[wwwdocs] gcc-12: Nvptx updates.

  [ was: Re: [wwwdocs][patch] gcc-12/changes.html: Document -misa update 
for nvptx ]

On 3/3/22 13:27, Tobias Burnus wrote:
> The current wording, https://gcc.gnu.org/gcc-12/changes.html#nvptx ,
> is outdated and (now wrongly) encourages to use -mptx=.
> 
> Updated as follows.

I've taken these changes as a base, revised and added some more items.

Any comments?

Also, feel free to instead comment on the full-text version below 
(copied from firefox after opening the page), that might be more readable.

Thanks,
- Tom

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

NVPTX

     The -march flag has been added. The -misa flag is now considered an 
alias of the -march flag.

     Support for PTX ISA target architectures sm_53, sm_70, sm_75 and 
sm_80 has been added. These can be specified using the -march flag.

     The default PTX ISA target architecture has been set back to sm_30, 
to fix support for sm_30 boards.

     The -march-map flag has been added. The -march-map value will be 
mapped to an valid -march flag value. For instance, -march-map=sm_50 
maps to -march=sm_35. This can be used to specify that generated code is 
to be executed on a board with at least some specific compute 
capability, without having to know the valid values for the -march flag.

     The -mptx flag has been added to specify the PTX ISA version for 
the generated code; permitted values are 3.1 (matches previous GCC 
versions), 6.0, 6.3, and 7.0. If not specified, the used version is the 
minimal version required for -march but at least 6.0.

     An mptx-3.1 multilib was added. This allows using older drivers 
which do not support PTX ISA version 6.0.

     The new __PTX_SM__ predefined macro allows code to check the PTX 
ISA target architecture being targeted by the compiler.

     The new __PTX_ISA_VERSION_MAJOR__ and __PTX_ISA_VERSION_MINOR__ 
predefined macros allows code to check the PTX ISA version being 
targeted by the compiler.

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

Hi!

Still catching up with GCC/nvptx back end changes...  %-)


In the following I'm not discussing the patch to document
"gcc-12: Nvptx updates", but rather one aspect of the
"gcc-12: Nvptx updates" themselves.  ;-)

On 2022-03-30T14:27:41+0200, Tom de Vries <tdevries@suse.de> wrote:
> +  <li>The <code>-march</code> flag has been added.  The <code>-misa</code>
> +    flag is now considered an alias of the <code>-march</code> flag.</li>
> +  <li>Support for PTX ISA target architectures <code>sm_53</code>,
> +    <code>sm_70</code>, <code>sm_75</code> and <code>sm_80</code> has been
> +    added.  These can be specified using the <code>-march</code> flag.</li>
> +  <li>The default PTX ISA target architecture has been set back
> +    to <code>sm_30</code>, to fix support for <code>sm_30</code> boards.</li>
> +  <li>The <code>-march-map</code> flag has been added.  The
> +    <code>-march-map</code> value will be mapped to an valid
> +    <code>-march</code> flag value.  For instance,
> +    <code>-march-map=sm_50</code> maps to <code>-march=sm_35</code>.
> +    This can be used to specify that generated code is to be executed on a
> +    board with at least some specific compute capability, without having to
> +    know the valid values for the <code>-march</code> flag.</li>

Regarding the following:

>    <li>The <code>-mptx</code> flag has been added to specify the PTX ISA version
>        for the generated code; permitted values are <code>3.1</code>
> -      (default, matches previous GCC versions) and <code>6.3</code>.
> +      (matches previous GCC versions), <code>6.0</code>, <code>6.3</code>,
> +      and <code>7.0</code>. If not specified, the used version is the minimal
> +      version required for <code>-march</code> but at least <code>6.0</code>.
>    </li>

For "the PTX ISA version [used is] at least '6.0'", per
<https://docs.nvidia.com/cuda/parallel-thread-execution/#release-notes>,
this means we now require "CUDA 9.0, driver r384" (or more recent).
Per <https://developer.nvidia.com/cuda-toolkit-archive>:
"CUDA Toolkit 9.0 (Sept 2017)", so ~4.5 years old.
Per <https://download.nvidia.com/XFree86/Linux-x86_64/>, I'm guessing a
similar timeframe for the imprecise "r384" Driver version stated in that
table.  That should all be fine (re not mandating use of all-too-recent
versions).

