Saturday, 31 July 2021

Re: [gccsdk] Sorry. Forgot

On Sat, Jul 31, 2021 at 02:08:38PM +0100, wra1th+gavin wrote:
> For the first time I have launched the command
> ./build-world (using Raspbian on a rpi4). It requires considerable patience,
> and faith because there is little to tell you that the build process is not
> stuck in a loop. It would be good if the web page gave a hint of how many
> hours the build-process is likely to take. It feels like inspecting the DNA
> of an ant by telescope from the moon.

In general build processes don't get stuck in loops - if they're generating
output it means they're making progress. If the build stops with an error,
that's more of a concern.

The time the build takes really depends on your hardware - if you have a
fast PC or server with many cores it's a lot quicker than if you have a
lightweight laptop or a development board like a Pi. It also depends a lot
on your storage - an SSD is likely to be a lot better than an SD card or a
HDD.

That said, it appears the build instructions on the website didn't describe
how to run a build that uses all the cores in your computer - in the case of
your Pi4 that means running one core not four. Using four cores isn't 4x as
fast (the GCCSDK build has a lot of autoconf which doesn't use all the
cores) but it's better.

I've now edited the page:
https://www.riscos.info/index.php/Using_GCCSDK#Building_the_cross-compiler
to describe how to run parallel builds, and to clarify how long it might
take.

Using that on a server with 72 threads reduces the build time from 67 to 16
minutes. Nowhere near a 72x speedup, but worth having.

Theo

_______________________________________________
GCCSDK mailing list gcc@gccsdk.riscos.info
Bugzilla: http://www.riscos.info/bugzilla/index.cgi
List Info: http://www.riscos.info/mailman/listinfo/gcc
Main Page: http://www.riscos.info/index.php/GCCSDK
Posted by at No comments:

[gccsdk] Sorry. Forgot

> An attempt was made to subscribe your address to the mailing list

Sorry. Forgot I was already subscribed. For the first time I have
launched the command
./build-world (using Raspbian on a rpi4). It requires considerable
patience, and faith because there is little to tell you that the build
process is not stuck in a loop. It would be good if the web page gave a
hint of how many hours the build-process is likely to take. It feels
like inspecting the DNA of an ant by telescope from the moon.

-- Gavin WRaith

_______________________________________________
GCCSDK mailing list gcc@gccsdk.riscos.info
Bugzilla: http://www.riscos.info/bugzilla/index.cgi
List Info: http://www.riscos.info/mailman/listinfo/gcc
Main Page: http://www.riscos.info/index.php/GCCSDK
Posted by at No comments:

Tuesday, 27 July 2021

[gccsdk] GCC 10 reply

Thanks Chris for the detailed information!
regads Stefan

Am Dienstag, 27. Juli 2021, 02:00:04 GMT+7 hat gcc-request@gccsdk.riscos.info <gcc-request@gccsdk.riscos.info> Folgendes geschrieben:


Today's Topics:

  1. Re: GCC 10 (Chris Gransden)


----------------------------------------------------------------------

Message: 1
Date: Mon, 26 Jul 2021 10:24:09 +0100
From: Chris Gransden <chrisg@care4free.net>
Subject: Re: [gccsdk] GCC 10
Content-Type: text/plain; charset=iso-8859-1

  Stefan Fr?hling <stefanfroehling@yahoo.de> wrote:
> Hello there!

> I have been told from Andrew Rawnsley that Iris is done with GCC10 and mostly ported by Lee Noa to create Iris.
> Where we can download the sources/binaries ?
> What is the status? What must be done to make it a RISC OS native compiler as we want to built an IDE around GCC and maybe support also other languages like BASIC, Python and Pascal.
> I want to find some programmers who are willing to port GCC 10 to RISC OS as now it is only available for Linux/Windows.
> (well with available I mean it is available at least for one or two or three RISC OS programmers?.)
> I appreciate any suggestions for a path forward so it can be made available for all RISC OS users soon.

> PS: Congratulation fot the good work on Iris and GCC10!

