You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@harmony.apache.org by Alexey Varlamov <al...@gmail.com> on 2007/04/24 17:52:23 UTC
[ia64] JIT & GC regressed
Folks,
While ia64 (aka IPF) development is suspended at this moment, I think
we should strive keeping this port alive. It was functional when I
tried it last mid-week, but did not even build today.
I've fixed broken Jitrino compilation, but still could run only
interpreter with manual switch to gc_cc. So we have the following
issues:
1) Jitrino silently dies after successfully compiling a few dozens of
methods during bootstrap - need closer investigation;
2) gc_gen is not ported to IPF yet - do we have plans/resources to do
it in some near future? Otherwise we better ifdef the default GC to
older one.
3) No CC is running on IPF and regressions are caught too late. Even
simple HWA would do, and I volunteer to compose x-lists for VM and
classlib tests if someone set up a CC.
--
Alexey
Re: [ia64] JIT & GC regressed
Posted by Xiao-Feng Li <xi...@gmail.com>.
On 4/27/07, Alexey Varlamov <al...@gmail.com> wrote:
> 2007/4/27, Xiao-Feng Li <xi...@gmail.com>:
> > Alexey, after a fresh installation of a SLES 10 to IPF, I found there
> > is no gcc (or other compiler) available, although it has gdb, etc
> > tools. I searched the web, but couldn't find a prebuilt binary. Did
> > you cross-compile it in a X86 machine, or where did you get it?
>
> I've got pre-installed SLES9 with all needed developers packages in
> place (+ Intel compiler aka icc). Do you mean SLES10 distro does not
> provide gcc at all? That's impossible IMO, related packages must be
> skipped during installation?
I also think that's too strange that the CD borrowed doesn't have gcc.
Will check if SUSE website provides an image that has gcc built-in. I
don't want to do cross-compilation unless it is necessary.
Thanks,
xiaofeng
> >
> > Thanks,
> > xiaofeng
> >
> > On 4/24/07, Alexey Varlamov <al...@gmail.com> wrote:
> > > Folks,
> > > While ia64 (aka IPF) development is suspended at this moment, I think
> > > we should strive keeping this port alive. It was functional when I
> > > tried it last mid-week, but did not even build today.
> > > I've fixed broken Jitrino compilation, but still could run only
> > > interpreter with manual switch to gc_cc. So we have the following
> > > issues:
> > > 1) Jitrino silently dies after successfully compiling a few dozens of
> > > methods during bootstrap - need closer investigation;
> > > 2) gc_gen is not ported to IPF yet - do we have plans/resources to do
> > > it in some near future? Otherwise we better ifdef the default GC to
> > > older one.
> > > 3) No CC is running on IPF and regressions are caught too late. Even
> > > simple HWA would do, and I volunteer to compose x-lists for VM and
> > > classlib tests if someone set up a CC.
> > >
> > > --
> > > Alexey
> > >
> >
> >
> > --
> > http://xiao-feng.blogspot.com
> >
>
--
http://xiao-feng.blogspot.com
Re: [ia64] JIT & GC regressed
Posted by Alexey Varlamov <al...@gmail.com>.
2007/4/27, Xiao-Feng Li <xi...@gmail.com>:
> Alexey, after a fresh installation of a SLES 10 to IPF, I found there
> is no gcc (or other compiler) available, although it has gdb, etc
> tools. I searched the web, but couldn't find a prebuilt binary. Did
> you cross-compile it in a X86 machine, or where did you get it?
I've got pre-installed SLES9 with all needed developers packages in
place (+ Intel compiler aka icc). Do you mean SLES10 distro does not
provide gcc at all? That's impossible IMO, related packages must be
skipped during installation?
>
> Thanks,
> xiaofeng
>
> On 4/24/07, Alexey Varlamov <al...@gmail.com> wrote:
> > Folks,
> > While ia64 (aka IPF) development is suspended at this moment, I think
> > we should strive keeping this port alive. It was functional when I
> > tried it last mid-week, but did not even build today.