Now, consider doing a GCC/nvptx offloading build with
'--with-cuda-driver' pointing to CUDA 9.0 (or more recent).  This means
that the libgomp nvptx plugin may now use CUDA Driver features of the
CUDA 9.0 distribution ("driver r384", etc.) -- because that's what it is
being 'configure'd and linked against.  (I say "may now use", because
we're currently not making a lot of effort to use "modern" CUDA Driver
features -- but we could, and probably should.  That's a separate
discussion, of course.)  It then follows that the libgomp nvptx plugin
has a hard dependency on CUDA Driver features of the CUDA 9.0
distribution ("driver r384", etc.).  That's dependency as in ABI: via
'*.so' symbol versions as well as internal CUDA interface configuration;
see <cuda.h> doing different '#define's for different
'__CUDA_API_VERSION' etc.)

Now assume one such dependency on "modern" CUDA Driver were not
implemented by:

> +  <li>An <code>mptx-3.1</code> multilib was added.  This allows using older
> +      drivers which do not support PTX ISA version 6.0.</li>

... this "old" CUDA Driver.  Then you do have the '-mptx-3.1' multilib to
use with "old" CUDA Driver -- but you cannot actually use the libgomp
nvptx plugin, because that's been built against "modern" CUDA Driver.

Same problem, generally, for 'nvptx-run' of the nvptx-tools, which has
similar CUDA Driver dependencies.

Now, that may currently be a latent problem only, because we're not
actually making use of "modern" CUDA Driver features.  But, I'd like to
resolve this "impedance mismatch", before we actually run into such
problems.

Already long ago Jakub put in changes to use '--without-cuda-driver' to
"Allow building GCC with PTX offloading even without CUDA being installed
(gcc and nvptx-tools patches)": "Especially for distributions it is
undesirable to need to have proprietary CUDA libraries and headers
installed when building GCC.", and I understand GNU/Linux distributions
all use that.  That configuration uses the GCC-provided
'libgomp/plugin/cuda/cuda.h', 'libgomp/plugin/cuda-lib.def' to manually
define the CUDA Driver ABI to use, and then 'dlopen("libcuda.so.1")'.
(Similar to what the libgomp GCN (and before: HSA) plugin is doing, for
example.)  Quite likely that our group (at work) are the only ones to
actually use '--with-cuda-driver'?

My proposal now is: we remove '--with-cuda-driver' (make its use a no-op,
per standard GNU Autoconf behavior), and offer '--without-cuda-driver'
only.  This shouldn't cause any user-visible change in behavior, so safe
without a prior deprecation phase.

Before I prepare the patches (GCC, nvptx-tools): any comments or
objections?


Grüße
 Thomas


>    <li>The new <code>__PTX_SM__</code> predefined macro allows code to check the
> -      compute model being targeted by the compiler.</li>
> +      PTX ISA target architecture being targeted by the compiler.</li>
> +  <li>The new <code>__PTX_ISA_VERSION_MAJOR__</code>
> +      and <code>__PTX_ISA_VERSION_MINOR__</code> predefined macros allows code
> +      to check the PTX ISA version being targeted by the compiler.</li>
>  </ul>
-----------------
Siemens Electronic Design Automation GmbH; Anschrift: Arnulfstraße 201, 80634 München; Gesellschaft mit beschränkter Haftung; Geschäftsführer: Thomas Heurung, Frank Thürauf; Sitz der Gesellschaft: München; Registergericht München, HRB 106955
  

Hi Thomas,
I apologise that it might complicate things, but one potential benefit of --with-cuda-driver
(i.e. linking the compiler against proprietary libraries) is that it would allow support for
-march=native on nvptx (i.e. the gcc driver can figure out what sm_xx is available on the
GPU(s) of the current machine, and pass that to cc1.  Like with all the microarchitectures
on other platforms (x86_64), figuring this out is not a trivial task for many end-users.

Of course, ideally I'd love to be able to figure out the PTX hardware specifications and
driver versions without using a third-party library, but I've no idea how this could be done
(portably across the platforms that support libcuda).  Perhaps dlopen at runtime?
Or calling out to (executing) nvptx-tools?

