Re: [LinuxBIOS] selection of compression algorithm

[Lu, Yinghai]

Copy_and_run is used by CAR and included in cache_as_ram_auto.c and
compiled with gcc. That should be changed to use lzma easily.

CONFIG_COMPRESSED_ROM_STREAM is used for pay load. --- stream_init will
do the uncompressed. ( stream_init is in ram)

CONFIG_COMPRESS is for linuxbios_ram.rom --- copy_and_run will do the
uncompress. ( copy_and_run is ROM, and used ram as stack and local
variable)

You may create one file only that is corresponding to lib/nrv2b.c...

YH

-----Original Message-----
From: [e-mail]
[mailto:[e-mail]] On Behalf Of Carl-Daniel
Hailfinger
Sent: Monday, May 22, 2006 12:29 PM
To: yhlu
Cc: Stefan Reinauer; LinuxBIOS
Subject: Re: [LinuxBIOS] selection of compression algorithm
> On 5/22/06, Stefan Reinauer  wrote:
> > We never actually compile nrv2b with romcc, do we? At the point we

run
> > nrv2b ram has to be enabled anyways - we uncompress the code

somewhere.
> > So our limits are far bigger than that.

yhlu wrote:
> in crt0.s, and it is used to uncompress linuxbios_ram when using

romcc.

You mean copy_and_run() is compiled with romcc? Can we change that?

What's the difference between CONFIG_COMPRESSED_ROM_STREAM and
CONFIG_COMPRESS? Can we use lzma compression for both?

Regards,
Carl-Daniel

[LinuxBIOS] selection of compression algorithm

[Carl-Daniel Hailfinger]

Lu, Yinghai wrote:
> Copy_and_run is used by CAR and included in cache_as_ram_auto.c and
> compiled with gcc. That should be changed to use lzma easily.
>
> CONFIG_COMPRESSED_ROM_STREAM is used for pay load. --- stream_init will
> do the uncompressed. ( stream_init is in ram)
>
> CONFIG_COMPRESS is for linuxbios_ram.rom --- copy_and_run will do the
> uncompress. ( copy_and_run is ROM, and used ram as stack and local
> variable)

Patch attached to clear up the naming confusion. Applies on top of my
older patch.
Please apply.

Regards,
Carl-Daniel

[LinuxBIOS] selection of compression algorithm

[Ronald G Minnich]

Carl-Daniel Hailfinger wrote:
> Lu, Yinghai wrote:
> > Copy_and_run is used by CAR and included in cache_as_ram_auto.c and
> > compiled with gcc. That should be changed to use lzma easily.
> >
> > CONFIG_COMPRESSED_ROM_STREAM is used for pay load. --- stream_init will
> > do the uncompressed. ( stream_init is in ram)
> >
> > CONFIG_COMPRESS is for linuxbios_ram.rom --- copy_and_run will do the
> > uncompress. ( copy_and_run is ROM, and used ram as stack and local
> > variable)
>
> Patch attached to clear up the naming confusion. Applies on top of my
> older patch.
> Please apply.

this has passed the abuild regression test and works on at least one 
platform, right?

ron

[LinuxBIOS] selection of compression algorithm

[Carl-Daniel Hailfinger]

Ronald G Minnich wrote:
> Carl-Daniel Hailfinger wrote:
> > Lu, Yinghai wrote:
> > > Copy_and_run is used by CAR and included in cache_as_ram_auto.c and
> > > compiled with gcc. That should be changed to use lzma easily.
> > >
> > > CONFIG_COMPRESSED_ROM_STREAM is used for pay load. --- stream_init will
> > > do the uncompressed. ( stream_init is in ram)
> > >
> > > CONFIG_COMPRESS is for linuxbios_ram.rom --- copy_and_run will do the
> > > uncompress. ( copy_and_run is ROM, and used ram as stack and local
> > > variable)
> >
> > Patch attached to clear up the naming confusion. Applies on top of my
> > older patch.
> > Please apply.
>
> this has passed the abuild regression test and works on at least one 
> platform, right?

I have no platform on which LinuxBIOS runs on, but it caused no regression
in abuild. (But a lot of targets do not seem to compile at all, even without
my patch. Does this depend on the gcc version? I have gcc 3.3.5)

Regards,
Carl-Daniel

[LinuxBIOS] selection of compression algorithm

[Ronald G Minnich]

Carl-Daniel Hailfinger wrote:
> I have no platform on which LinuxBIOS runs on, but it caused no regression
> in abuild. (But a lot of targets do not seem to compile at all, even without
> my patch. Does this depend on the gcc version? I have gcc 3.3.5)

don't know, I will try to test your patch via abuild but I don't want to 
commit until then.

ron

[LinuxBIOS] selection of compression algorithm

[Carl-Daniel Hailfinger]

Ronald G Minnich wrote:
> Carl-Daniel Hailfinger wrote:
> > I have no platform on which LinuxBIOS runs on, but it caused no
> > regression
> > in abuild. (But a lot of targets do not seem to compile at all, even
> > without
> > my patch. Does this depend on the gcc version? I have gcc 3.3.5)
>
> don't know, I will try to test your patch via abuild but I don't want to
> commit until then.