> > I've fixed broken Jitrino compilation, but still could run only
> > interpreter with manual switch to gc_cc. So we have the following
> > issues:
> > 1) Jitrino silently dies after successfully compiling a few dozens of
> > methods during bootstrap - need closer investigation;
> > 2) gc_gen is not ported to IPF yet - do we have plans/resources to do
> > it in some near future? Otherwise we better ifdef the default GC to
> > older one.
> > 3) No CC is running on IPF and regressions are caught too late. Even
> > simple HWA would do, and I volunteer to compose x-lists for VM and
> > classlib tests if someone set up a CC.
> >
> > --
> > Alexey
> >
>
>
> --
> http://xiao-feng.blogspot.com
>
Re: [ia64] JIT & GC regressed
Posted by Xiao-Feng Li <xi...@gmail.com>.
Alexey, after a fresh installation of a SLES 10 to IPF, I found there
is no gcc (or other compiler) available, although it has gdb, etc
tools. I searched the web, but couldn't find a prebuilt binary. Did
you cross-compile it in a X86 machine, or where did you get it?
Thanks,
xiaofeng
On 4/24/07, Alexey Varlamov <al...@gmail.com> wrote:
> Folks,
> While ia64 (aka IPF) development is suspended at this moment, I think
> we should strive keeping this port alive. It was functional when I
> tried it last mid-week, but did not even build today.
> I've fixed broken Jitrino compilation, but still could run only
> interpreter with manual switch to gc_cc. So we have the following
> issues:
> 1) Jitrino silently dies after successfully compiling a few dozens of
> methods during bootstrap - need closer investigation;
> 2) gc_gen is not ported to IPF yet - do we have plans/resources to do
> it in some near future? Otherwise we better ifdef the default GC to
> older one.
> 3) No CC is running on IPF and regressions are caught too late. Even
> simple HWA would do, and I volunteer to compose x-lists for VM and
> classlib tests if someone set up a CC.
>
> --
> Alexey
>
--
http://xiao-feng.blogspot.com
Re: [ia64] JIT & GC regressed
Posted by Xiao-Feng Li <xi...@gmail.com>.
On 4/25/07, Alexey Varlamov <al...@gmail.com> wrote:
> 2007/4/25, Xiao-Feng Li <xi...@gmail.com>:
> > On 4/24/07, Alexey Varlamov <al...@gmail.com> wrote:
> > > Folks,
> > > While ia64 (aka IPF) development is suspended at this moment, I think
> > > we should strive keeping this port alive. It was functional when I
> > > tried it last mid-week, but did not even build today.
> > > I've fixed broken Jitrino compilation, but still could run only
> > > interpreter with manual switch to gc_cc. So we have the following
> > > issues:
> > > 1) Jitrino silently dies after successfully compiling a few dozens of
> > > methods during bootstrap - need closer investigation;
> > > 2) gc_gen is not ported to IPF yet - do we have plans/resources to do
> > > it in some near future? Otherwise we better ifdef the default GC to
> > > older one.
> >
> > I assume it's trivial to make GCv5 work in IPF since it's running on
> > EM64T very well. I don't see real difference there from GC's POV.
> That was my assumption, as well ;)
>
> > Sorry I was just not aware of this IPF requirement before.
> > Would you let me know the environment you are using? Windows or Linux?
>
> This is not a requirement per se, as not everybody has spare Itanium
> box at hand - but if you can do this port, very nice! Currently Linux
> only is supported, my env is SLES9/gcc 3.3.3.
Ok, I got an IPF with very old Linux installed. I will have a try
then. Thanks, xiaofeng
> --
> Regards,
> Alexey
>
> >
> > Thanks,
> > xiaofeng
> >
> > > 3) No CC is running on IPF and regressions are caught too late. Even
> > > simple HWA would do, and I volunteer to compose x-lists for VM and
> > > classlib tests if someone set up a CC.