Cheers,
Roger
--

> -----Original Message-----
> From: Thomas Schwinge <thomas@codesourcery.com>
> Sent: 05 April 2022 16:14
> To: Tom de Vries <tdevries@suse.de>; Jakub Jelinek <jakub@redhat.com>
> Cc: gcc-patches@gcc.gnu.org; Tobias Burnus <tobias@codesourcery.com>;
> Roger Sayle <roger@nextmovesoftware.com>
> Subject: Proposal to remove '--with-cuda-driver' (was: [wwwdocs][patch] gcc-
> 12: Nvptx updates)
> 
> Hi!
> 
> Still catching up with GCC/nvptx back end changes...  %-)
> 
> 
> In the following I'm not discussing the patch to document
> "gcc-12: Nvptx updates", but rather one aspect of the
> "gcc-12: Nvptx updates" themselves.  ;-)
> 
> On 2022-03-30T14:27:41+0200, Tom de Vries <tdevries@suse.de> wrote:
> > +  <li>The <code>-march</code> flag has been added.  The <code>-
> misa</code>
> > +    flag is now considered an alias of the <code>-march</code>
> > + flag.</li>  <li>Support for PTX ISA target architectures <code>sm_53</code>,
> > +    <code>sm_70</code>, <code>sm_75</code> and <code>sm_80</code>
> has been
> > +    added.  These can be specified using the <code>-march</code>
> > + flag.</li>  <li>The default PTX ISA target architecture has been set back
> > +    to <code>sm_30</code>, to fix support for <code>sm_30</code>
> > + boards.</li>  <li>The <code>-march-map</code> flag has been added.  The
> > +    <code>-march-map</code> value will be mapped to an valid
> > +    <code>-march</code> flag value.  For instance,
> > +    <code>-march-map=sm_50</code> maps to <code>-
> march=sm_35</code>.
> > +    This can be used to specify that generated code is to be executed on a
> > +    board with at least some specific compute capability, without having to
> > +    know the valid values for the <code>-march</code> flag.</li>
> 
> Regarding the following:
> 
> >    <li>The <code>-mptx</code> flag has been added to specify the PTX ISA
> version
> >        for the generated code; permitted values are <code>3.1</code>
> > -      (default, matches previous GCC versions) and <code>6.3</code>.
> > +      (matches previous GCC versions), <code>6.0</code>, <code>6.3</code>,
> > +      and <code>7.0</code>. If not specified, the used version is the minimal
> > +      version required for <code>-march</code> but at least
> <code>6.0</code>.
> >    </li>
> 
> For "the PTX ISA version [used is] at least '6.0'", per
> <https://docs.nvidia.com/cuda/parallel-thread-execution/#release-notes>,
> this means we now require "CUDA 9.0, driver r384" (or more recent).
> Per <https://developer.nvidia.com/cuda-toolkit-archive>:
> "CUDA Toolkit 9.0 (Sept 2017)", so ~4.5 years old.
> Per <https://download.nvidia.com/XFree86/Linux-x86_64/>, I'm guessing a
> similar timeframe for the imprecise "r384" Driver version stated in that table.
> That should all be fine (re not mandating use of all-too-recent versions).
> 
> Now, consider doing a GCC/nvptx offloading build with '--with-cuda-driver'
> pointing to CUDA 9.0 (or more recent).  This means that the libgomp nvptx
> plugin may now use CUDA Driver features of the CUDA 9.0 distribution ("driver
> r384", etc.) -- because that's what it is being 'configure'd and linked against.  (I
> say "may now use", because we're currently not making a lot of effort to use
> "modern" CUDA Driver features -- but we could, and probably should.  That's a
> separate discussion, of course.)  It then follows that the libgomp nvptx plugin
> has a hard dependency on CUDA Driver features of the CUDA 9.0 distribution
> ("driver r384", etc.).  That's dependency as in ABI: via '*.so' symbol versions as
> well as internal CUDA interface configuration; see <cuda.h> doing different
> '#define's for different '__CUDA_API_VERSION' etc.)
> 
> Now assume one such dependency on "modern" CUDA Driver were not
> implemented by:
> 
> > +  <li>An <code>mptx-3.1</code> multilib was added.  This allows using older
> > +      drivers which do not support PTX ISA version 6.0.</li>
> 
> ... this "old" CUDA Driver.  Then you do have the '-mptx-3.1' multilib to use with
> "old" CUDA Driver -- but you cannot actually use the libgomp nvptx plugin,
> because that's been built against "modern" CUDA Driver.
> 
> Same problem, generally, for 'nvptx-run' of the nvptx-tools, which has similar
> CUDA Driver dependencies.
> 
> Now, that may currently be a latent problem only, because we're not actually
> making use of "modern" CUDA Driver features.  But, I'd like to resolve this
> "impedance mismatch", before we actually run into such problems.
> 
> Already long ago Jakub put in changes to use '--without-cuda-driver' to "Allow
> building GCC with PTX offloading even without CUDA being installed (gcc and
> nvptx-tools patches)": "Especially for distributions it is undesirable to need to
> have proprietary CUDA libraries and headers installed when building GCC.", and I
> understand GNU/Linux distributions all use that.  That configuration uses the
> GCC-provided 'libgomp/plugin/cuda/cuda.h', 'libgomp/plugin/cuda-lib.def' to
> manually define the CUDA Driver ABI to use, and then 'dlopen("libcuda.so.1")'.
> (Similar to what the libgomp GCN (and before: HSA) plugin is doing, for
> example.)  Quite likely that our group (at work) are the only ones to actually use
> '--with-cuda-driver'?
> 
> My proposal now is: we remove '--with-cuda-driver' (make its use a no-op, per
> standard GNU Autoconf behavior), and offer '--without-cuda-driver'
> only.  This shouldn't cause any user-visible change in behavior, so safe without a
> prior deprecation phase.
> 
> Before I prepare the patches (GCC, nvptx-tools): any comments or objections?
> 
> 
> Grüße
>  Thomas
> 
> 
> >    <li>The new <code>__PTX_SM__</code> predefined macro allows code to
> check the
> > -      compute model being targeted by the compiler.</li>
> > +      PTX ISA target architecture being targeted by the
> > + compiler.</li>  <li>The new <code>__PTX_ISA_VERSION_MAJOR__</code>
> > +      and <code>__PTX_ISA_VERSION_MINOR__</code> predefined macros
> allows code
> > +      to check the PTX ISA version being targeted by the
> > + compiler.</li>
> >  </ul>
> -----------------
> Siemens Electronic Design Automation GmbH; Anschrift: Arnulfstraße 201,
> 80634 München; Gesellschaft mit beschränkter Haftung; Geschäftsführer:
> Thomas Heurung, Frank Thürauf; Sitz der Gesellschaft: München; Registergericht
> München, HRB 106955
  

