<div dir="ltr"><br><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">---------- Forwarded message ---------<br>From: <strong class="gmail_sendername" dir="auto">Jussi Kivilinna</strong> <span dir="auto"><<a href="mailto:jussi.kivilinna@iki.fi">jussi.kivilinna@iki.fi</a>></span><br>Date: Fri, Nov 4, 2022 at 7:50 PM<br>Subject: Re: Fwd: cipher-gcm-armv7-neon.S issue<br>To: Dmytro Kovalov <<a href="mailto:dmytro.a.kovalov@globallogic.com">dmytro.a.kovalov@globallogic.com</a>><br></div><br><br><div>Hello,<br>
<br>
Please send questions about libgcrypt to "<a href="mailto:gcrypt-devel@gnupg.org" target="_blank">gcrypt-devel@gnupg.org</a>"<br>
<br>
-Jussi<br>
<br>
On 4.11.2022 19.33, Dmytro Kovalov wrote:<br>
> <br>
> Hello Jussi,<br>
> <br>
> Little update.<br>
> I found the root cause of thecipher-gcm-armv7-neon.S (sha512-armv7-neon.S) issue .<br>
> It was an extra '$' symbol.<br>
> <br>
> This source describe diffs between AT&T and INTEL dialect<br>
> <a href="https://imada.sdu.dk/~kslarsen/dm546/Material/IntelnATT.htm#:~:text=In%20Intel%20syntax%20the%20first,in%20this%20situation%20is%20obvious" target="_blank">https://imada.sdu.dk/~kslarsen/dm546/Material/IntelnATT.htm#:~:text=In%20Intel%20syntax%20the%20first,in%20this%20situation%20is%20obvious</a> <<a href="https://imada.sdu.dk/~kslarsen/dm546/Material/IntelnATT.htm#:~:text=In%20Intel%20syntax%20the%20first,in%20this%20situation%20is%20obvious" target="_blank">https://imada.sdu.dk/~kslarsen/dm546/Material/IntelnATT.htm#:~:text=In%20Intel%20syntax%20the%20first,in%20this%20situation%20is%20obvious</a>>.<br>
> and the AT&T syntax is different from either Intel or libgcrypt asm source code.<br>
> I mean from all rules described on source there only *Prefixes* rule partly uses by libgcrypt code (two '%' and $ are from AT&T )<br>
> The others (*Memory Operands, **Direction of Operands, **Suffixes***) haven't applied to libgcrypt code.<br>
> <br>
> Could you please comment difference between AT&T Intel and gcc approaches ?<br>
> I am a bit confused , but can confirm that at least clang build (sha512-neon-arm) works now after removal of '%' '$' on the arm machine.<br>
> <br>
> Best Regards,<br>
> Dmytro<br>
> ---------- Forwarded message ---------<br>
> From: *Dmytro Kovalov* <<a href="mailto:dmytro.a.kovalov@globallogic.com" target="_blank">dmytro.a.kovalov@globallogic.com</a> <mailto:<a href="mailto:dmytro.a.kovalov@globallogic.com" target="_blank">dmytro.a.kovalov@globallogic.com</a>>><br>
> Date: Fri, Nov 4, 2022 at 5:50 PM<br>
> Subject: cipher-gcm-armv7-neon.S issue<br>
> To: Jussi Kivilinna <<a href="mailto:jussi.kivilinna@iki.fi" target="_blank">jussi.kivilinna@iki.fi</a> <mailto:<a href="mailto:jussi.kivilinna@iki.fi" target="_blank">jussi.kivilinna@iki.fi</a>>><br>
> <br>
> <br>
> Hello Jussi,<br>
> <br>
> Just found an issue: the cipher-gcm-armv7-neon.S failed on clang compile stage.<br>
> The fix is just disable NEON on version 1.8.6 no such issue because of file absence.<br>
> <br>
> Regards,<br>
> Dmytro.<br>
</div>
</div></div>