> > >
> > > --
> > > Alexey
> > >
> >
> >
> > --
> > http://xiao-feng.blogspot.com
> >
>
--
http://xiao-feng.blogspot.com
Re: [ia64] JIT & GC regressed
Posted by Alexey Varlamov <al...@gmail.com>.
2007/4/25, Xiao-Feng Li <xi...@gmail.com>:
> On 4/24/07, Alexey Varlamov <al...@gmail.com> wrote:
> > Folks,
> > While ia64 (aka IPF) development is suspended at this moment, I think
> > we should strive keeping this port alive. It was functional when I
> > tried it last mid-week, but did not even build today.
> > I've fixed broken Jitrino compilation, but still could run only
> > interpreter with manual switch to gc_cc. So we have the following
> > issues:
> > 1) Jitrino silently dies after successfully compiling a few dozens of
> > methods during bootstrap - need closer investigation;
> > 2) gc_gen is not ported to IPF yet - do we have plans/resources to do
> > it in some near future? Otherwise we better ifdef the default GC to
> > older one.
>
> I assume it's trivial to make GCv5 work in IPF since it's running on
> EM64T very well. I don't see real difference there from GC's POV.
That was my assumption, as well ;)
> Sorry I was just not aware of this IPF requirement before.
> Would you let me know the environment you are using? Windows or Linux?
This is not a requirement per se, as not everybody has spare Itanium
box at hand - but if you can do this port, very nice! Currently Linux
only is supported, my env is SLES9/gcc 3.3.3.
--
Regards,
Alexey
>
> Thanks,
> xiaofeng
>
> > 3) No CC is running on IPF and regressions are caught too late. Even
> > simple HWA would do, and I volunteer to compose x-lists for VM and
> > classlib tests if someone set up a CC.
> >
> > --
> > Alexey
> >
>
>
> --
> http://xiao-feng.blogspot.com
>
Re: [ia64] JIT & GC regressed
Posted by Xiao-Feng Li <xi...@gmail.com>.
On 4/24/07, Alexey Varlamov <al...@gmail.com> wrote:
> Folks,
> While ia64 (aka IPF) development is suspended at this moment, I think
> we should strive keeping this port alive. It was functional when I
> tried it last mid-week, but did not even build today.
> I've fixed broken Jitrino compilation, but still could run only
> interpreter with manual switch to gc_cc. So we have the following
> issues:
> 1) Jitrino silently dies after successfully compiling a few dozens of
> methods during bootstrap - need closer investigation;
> 2) gc_gen is not ported to IPF yet - do we have plans/resources to do
> it in some near future? Otherwise we better ifdef the default GC to
> older one.
I assume it's trivial to make GCv5 work in IPF since it's running on
EM64T very well. I don't see real difference there from GC's POV.
Sorry I was just not aware of this IPF requirement before.
Would you let me know the environment you are using? Windows or Linux?
Thanks,
xiaofeng
> 3) No CC is running on IPF and regressions are caught too late. Even
> simple HWA would do, and I volunteer to compose x-lists for VM and
> classlib tests if someone set up a CC.
>
> --
> Alexey
>
--
http://xiao-feng.blogspot.com
Re: [ia64] JIT & GC regressed
Posted by Xiao-Feng Li <xi...@gmail.com>.
Now I made it work by compiling only GCv5 module. The only change is
to include IPF to GCv5 build file and a few lines of asm change. Since
the SVN is frozen, I will commit it after it is open to commit again.
Thanks,
xiaofeng
On 4/29/07, Alexey Varlamov <al...@gmail.com> wrote:
> Yes, it is compiles w/o questions on gcc 3.3.3 - and apparently you
> have more modern version of gcc toolset with more restrictive
> compliance checks.
> Actually those "-Werr" are for cpp compilation, compilation of
> separate asm sources is ruled by different flags, see
> build\targets\build.native.xml. I guess removal of "--fatal-warnings"
> switch in there should help.