On 4/5/22 17:14, Thomas Schwinge wrote:
> Hi!
> 
> Still catching up with GCC/nvptx back end changes...  %-)
> 
> 
> In the following I'm not discussing the patch to document
> "gcc-12: Nvptx updates", but rather one aspect of the
> "gcc-12: Nvptx updates" themselves.  ;-)
> 
> On 2022-03-30T14:27:41+0200, Tom de Vries <tdevries@suse.de> wrote:
>> +  <li>The <code>-march</code> flag has been added.  The <code>-misa</code>
>> +    flag is now considered an alias of the <code>-march</code> flag.</li>
>> +  <li>Support for PTX ISA target architectures <code>sm_53</code>,
>> +    <code>sm_70</code>, <code>sm_75</code> and <code>sm_80</code> has been
>> +    added.  These can be specified using the <code>-march</code> flag.</li>
>> +  <li>The default PTX ISA target architecture has been set back
>> +    to <code>sm_30</code>, to fix support for <code>sm_30</code> boards.</li>
>> +  <li>The <code>-march-map</code> flag has been added.  The
>> +    <code>-march-map</code> value will be mapped to an valid
>> +    <code>-march</code> flag value.  For instance,
>> +    <code>-march-map=sm_50</code> maps to <code>-march=sm_35</code>.
>> +    This can be used to specify that generated code is to be executed on a
>> +    board with at least some specific compute capability, without having to
>> +    know the valid values for the <code>-march</code> flag.</li>
> 
> Regarding the following:
> 
>>     <li>The <code>-mptx</code> flag has been added to specify the PTX ISA version
>>         for the generated code; permitted values are <code>3.1</code>
>> -      (default, matches previous GCC versions) and <code>6.3</code>.
>> +      (matches previous GCC versions), <code>6.0</code>, <code>6.3</code>,
>> +      and <code>7.0</code>. If not specified, the used version is the minimal
>> +      version required for <code>-march</code> but at least <code>6.0</code>.
>>     </li>
> 
> For "the PTX ISA version [used is] at least '6.0'", per
> <https://docs.nvidia.com/cuda/parallel-thread-execution/#release-notes>,
> this means we now require "CUDA 9.0, driver r384" (or more recent).

