation, will soon be available directly
> at cfarm.net. This is still in progress.
>
> DNS records for fsffrance.org will no longer be updated, and new machines
> will only be added under the cfarm.net domain.

+1 for updating the list of machines, and the instructions at the head
of the page. .

And I think "Announcement" is mis-spelled. It is missing a "n":

> CFarm Annoucements via cfarm-announces

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] GCC119, zip and unzip programs

[Jeffrey Walton via cfarm-users]

On Mon, Oct 2, 2023 at 12:13 PM David Edelsohn  wrote:
>
> I have installed the unzip package.

Thank you sir!

> On Mon, Oct 2, 2023 at 12:06 PM Jeffrey Walton via cfarm-users 
>  wrote:
>>
>> Hi Everyone,
>>
>> Is it possible to install zip and unzip programs on GCC119?
>>
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] GCC119, zip and unzip programs

[Jeffrey Walton via cfarm-users]

Hi Everyone,

Is it possible to install zip and unzip programs on GCC119?

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] GCC119 and XLC/C++

[Jeffrey Walton via cfarm-users]

On Sun, Oct 1, 2023 at 12:39 PM mirabilos via cfarm-users
 wrote:
>
> Noah Misch via cfarm-users dixit:
>
> >/opt/IBM/openxlC/17.1.1/bin/ibm-clang_r is the new system's XLC.
>
> Huh, so it’s also “just another LLVM+Clang”, not the old xlC?

It sounds like it.

No doubt Clang will define __xlC__ but fail to compile a program
written for xlC, like one using __vcipher, __vcipherlast, __vshasigmaw
and __vshasigmad (AES and SHA intrinsics).

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] GCC119 and XLC/C++

[Jeffrey Walton via cfarm-users]

On Sun, Oct 1, 2023 at 11:33 AM Noah Misch  wrote:
>
> On Sun, Oct 01, 2023 at 11:27:28AM -0400, Jeffrey Walton via cfarm-users 
> wrote:
> > I _thought_ GCC119 had IBM's XLC/C++ installed.
> >
> > gcc119:$ command -v xlC
> > gcc119:$
> >
> > And:
> >
> > gcc119:$ find /opt -name xlC
> > gcc119:$
> >
> > Is that no longer the case?
>
> A new AIX 7.3 box replaced the system formerly known as gcc119.
> /opt/IBM/openxlC/17.1.1/bin/ibm-clang_r is the new system's XLC.

Ok, thanks.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] GCC119 and XLC/C++

[Jeffrey Walton via cfarm-users]

Hi Everyone,

I _thought_ GCC119 had IBM's XLC/C++ installed.

gcc119:$ command -v xlC
gcc119:$

And:

gcc119:$ find /opt -name xlC
gcc119:$

Is that no longer the case?

Thanks in advance,

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] cfarm29 access

[Jeffrey Walton via cfarm-users]

Hi Everyone,

I'm trying to connect to cfarm29. According to [0], I _think_ I need
to use cfarm29.fsffrance.org. (There are no instructions for cfarmXX
machines). But:

$ ssh cfarm29.fsffrance.org
ssh: Could not resolve hostname cfarm29.fsffrance.org: Name or
service not known

Does anyone know what I am doing incorrectly?

Thanks in advance.

Jeff

[0] https://cfarm.tetaneutral.net/machines/list/
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] Two problems on the gcc119 machine

[Jeffrey Walton via cfarm-users]

On Mon, Sep 19, 2022 at 2:43 PM Paul H. Hargrove via cfarm-users
 wrote:
>
> Regarding `vi`, you may wish to try setting environment variable 
> `TMPDIR=/tmp` or `TMPDIR=$HOME`.

/tmp is sometimes off-limits to users. I think TMPDIR=$HOME/tmp would
probably be the best choice. I started using $HOME/tmp in my scripts
after I encountered a read-only partition that we could usually write
to in other environments.

> While I've not tested gcc119 specifically, this is usually the right way to 
> control the location used for temporary files.
> Of course, given that this is AIX, YMMV.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] Unable to git pull on gcc112

[Jeffrey Walton via cfarm-users]

On Tue, Mar 29, 2022 at 11:01 PM Jacob Bachmeyer  wrote:
>
> Jeffrey Walton via cfarm-users wrote:
> > Hi Everyone,
> >
> > I'm trying a git pull. When I use https://github.com/weidai11/cryptopp/:
> >
> > $ git pull
> > fatal: unable to access 'https://github.com/weidai11/cryptopp/':
> > Problem with the SSL CA c
> > ert (path? access rights?)
>
> If I read this correctly, Git is not able to verify GitHub's TLS
> certificate for some reason.  I will take a stab in the dark and guess
> that a needed root CA certificate is missing.
>
> > When I try git://github.com/weidai11/cryptopp/:
> >
> > $ git pull
> > fatal: remote error:
> >  The unauthenticated git protocol on port 9418 is no longer supported.
> >
>
> While this does not help you, the root of this latter problem seems to
> be that GitHub has decided to deliberately break compatibility with one
> of Git's standard features using "security" as an excuse.  This is, of
> course, ridiculous for public repositories, since public repositories
> are, well, public.

Thanks Jacob.

I wonder how many older CFarm machines it is going to break. Maybe the
farm can setup a Git proxy. Does one even exist?

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] Unable to git pull on gcc112

[Jeffrey Walton via cfarm-users]

Hi Everyone,

I'm trying a git pull. When I use https://github.com/weidai11/cryptopp/:

$ git pull
fatal: unable to access 'https://github.com/weidai11/cryptopp/':
Problem with the SSL CA c
ert (path? access rights?)

When I try git://github.com/weidai11/cryptopp/:

$ git pull
fatal: remote error:
 The unauthenticated git protocol on port 9418 is no longer supported.

I think we need an updated Git for the machine.

$ git --version
git version 1.8.3.1

What do we try next?

Thanks in advance.
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] Does gcc304 sync certificates?

[Jeffrey Walton via cfarm-users]

On Sat, Mar 12, 2022 at 2:24 PM Daniel Widdis via cfarm-users
 wrote:
> ...
> I’ve never been able to log into the macOS M1 machine, gcc304.  It asks for a 
> password (which implies the certificate isn’t synced):

It is probably the RSA key. OpenSSH made a change recently concerning
RSA keys. Also see https://www.openssh.com/releasenotes.html.

Add an ecdsa or ed25519 key to your authorized_keys.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] Install libtool on GCC119

[Jeffrey Walton via cfarm-users]

On Sun, Feb 13, 2022 at 9:36 PM Jeffrey Walton  wrote:
>
> ...
> gcc119:cryptopp$ command -v libtool
> gcc119:cryptopp$
>
> It looks like libtool is missing.
>
> Would it be possible to have libtool installed, please?

Ping...
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] Install libtool on GCC119

[Jeffrey Walton via cfarm-users]

Hi Everyone,

I'm trying to test an Autotools project on GCC119. Bootstrap is failing:

Running bootstrap
  - Running aclocal
  - Running autoupdate
  - Running autoreconf
autoreconf failed, running again.
configure.ac:2587: warning: macro 'AM_ASFLAGS' not found in library
configure.ac:2588: warning: macro 'AM_LDFLAGS' not found in library
configure.ac:146: error: possibly undefined macro: AC_PROG_LIBTOOL
 If this token and others are legitimate, please use m4_pattern_allow.
 See the Autoconf documentation.
autoreconf: /opt/freeware/bin/autoconf failed with exit status: 1
autoreconf failed, again.
bootstrap failed.

gcc119:cryptopp$ command -v libtool
gcc119:cryptopp$

It looks like libtool is missing.

Would it be possible to have libtool installed, please?

Thanks in advance.
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] [cfarm-announces] clock skews

[Jeffrey Walton via cfarm-users]

On Sat, Jan 8, 2022 at 1:30 AM Bruno Haible via cfarm-users
 wrote:
>
> Peter Gutmann wrote:
> > A quick
> > workaround if you're using zipped archives is to add -DD to the command 
> > line.
>
> What do you mean? Neither 'tar' not 'make' takes a '-DD' option.

I believe that would be the -D and -DD option to unzip (1). The man page says:

-D

skip restoration of timestamps for extracted items. Normally, unzip
tries to restore all meta-information for extracted items that are
supplied in the Zip archive (and do not require privileges or impose a
security risk). By specifying -D, unzip is told to suppress
restoration of timestamps for directories explicitly created from Zip
archive entries. This option only applies to ports that support
setting timestamps for directories (currently ATheOS, BeOS, MacOS,
OS/2, Unix, VMS, Win32, for other unzip ports, -D has no effect). The
duplicated option -DD forces suppression of timestamp restoration for
all extracted entries (files and directories). This option results in
setting the timestamps for all extracted entries to the current time.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] Sparc machine access

[Jeffrey Walton via cfarm-users]

On Mon, Nov 29, 2021 at 11:32 PM Bart Van Assche via cfarm-users
 wrote:
> ...
> I'm using an RSA key but can't log in to gcc210 nor to gcc211. OpenSUSE
> (OpenSSH_8.8p1) displays the following message:
>
> $ ssh gcc210.fsffrance.org
> Unable to negotiate with 213.178.77.185 port 22: no matching host key
> type found. Their offer: ssh-rsa,ssh-dss
>
> And Ubuntu (OpenSSH_8.4p1) displays the following message:
>
> $ ssh gcc210.fsffrance.org
> Unable to negotiate with 213.178.77.185 port 22: no matching key
> exchange method found. Their offer:
> gss-group1-sha1-toWM5Slw5Ew8Mqkay+al2g==,diffie-hellman-group-exchange-sha1,diffie-hellman-group14-sha1,diffie-hellman-group1-sha1
>
> Am I perhaps missing something?