>
> BTW, at first glance, this particular code is dead; seems it was
> copy-pasted from interpreter for some purpose but now is unused - so
> you can also try dropping it altoghether from compilation. Well,
> you'll meet it again in interpreter - but that will be certain
> progress anyway.
> Unfortunately I'll be offline the next week and will not be able to
> assist you - in case you are going to work during China holiday ;).
>
> 2007/4/28, Xiao-Feng Li <xi...@gmail.com>:
> > Alexey, can you build Harmony now? After big efforts installing all
> > the necessary supporting software, I built classlib successfully, but
> > failed at drlvm with following assembler message:
> >
> > build.native.asm:
> > [exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:
> > Assembler messages:
> > [exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:28:
> > Warning: Missing .endp after previous .proc
> > [exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:29:
> > Warning: Missing .endp after previous .proc
> > [exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:30:
> > Warning: Missing .endp after previous .proc
> > [exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:31:
> > Warning: Missing .endp after previous .proc
> > [exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:41:
> > Warning: .save outside of prologue
> > [exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:47:
> > Warning: .save outside of prologue
> > [exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:59:
> > Warning: .save outside of prologue
> > [exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:216:
> > Warning: `invokeJNI#' was not specified with previous .proc
> > [exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:216:
> > Warning: `vm_invoke_native_array_stub_int' should be an operand to
> > this .endp
> > [exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:218:
> > Error: 9 warnings, treating warnings as errors
> >
> > I've commented the only -Werr I found, but still hit it. Have you met
> > the problem?
> >
> > Thanks,
> > xiaofeng
> >
> > On 4/24/07, Alexey Varlamov <al...@gmail.com> wrote:
> > > Folks,
> > > While ia64 (aka IPF) development is suspended at this moment, I think
> > > we should strive keeping this port alive. It was functional when I
> > > tried it last mid-week, but did not even build today.
> > > I've fixed broken Jitrino compilation, but still could run only
> > > interpreter with manual switch to gc_cc. So we have the following
> > > issues:
> > > 1) Jitrino silently dies after successfully compiling a few dozens of
> > > methods during bootstrap - need closer investigation;
> > > 2) gc_gen is not ported to IPF yet - do we have plans/resources to do
> > > it in some near future? Otherwise we better ifdef the default GC to
> > > older one.
> > > 3) No CC is running on IPF and regressions are caught too late. Even
> > > simple HWA would do, and I volunteer to compose x-lists for VM and
> > > classlib tests if someone set up a CC.
> > >
> > > --
> > > Alexey
> > >
> >
> >
> > --
> > http://xiao-feng.blogspot.com
> >
>
--
http://xiao-feng.blogspot.com
Re: [ia64] JIT & GC regressed
Posted by Alexey Varlamov <al...@gmail.com>.
Yes, it is compiles w/o questions on gcc 3.3.3 - and apparently you
have more modern version of gcc toolset with more restrictive
compliance checks.
Actually those "-Werr" are for cpp compilation, compilation of
separate asm sources is ruled by different flags, see
build\targets\build.native.xml. I guess removal of "--fatal-warnings"
switch in there should help.
BTW, at first glance, this particular code is dead; seems it was
copy-pasted from interpreter for some purpose but now is unused - so
you can also try dropping it altoghether from compilation. Well,
you'll meet it again in interpreter - but that will be certain
progress anyway.
Unfortunately I'll be offline the next week and will not be able to
assist you - in case you are going to work during China holiday ;).