Well, that would be the case if there was no -mptx=3.1.

> Per <https://developer.nvidia.com/cuda-toolkit-archive>:
> "CUDA Toolkit 9.0 (Sept 2017)", so ~4.5 years old.
> Per <https://download.nvidia.com/XFree86/Linux-x86_64/>, I'm guessing a

I just see a list with version numbers there, I'm not sure what 
information you're referring to.

> similar timeframe for the imprecise "r384" Driver version stated in that
> table.  That should all be fine (re not mandating use of all-too-recent
> versions).
> 

I don't know what an imprecise driver is.

> Now, consider doing a GCC/nvptx offloading build with
> '--with-cuda-driver' pointing to CUDA 9.0 (or more recent).  This means
> that the libgomp nvptx plugin may now use CUDA Driver features of the
> CUDA 9.0 distribution ("driver r384", etc.) -- because that's what it is
> being 'configure'd and linked against.  (I say "may now use", because
> we're currently not making a lot of effort to use "modern" CUDA Driver
> features -- but we could, and probably should.  That's a separate
> discussion, of course.)  It then follows that the libgomp nvptx plugin
> has a hard dependency on CUDA Driver features of the CUDA 9.0
> distribution ("driver r384", etc.).  That's dependency as in ABI: via
> '*.so' symbol versions as well as internal CUDA interface configuration;
> see <cuda.h> doing different '#define's for different
> '__CUDA_API_VERSION' etc.)
> 
> Now assume one such dependency on "modern" CUDA Driver were not
> implemented by:
> 

Thanks for reminding me, I forgot about this configure option.

>> +  <li>An <code>mptx-3.1</code> multilib was added.  This allows using older
>> +      drivers which do not support PTX ISA version 6.0.</li>
> 
> ... this "old" CUDA Driver.  Then you do have the '-mptx-3.1' multilib to
> use with "old" CUDA Driver -- but you cannot actually use the libgomp
> nvptx plugin, because that's been built against "modern" CUDA Driver.
> 

I remember the following problem: using -with-cuda-driver to specify 
what cuda driver interface (version) you want to link the libgomp plugin 
against, and then using an older driver in combination with that libgomp 
plugin.   We may run into trouble, typically at libgomp plugin load 
time, with an error mentioning an unresolved symbol or some abi symbol 
version being not sufficient.

So, do I understand it correctly that your point is that using -mptx=3.1 
doesn't fix that problem?

> Same problem, generally, for 'nvptx-run' of the nvptx-tools, which has
> similar CUDA Driver dependencies.
> 
> Now, that may currently be a latent problem only, because we're not
> actually making use of "modern" CUDA Driver features.  But, I'd like to
> resolve this "impedance mismatch", before we actually run into such
> problems.
> 

It would be helpful for me if you would come up with an example of a 
modification to the libgomp plugin that would cause trouble in 
combination with mptx=3.1.

> Already long ago Jakub put in changes to use '--without-cuda-driver' to
> "Allow building GCC with PTX offloading even without CUDA being installed
> (gcc and nvptx-tools patches)": "Especially for distributions it is
> undesirable to need to have proprietary CUDA libraries and headers
> installed when building GCC.", and I understand GNU/Linux distributions
> all use that.  That configuration uses the GCC-provided
> 'libgomp/plugin/cuda/cuda.h', 'libgomp/plugin/cuda-lib.def' to manually
> define the CUDA Driver ABI to use, and then 'dlopen("libcuda.so.1")'.
> (Similar to what the libgomp GCN (and before: HSA) plugin is doing, for
> example.)  Quite likely that our group (at work) are the only ones to
> actually use '--with-cuda-driver'?
> 

Right, I see in my scripts that I don't use --with-cuda-driver, possibly 
because of years-ago running into issues when changing drivers forth and 
back.