Maybe related from the OpenSSH 8.8 release notes
(https://www.openssh.com/releasenotes.html):

This release disables RSA signatures using the
SHA-1 hash algorithm by default...

Incompatibility is more likely when connecting to
older SSH implementations that have not been
upgraded or have not closely tracked
improvements in the SSH protocol. For these
cases, it may be necessary to selectively
re-enable RSA/SHA1...

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] New Apple Mac M1 machine (gcc304)

[Jeffrey Walton via cfarm-users]

On Wed, Mar 24, 2021 at 10:50 AM Jonas Maebe via cfarm-users
 wrote:
>
> On 23/03/2021 01:31, Assaf Gordon via cfarm-users wrote:
> > - will it compromise SIP (
> > https://en.wikipedia.org/wiki/System_Integrity_Protection ) ?
>
> Note that keeping SIP enabled completely decimates compiler regression
> testing performance, because it means that every time you execute a
> compiled binary for the first time,
> 1) it gets checked for malware (XprotectService)
> 2) its code signature gets checked (syspolicyd, trustd, tccd) [1]
>
> Both 1) and 2) happen in single-threaded processes that handle only a
> single binary at a time. Moreover, if a network connection is available,
> checking a code signature involves checking with Apple's root
> certificate servers (to verify that the used certificate has not been
> revoked) [2]. The combination of these points is that the system spends
> way more time checking for malware and verifying certificates than
> executing test programs.

Yeah, but the other side to disabling SIP is a bunch of broken
packages. Libgcrypt, Nettle, GnuPG and friends can't get through their
self tests because they are not being tested on a SIP-enabled machine.
Some of the breaks have been known for over a year...

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] New Apple Mac M1 machine (gcc304)

[Jeffrey Walton via cfarm-users]

> 4. Seems like the MiniMac-M1 crashed and restarted itself about two
> hours ago.

That problem was reported when the M1 was first released. I understand
the problem has been corrected in an OS X update.

It looks like the M1 needs to be updated. OS X 10.2.3 is the latest:

@GCC-Farm-MiniMac ~ % sw_vers
ProductName:macOS
ProductVersion: 11.2
BuildVersion:   20D64

> 2. gnu libtool is already available as "glibtool" (with HomeBrew).
> It doesn't seems there's a "g-prefix-less" version installed.
> "brew info libtool" says:
> ==> Caveats
> In order to prevent conflicts with Apple's own libtool we have
> prepended a "g" so, you have instead: glibtool and glibtoolize.

Oh yeah... You will confuse the hell out of Autotools if you install
libtool without the g prefix. Apparently Autotools does not do a
simple grep to see if libtool is Apple's one.

% libtool -V
Apple Inc. version cctools-977.1

Autotools also ignores envars LIBTOOL and LIBTOOLIZE, so there's no
way easy way to fix it (the missing g prefix).

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] New Apple Mac M1 machine (gcc304)

[Jeffrey Walton via cfarm-users]

On Mon, Mar 22, 2021 at 12:49 PM Bruno Haible via cfarm-users
 wrote:
>
> > 6. There is no way to do interactive debugging, AFAICS:
>
> Post-mortem debugging, through core dumps, does not work either [1][2].

If you are familiar with Zsh and don't mind experimenting with it,
then see this thread:
https://developer.apple.com/forums/thread/671753.

I don't know if it is useful or not. But it is a different answer than
what you come across on the web.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] [cfarm-announces] New Apple Mac M1 machine (gcc304)

[Jeffrey Walton via cfarm-users]

On Sat, Mar 20, 2021 at 5:54 PM CFarm Annoucements via cfarm-announces
 wrote:
>
> We are happy to announce the availability of a new MiniMac M1 (with
> Apple-Silicon arm64 cpu architecture).
>
> The machine is running OS X in a default configuration, please read the
> login message carefully for guidance.
>
> You can connect over SSH at gcc304.fsffrance.org using custom SSH port 2409.
> See https://cfarm.tetaneutral.net/machines/list/ for details.
>
> Disk space on this machine is tight. Please remember to clean up unused
> build trees before you log out.
>
> Hosting is provided by House Gordon Software Company LTD in Calgary,
> Canada, many thanks to them!
>
> We are always looking for more machines to expand the farm, don't hesitate
> to get in touch if you are able to provide (and host) interesting
> hardware.

That's a nice addition to the compile farm. Thanks for providing it.

For those who are not aware, it is an AMRv8.2-a machine. It has cpu
extensions like SHA512 and SHA3:

% sysctl -a | grep brand
machdep.cpu.brand_string: Apple M1

and:

% sysctl -a | grep armv8
hw.optional.armv8_1_atomics: 1
hw.optional.armv8_crc32: 1
hw.optional.armv8_2_fhm: 1
hw.optional.armv8_2_sha512: 1
hw.optional.armv8_2_sha3: 1

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] GCC220 and error: 'immintrin.h' file not found

[Jeffrey Walton via cfarm-users]

Hi Everyone,

I'm doing some testing on GCC220, which is OpenBSD 6.8. I'm catching a
compile error:

misc.cpp:24:25: error: immintrin.h: No such file or directory
...
misc.cpp:81: error: '__m128i' was not declared in this scope
...
misc.cpp:83: error: expected type-specifier before '__m128i'
...
misc.cpp:84: error: '_mm_castsi128_ps' was not declared in this scope
...

It looks like  is missing and the SSE gear is not
included by other means.

I found several similar issues on the web and a similar thread at
https://lists.freebsd.org/pipermail/freebsd-current/2018-July/070290.html.
The thread says the headers need to be installed. I'm not a BSD person
so I am not sure how correct that is.

Can someone verify all the headers are installed on the system?

Thanks in advance,

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] MIPS machines with MSA?

[Jeffrey Walton via cfarm-users]

Hi Everyone,

Does anyone know where to find a MIPS machine with MSA for testing?

GCC22, GCC23 or GCC24 are mips64 machines, but they lack MSA (thanks AP).

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] GCC 10.10.2 on POWER8

[Jeffrey Walton via cfarm-users]

On Mon, Dec 21, 2020 at 3:51 PM Joseph Benden via cfarm-users
 wrote:
>
> What about the gcc203 machine?
>
> Linux gcc203 5.9.0-4-powerpc64 #1 SMP Debian 5.9.11-1 (2020-11-27) ppc64
> jbenden@gcc203:~$ gcc -v
> Using built-in specs.
> COLLECT_GCC=gcc
> COLLECT_LTO_WRAPPER=/usr/lib/gcc/powerpc64-linux-gnu/10/lto-wrapper
> Target: powerpc64-linux-gnu
> Configured with: ../src/configure -v --with-pkgversion='Debian 10.2.1-1'
> --with-bugurl=file:///usr/share/doc/gcc-10/README.Bugs
> --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++ --prefix=/usr
> --with-gcc-major-version-only --program-suffix=-10
> --program-prefix=powerpc64-linux-gnu- --enable-shared
> --enable-linker-build-id --libexecdir=/usr/lib
> --without-included-gettext --enable-threads=posix --libdir=/usr/lib
> --enable-nls --enable-bootstrap --enable-clocale=gnu
> --enable-libstdcxx-debug --enable-libstdcxx-time=yes
> --with-default-libstdcxx-abi=new --enable-gnu-unique-object
> --disable-libquadmath --disable-libquadmath-support --enable-plugin
> --enable-default-pie --with-system-zlib
> --enable-libphobos-checking=release --with-target-system-zlib=auto
> --enable-objc-gc=auto --enable-secureplt --disable-softfloat
> --enable-targets=powerpc64-linux,powerpc-linux --enable-multiarch
> --disable-werror --with-long-double-128 --enable-multilib
> --enable-checking=release --build=powerpc64-linux-gnu
> --host=powerpc64-linux-gnu --target=powerpc64-linux-gnu
> --with-build-config=bootstrap-lto-lean --enable-link-mutex
> Thread model: posix
> Supported LTO compression algorithms: zlib zstd
> gcc version 10.2.1 20201207 (Debian 10.2.1-1)

Perfect, thanks. I can duplicate the issue on GCC203.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] GCC 10.10.2 on POWER8

[Jeffrey Walton via cfarm-users]

On Mon, Dec 21, 2020 at 3:44 PM David Edelsohn  wrote:
>
> What does GCC 10.10.2 mean?  There is no GCC 10.10.2 release.  Do you
> mean GCC 10.2?

I guess I am parsing the version number wrong from the package name.
Here is the package name:

g++-10_10.2.1-1 gcc-10_10.2.1-1

> There are no Power systems in the GNU Compile Farm running Debian,
> only CentOS and Fedora.

Yeah, I was hoping it is a bad interaction with the compiler and we
can duplicate it with the same compiler.

> GCC for PPC64LE defaults to Power8, so there should be no illegal
> instructions.  Is the Debian system running a correct version of the
> Linux kernel for Power8?  There could be some sort of an unaligned
> access or other special situation with a Power8 instruction that the
> Debian  kernel is not sufficiently recent or missing a patch to handle
> the fixup.

I'm not sure about the details of the Debian machine.

Thanks,

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] gcc22 login troubles?

[Jeffrey Walton via cfarm-users]

Hi Everyone,

I'm having trouble logging in to gcc22. Other machines are OK.

gcc22 looks a little suspicious. It looks like both my ECDSA and RSA
keys were rejected.

Is anyone else having trouble?