2007/4/28, Xiao-Feng Li <xi...@gmail.com>:
> Alexey, can you build Harmony now? After big efforts installing all
> the necessary supporting software, I built classlib successfully, but
> failed at drlvm with following assembler message:
>
> build.native.asm:
> [exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:
> Assembler messages:
> [exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:28:
> Warning: Missing .endp after previous .proc
> [exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:29:
> Warning: Missing .endp after previous .proc
> [exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:30:
> Warning: Missing .endp after previous .proc
> [exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:31:
> Warning: Missing .endp after previous .proc
> [exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:41:
> Warning: .save outside of prologue
> [exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:47:
> Warning: .save outside of prologue
> [exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:59:
> Warning: .save outside of prologue
> [exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:216:
> Warning: `invokeJNI#' was not specified with previous .proc
> [exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:216:
> Warning: `vm_invoke_native_array_stub_int' should be an operand to
> this .endp
> [exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:218:
> Error: 9 warnings, treating warnings as errors
>
> I've commented the only -Werr I found, but still hit it. Have you met
> the problem?
>
> Thanks,
> xiaofeng
>
> On 4/24/07, Alexey Varlamov <al...@gmail.com> wrote:
> > Folks,
> > While ia64 (aka IPF) development is suspended at this moment, I think
> > we should strive keeping this port alive. It was functional when I
> > tried it last mid-week, but did not even build today.
> > I've fixed broken Jitrino compilation, but still could run only
> > interpreter with manual switch to gc_cc. So we have the following
> > issues:
> > 1) Jitrino silently dies after successfully compiling a few dozens of
> > methods during bootstrap - need closer investigation;
> > 2) gc_gen is not ported to IPF yet - do we have plans/resources to do
> > it in some near future? Otherwise we better ifdef the default GC to
> > older one.
> > 3) No CC is running on IPF and regressions are caught too late. Even
> > simple HWA would do, and I volunteer to compose x-lists for VM and
> > classlib tests if someone set up a CC.
> >
> > --
> > Alexey
> >
>
>
> --
> http://xiao-feng.blogspot.com
>
Re: [ia64] JIT & GC regressed
Posted by Xiao-Feng Li <xi...@gmail.com>.
Alexey, can you build Harmony now? After big efforts installing all
the necessary supporting software, I built classlib successfully, but
failed at drlvm with following assembler message:
build.native.asm:
[exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:
Assembler messages:
[exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:28:
Warning: Missing .endp after previous .proc
[exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:29:
Warning: Missing .endp after previous .proc
[exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:30:
Warning: Missing .endp after previous .proc
[exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:31:
Warning: Missing .endp after previous .proc
[exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:41:
Warning: .save outside of prologue
[exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:47:
Warning: .save outside of prologue
[exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:59:
Warning: .save outside of prologue
[exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:216:
Warning: `invokeJNI#' was not specified with previous .proc
[exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:216:
Warning: `vm_invoke_native_array_stub_int' should be an operand to
this .endp
[exec] /harmony-4-27/working_vm/vm/vmcore/src/util/ipf/base/invoke_native_stub_ipf.asm:218:
Error: 9 warnings, treating warnings as errors
I've commented the only -Werr I found, but still hit it. Have you met
the problem?
Thanks,
xiaofeng
On 4/24/07, Alexey Varlamov <al...@gmail.com> wrote:
> Folks,
> While ia64 (aka IPF) development is suspended at this moment, I think
> we should strive keeping this port alive. It was functional when I
> tried it last mid-week, but did not even build today.
> I've fixed broken Jitrino compilation, but still could run only
> interpreter with manual switch to gc_cc. So we have the following
> issues:
> 1) Jitrino silently dies after successfully compiling a few dozens of
> methods during bootstrap - need closer investigation;
> 2) gc_gen is not ported to IPF yet - do we have plans/resources to do
> it in some near future? Otherwise we better ifdef the default GC to
> older one.
> 3) No CC is running on IPF and regressions are caught too late. Even
> simple HWA would do, and I volunteer to compose x-lists for VM and
> classlib tests if someone set up a CC.
>
> --
> Alexey
>
--
http://xiao-feng.blogspot.com