> My proposal now is: we remove '--with-cuda-driver' (make its use a no-op,
> per standard GNU Autoconf behavior), and offer '--without-cuda-driver'
> only.  This shouldn't cause any user-visible change in behavior, so safe
> without a prior deprecation phase.
> 

I think the dlopen use-case is the most flexible, and I don't see any 
user benefit from using --with-cuda-driver, so I don't see a problem 
with removing --with-cuda-driver for the user.

I did wonder about keeping it available in some form, say rename to 
--maintainer-mode-with-cuda-driver.  This could be useful for debugging 
/ comparison purposes.  But it would mean having to test it when making 
relevant changes, which is maintenance burden for a feature not visible 
to the user, so I guess that's not worth it.

So, I'm fine with removing.

Thanks,
- Tom
  

Hi Tom!

On 2022-04-06T11:57:57+0200, Tom de Vries <tdevries@suse.de> wrote:
> On 4/5/22 17:14, Thomas Schwinge wrote:
>> Regarding the following:
>>
>> On 2022-03-30T14:27:41+0200, Tom de Vries <tdevries@suse.de> wrote:
>>>     <li>The <code>-mptx</code> flag has been added to specify the PTX ISA version
>>>         for the generated code; permitted values are <code>3.1</code>
>>> -      (default, matches previous GCC versions) and <code>6.3</code>.
>>> +      (matches previous GCC versions), <code>6.0</code>, <code>6.3</code>,
>>> +      and <code>7.0</code>. If not specified, the used version is the minimal
>>> +      version required for <code>-march</code> but at least <code>6.0</code>.
>>>     </li>
>>
>> For "the PTX ISA version [used is] at least '6.0'", per
>> <https://docs.nvidia.com/cuda/parallel-thread-execution/#release-notes>,
>> this means we now require "CUDA 9.0, driver r384" (or more recent).
>
> Well, that would be the case if there was no -mptx=3.1.

When considering *using* GCC/nvptx, the '-mptx-3.1' multilib may be used
with "old" hardware/CUDA Driver versions, correct.  When considering
*building* GCC/nvptx, we do require CUDA 9.0, as otherwise the default
multilib can't be built (unless you disable 'ptxas' verification).

>> Per <https://developer.nvidia.com/cuda-toolkit-archive>:
>> "CUDA Toolkit 9.0 (Sept 2017)", so ~4.5 years old.
>> Per <https://download.nvidia.com/XFree86/Linux-x86_64/>, I'm guessing a
>
> I just see a list with version numbers there, I'm not sure what
> information you're referring to.

I'd assumed that from that URL as well as structure of these version
numbers, you could tell these are Nvidia Driver releases (including
bundled CUDA Driver libraries).

>> similar timeframe for the imprecise "r384" Driver version stated in that
>> table.  That should all be fine (re not mandating use of all-too-recent
>> versions).
>
> I don't know what an imprecise driver is.

Not "imprecise driver" but "imprecise [...] version".  For example,
<https://docs.nvidia.com/cuda/parallel-thread-execution/#release-notes>
talks about "driver r384", but such a version doesn't exist; it's rather
384.130, or 384.111, or 384.98, etc.


>> Now, consider doing a GCC/nvptx offloading build with
>> '--with-cuda-driver' pointing to CUDA 9.0 (or more recent).  This means
>> that the libgomp nvptx plugin may now use CUDA Driver features of the
>> CUDA 9.0 distribution ("driver r384", etc.) -- because that's what it is
>> being 'configure'd and linked against.  (I say "may now use", because
>> we're currently not making a lot of effort to use "modern" CUDA Driver
>> features -- but we could, and probably should.  That's a separate
>> discussion, of course.)  It then follows that the libgomp nvptx plugin
>> has a hard dependency on CUDA Driver features of the CUDA 9.0
>> distribution ("driver r384", etc.).  That's dependency as in ABI: via
>> '*.so' symbol versions as well as internal CUDA interface configuration;
>> see <cuda.h> doing different '#define's for different
>> '__CUDA_API_VERSION' etc.)
>>
>> Now assume one such dependency on "modern" CUDA Driver were not
>> implemented by:
>
> Thanks for reminding me, I forgot about this configure option.