Jeff



debug1: Authentications that can continue: publickey,password
debug1: Offering public key: /c/Users/Jeff/.ssh/id_ecdsa ECDSA
SHA256:GxE2scawbwH/lzz1XmWKqP0NlQyBryFeuTgmzVbssfs explicit
debug3: send packet: type 50
debug2: we sent a publickey packet, wait for reply
debug3: receive packet: type 51
debug1: Authentications that can continue: publickey,password
debug1: Offering public key: /c/Users/Jeff/.ssh/id_rsa RSA
SHA256:METoVKRhTmhvcs9s2CUnaeyfLUqCeh06YsLLbrOXelo explicit
debug3: send packet: type 50
debug2: we sent a publickey packet, wait for reply
debug3: receive packet: type 51
debug1: Authentications that can continue: publickey,password
debug2: we did not send a packet, disable method
debug3: authmethod_lookup password
debug3: remaining preferred:
debug3: authmethod_is_enabled password
debug1: Next authentication method: password
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] getopt on AIX

[Jeffrey Walton via cfarm-users]

On Sun, Jul 5, 2020 at 3:07 PM zatrazz  wrote:
>
> On Sun, Jul 5, 2020 at 4:00 PM Bruno Haible via cfarm-users 
>  wrote:
>>
>> > b2sum-src/b2sum.c:266:26: error: array type has incomplete element
>> > type 'struct option'
>> >  static struct option long_options[] = {
>> >   ^~~~
>>
>> When you use 'struct option', you need to
>>
>>   1) use Gnulib module 'getopt-gnu' - see the documentation
>>  https://www.gnu.org/software/gnulib/manual/html_node/getopt.html
>>  https://www.gnu.org/software/gnulib/manual/html_node/getopt_002eh.html
>>  
>> https://www.gnu.org/software/gnulib/manual/html_node/getopt_005flong.html
>>
>>   2) #include , not  - see the documentation
>>  https://www.gnu.org/software/gnulib/MODULES.html#module=getopt-gnu
>
> And getopt(3) with long options is a GNU extension implemented by glibc, some
> other libc also provides similar interfaces (such as FreeBSD with 
> getopt_long).
>
> However it seems that AIX does not have a similar interface so the best option
> is indeed to use gnulib.

Ack, thanks Bruno and zatrazz.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] getopt on AIX

[Jeffrey Walton via cfarm-users]

Hi Everyone,

I'm testing on gcc119, which is the AIX ppc64be machine. One of the
source files fails to compile (see below).

According to the getopt(3) man page, the function is available. The
man page says the only include needed is .

I don't use getopt for my programs so I don't have much experience with it.

Any ideas how to use getopt(3) on AIX?

Thanks in advance.

=

Error when including 

b2sum-src/b2sum.c:266:26: error: array type has incomplete element
type 'struct option'
 static struct option long_options[] = {
  ^~~~
b2sum-src/b2sum.c:267:18: error: 'no_argument' undeclared (first use
in this function)
   { "help",  no_argument, 0,  0  },
  ^~~

Error when including :

b2sum-src/b2sum.c:24:10: fatal error: getopt.h: No such file or directory
 #include 
  ^~
compilation terminated.
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] GCC135/Power9 performance?

[Jeffrey Walton via cfarm-users]

On Sun, Apr 12, 2020 at 10:20 AM Andy Polyakov via cfarm-users
 wrote:
>
> >>> GCC135 is a Power9 machine. Benchmarking on the machine shows
> >>> performance is off. For example, here are some numbers for AES in ECB
> >>> mode:
> >>>
> >>> GCC112 (Linux, ppc64le, 3.7 GHz, GCC 8.2):
> >>>* 1.12 cpb, 2851 MB/s
> >>>
> >>> GCC119 (AIX, ppc64be, 4.1 GHz, GCC 8.2):
> >>>* 0.54 cpb, 7242 MB/s
>
> Just in case, AIX is misreporting resource usage (getrusage), and
> presented result appears to be aligned with the said misrepresentation.
> I mean if you try to calculate performance based on the skewed getrusage
> values, you'll customarily observe ~2 "better" cycles-per-byte results.
> In essence you should either use wallclock or disregard AIX results. But
> don't ask me why AIX does it, as I don't know...

Yeah, we use simple wall clock via clock() calls. It seems to perform
most reliably on most platforms as long as high precision is not
needed.

> >>> GCC135 (Linux, ppc64le, 3.8 GHz, GCC 8.3):
> >>>* 1.94 cpb, 1815 MB/s
> >>
> >> What source code did you use for this test?
> >
> > I used Crypto++ (https://github.com/weidai11/cryptopp) for the test.
> >
> > I also spoke with Andy Polyakov. OpenSSL is observing the same issue.
>
> As already said POWER9 is "allergic" to mixing scalar and vector
> instructions. And since you will always have scalar instructions in the
> mix, most notably to calculate effective addresses, vector code is
> effectively bound to perform suboptimally. It's just the way POWER9 is,
> ...

Yeah, I've been thinking about that. How would the following perform
for loop control?

# Run 10 iterations
vector unsigned int x, l, s;
x = vec_spalt(1);
l = vec_splat(10);
c = vec_spalt(1);
while (vec_all_ne(x, y))
{
...
x = vec_add(x, s);
}

The return value from vec_all_ne is an int. Will using a vector as
loop control improve performance.

> What
> one can do is to calculate as much effective addresses as possible in
> advance and group those instructions, as opposite to spreading them
> throughout loop. And of course, if you rely on compiler intrinsics, you
> are at compiler's mercy, and is not exactly in position to control
> effective address calculations (and complain ;-).

Yeah, it is a shame intrinsics are second class citizens.

Clang calculates effective addresses using indexes (pointer math)
while GCC and XLC use offsets (integer math). It makes it harder to
follow the advice.

I thought Power9 was going to make things easier due to the vector
char and vector short loads, but they don't matter much when the
machine runs code more slowly.

> >>> All algorithms show a similar slowdown. SHA is so slow I am
> >>> considering disabling in-core crypto for SHA and going back to the
> >>> integer unit.
>
> While IBM screwed up vector-scalar mix, they did improve scalar
> performance in POWER9, significantly[!]. So that the gap between between
> vector and equivalent scalar implementations gets reduced from both
> ends. I mean vector is slower, and scalar is faster, both in comparison
> to POWER8 that is. And it's more so from scalar end. Properly optimized
> (a.k.a. hand-written) vector SHA is faster than scalar, but by mere 12%,
> so that if you let compiler calculate effective addresses, it shouldn't
> come as surprise if vector turns out slower than scalar. And again, it's
> just the way POWER9 is, just accept it.
>
> >>> What is different about GCC135? Is the Power9 hardware really that slow?
> >>
> >> Generally no: https://www.ibm.com/downloads/cas/K90RQOW8
>
> One should make distinction between multi-user/capacity suite-specific
> benchmarks (like SPEC/SAP/Oracle/etc.) and cryptographic primitive
> cycles-per-byte benchmarks for single thread on idle system. In addition
> cryptographic algorithms are to certain degree special case even in
> single-thread context, because they customarily have relatively short
> dependencies between steps, which results on all kinds of special and
> non-formalize-able relations between compiler (or assembler programmer)
> and hardware :-) In other words better SPECrates are not guaranteed
> indication of better cycles-per-byte for crypto primitives.

Thanks.
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] GCC135/Power9 performance?

[Jeffrey Walton via cfarm-users]

On Sun, Apr 12, 2020 at 5:11 AM Noah Misch  wrote:
>
> On Sun, Apr 12, 2020 at 12:03:13AM -0400, Jeffrey Walton via cfarm-users 
> wrote:
> > GCC135 is a Power9 machine. Benchmarking on the machine shows
> > performance is off. For example, here are some numbers for AES in ECB
> > mode:
> >
> > GCC112 (Linux, ppc64le, 3.7 GHz, GCC 8.2):
> >* 1.12 cpb, 2851 MB/s
> >
> > GCC119 (AIX, ppc64be, 4.1 GHz, GCC 8.2):
> >* 0.54 cpb, 7242 MB/s
> >
> > GCC135 (Linux, ppc64le, 3.8 GHz, GCC 8.3):
> >* 1.94 cpb, 1815 MB/s
>
> What source code did you use for this test?

I used Crypto++ (https://github.com/weidai11/cryptopp) for the test.

I also spoke with Andy Polyakov. OpenSSL is observing the same issue.

> > All algorithms show a similar slowdown. SHA is so slow I am
> > considering disabling in-core crypto for SHA and going back to the
> > integer unit.
> >
> > What is different about GCC135? Is the Power9 hardware really that slow?
>
> Generally no: https://www.ibm.com/downloads/cas/K90RQOW8
>
> Your source code or compiler may tune poorly for POWER9.  I assume you checked
> that gcc135 was lightly loaded at the time of your test.

It has been happening for at least a year.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] GCC135/Power9 performance?

[Jeffrey Walton via cfarm-users]

Hi Everyone,

GCC135 is a Power9 machine. Benchmarking on the machine shows
performance is off. For example, here are some numbers for AES in ECB
mode:

GCC112 (Linux, ppc64le, 3.7 GHz, GCC 8.2):
   * 1.12 cpb, 2851 MB/s

GCC119 (AIX, ppc64be, 4.1 GHz, GCC 8.2):
   * 0.54 cpb, 7242 MB/s

GCC135 (Linux, ppc64le, 3.8 GHz, GCC 8.3):
   * 1.94 cpb, 1815 MB/s

All algorithms show a similar slowdown. SHA is so slow I am
considering disabling in-core crypto for SHA and going back to the
integer unit.

What is different about GCC135? Is the Power9 hardware really that slow?

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] What is the name of the compiler on GCC118

[Jeffrey Walton via cfarm-users]

Hi Everyone,

What is the name of the compiler on GCC118?