Yes, it was a gcc bug. gcc 4.1.0 has no problems compiling it.
I'll continue testing and developing with gcc 4.1.0 from now on.

Regards,
Carl-Daniel

[LinuxBIOS] selection of compression algorithm

[Stefan Reinauer]

* Carl-Daniel Hailfinger  [060523 20:09]:
> Ronald G Minnich wrote:
> > Carl-Daniel Hailfinger wrote:
> > > I have no platform on which LinuxBIOS runs on, but it caused no
> > > regression
> > > in abuild. (But a lot of targets do not seem to compile at all, even
> > > without
> > > my patch. Does this depend on the gcc version? I have gcc 3.3.5)
> >
> > don't know, I will try to test your patch via abuild but I don't want to
> > commit until then.
>
> Yes, it was a gcc bug. gcc 4.1.0 has no problems compiling it.
> I'll continue testing and developing with gcc 4.1.0 from now on.

Can we get it compiling with 3.3.5 as well? What's the bug about?
I'm using 4.1 as a reference compiler at the moment, but lots of people
out there have older compilers I think...

Stefan

[LinuxBIOS] selection of compression algorithm

[Carl-Daniel Hailfinger]

Stefan Reinauer wrote:
> Carl-Daniel Hailfinger  [060523 20:09]:
> > Ronald G Minnich wrote:
> > > Carl-Daniel Hailfinger wrote:
> > > > I have no platform on which LinuxBIOS runs on, but it caused no
> > > > regression
> > > > in abuild. (But a lot of targets do not seem to compile at all, even
> > > > without
> > > > my patch. Does this depend on the gcc version? I have gcc 3.3.5)
> > >
> > > don't know, I will try to test your patch via abuild but I don't want to
> > > commit until then.
> >
> > Yes, it was a gcc bug. gcc 4.1.0 has no problems compiling it.
> > I'll continue testing and developing with gcc 4.1.0 from now on.
>
> Can we get it compiling with 3.3.5 as well? What's the bug about?
> I'm using 4.1 as a reference compiler at the moment, but lots of people
> out there have older compilers I think...

LinuxBIOSv2-2310/src/cpu/amd/model_gx2/cpubug.c:199:75: missing terminating ' character
LinuxBIOSv2-2310/src/cpu/amd/model_gx2/cpubug.c:205:70: missing terminating ' character

I can retest once that is fixed.

Regards,
Carl-Daniel

[LinuxBIOS] selection of compression algorithm

[Stefan Reinauer]

* Carl-Daniel Hailfinger  [060524 02:06]:
> Stefan Reinauer wrote:
> > Carl-Daniel Hailfinger  [060523 20:09]:
> > > Ronald G Minnich wrote:
> > > > Carl-Daniel Hailfinger wrote:
> > > > > I have no platform on which LinuxBIOS runs on, but it caused no
> > > > > regression
> > > > > in abuild. (But a lot of targets do not seem to compile at all, even
> > > > > without
> > > > > my patch. Does this depend on the gcc version? I have gcc 3.3.5)
> > > >
> > > > don't know, I will try to test your patch via abuild but I don't want to
> > > > commit until then.
> > >
> > > Yes, it was a gcc bug. gcc 4.1.0 has no problems compiling it.
> > > I'll continue testing and developing with gcc 4.1.0 from now on.
> >
> > Can we get it compiling with 3.3.5 as well? What's the bug about?
> > I'm using 4.1 as a reference compiler at the moment, but lots of people
> > out there have older compilers I think...
>
> LinuxBIOSv2-2310/src/cpu/amd/model_gx2/cpubug.c:199:75: missing terminating ' character
> LinuxBIOSv2-2310/src/cpu/amd/model_gx2/cpubug.c:205:70: missing terminating ' character

Indeed weird! This is in the middle of #ifdef 0?