OK, good.  ;-)

>>> +  <li>An <code>mptx-3.1</code> multilib was added.  This allows using older
>>> +      drivers which do not support PTX ISA version 6.0.</li>
>>
>> ... this "old" CUDA Driver.  Then you do have the '-mptx-3.1' multilib to
>> use with "old" CUDA Driver -- but you cannot actually use the libgomp
>> nvptx plugin, because that's been built against "modern" CUDA Driver.
>
> I remember the following problem: using -with-cuda-driver to specify
> what cuda driver interface (version) you want to link the libgomp plugin
> against, and then using an older driver in combination with that libgomp
> plugin.   We may run into trouble, typically at libgomp plugin load
> time, with an error mentioning an unresolved symbol or some abi symbol
> version being not sufficient.

Right.

> So, do I understand it correctly that your point is that using -mptx=3.1
> doesn't fix that problem?

Right.

>> Same problem, generally, for 'nvptx-run' of the nvptx-tools, which has
>> similar CUDA Driver dependencies.
>>
>> Now, that may currently be a latent problem only, because we're not
>> actually making use of "modern" CUDA Driver features.  But, I'd like to
>> resolve this "impedance mismatch", before we actually run into such
>> problems.
>
> It would be helpful for me if you would come up with an example of a
> modification to the libgomp plugin that would cause trouble in
> combination with mptx=3.1.

For example, something like the following scenario -- made up, so details
may be wrong.

Consider GCC's libgomp nvptx plugin is built with '--with-cuda-driver'
pointing to a modern CUDA release.  The 'configure' script finds the
'cuMemPrefetchAsync' function available (CUDA 8.0+?), enables respective
hypothetical code in the libgomp nvptx plugin, and thus
'libgomp-plugin-nvptx.so' now has a load-time dependency on 'libcuda.so'
providing 'cuMemPrefetchAsync': the plugin will fail to load if that's
not available.  If you now use this plugin on an "old" system (old CUDA
Driver version), the '-mptx-3.1' multilib that is meant to keep tings
working for such "old" configurations, doesn't help, because the plugin
won't load.  (Orthogonal aspects.)

..., and I've since discovered that we use 'libgomp/plugin/cuda-lib.def'
not just for 'PLUGIN_NVPTX_DYNAMIC' (that is, '--without-cuda-driver')
configurations, but also for '!PLUGIN_NVPTX_DYNAMIC' (that is,
'--with-cuda-driver') configurations.  So, instead of 'configure'-time
detection of 'cuMemPrefetchAsync' availability and using
'CUDA_CALL (cuMemPrefetchAsync)', 'libgomp/plugin/cuda-lib.def' should
specify 'CUDA_ONE_CALL_MAYBE_NULL (cuMemPrefetchAsync)', which for
'!PLUGIN_NVPTX_DYNAMIC' configurations would make sure to provide a
'weak' prototype for 'cuMemPrefetchAsync', and then ought to check
'CUDA_CALL_EXISTS (cuMemPrefetchAsync)' before doing
'CUDA_CALL (cuMemPrefetchAsync)'.

So: all is good, my scenario can be made work correctly even for the
'--with-cuda-driver' case: at run-time we see whether
'cuMemPrefetchAsync' is available, and only call if it is -- no matter
how the CUDA Driver library gets linked ('-lcuda') or loaded ('dlopen').


Anyway, despite that, we seem to agree that '--with-cuda-driver' is not
very useful, and may be removed:

>> Already long ago Jakub put in changes to use '--without-cuda-driver' to
>> "Allow building GCC with PTX offloading even without CUDA being installed
>> (gcc and nvptx-tools patches)": "Especially for distributions it is
>> undesirable to need to have proprietary CUDA libraries and headers
>> installed when building GCC.", and I understand GNU/Linux distributions
>> all use that.  That configuration uses the GCC-provided
>> 'libgomp/plugin/cuda/cuda.h', 'libgomp/plugin/cuda-lib.def' to manually
>> define the CUDA Driver ABI to use, and then 'dlopen("libcuda.so.1")'.
>> (Similar to what the libgomp GCN (and before: HSA) plugin is doing, for
>> example.)  Quite likely that our group (at work) are the only ones to
>> actually use '--with-cuda-driver'?
>
> Right, I see in my scripts that I don't use --with-cuda-driver, possibly
> because of years-ago running into issues when changing drivers forth and
> back.
>
>> My proposal now is: we remove '--with-cuda-driver' (make its use a no-op,
>> per standard GNU Autoconf behavior), and offer '--without-cuda-driver'
>> only.  This shouldn't cause any user-visible change in behavior, so safe
>> without a prior deprecation phase.
>
> I think the dlopen use-case is the most flexible, and I don't see any
> user benefit from using --with-cuda-driver, so I don't see a problem
> with removing --with-cuda-driver for the user.

ACK, thanks.

> I did wonder about keeping it available in some form, say rename to
> --maintainer-mode-with-cuda-driver.  This could be useful for debugging
> / comparison purposes.  But it would mean having to test it when making
> relevant changes, which is maintenance burden for a feature not visible
> to the user, so I guess that's not worth it.
>
> So, I'm fine with removing.

Based on the point you made above, I realized that it may be beneficial
to "keep the underlying functionality available for the developers":
"if you develop CUDA API-level changes in the libgomp nvptx plugin, it's
likely to be easier to just use the full CUDA toolkit 'cuda.h' and
directly link against libcuda (so that you've got all symbols etc.
available), and only once you know what exactly you need, update GCC's
'include/cuda/cuda.h' and 'libgomp/plugin/cuda-lib.def'".  (See the
thread "libgomp nvptx plugin: Split 'PLUGIN_NVPTX_DYNAMIC' into
'PLUGIN_NVPTX_INCLUDE_SYSTEM_CUDA_H' and 'PLUGIN_NVPTX_LINK_LIBCUDA'".)

Do we agree that it's OK to remove the user-visiable '--with-cuda-driver'
etc. options, and do not introduce any new
'--enable-maintainer-mode-with-cuda-driver' (or similar) option, and
instead let this functionality be available to developers only, via
manually editing 'libgomp/plugin/Makefrag.am'?

Happy to submit an illustrative patch, if that helps.


Grüße
 Thomas
-----------------
Siemens Electronic Design Automation GmbH; Anschrift: Arnulfstraße 201, 80634 München; Gesellschaft mit beschränkter Haftung; Geschäftsführer: Thomas Heurung, Frank Thürauf; Sitz der Gesellschaft: München; Registergericht München, HRB 106955
  

Hi Roger!

On 2022-04-05T19:46:08+0100, "Roger Sayle" <roger@nextmovesoftware.com> wrote:
> I apologise that it might complicate things

No worries; that's why I asked.  :-)

> one potential benefit of --with-cuda-driver
> (i.e. linking the compiler against proprietary libraries) is that it would allow support for
> -march=native on nvptx (i.e. the gcc driver can figure out what sm_xx is available on the
> GPU(s) of the current machine, and pass that to cc1.  Like with all the microarchitectures
> on other platforms (x86_64), figuring this out is not a trivial task for many end-users.

Ah, interesting idea!  (I suppose still not too relevant right now, but
once we've got GCC/nvptx multilibs making better use of modern PTX
features, I certainly do see the point.)

> Of course, ideally I'd love to be able to figure out the PTX hardware specifications and
> driver versions without using a third-party library, but I've no idea how this could be done
> (portably across the platforms that support libcuda).  Perhaps dlopen at runtime?
> Or calling out to (executing) nvptx-tools?

Indeed I suppose I'd do this via 'dlopen', and not directly link GCC
against 'libcuda.so' etc.: so that the GCC build may also be used on a
system where no 'libcuda.so' is available.

So -- sorry ;-) -- that's not a show-stopper for removing the
'--with-cuda-driver' etc. 'configure'-time options.


Grüße
 Thomas
-----------------
Siemens Electronic Design Automation GmbH; Anschrift: Arnulfstraße 201, 80634 München; Gesellschaft mit beschränkter Haftung; Geschäftsführer: Thomas Heurung, Frank Thürauf; Sitz der Gesellschaft: München; Registergericht München, HRB 106955