> regards Stefan
> from the Cloverleaf RISC OS project
> www.riscoscloverleaf.com


The GCCK 10.2.0 native compiler hasn't been released yet. Once you have the
autobuilder set up you can build it yourself.

mkdir gccsdk
cd gccsdk
svn co svn://svn.riscos.info/gccsdk/trunk/autobuilder/ autobuilder
svn co svn://svn.riscos.info/gccsdk/trunk/gcc4/ gcc4

mkdir build
cd build


Create a file call build-setvars containing the following contents.
(Assuming the autobuilder has been installed in /home/gccsdk.)

      GCCSDK_INSTALL_CROSSBIN=/home/gccsdk/cross/bin
      GCCSDK_INSTALL_ENV=/home/gccsdk/env
      RO_SHAREDLIBS=yes
      RO_USE_ARMEABIHF=yes
      AB_ELFBUILD=yes

rm -f /home/gccsdk/autobuilder/develop/gcc/gcc.dwarf2out.c.p
../autobuilder/build -v gcc

The packages for the native compiler will be created in
/home/gccsdk/autobuilder/autobuilder_packages/arm/Development.

Using GCC natively on RISC OS is very slow. It's OK for simple stuff.

I had to set a 32000k wimplot otherwise everything froze.

Using GCC 10.2.0 with the autobuilder successully builds e.g. Qupzilla,
RPCEmu, ffmpeg and  mplayer.




------------------------------

_______________________________________________
gcc mailing list

End of gcc Digest, Vol 158, Issue 3
***********************************
Posted by at No comments:

Monday, 26 July 2021

Re: [gccsdk] GCC 10

In article <2131397311.232135.1626596425379@mail.yahoo.com>,
Stefan Fröhling <stefanfroehling@yahoo.de> wrote:
> Hello there!

> I have been told from Andrew Rawnsley that Iris is done with GCC10 and mostly ported by Lee Noa to create Iris.
> Where we can download the sources/binaries ?
> What is the status? What must be done to make it a RISC OS native compiler as we want to built an IDE around GCC and maybe support also other languages like BASIC, Python and Pascal.
> I want to find some programmers who are willing to port GCC 10 to RISC OS as now it is only available for Linux/Windows.
> (well with available I mean it is available at least for one or two or three RISC OS programmersŒ.)
> I appreciate any suggestions for a path forward so it can be made available for all RISC OS users soon.

> PS: Congratulation fot the good work on Iris and GCC10!

> regards Stefan
> from the Cloverleaf RISC OS project
> www.riscoscloverleaf.com


The GCCK 10.2.0 native compiler hasn't been released yet. Once you have the
autobuilder set up you can build it yourself.

mkdir gccsdk
cd gccsdk
svn co svn://svn.riscos.info/gccsdk/trunk/autobuilder/ autobuilder
svn co svn://svn.riscos.info/gccsdk/trunk/gcc4/ gcc4

mkdir build
cd build


Create a file call build-setvars containing the following contents.
(Assuming the autobuilder has been installed in /home/gccsdk.)

GCCSDK_INSTALL_CROSSBIN=/home/gccsdk/cross/bin
GCCSDK_INSTALL_ENV=/home/gccsdk/env
RO_SHAREDLIBS=yes
RO_USE_ARMEABIHF=yes
AB_ELFBUILD=yes

rm -f /home/gccsdk/autobuilder/develop/gcc/gcc.dwarf2out.c.p
../autobuilder/build -v gcc

The packages for the native compiler will be created in
/home/gccsdk/autobuilder/autobuilder_packages/arm/Development.

Using GCC natively on RISC OS is very slow. It's OK for simple stuff.

I had to set a 32000k wimplot otherwise everything froze.

Using GCC 10.2.0 with the autobuilder successully builds e.g. Qupzilla,
RPCEmu, ffmpeg and mplayer.


_______________________________________________
GCCSDK mailing list gcc@gccsdk.riscos.info
Bugzilla: http://www.riscos.info/bugzilla/index.cgi
List Info: http://www.riscos.info/mailman/listinfo/gcc
Main Page: http://www.riscos.info/index.php/GCCSDK
Posted by at No comments:

Sunday, 18 July 2021

[gccsdk] GCC 10

Hello there!

I have been told from Andrew Rawnsley that Iris is done with GCC10 and mostly ported by Lee Noa to create Iris.
Where we can download the sources/binaries ?
What is the status? What must be done to make it a RISC OS native compiler as we want to built an IDE around GCC and maybe support also other languages like BASIC, Python and Pascal.
I want to find some programmers who are willing to port GCC 10 to RISC OS as now it is only available for Linux/Windows.
(well with available I mean it is available at least for one or two or three RISC OS programmers….)
I appreciate any suggestions for a path forward so it can be made available for all RISC OS users soon.