[LinuxBIOS] selection of compression algorithm

[Carl-Daniel Hailfinger]

Stefan Reinauer wrote:
> Carl-Daniel Hailfinger  [060524 02:06]:
> > Stefan Reinauer wrote:
> > > Carl-Daniel Hailfinger  [060523 20:09]:
> > > > Ronald G Minnich wrote:
> > > > > Carl-Daniel Hailfinger wrote:
> > > > > > I have no platform on which LinuxBIOS runs on, but it caused no
> > > > > > regression
> > > > > > in abuild. (But a lot of targets do not seem to compile at all, even
> > > > > > without
> > > > > > my patch. Does this depend on the gcc version? I have gcc 3.3.5)
> > > > >
> > > > > don't know, I will try to test your patch via abuild but I don't want to
> > > > > commit until then.
> > > >
> > > > Yes, it was a gcc bug. gcc 4.1.0 has no problems compiling it.
> > > > I'll continue testing and developing with gcc 4.1.0 from now on.
> > >
> > > Can we get it compiling with 3.3.5 as well? What's the bug about?
> > > I'm using 4.1 as a reference compiler at the moment, but lots of people
> > > out there have older compilers I think...
> >
> > LinuxBIOSv2-2310/src/cpu/amd/model_gx2/cpubug.c:199:75: missing terminating ' character
> > LinuxBIOSv2-2310/src/cpu/amd/model_gx2/cpubug.c:205:70: missing terminating ' character
>
> Indeed weird! This is in the middle of #ifdef 0?

Yes.

Tom Sylla @ AMD just wrote off-list:
> You can safely just remove those ticks. They are in comments in an
> if 0'ed section of reference code.

Regards,
Carl-Daniel

[LinuxBIOS] selection of compression algorithm

[Per Hallsmark]

well, perhaps not so weird... the #ifdef 0 is weird. Shouldn't it be #if 
0 instead to comment out stuff? :-)

Stefan Reinauer wrote:
> Carl-Daniel Hailfinger  [060524 02:06]:
> > Stefan Reinauer wrote:
> > > Carl-Daniel Hailfinger  [060523 20:09]:
> > > > Ronald G Minnich wrote:
> > > > > Carl-Daniel Hailfinger wrote:
> > > > > > I have no platform on which LinuxBIOS runs on, but it caused no
> > > > > > regression
> > > > > > in abuild. (But a lot of targets do not seem to compile at all, even
> > > > > > without
> > > > > > my patch. Does this depend on the gcc version? I have gcc 3.3.5)
> > > > >
> > > > > don't know, I will try to test your patch via abuild but I don't want to
> > > > > commit until then.
> > > >
> > > > Yes, it was a gcc bug. gcc 4.1.0 has no problems compiling it.
> > > > I'll continue testing and developing with gcc 4.1.0 from now on.
> > >
> > > Can we get it compiling with 3.3.5 as well? What's the bug about?
> > > I'm using 4.1 as a reference compiler at the moment, but lots of people
> > > out there have older compilers I think...
> >
> > LinuxBIOSv2-2310/src/cpu/amd/model_gx2/cpubug.c:199:75: missing terminating ' character
> > LinuxBIOSv2-2310/src/cpu/amd/model_gx2/cpubug.c:205:70: missing terminating ' character
>
> Indeed weird! This is in the middle of #ifdef 0?

[LinuxBIOS] selection of compression algorithm

[Stefan Reinauer]

* Per Hallsmark  [060524 16:53]:
> well, perhaps not so weird... the #ifdef 0 is weird. Shouldn't it be #if 0
> instead to comment out stuff? :-)

heh.. No doubt. But the code says #if 0 indeed. It was just me who
messed the mail up.

Stefan

[LinuxBIOS] selection of compression algorithm

[Stefan Reinauer]

* Carl-Daniel Hailfinger  [060523 19:49]:
> Ronald G Minnich wrote:
> > Carl-Daniel Hailfinger wrote:
> > > Lu, Yinghai wrote:
> > > > Copy_and_run is used by CAR and included in cache_as_ram_auto.c and
> > > > compiled with gcc. That should be changed to use lzma easily.
> > > >
> > > > CONFIG_COMPRESSED_ROM_STREAM is used for pay load. --- stream_init will
> > > > do the uncompressed. ( stream_init is in ram)
> > > >
> > > > CONFIG_COMPRESS is for linuxbios_ram.rom --- copy_and_run will do the
> > > > uncompress. ( copy_and_run is ROM, and used ram as stack and local
> > > > variable)
> > >
> > > Patch attached to clear up the naming confusion. Applies on top of my
> > > older patch.
> > > Please apply.
> >
> > this has passed the abuild regression test and works on at least one 
> > platform, right?
>
> I have no platform on which LinuxBIOS runs on, but it caused no regression
> in abuild. (But a lot of targets do not seem to compile at all, even without
> my patch. Does this depend on the gcc version? I have gcc 3.3.5)

Is it different from this one?

http://snapshots.linuxbios.org/stats/abuild-LinuxBIOSv2-2310.log

This one seems to occur a couple of times, who's the culprit?
  crt0.S:(.rom.text+0x350b): undefined reference to `printk_debug'

Stefan