$ find / -name 'g++' 2>/dev/null
/usr/share/bash-completion/completions/g++
$

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] Please check your memory usage on GCC117

[Jeffrey Walton via cfarm-users]

On Sun, Oct 13, 2019 at 11:47 AM Noah Misch  wrote:
>
> On Sun, Oct 13, 2019 at 09:32:48AM -0400, Jeffrey Walton via cfarm-users 
> wrote:
> > On Sun, Oct 13, 2019 at 9:31 AM Segher Boessenkool 
> >  wrote:
> > > On Sun, Oct 13, 2019 at 08:42:14AM -0400, Jeffrey Walton via cfarm-users 
> > > wrote:
> > > > $ make
> > > > g++ -DNDEBUG -g2 -O3 -fPIC -pthread -pipe -c xts.cpp
> > > > virtual memory exhausted: Cannot allocate memory
> > > > GNUmakefile:1625: recipe for target 'xts.o' failed
> > > > make: *** [xts.o] Error 1
>
> > My own job was a 'make' which tried to compile 1 c++ source file.
> >
> > As I told you I seriously doubt it took up 8 GB of RAM.
>
> You can rule that out by setting RLIMIT_AS (e.g. "(ulimit -v 4194304; make)" 
> for
> a 4G limit).  It's possible to exceed the limit when RSS is low, but not
> exceeding the limit does confirm every kind of memory usage is low.  It would
> not surprise me if some -g2 -O3 compilations take 8G for a single file.

/usr/bin/time -v make xts.o shows the max memory usage at 195K.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] Please check your memory usage on GCC117

[Jeffrey Walton via cfarm-users]

On Sun, Oct 13, 2019 at 9:31 AM Segher Boessenkool
 wrote:
>
> On Sun, Oct 13, 2019 at 08:42:14AM -0400, Jeffrey Walton via cfarm-users 
> wrote:
> > I'm catching a OOM exception while trying to compile a source file on 
> > GCC117:
> >
> > $ make
> > g++ -DNDEBUG -g2 -O3 -fPIC -pthread -pipe -c xts.cpp
> > virtual memory exhausted: Cannot allocate memory
> > GNUmakefile:1625: recipe for target 'xts.o' failed
> > make: *** [xts.o] Error 1
> >
> > GCC117 has 16GB of RAM but no virtual memory because there is no page
> > file. (If I am parsing an off-line conversation correctly). My guess
> > is the OS or someone's application has a memory leak.
> >
> > Please check their jobs and ensure things look OK.
>
> As I told you, this box has only 16GB, and no swap.  Only about half of
> memory is used, and not from automatic tests (I would ruthlessly kill
> those).
>
> Check your *own* job, it must have used 7GB or 8GB itself, making it the
> biggest job running on this shared machine.

My own job was a 'make' which tried to compile 1 c++ source file.

As I told you I seriously doubt it took up 8 GB of RAM.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] Please check your memory usage on GCC117

[Jeffrey Walton via cfarm-users]

Hi Everyone,

I'm catching a OOM exception while trying to compile a source file on GCC117:

$ make
g++ -DNDEBUG -g2 -O3 -fPIC -pthread -pipe -c xts.cpp
virtual memory exhausted: Cannot allocate memory
GNUmakefile:1625: recipe for target 'xts.o' failed
make: *** [xts.o] Error 1

GCC117 has 16GB of RAM but no virtual memory because there is no page
file. (If I am parsing an off-line conversation correctly). My guess
is the OS or someone's application has a memory leak.

Please check their jobs and ensure things look OK.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] Looking for the latest GCC for PowerPC

[Jeffrey Walton via cfarm-users]

Hi Everyone,

I'm looking for the latest GCC for PowerPC. I think I need something
from this week or last week.

GCC135 has "gcc version 8.2.1 20180813" but I am looking for something
newer. The other PowerPCs seem to be older, like GCC 6.2 or 8.2.0.

Does anyone know where to find the latest GCC?

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] GCC 8.2 UBsan and runtime error: signed integer overflow

[Jeffrey Walton via cfarm-users]

On Tue, Nov 27, 2018 at 3:51 PM Jeffrey Walton  wrote:
>
> I've got some self tests failing on PowerPC with GCC 8.2. The program
> was compiled with -O1 and -fsanitize=undefined. The program uses
> Altivec. A sample finding is shown below.
>
> ppc_simd.h:965 is a vec_add, and it happens on a unsigned int vector.
> We don't use signed vector types. I don't believe overflow can happen
> and the algorithm depends on unsigned wrap.

In case it matters:

gcc135:~$ /opt/at12.0/bin/c++ --version
c++ (GCC) 8.2.1 20180813 (Advance-Toolchain-at12.0) [revision 263510]

And a reproducer does not reproduce:

$ cat test.cxx
#include 
#undef vector
#undef pixel
#undef bool

typedef __vector unsigned int uint32x4_p;

int main(int argc, char* argv[])
{
  uint32x4_p a = { 0x, 0x, 0x, 0x };
  uint32x4_p b = { 2, 2, 2, 2};

  uint32x4_p c = vec_add(a, b);

  return 0;
}
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] GCC 8.2 UBsan and runtime error: signed integer overflow

[Jeffrey Walton via cfarm-users]

Hi Everyone,

I've got some self tests failing on PowerPC with GCC 8.2. The program
was compiled with -O1 and -fsanitize=undefined. The program uses
Altivec. A sample finding is shown below.

ppc_simd.h:965 is a vec_add, and it happens on a unsigned int vector.
We don't use signed vector types. I don't believe overflow can happen
and the algorithm depends on unsigned wrap.

I'm only finding two hits on search [1] and neither are a GCC bug
report. The code is clean using Clang 7.0 and sanitizers.

Any ideas what might be going on?

Jeff

[1] 
https://www.google.com/search?q=GCC+ubsan+"vec_add"++"runtime+error%3A+signed+integer+overflow;

$ ./cryptest.exe tv speck
...

Testing SymmetricCipher algorithm SPECK-64/ECB.
ppc_simd.h:965:38: runtime error: signed integer overflow: 544497990 +
1936613733 cannot be represented in type 'int'
ppc_simd.h:982:38: runtime error: signed integer overflow: 2066890867
- -184737755 cannot be represented in type 'int'
...

Test FAILED.
incorrectly decrypted: 43F7B0E539C4CA8A6D54A9...
...
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] Migrate away from openSUSE 42 on gcc118

[Jeffrey Walton via cfarm-users]

On Tue, Nov 13, 2018 at 8:45 AM Martin Jambor  wrote:
>
> On Mon, Nov 12 2018, Jeffrey Walton via cfarm-users wrote:
> >
> > GCC118 is an openSUSE machine. openSUSE does not support ARMv8.1
> > extensions like CRC and Crypto. Attempts to use CRC, PMULL, AES and
> > SHA result in compile failures.
> >
> > Would it be possible to migrate to something newer for GCC118?
> >
> > I believe Fedora would be a nice choice nowadays if you want to
> > provide a variety of OSes and support Aarch64. Fedora's DNF system
> > upgrade works well, and I have never had a problem with distro
> > upgrades.
>
> If it is enough, upgrading it to openSUSE Leap 15 should be fairly easy
> (and I can do it if you trust me with root access).  Reinstalling a new
> distro on these particular Seattle boards can be a bit challenging,
> unless the local network is set up for ARM PXE boot.

Thanks Martin.

I forgot to mention... The farm has four Mustangs boards. The
processors on the Mustang boards lack CRC and Crypto. openSUSE may
make a good choice for the Mustang's since it does not support the
extensions, either.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] Migrate away from openSUSE 42 on gcc118

[Jeffrey Walton via cfarm-users]

Hi Everyone,

GCC118 is an openSUSE machine. openSUSE does not support ARMv8.1
extensions like CRC and Crypto. Attempts to use CRC, PMULL, AES and
SHA result in compile failures.

Would it be possible to migrate to something newer for GCC118?

I believe Fedora would be a nice choice nowadays if you want to
provide a variety of OSes and support Aarch64. Fedora's DNF system
upgrade works well, and I have never had a problem with distro
upgrades.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] CPU speeds of GCC135?

[Jeffrey Walton via cfarm-users]

On Sun, Oct 28, 2018 at 8:42 AM Andy Polyakov via cfarm-users
 wrote:
>
> > The initial benchmarks are kind of flat when using 3.8 GHz as the
> > frequency. I think the problem is, we are not working the machine hard
> > enough so the cpu's are reluctant to move from a low energy state.
>
> I'd say it's more likely because POWER9 appears to be "allergic" to
> mixtures of vector and scalar instructions. And since you are likely to
> reference memory you will always have scalar instructions at least to
> calculate effective addresses. Normalized[!] difference to POWER8 can be
> anywhere from "little" to a "lot". Example of "little" can be ~15% in
> SHA512(*) and VSX Chacha20. Example of "lot" is ~50% for pre-VSX
> Chacha20 implementation where one interleaves scalar and vector in more
> or less equal proportion. Though on the other hand pure scalar code is
> normally faster...

Thanks Andy.

That's what we are seeing. AES and SHA slowed down, and the ChaChaR
sped-up (even the SIMD version of ChaCha benefited).

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] CPU speeds of GCC135?

[Jeffrey Walton via cfarm-users]

On Sat, Oct 27, 2018 at 8:28 AM Jeffrey Walton  wrote:
>
> Hi Everyone,
>
> I want to run some benchmarks on GCC135, which is the ppc64-le machine
> with Power9 hardware.
>
> /proc/cpuinfo is showing me some suspiciously low numbers, like 2166
> MHz. I'm guessing the machine is closer ot 3 GHz and probably higher.
> (GCC119 is 4.1 GHz).
>
> Does anyone know what the nominal and max CPU speeds are?

