[PATCH 0/3] Enable full libgfortran for AMD GCN

classic Classic list List threaded Threaded
7 messages Options
Reply | Threaded
Open this post in threaded view
|

[PATCH 0/3] Enable full libgfortran for AMD GCN

Andrew Stubbs-4
This short patch series is intended to allow GCN kernels to print to
stdout using Fortran "print" and "write" (C printf, write, etc. already
work).

The first two patches implement dependencies that libgfortran needs.
I'll also be posting a newlib patch shortly.

The third patch enables the rest of libgfortran.

--
Andrew Stubbs
Mentor Graphics / CodeSourcery
Reply | Threaded
Open this post in threaded view
|

Re: [PATCH 1/3] Create GCN-specific gthreads

Andrew Stubbs-4
Ping.

I can probably approve this myself, as it only affects GCN, but I'd
appreciate a second opinion.

Thanks

Andrew

On 07/06/2019 15:39, Andrew Stubbs wrote:

> This patch creates a new gthread model for AMD GCN devices.
>
> For now, there's just enough support for libgfortran to use mutexes in
> its I/O routines. The rest can be added at a later time, if at all.
>
> Notes:
>
>   * GCN GPUs do not support dynamic creation and deletion of threads, so
>     there can be no implementation for those functions. (There may be
>     many threads, of course, but they are hardware managed and must be
>     launched all at once.)
>
>   * It would be possible to implement support for EMUTLS, but I have no
>     wish to do so at this time, and it isn't likely to be needed by
>     OpenMP or OpenACC offload kernels, so those functions are also stub
>     implementations.
>
> OK to commit?
>

Reply | Threaded
Open this post in threaded view
|

Re: [PATCH 2/3] Stub implementation of unwinding for AMD GCN.

Andrew Stubbs-4
In reply to this post by Andrew Stubbs-4
Ping.

I can probably approve this myself, as it only affects GCN, but I'd
appreciate a second opinion.

Thanks

Andrew

On 07/06/2019 15:40, Andrew Stubbs wrote:
> This patch provides the "_Unwind_Backtrace" and "_Unwind_GetIPInfo"
> symbols required to link programs using libgfortran.
>
> I do not wish to implement proper backtracing at this time (I have other
> things to work on), and IIUC none of the existing implementations will
> Just Work.
>
> OK to commit?
>

Reply | Threaded
Open this post in threaded view
|

Re: [PATCH 1/3] Create GCN-specific gthreads

Jeff Law
In reply to this post by Andrew Stubbs-4
On 6/19/19 2:57 AM, Andrew Stubbs wrote:
> Ping.
>
> I can probably approve this myself, as it only affects GCN, but I'd
> appreciate a second opinion.
Yes, this would fall under things you could approve yourself.  Thanks
for double-checking.

jeff
Reply | Threaded
Open this post in threaded view
|

Re: [PATCH 3/3] Enable full libgfortran library for AMD GCN

Jeff Law
In reply to this post by Andrew Stubbs-4
On 6/7/19 8:40 AM, Andrew Stubbs wrote:
> This patch basically reverts the previous patch to put AMD GCN in
> "minimal" mode.
>
> OK to commit?
OK
jeff

Reply | Threaded
Open this post in threaded view
|

Re: [PATCH 1/3] Create GCN-specific gthreads

Andrew Stubbs
In reply to this post by Jeff Law
On 19/06/2019 17:04, Jeff Law wrote:
> On 6/19/19 2:57 AM, Andrew Stubbs wrote:
>> Ping.
>>
>> I can probably approve this myself, as it only affects GCN, but I'd
>> appreciate a second opinion.
> Yes, this would fall under things you could approve yourself.  Thanks
> for double-checking.

Sorry, I meant I'd like another opinion on the patch contents. I'm not
confident that I didn't miss something.

Andrew
Reply | Threaded
Open this post in threaded view
|

Re: [PATCH 2/3] Stub implementation of unwinding for AMD GCN.

Andrew Stubbs-4
In reply to this post by Andrew Stubbs-4
On 19/06/2019 17:04, Jeff Law wrote:
> On 6/19/19 2:58 AM, Andrew Stubbs wrote:
>> Ping.
>>
>> I can probably approve this myself, as it only affects GCN, but I'd
>> appreciate a second opinion.
> Similarly this is fine to self-approve.  Thanks.

Sorry, same again, I meant I'd like another opinion on the patch contents.

Andrew