PS: Congratulation fot the good work on Iris and GCC10!

regards Stefan
from the Cloverleaf RISC OS project
www.riscoscloverleaf.com


Posted by at No comments:

Sunday, 11 July 2021

Re: [Rpcemu] Apple Silicon build of RPCEmu 0.9.3

Timothy Coltman, on 11 Jul, wrote:

> Hello all
>
> I've built a new version of RPCEmu 0.9.3 for Apple Macs. The interpreter
build is now a universal binary, which means it will run natively on both
Intel and Apple Silicon (arm64) Macs.

[snip]

Thank you for this, it is most welcome.

As tested on a M1 MacBook Air the native interpreter runs faster than the
previous 0.9.3a under Rosetta2. The recompiler looks very slow, as seen by
very slow screen redraws, but it benchmarks better than that might imply.


MIPs romark drystone
Previous 0.9.3a
interpreter/Intel 130 119000 312000

New 0.9.3b
interpreter/Native 184 195000 418000
recompiler/Intel 280 429000 914000

For comparison on an iMacPro
interpreter 80 110000 181000
recompiler 280 1393000 1001000

Thanks again.
--
David Pitt

_______________________________________________
RPCEmu mailing list
RPCEmu@riscos.info
http://www.riscos.info/cgi-bin/mailman/listinfo/rpcemu
Posted by at No comments:

[Rpcemu] Apple Silicon build of RPCEmu 0.9.3

Hello all

I've built a new version of RPCEmu 0.9.3 for Apple Macs.  The interpreter build is now a universal binary, which means it will run natively on both Intel and Apple Silicon (arm64) Macs.  It can be downloaded here:


The recompiler is not a universal binary, as arm64 is not a supported architecture for the dynamic code generation; if you run it on anything using Apple Silicon, it will still run under Rosetta 2.  I did look at the recompiler code to see how viable it was to implement an arm64 version, but unfortunately that's way beyond my capabilities.

Cheers
Tim

Posted by at No comments:

Sunday, 4 July 2021

Re: [gccsdk] Investigating failure with libxmp tests

Hi

It looks like the issue is related to the use of fdopen(fd, "w+b") for the temporary file and reading from it after it has been written to. I have been able to come up with a couple of workarounds for libxmp:
 
 - Closing the file handle and opening it again.
 - Disabling the file buffering before writing to it.

However the problem seems to be with UnixLib itself and its implementation of file buffering.

Regards
Cameron


On Mon, 7 Jun 2021 at 21:45, Cameron Cawley <ccawley2011@gmail.com> wrote:
Hi

In this case, "module" refers to a tracker music file like MOD or S3M, rather than a shared object. All of the depackers are built into the main library, apart from the rar and mo3 depackers, which call an external process using either fork/execvp or popen.

Regards
Cameron


On Mon, 7 Jun 2021 at 20:52, Lee Noar <lee.noar@sky.com> wrote:
On 07/06/2021 19:42, Cameron Cawley wrote:
> Hi
>
> I'm currently investigating why a number of tests in libxmp are failing
> when running on RISC OS. All of them are related to the depackers, which
> use an intermediate temporary file created using mkstemp() and fdopen(),
> however although depacking seems to work, the output file doesn't always
> work when reading it back in. I'm at a bit of a loss as to what's wrong,
> so I would appreciate it if someone more knowledgeable could advise.
>
> For reference: https://github.com/libxmp/libxmp/issues/369
> <https://github.com/libxmp/libxmp/issues/369>

I'm probably way off base, but after a quick look at the stdout, I see a
lot of the fails complain about failing to load a module, could that be
relevant? Is this a dl_open type plugin library that is failing to load?
If not, then what is the nature of the module that is failing to load?

Lee.

_______________________________________________
GCCSDK mailing list gcc@gccsdk.riscos.info
Bugzilla: http://www.riscos.info/bugzilla/index.cgi
List Info: http://www.riscos.info/mailman/listinfo/gcc
Main Page: http://www.riscos.info/index.php/GCCSDK
Posted by at No comments:
Subscribe to: Posts (Atom)