Thanks everyone.

The initial benchmarks are kind of flat when using 3.8 GHz as the
frequency. I think the problem is, we are not working the machine hard
enough so the cpu's are reluctant to move from a low energy state.

For us, the benchmarks are boring vectorized algorithms. They are not
multi-core or computationally intense like mathematical modeling.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] CPU speeds of GCC135?

[Jeffrey Walton via cfarm-users]

Hi Everyone,

I want to run some benchmarks on GCC135, which is the ppc64-le machine
with Power9 hardware.

/proc/cpuinfo is showing me some suspiciously low numbers, like 2166
MHz. I'm guessing the machine is closer ot 3 GHz and probably higher.
(GCC119 is 4.1 GHz).

Does anyone know what the nominal and max CPU speeds are?

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] What hardware in GCC22

[Jeffrey Walton via cfarm-users]

On Sat, Oct 20, 2018 at 11:23 AM Andrew Pinski  wrote:
>
> MSA is an optional extension to MIPSr4 (I think r4 did not exist either, r5 
> does).  Octeon 3 implements r5 but not MSA.  It does have the virtualization 
> extension though.
>
> After octeon 3, cavium/marvell is no longer working on MIPS cores.  I dont 
> know of any other mainstream source for MIPS cores either.

Thanks again Andrew. I think I found a reference at
https://en.wikipedia.org/wiki/List_of_MIPS_architecture_processors .

According to the chart under Imagination Technologies, it looks like
the P5600 released in 2013 is one that supports MSA. Its slim pickins'
if the list is accurate.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] What hardware in GCC22

[Jeffrey Walton via cfarm-users]

On Sat, Oct 20, 2018 at 10:35 AM Andrew Pinski  wrote:
>
> Octeon 2 does not have MSA. Octeon 2 does not have any SIMD extensions 
> implemented.  Octeon 2 does not even have a floating point unit.
> MSA is newer than MIPS64r2/3.

Thanks Andrew. It looks like I hit another dead end.

I'll look for newer hardware. Does MIPS64r4 sound about right? (I wish
MIPS made this easy like Intel and ARM).

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] Modern GCC for mips64 machines

[Jeffrey Walton via cfarm-users]

On Thu, Oct 18, 2018 at 11:31 PM Zach van Rijn  wrote:
> On Thu, 2018-10-18 at 19:29 -0400, Jeffrey Walton via cfarm-
> users wrote:
> > ...
> > I'm interested in testing a build under a more modern GCC
> > built on mips64.
> > ...
>
> If you have any interest in trying out pre-built toolchains,
> which you can use out of your home directory, there are a few
> available (including MIPS64) here, for GCC 8.2.0, 7.3.0, and
> 6.4.0 (the latter are linked toward the bottom of the page):

OK, thanks. So I was able to compile using the instruction but the
resulting program is a dud:

$ ./test.exe
-bash: ./test.exe: No such file or directory

The file is present, its executable, its mips64, etc.

Running with LD_DEBUG=all it looks like binaries from different
runtimes are being used together. For example, the first linked
library loaded is libtinfo and it was built with a different runtime.
This is not surprising and probably causing the unusual result. I
think this is probably a dead end.

Let me send a request to cfarm-admins to see if someone who is
experienced has time to build a compiler.

Thanks again for the help.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] What hardware in GCC22

[Jeffrey Walton via cfarm-users]

Hi Everyone,

https://cfarm.tetaneutral.net/machines/list/ says GCC22 to GCC24 is
mips64. The cpuinfo is below from GCC23. I can't tell if this machine
has MIPS SIMD Architecture (MSA) extensions.

I know the Cavium Octeon II/CN6120 should lead me to a spec sheet but
I am having trouble finding the information. I think the mips64r2
implies MSA but I could be mistaken.

Does the Octeon II/CN6120 have MSA? If not, then what should I be
looking for when reading through marketing material and technical
descriptions?

Thanks

+++
$ cat /proc/cpuinfo
system type : UBNT_E200 (CN6120p1.1-1000-NSP)
machine : Unknown
processor   : 0
cpu model   : Cavium Octeon II V0.1
BogoMIPS: 2000.00
wait instruction: yes
microsecond timers  : yes
tlb_entries : 128
extra interrupt vector  : yes
hardware watchpoint : yes, count: 2, address/irw mask: [0x0ffc, 0x0ffb]
isa : mips1 mips2 mips3 mips4 mips5 mips64r2
ASEs implemented:
shadow register sets: 1
kscratch registers  : 3
core: 0
VCED exceptions : not available
VCEI exceptions : not available

processor   : 1
cpu model   : Cavium Octeon II V0.1
BogoMIPS: 2000.00
wait instruction: yes
microsecond timers  : yes
tlb_entries : 128
extra interrupt vector  : yes
hardware watchpoint : yes, count: 2, address/irw mask: [0x0ffc, 0x0ffb]
isa : mips1 mips2 mips3 mips4 mips5 mips64r2
ASEs implemented:
shadow register sets: 1
kscratch registers  : 3
core: 1
VCED exceptions : not available
VCEI exceptions : not available
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] Modern GCC for mips64 machines

[Jeffrey Walton via cfarm-users]

Hi Everyone,

I'm interested in testing a build under a more modern GCC built on mips64.
Eventually I'd like to try the MIPS builtins/intrinsics like discussed
at 
https://s3-eu-west-1.amazonaws.com/downloads-mips/documents/MD00927-SIMD-programming-whitepaper.pdf
.

From gcc22:

$ gcc --version
gcc (Debian 4.6.3-14) 4.6.3

And:

$ ls -l /opt/cfarm/gcc-latest
lrwxrwxrwx 1 iulius iulius 36 Dec 10  2017 /opt/cfarm/gcc-latest
-> /home/iulius/autobuild/bin/gcc-7.2.0

But:

$ ls /home/iulius/autobuild/bin/gcc-7.2.0
ls: cannot access /home/iulius/autobuild/bin/gcc-7.2.0: No such
file or directory

It looks like there may be something available.

Any ideas on how I can access a newer compiler?

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] Is it possible to get an x86_64 or aarch64 machine running AIX?

[Jeffrey Walton via cfarm-users]

On Sun, Oct 7, 2018 at 12:28 AM CM Graff via cfarm-users
 wrote:
> ...
> So my question is this: Is it possible to have a virtual machine (or
> maybe even real hardware) running AIX on x86_64 at the farm? hlibc
> also supports arm v7, aarch64 and i386. so I'd be interested in any of
> these. Perhaps out of that list aarch64 and x86_64 would be more
> useful though.

I might be mis-parsing things, but why try to emulate when you have
real hardware?

The farm has GCC112 and GCC119, which are POWER8. GCC112 is Linux
(little-endian), and GCC119 is AIX (big-endian). There are some POWER7
machines also but I don't usually use them.

The farm also has GCC117 and GCC118. The are Aarch64 machines with
ARMv8.1, and they have the crypto extensions. There are some Mustang
boards that are AArch64, but they lack the crypto extensions.

Also see https://cfarm.tetaneutral.net/machines/list/ .

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] Enable one ISA for my code, and another ISA for GCC code?

[Jeffrey Walton via cfarm-users]

On Wed, Aug 15, 2018 at 7:02 AM, Jonathan Wakely  wrote:
> On Wed, 15 Aug 2018 at 11:33, Jeffrey Walton via cfarm-users
>  wrote:
>>
>> I'm catching a SIGILL on POWER7 ppc64-le machine. I compiled the
>> source with -mcpu=power8 so I could enable some builtins I may need.
>> Use of the builtins is guarded at runtime. (This is the classic distro
>> model, where most source code is compiled for a minimum machine but
>> higher ISAs are guarded at runtime).
>>
>> GCC converted some C code into POWER8 code. The SIGILL is happening in
>> the function prologue due to use of mtfprwz. Apparently mtfprwz is a
>> ISA 2.07B/POWER8 instruction.
>>
>> My question is, how do I tell GCC to use a different ISA than I am using?
>
> Doesn't this belong on gcc-h...@gcc.gnu.org not the cfarm users list?

Oh, my bad... I saw the GCC but it was the compile farm and not GCC-help.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] Enable one ISA for my code, and another ISA for GCC code?

[Jeffrey Walton via cfarm-users]

I'm catching a SIGILL on POWER7 ppc64-le machine. I compiled the
source with -mcpu=power8 so I could enable some builtins I may need.
Use of the builtins is guarded at runtime. (This is the classic distro
model, where most source code is compiled for a minimum machine but
higher ISAs are guarded at runtime).

GCC converted some C code into POWER8 code. The SIGILL is happening in
the function prologue due to use of mtfprwz. Apparently mtfprwz is a
ISA 2.07B/POWER8 instruction.

My question is, how do I tell GCC to use a different ISA than I am using?

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] GCC119 and GCC support for init_priority

[Jeffrey Walton via cfarm-users]

Hi Everyone,

GCC119 is the AIX ppc64-be iron. GCC is configured to accept
init_priority. As far as I know AIX linker does not support it.

I believe this is the manual page with the options:
https://www.ibm.com/support/knowledgecenter/en/SSGH4D_15.1.0/com.ibm.xlf151.aix.doc/compiler_ref/fcat_link.html

Can anyone confirm the AIX linker supports initialization priorities?

