Want to test Racket/Chez on powerpc64le?

Want to test Racket/Chez on powerpc64le?

[Philip McGrath]

Hi folks,

I have a patch series at https://gitlab.com/philip1/guix-patches/-/tree/zuo for the Racket 8.5.900 release candidate for the imminent Racket 8.6 release.[1] The patch series also enables Racket CS and the 'chez-scheme-for-racket' package on systems they haven't supported until now, such as powerpc64le-linux—at least, I think it does: I'm hoping some of you can test it. The riscv64-linux, mips64el-linux, and i586-gnu are similarly situated: I ask particularly about powerpc64le-linux because Matthew Flatt has tested it upstream (thanks to the GCC Compile Farm).

If you just want to help quickly, try this:

guix time-machine --url=https://gitlab.com/philip1/guix-patches --branch=zuo --disable-authentication -- build chez-scheme-for-racket racket

and let me know how it goes, either here or at [2]. Allow about an hour for the build. (So, not *that* quickly …)

Some more background about what's new:

The systems I listed above are ones for which Racket's variant of Chez Scheme doesn't (yet!) have backends to generate machine code. With these changes, they instead can now use "pbarch" backends added to Racket's variant of Chez Scheme: a "portable bytecode" mode specialized to word size and endianness, with core bytecode additionally compiled to C.

I've cross-compiled 'chez-scheme-for-racket' and 'racket-vm-cs' successfully for the Linux-based systems. (There's still a problem with i586-pc-gnu, probably some missing/different C flag.) Unfortunately, I ran into a QEMU error [3] when I tried an emulated build, possibly related to Racket BC's use of SIGSEGV to implement the GC write barrier, and I don't have access to any of the relevant hardware myself. So, I'm hoping people who do might be able to give it a try!

Thanks,
Philip

[1]: https://racket.discourse.group/t/racket-v8-6-release-thread/1091
[2]: https://racket.discourse.group/t/chez-for-architectures-without-native-backends/950
[3]: https://racket.discourse.group/t/chez-for-architectures-without-native-backends/950/13

Re: Want to test Racket/Chez on powerpc64le?

[Efraim Flashner]

[-- Attachment #1: Type: text/plain, Size: 2856 bytes --]

On Mon, Aug 01, 2022 at 03:46:21PM -0400, Philip McGrath wrote:
> Hi folks,
> 
> I have a patch series at https://gitlab.com/philip1/guix-patches/-/tree/zuo for the Racket 8.5.900 release candidate for the imminent Racket 8.6 release.[1] The patch series also enables Racket CS and the 'chez-scheme-for-racket' package on systems they haven't supported until now, such as powerpc64le-linux—at least, I think it does: I'm hoping some of you can test it. The riscv64-linux, mips64el-linux, and i586-gnu are similarly situated: I ask particularly about powerpc64le-linux because Matthew Flatt has tested it upstream (thanks to the GCC Compile Farm).
> 
> If you just want to help quickly, try this:
> 
> guix time-machine --url=https://gitlab.com/philip1/guix-patches --branch=zuo --disable-authentication -- build chez-scheme-for-racket racket
> 
> and let me know how it goes, either here or at [2]. Allow about an hour for the build. (So, not *that* quickly …)
> 
> Some more background about what's new:
> 
> The systems I listed above are ones for which Racket's variant of Chez Scheme doesn't (yet!) have backends to generate machine code. With these changes, they instead can now use "pbarch" backends added to Racket's variant of Chez Scheme: a "portable bytecode" mode specialized to word size and endianness, with core bytecode additionally compiled to C.
> 
> I've cross-compiled 'chez-scheme-for-racket' and 'racket-vm-cs' successfully for the Linux-based systems. (There's still a problem with i586-pc-gnu, probably some missing/different C flag.) Unfortunately, I ran into a QEMU error [3] when I tried an emulated build, possibly related to Racket BC's use of SIGSEGV to implement the GC write barrier, and I don't have access to any of the relevant hardware myself. So, I'm hoping people who do might be able to give it a try!
> 
> Thanks,
> Philip
> 
> [1]: https://racket.discourse.group/t/racket-v8-6-release-thread/1091
> [2]: https://racket.discourse.group/t/chez-for-architectures-without-native-backends/950
> [3]: https://racket.discourse.group/t/chez-for-architectures-without-native-backends/950/13

In general I should be able to help with the riscv64-linux for testing.
Unfortunately I'm working remotely for the next few weeks and a power
outage at home brought down my machines, so I won't be able to for a
while yet.

For mips64el, Guix used the Loongson2F as a build machine and target
while Debian targets the Loongson3A. I do have one of the retired Lemote
machines we used, but I haven't had it powered up for a few years and
isn't currently prepared for building packages.

-- 
Efraim Flashner   <efraim@flashner.co.il>   אפרים פלשנר
GPG key = A28B F40C 3E55 1372 662D  14F7 41AA E7DC CA3D 8351
Confidentiality cannot be guaranteed on emails sent or received unencrypted

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: Want to test Racket/Chez on powerpc64le?

[Philip McGrath]

Hi,

On Mon, Aug 1, 2022, at 5:02 PM, Efraim Flashner wrote:
> On Mon, Aug 01, 2022 at 03:46:21PM -0400, Philip McGrath wrote:
>> Hi folks,
>> 
>> I have a patch series at https://gitlab.com/philip1/guix-patches/-/tree/zuo for the Racket 8.5.900 release candidate for the imminent Racket 8.6 release.[1] The patch series also enables Racket CS and the 'chez-scheme-for-racket' package on systems they haven't supported until now, such as powerpc64le-linux—at least, I think it does: I'm hoping some of you can test it. The riscv64-linux, mips64el-linux, and i586-gnu are similarly situated: I ask particularly about powerpc64le-linux because Matthew Flatt has tested it upstream (thanks to the GCC Compile Farm).
>> 
>> If you just want to help quickly, try this:
>> 
>> guix time-machine --url=https://gitlab.com/philip1/guix-patches --branch=zuo --disable-authentication -- build chez-scheme-for-racket racket
>> 
>> and let me know how it goes, either here or at [2]. Allow about an hour for the build. (So, not *that* quickly …)
>> 
>> [...]
>> 
>> [2]: https://racket.discourse.group/t/chez-for-architectures-without-native-backends/950

FYI, I've sent the series as <https://issues.guix.gnu.org/57050>, and I also updated the branch above. Testing on real hardware would still be very welcome, though!

>> (There's still a problem with i586-pc-gnu, probably some missing/different C flag.)

I ended up adding support for the Hurd as a native backend.

-Philip