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] sigsegv in std::locale::operator=(std::locale const&)

[Jeffrey Walton via cfarm-users]

I'm catching a crash in std::locale::operator=(std::locale const&):

0xd0a325e0 in std::locale::operator=(std::locale const&) ()
   from 
/opt/freeware/lib/gcc/powerpc-ibm-aix7.2.0.0/7.2.0/libstdc++.a(libstdc++.so.6)
(gdb) bt
#0  0xd0a325e0 in std::locale::operator=(std::locale const&) ()
   from 
/opt/freeware/lib/gcc/powerpc-ibm-aix7.2.0.0/7.2.0/libstdc++.a(libstdc++.so.6)
#1  0xd0a389fc in std::ios_base::_M_init() ()
   from 
/opt/freeware/lib/gcc/powerpc-ibm-aix7.2.0.0/7.2.0/libstdc++.a(libstdc++.so.6)
#2  0xd0991d08 in std::basic_ios >::init ()
   from 
/opt/freeware/lib/gcc/powerpc-ibm-aix7.2.0.0/7.2.0/libstdc++.a(libstdc++.so.6)
...

I believe the issue is the same one discussed at
https://stackoverflow.com/q/43490185/608639 , but there are no
answers. I've tried setting LC_ALL using several common locales but it
did not help.

This may be related, but it seems like a stretch:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=59439 (maybe the wrong
locks in this instance?)

Does anyone have an idea?

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] Please install CMake on GCC119

[Jeffrey Walton via cfarm-users]

Hi Everyone,

We have the (dis)pleasure of maintaining CMake source files. If it is
not too much trouble could someone install Cmake on GCC119.

CMake is kind of fragile to begin with. I think it would be prudent to
test on AIX to see how things break and then fix the breaks.

Thanks in advance.
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] GCC22 (mips64) and Permission denied

[Jeffrey Walton via cfarm-users]

Hi Everyone,

I'm having trouble accessing GCC22 (and 23 and 24), which is a mip64
machine. Also see https://cfarm.tetaneutral.net/machines/list/ .

Does anyone know if there's a jump box or alternate port for the
machines? Maybe something else?

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] Nice on GCC202?

[Jeffrey Walton via cfarm-users]

On Thu, Mar 8, 2018 at 6:19 PM, Baptiste Jonglez
<bapti...@bitsofnetworks.org> wrote:
> Hi,
>
> On 08-03-18, Jeffrey Walton via cfarm-users wrote:
>> Hi Everyone,
>>
>> GCC202 is a Sparc64 machine. It has 16 cpus and each cpu has 16 cores
>> for a total of 256 virtual cpu's.
>
> Where did you see those 256 "virtual cpu's"?  According to htop,
> /proc/cpuinfo and https://cfarm.tetaneutral.net/machines/list/ it only has
> 16 "threads".

/sys/devices/system/cpu

> As far as I can remember, it's a virtual machine with only 2 CPU cores of
> the physical machine, each core having 8 "threads", although it seems to
> be quite different from Intel Hyperthreading.  Linux does not understand
> such subtleties and just sees each "thread" as a separate CPU.

/proc/cpuinfo shows 16 cpu's.

Related, do you know where the CPU freq is tucked away on Sparc64? It
is not in the places I usually look, like /proc/cpuinfo and
/sys/devices/system/cpu.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] Nice on GCC202?

[Jeffrey Walton via cfarm-users]

Hi Everyone,

GCC202 is a Sparc64 machine. It has 16 cpus and each cpu has 16 cores
for a total of 256 virtual cpu's.

My question is, what is nice? Should I use 8 cores, or can I use 128 of them?

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] gcc112/ppc64-le, and undefined reference to __cxa_thread_atexit_impl

[Jeffrey Walton via cfarm-users]

On Tue, Feb 20, 2018 at 6:54 AM, Jonathan Wakely  wrote:
> ...
>>> That is special.  /opt/cfarm is not installed as an alternate system
>>> compiler.  It's not even the latest compiler.  If you choose to
>>> override the defaults and refer to a random compiler whose origin you
>>> don't know, you may encounter unintended consequences, as you have
>>> discovered.
>>
>> Oh, my bad. I thought the latest stuff was installed in /opt/cfarm.
>>
>> So I am clear: we are _not_ supposed to use stuff in /opt/cfarm. Is
>> that correct?
>
> That's not what he said, he said "If you choose to override the
> defaults and refer to a random compiler whose origin you don't know,
> you may encounter unintended consequences".

Now I am even more confused.

The admins install stiff in /opt/cfarm. I used to look in /usr/local
for the upgrades, but I was told to look in /opt/cfarm instead.

How is that randomized shit that nobody knows anything about?

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] gcc112/ppc64-le, and undefined reference to __cxa_thread_atexit_impl

[Jeffrey Walton via cfarm-users]

>>> On Mon, Feb 19, 2018 at 02:49:47PM -0500, Jeffrey Walton via cfarm-users 
>>> wrote:
>>>> /opt/cfarm/gcc7rc2/bin/g++ -o cryptest.exe -DNDEBUG -g2 -O3 -fPIC
>>>> -pthread -pipe adhoc.o test.o bench1.o bench2.o validat0.o validat1.o
>>>> validat2.o validat3.o validat4.o datatest.o regtest1.o regtest2.o
>>>> regtest3.o dlltest.o fipsalgt.o ./libcryptopp.a
>>>> /home/guerby/opt/cfarm/gcc7rc2/bin/../lib/gcc/powerpc64le-unknown-linux-gnu/7.0.1/../../../../lib64/libstdc++.so:
>>>> undefined reference to `__cxa_thread_atexit_impl@GLIBC_2.18'
>>>> collect2: error: ld returned 1 exit status
>>>> make: *** [cryptest.exe] Error 1
>>>
>>> How/why do you get anything from /home/guerby?  /opt/cfarm/gcc7rc2 was
>>> build on a different machine (and it is only there so people can use it
>>> as an easy starting point for building ada: the system compiler on 112
>>> does not support that) so there is no way it refers to there.
>>
>> Hmm, I'm not sure. I'm on GCC112, and it is in /opt/cfarm:
>>
>> $ ls /opt/cfarm/
>> autoconf-latestftp krb5-latest  mpfr-latest python3-latest
>> bison-latest   gcc7rc2 m4-latestopenssl-latest  release
>> cyrus-sasl-latest  gcc-latest  make-latest  perl5-latestzlib-latest
>> db-latest  gmp-latest  mpc-latest   python2-latest
>>
>>> I know nothing about the build system you used, but it seems buggy.
>>
>> I used a GNUmakefile:
>>
>> CXX=/opt/cfarm/gcc7rc2/bin/g++ make -j 20
>>
>> Nothing special, as far as I know. Just using the latest compiler.
>
> Why are you using /opt/cfarm instead of the default system compiler?

I'm trying to reproduce a problem under Fedora 28.

The problem is not present under the default compiler, which is GCC 4.8.5.

> That is special.  /opt/cfarm is not installed as an alternate system
> compiler.  It's not even the latest compiler.  If you choose to
> override the defaults and refer to a random compiler whose origin you
> don't know, you may encounter unintended consequences, as you have
> discovered.

Oh, my bad. I thought the latest stuff was installed in /opt/cfarm.

So I am clear: we are _not_ supposed to use stuff in /opt/cfarm. Is
that correct?

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] gcc112/ppc64-le, and undefined reference to __cxa_thread_atexit_impl

[Jeffrey Walton via cfarm-users]

On Mon, Feb 19, 2018 at 4:12 PM, Segher Boessenkool
<seg...@kernel.crashing.org> wrote:
> Hi,
>
> On Mon, Feb 19, 2018 at 02:49:47PM -0500, Jeffrey Walton via cfarm-users 
> wrote:
>> /opt/cfarm/gcc7rc2/bin/g++ -o cryptest.exe -DNDEBUG -g2 -O3 -fPIC
>> -pthread -pipe adhoc.o test.o bench1.o bench2.o validat0.o validat1.o
>> validat2.o validat3.o validat4.o datatest.o regtest1.o regtest2.o
>> regtest3.o dlltest.o fipsalgt.o ./libcryptopp.a
>> /home/guerby/opt/cfarm/gcc7rc2/bin/../lib/gcc/powerpc64le-unknown-linux-gnu/7.0.1/../../../../lib64/libstdc++.so:
>> undefined reference to `__cxa_thread_atexit_impl@GLIBC_2.18'
>> collect2: error: ld returned 1 exit status
>> make: *** [cryptest.exe] Error 1
>
> How/why do you get anything from /home/guerby?  /opt/cfarm/gcc7rc2 was
> build on a different machine (and it is only there so people can use it
> as an easy starting point for building ada: the system compiler on 112
> does not support that) so there is no way it refers to there.

Hmm, I'm not sure. I'm on GCC112, and it is in /opt/cfarm:

$ ls /opt/cfarm/
autoconf-latestftp krb5-latest  mpfr-latest python3-latest
bison-latest   gcc7rc2 m4-latestopenssl-latest  release
cyrus-sasl-latest  gcc-latest  make-latest  perl5-latestzlib-latest
db-latest  gmp-latest  mpc-latest   python2-latest

> I know nothing about the build system you used, but it seems buggy.

I used a GNUmakefile:

CXX=/opt/cfarm/gcc7rc2/bin/g++ make -j 20

Nothing special, as far as I know. Just using the latest compiler.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] Please cleanup your /home on farm machines

[Jeffrey Walton via cfarm-users]

On Fri, Jan 19, 2018 at 11:15 AM, David Edelsohn via cfarm-users
 wrote:
> On Fri, Jan 19, 2018 at 10:47 AM, Segher Boessenkool via cfarm-users
>  wrote:
>> On Fri, Jan 19, 2018 at 03:05:48PM +0100, Martin Guy via cfarm-users wrote:
>>> find /home \( \
>>> \( -name '*.o' -a \! -atime +7 \) \
>>
>> I veto this 400%.  People have valuable .o files, and this is not very
>> effective anyway (other files take up more space).
>
> I agree with Segher.
>
> Automatically deleting files is completely unacceptable and contrary
> to the purpose of the GNU Compile Farm.  The Compile Farm requires
> that users are considerate of each other and the shared resources that
> are provided gratis.

Is there a way to make the service "opt-in" through the control panel?
Is it a change the farm is willing to entertain?

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] Please cleanup your /home on farm machines

[Jeffrey Walton via cfarm-users]

On Fri, Jan 19, 2018 at 9:02 AM, Martin Guy via cfarm-users
<cfarm-users@lists.tetaneutral.net> wrote:
> On 18/01/2018, Jeffrey Walton via cfarm-users
> <cfarm-users@lists.tetaneutral.net> wrote:
>> I can't speak for others, but I would not mind being subject to a
>> script that removes temporary object files that are over N days old.
>> In fact I would welcome it because I am forgetful at times.
>
> Traditionally, such a script is called "fileclean", essentialy a
> humungous "find" that deletes *.tmp ~*, *.o older than a week, core
> older than a day and so on.

Sounds perfect.

It makes me wonder why a problem that existed in 1987 still exists to day :)

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] Please cleanup your /home on farm machines

[Jeffrey Walton via cfarm-users]

On Thu, Jan 18, 2018 at 6:17 AM, Baptiste Jonglez via cfarm-users
 wrote:
> Hi everyone,
>
> Let's start the new year with a positive action for the compile farm: home 
> cleanup!
>
> The farm machines with the most critical disk usage are:
>
> - gcc110 (99% used out of 1.6 TB)
> - gcc20  (98% used out of 826 GB)
> - gcc13  (95% used out of 459 GB)
> - gcc112 (92% used out of 1.8 TB)
> - gcc16  (85% used out of 459 GB)
>
> Please check each of these machines: if you have unused data in your home
> directory, it should be deleted!

I can't speak for others, but I would not mind being subject to a
script that removes temporary object files that are over N days old.
In fact I would welcome it because I am forgetful at times.

I don't know how long a "log running" job is for some users. Maybe you
can exclude users who are logged in, and clean objects that meet
criteria for the rest.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] gcc67 down ? was: gcc67 instability?

[Jeffrey Walton via cfarm-users]

On Sun, Dec 31, 2017 at 2:08 AM, Mike Rosset via cfarm-users
 wrote:
> Nicolas,
>
> Will those machines eventually come back up?

+1, those are nice test machines. The 1700x is the only test machines
with IA-32 SHA instructions.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] Installing emacs on GCC119/AIX

[Jeffrey Walton via cfarm-users]

On Wed, Oct 11, 2017 at 1:33 PM, David Edelsohn  wrote:
> ...
>>> GNU Emacs already is installed on gcc119 in /opt/freeware.
>>
>> It is, but when I try to run it, I get an error:
>>
>> jamborm@power8-aix:~/gcc$ which emacs
>> /opt/freeware/bin//emacs
>> jamborm@power8-aix:~/gcc$ emacs
>> exec(): 0509-036 Cannot load program emacs because of the following errors:
>> 0509-130 Symbol resolution failed for 
>> /opt/freeware/lib/libglib-2.0.a(libglib-2.0.so.0) because:
>> 0509-136   Symbol __dbargs (number 187) is not exported from
>>dependent module 
>> /opt/freeware/lib/libintl.a(libintl.so.8).
>> 0509-136   Symbol __dbsubc (number 195) is not exported from
>>dependent module 
>> /opt/freeware/lib/libintl.a(libintl.so.8).
>> 0509-136   Symbol __dbsubg (number 196) is not exported from
>>dependent module 
>> /opt/freeware/lib/libintl.a(libintl.so.8).
>> 0509-136   Symbol __dbsubn (number 197) is not exported from
>>dependent module 
>> /opt/freeware/lib/libintl.a(libintl.so.8).
>> 0509-026 System error: Error 0
>> 0509-192 Examine .loader section symbols with the
>>  'dump -Tv' command.
>>
>> Or am I the only one getting that error?
>
> I upgraded glib2 and Emacs.  Emacs-25.1 seems to be installed and
> working now on gcc119.

Thank you very much, David.
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] Installing emacs on GCC119/AIX

[Jeffrey Walton via cfarm-users]

On Wed, Oct 11, 2017 at 11:39 AM, CM Graff via cfarm-users
<cfarm-users@lists.tetaneutral.net> wrote:
> Continued mention of the need for nearly useless AIX cross platform
> testing while not addressing the need for useful netbsd and freebsd
> cross platform testing is illogical.

I mat be mis-parsing things, but you seem to be conflating two issues.
First, you say AIX is useless. But testing on AIX with GCC and XLC
revealed several issues in Crypto++, which is a free and open software
project: https://github.com/weidai11/cryptopp/issues?q=is%3Aissue%20aix
. So I don't think its useless.

The second is, you want NetBSD and FreeBSD machines. Maybe even a
DragonFly that continually runs out of disk space due to its Hammer
filesystem :) That seems like a separate problem. Perhaps you can
submit a request for two additional VM's - one for NetBSD/PPC and
another for FreeBSD/PPC.

Testing on Solaris often produces similar findings. Its like Wakely
said, I'd love to get my hands on some Solaris machines for testing. I
have a Solaris 11 running on a G5; I need Ultrasparc and Solaris 10
for testing.

Jeff

> On 10/11/17, Toebs Douglass via cfarm-users
> <cfarm-users@lists.tetaneutral.net> wrote:
>> On 11/10/17 16:22, Jeffrey Walton via cfarm-users wrote:
>>> I've also found that using unique compilers (XL/C) and unique
>>> platforms (AIX) helps make the code more portable and reveal some
>>> assumptions made in code.
>>
>> I too have always found this, and I regard using multiple platforms and
>> compilers as an essential step in software quality.
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] gcc119 - xlC version?

[Jeffrey Walton via cfarm-users]

On Wed, Sep 20, 2017 at 9:03 PM, Steven R Loomis via cfarm-users
 wrote:
>
> Hello,
>  Does the gcc119 system have the xlC fix packs present here?
> http://www-01.ibm.com/support/docview.wss?uid=swg27038605#tab_131X

GCC119 (AIX, big endian):

$ xlc -qversion
IBM XL C/C++ for AIX, V13.1.3 (5725-C72, 5765-J07)
Version: 13.01.0003.

GCC112 (Linux, little endian):

$ xlc -qversion
IBM XL C/C++ for Linux, V13.1.5 (5725-C73, 5765-J08)
Version: 13.01.0005.0001

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] Where are the mailing list archives located?

[Jeffrey Walton via cfarm-users]

I'm having trouble finding the mailing list archive and the mailing
list sign-up page from the web pages.

The CFarm wiki page (https://gcc.gnu.org/wiki/CompileFarm) says to go
to https://cfarm.tetaneutral.net/, but the tetaneutral.net pages do
not list it.

https://gna.org/projects/gcc-cfarm/ appears to be down.

Where are the mailing list archives located?

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] GCC119/AIX disassembler?

[Jeffrey Walton via cfarm-users]

On Wed, Sep 13, 2017 at 1:00 AM, Markus Trippelsdorf
<mar...@trippelsdorf.de> wrote:
> On 2017.09.13 at 00:54 -0400, Jeffrey Walton wrote:
>> On Wed, Sep 13, 2017 at 12:51 AM, Markus Trippelsdorf
>> <mar...@trippelsdorf.de> wrote:
>> > On 2017.09.13 at 00:42 -0400, Jeffrey Walton via cfarm-users wrote:
>> >> Hi Everyone,
>> >>
>> >> Please forgive my ignorance. What is (are?) the disassemblers
>> >> available on GCC119, which is an AIX machine?
>> >
>> > In another thread you have stated that you don't read PPC assembly, so
>> > of what use would be a disassembler?
>>
>> I want to examine the code generated by XL/C and compare it GCC.
>>
>> Below is the sequence I am seeing laid out by GCC on GCC112, which is
>> Linux LE. I am interested in learning how the code changes when moving
>> to XL/C on a buig endian machine.
>
> Well, XL/C also supports the -S option.

Thanks. I was not aware it was a disassembler, too.

Thanks again.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] GCC112 and missing hwcaps

[Jeffrey Walton via cfarm-users]

Hi Everyone,

GCC112 is a Power8 machine, but hwcaps is missing features present on
the cpu like in-core crypto.

CPU feature probing shows the features do in fact exist. It looks like
an old hwcap.h file (shown below).

Jeff

$ cat /usr/include/bits/hwcap.h

/* Defines for bits in AT_HWCAP and AT_HWCAP2.
   Copyright (C) 2012-2013 Free Software Foundation, Inc.
   This file is part of the GNU C Library.

   The GNU C Library is free software; you can redistribute it and/or
   modify it under the terms of the GNU Lesser General Public
   License as published by the Free Software Foundation; either
   version 2.1 of the License, or (at your option) any later version.

   The GNU C Library is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
   Lesser General Public License for more details.

   You should have received a copy of the GNU Lesser General Public
   License along with the GNU C Library; if not, see
   .  */

#ifndef _SYS_AUXV_H
# error "Never include  directly; use  instead."
#endif

/* The bit numbers must match those in the kernel's asm/cputable.h.  */

/* Feature definitions in AT_HWCAP.  */
#define PPC_FEATURE_32  0x8000 /* 32-bit mode. */
#define PPC_FEATURE_64  0x4000 /* 64-bit mode. */
#define PPC_FEATURE_601_INSTR   0x2000 /* 601 chip, Old POWER ISA.  */
#define PPC_FEATURE_HAS_ALTIVEC 0x1000 /* SIMD/Vector Unit.  */
#define PPC_FEATURE_HAS_FPU 0x0800 /* Floating Point Unit.  */
#define PPC_FEATURE_HAS_MMU 0x0400 /* Memory Management Unit.  */
#define PPC_FEATURE_HAS_4xxMAC  0x0200 /* 4xx Multiply Accumulator.  */
#define PPC_FEATURE_UNIFIED_CACHE   0x0100 /* Unified I/D cache.  */
#define PPC_FEATURE_HAS_SPE 0x0080 /* Signal Processing ext.  */
#define PPC_FEATURE_HAS_EFP_SINGLE  0x0040 /* SPE Float.  */
#define PPC_FEATURE_HAS_EFP_DOUBLE  0x0020 /* SPE Double.  */
#define PPC_FEATURE_NO_TB   0x0010 /* 601/403gx have no timebase */
#define PPC_FEATURE_POWER4  0x0008 /* POWER4 ISA 2.00 */
#define PPC_FEATURE_POWER5  0x0004 /* POWER5 ISA 2.02 */
#define PPC_FEATURE_POWER5_PLUS 0x0002 /* POWER5+ ISA 2.03 */
#define PPC_FEATURE_CELL_BE 0x0001 /* CELL Broadband Engine */
#define PPC_FEATURE_BOOKE   0x8000 /* ISA Category Embedded */
#define PPC_FEATURE_SMT 0x4000 /* Simultaneous
  Multi-Threading */
#define PPC_FEATURE_ICACHE_SNOOP0x2000
#define PPC_FEATURE_ARCH_2_05   0x1000 /* ISA 2.05 */
#define PPC_FEATURE_PA6T0x0800 /* PA Semi 6T Core */
#define PPC_FEATURE_HAS_DFP 0x0400 /* Decimal FP Unit */
#define PPC_FEATURE_POWER6_EXT  0x0200 /* P6 + mffgpr/mftgpr */
#define PPC_FEATURE_ARCH_2_06   0x0100 /* ISA 2.06 */
#define PPC_FEATURE_HAS_VSX 0x0080 /* P7 Vector Extension.  */
#define PPC_FEATURE_PSERIES_PERFMON_COMPAT  0x0040
#define PPC_FEATURE_TRUE_LE 0x0002
#define PPC_FEATURE_PPC_LE  0x0001

/* Feature definitions in AT_HWCAP2.  */
#define PPC_FEATURE2_ARCH_2_07 0x8000 /* ISA 2.07 */
#define PPC_FEATURE2_HAS_HTM   0x4000 /* Hardware Transactional
 Memory */
#define PPC_FEATURE2_HAS_DSCR  0x2000 /* Data Stream Control
 Register */
#define PPC_FEATURE2_HAS_EBB   0x1000 /* Event Base Branching */
#define PPC_FEATURE2_HAS_ISEL  0x0800 /* Integer Select */
#define PPC_FEATURE2_HAS_TAR   0x0400 /* Target Address Register */
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] GCC119 (AIX), GDB 7.9.1, and unable to display VMX regs

[Jeffrey Walton via cfarm-users]

Hi Everyone,

I'm working on GCC119 which is an AIX machine. Its a Power 8, so it
has the VMX unit.

I'm trying to use the in-core crypto. The instructions are gets
processed by the VMX unit. I need to see the reg values, but GDB
cannot print them (see below).

Does anyone know how to display the registers when working on AIX?

Thanks in advance.

==

(gdb) info all-registers
r0 0x1032   0x1032
r1 0x2ff22a60   0x2ff22a60
...
xer0x3d 0x3d
fpscr  0x0  0x0
vr0{uint128 = , v4_float = {,
, , }, v4_int32 = {,
, , }, v8_int16 = {,
, , , , ,
, }, v16_int8 = {
 }}

...

[repeated]

...
vr31   {uint128 = , v4_float = {,
, , }, v4_int32 = {,
, , }, v8_int16 = {,
, , , , ,
, }, v16_int8 = {
 }}
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] Installing emacs on GCC119/AIX

[Jeffrey Walton via cfarm-users]

On Thu, Aug 31, 2017 at 8:12 AM, David Edelsohn <dje@gmail.com> wrote:
> On Thu, Aug 31, 2017 at 7:30 AM, Jeffrey Walton via cfarm-users
> ...
> If you want to test building GNU Emacs for some reason, go ahead, but
> your build of Emacs will not be installed for general use.

Oh, thank you. I checked in /opt/cfarm but did not see emacs.

Is emacs always installed in /opt/freeware? Is it part of the standard build?

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] Installing emacs on GCC119/AIX

[Jeffrey Walton via cfarm-users]

Hi Everyone,

I'd like to install emacs on GCC119/AIX.

I'm not familiar with AIX or installing into /opt/cfarm. In
particular, I'm not sure if AIX suffers the ridiculous runtime path
problems that plague Linux; and how to fix runtime path problems on
AIX if they exist.

Does the configure line below need anything added or taken away for AIX?

Thanks in adavnce.

./configure --with-xml2 --with-zlib \
  --without-x --without-sound --without-xpm \
  --without-jpeg --without-tiff --without-gif --without-png --without-rsvg \
  --without-imagemagick --without-xft --without-libotf --without-m17n-flt \
  --without-xaw3d --without-toolkit-scroll-bars --without-gpm --without-dbus \
  --without-gconf --without-gsettings --without-makeinfo \
  --without-compress-install \
  --prefix=/opt/cfarm
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] Please install Valgrind on GCC117

[Jeffrey Walton via cfarm-users]

On Tue, Aug 15, 2017 at 3:11 PM, Aymeric via cfarm-users
 wrote:
> ...
> gcc117 root filesystem is in read-only, need to be rebooted to recover
> (btrfs probably went wrong)
>
> As /home is still available in R/W, if you've data you want to keep,
> please backup them, I will reboot it tomorrow (around 19:00 UTC) during
> the upgrade of gcc12,13 et 14.

Ack, thanks. Bounce it when it is convenient for you. I can always log back in.

You might want to build Valgrind from sources and avoid distro
packages. I have 8 ARM dev boards (including two Aarch64), and there's
almost always VEX problems using prebuilt packages. Every ARM dev
board I have has Valgrind built from sources due to VEX updates
(http://valgrind.org/downloads/repository.html).

If you use the distro package, then some folks may come back in the
future asking for a newer version due to VEX parsing/decoding errors.
You may as well side step it now so its not a problem in the future.

Thanks for the help.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] Please install Valgrind on GCC117

[Jeffrey Walton via cfarm-users]

Hi Everyone,

I have another small request. Please install Valgrind on GCC117.

We are catching a Address Sanitizer finding using gcc-latest. We want
to take a look at it with Valgrind. (And we don't witness the finding
on other platforms).

Thanks in advance.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] Install NTP client on GCC117 and GCC118

[Jeffrey Walton via cfarm-users]

It looks like GCC117 has about a 7 minute drift:

gcc117$ date
Sun Aug  6 12:13:19 EDT 2017

hikey$ date
Sun, Aug  6, 2017 12:06:28 PM

My makefile does not build out of date targets that have been scp'd to
the machine.

Would it be possible to install a ntp client on GCC117 and GCC 118?

Thanks in advance.
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] gcc202: New UltraSparc T5 machine

[Jeffrey Walton via cfarm-users]

On Thu, Jul 27, 2017 at 4:34 PM, Segher Boessenkool via cfarm-users
 wrote:
> ...
>> gcc110 -- good:
>>
>>   $ ssh gcc110.fsffrance.org
>>   Last login: Sat Jul  1 14:18:27 2017 from 70.79.163.252
>>   [kaz@gcc1-power7 ~]$
>>
>> gcc202 -- bad:
>>
>>   $ ssh gcc202.fsffrance.org
>>   k...@gcc202.fsffrance.org's password:
>>
>> The ~/.ssh paraphernalia appears not to be propagated to the machine.
>
> Works fine for me.  It doesn't accept DSA, maybe that is your problem?

It might be worth mentioning the OpenSSH folks disabled DSA by default
at 7.0. That occurred two or three years ago. Newer OSes probably have
it disabled by default, while older distros probably have it enabled
by default.

Also see http://www.openssh.com/legacy.html.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] Upgrade on gcc12, gcc13 and gcc14

[Jeffrey Walton via cfarm-users]

On Sat, Jul 22, 2017 at 4:16 AM, Aymeric via cfarm-users
 wrote:

> I plan to upgrade during the third week of August (w33) :
>
> - gcc12 and gcc13 to Debian 7 (they are running Debian 5) ;
> - gcc14 to Debian 8 (running Debian 5).

+1

> After that we won't have any Debian 5 on the farm, if you still required
> a Debian 5 please let me know asap so we can at least keep one running.

Debian 5 is very old (relatively speaking); and GCC12 and GCC13 are on
commodity hardware (x86_64).

I think its of greater benefit to the community to perform the
upgrade. Its probably easier to maintain Debian 8 since its modern,
too. Those who want Debian 5 on x86_64 can easily stand up a local VM.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

[cfarm-users] cfarm.tetaneutral.net and requests

[Jeffrey Walton via cfarm-users]

Hi Everyone,

According to Usage at https://gcc.gnu.org/wiki/CompileFarm#Usage, we
are supposed to file requests at https://cfarm.tetaneutral.net/. I
logged into the service, but I no longer see the request/issue page.

What is the process for filing requests?

(I like the new interface, and the ability to add new SSH keys, btw. Good job).

Thanks in advance,

Jeffrey Walton
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users