<div dir="ltr"><div class="gmail_extra"><div class="gmail_quote">On 25 May 2018 at 13:34, Hal Finkel via cfe-dev <span dir="ltr"><<a href="mailto:cfe-dev@lists.llvm.org" target="_blank">cfe-dev@lists.llvm.org</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div text="#000000" bgcolor="#FFFFFF"><span class="">
<p><br>
</p>
<div class="m_6348209695862640419moz-cite-prefix">On 05/25/2018 02:40 PM, Richard Smith
wrote:<br>
</div>
<blockquote type="cite">
<div dir="ltr">
<div class="gmail_extra">
<div class="gmail_quote">On 25 May 2018 at 12:15, Hal Finkel
via cfe-dev <span dir="ltr"><<a href="mailto:cfe-dev@lists.llvm.org" target="_blank">cfe-dev@lists.llvm.org</a>></span>
wrote:<br>
<blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
<div bgcolor="#FFFFFF"><span class="m_6348209695862640419gmail-">
<p><br>
</p>
<div class="m_6348209695862640419gmail-m_-7615144704090997817moz-cite-prefix">On
05/25/2018 02:01 PM, Friedman, Eli via cfe-dev
wrote:<br>
</div>
<blockquote type="cite">
<div class="m_6348209695862640419gmail-m_-7615144704090997817moz-cite-prefix">On
5/25/2018 11:46 AM, JF Bastien wrote:<br>
</div>
<blockquote type="cite"> <br>
<div><br>
<blockquote type="cite">
<div>On May 25, 2018, at 11:38 AM, Friedman,
Eli <<a href="mailto:efriedma@codeaurora.org" target="_blank">efriedma@codeaurora.org</a>>
wrote:</div>
<br class="m_6348209695862640419gmail-m_-7615144704090997817Apple-interchange-newline">
<div>
<div bgcolor="#FFFFFF">
<div class="m_6348209695862640419gmail-m_-7615144704090997817moz-cite-prefix">On
5/25/2018 11:29 AM, JF Bastien via
cfe-dev wrote:<br>
</div>
<blockquote type="cite">
<div>
<div>
<ol class="m_6348209695862640419gmail-m_-7615144704090997817MailOutline">
<li>Teach the target
infrastructure that hardware
interference size is something
they can specify (in tablegen
files somewhere).</li>
<li>Allow overriding the value in
sub-targets using -march or
-mcpu (the sub-target defines
the numeric value, and the user
gets the overriden one by using
-march or -mcpu).</li>
</ol>
</div>
</div>
</blockquote>
<br>
We can't change the value based on -mcpu.
We generally allow mixing code built with
different values of -mcpu. And any code
which is linked together must use the same
value for hardware_destructive_interfere<wbr>nce_size,
or else we violate ODR.<br>
</div>
</div>
</blockquote>
<div><br>
</div>
<div>Interesting point. The case I’d like to
cover is one where the developer wants to get
the exact right value for their particular
CPU, instead of a conservative answer with
extra padding. How do you think we should meet
this use case?</div>
<br>
</div>
</blockquote>
<br>
Go back to the standards committee and ask for a
function that isn't constexpr? I can't think of any
other reasonable solution.<br>
</blockquote>
<br>
</span> Unfortunately, to define structure layouts they
need to be constant.<br>
<br>
The best solution I've thought of is to extend the
abi_tag support to force the mangling of interfaces
depending on values of these constructs to be different.<br>
</div>
</blockquote>
<div><br>
</div>
<div>abi_tag is not an effective way of maintaining ABI,
because it needs to be "viral" / transitive, and can't be
(at least, not without huge developer effort).</div>
</div>
</div>
</div>
</blockquote>
<br></span>
Interesting. I had thought that abi_tag was transitive.<br>
<br>
It occurs to me that Transitive ABI Infection Mechanism (TAIM) has a
reasonable acronym. :-) - I suspect that's what we need in this
case.</div></blockquote><div><br></div><div>That's not possible, because classes can be forward-declared, and you need to know what fields and base classes they have to transitively propagate the taint.</div><div><br></div><div>GCC tries make it possible to transitively propagate the taint manually: they have a warning for a type that uses a tainted type and isn't itself declared with the abi_tag attribute. But in practice (at least for the "C++11 ABI" abi_tag) the taint ends up affecting sufficiently many classes as to make using the attribute that way infeasible to all but the most dedicated.</div><div><br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div text="#000000" bgcolor="#FFFFFF"><span class=""><blockquote type="cite">
<div dir="ltr">
<div class="gmail_extra">
<div class="gmail_quote">
<div>Perhaps we could add an attribute
to hardware_{con,de}structive_<wbr>interference_size that
produces a warning if they are used outside the main
source file?</div>
</div>
</div>
</div>
</blockquote>
<br></span>
I thought about suggesting this, but didn't, because I suspect that
many/most uses will be in header files, just project-internal header
files (because they'll be defining structure layouts, padding
arrays, etc.). I think that such a warning will be pretty noisy,
unfortunately.<span class=""><br>
<br>
<blockquote type="cite">
<div dir="ltr">
<div class="gmail_extra">
<div class="gmail_quote">
<div> We'd also need to make them non-inline, which is an
observable conformance break, but seems unlikely to be
important compared to the other issues.</div>
</div>
</div>
</div>
</blockquote>
<br></span>
Good point. Do you think that we should file a DR about this? I
imagine that most everyone is going to be in the same boat in this
regard.<br></div></blockquote><div><br></div><div>Yes, we probably should. Making them non-inline would presumably mean that we can tell the user it's their fault if the value differs between translation units and they use it in an ODR-relevant context. That doesn't solve the problem, but it does make it not our problem to solve, to a certain extent. We're still left with this being a dangerous feature, but I think that's really unavoidable if we want these things to be compile-time constants.</div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div text="#000000" bgcolor="#FFFFFF">
-Hal<span class=""><br>
<br>
<blockquote type="cite">
<div dir="ltr">
<div class="gmail_extra">
<div class="gmail_quote">
<div> </div>
<blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
<div bgcolor="#FFFFFF"> -Hal <br>
<br>
<blockquote type="cite"><span class="m_6348209695862640419gmail-"> <br>
-Eli<br>
<pre class="m_6348209695862640419gmail-m_-7615144704090997817moz-signature" cols="72">--
Employee of Qualcomm Innovation Center, Inc.
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project</pre>
<br>
<fieldset class="m_6348209695862640419gmail-m_-7615144704090997817mimeAttachmentHeader"></fieldset>
<br>
</span>
<pre>______________________________<wbr>_________________
cfe-dev mailing list
<a class="m_6348209695862640419gmail-m_-7615144704090997817moz-txt-link-abbreviated" href="mailto:cfe-dev@lists.llvm.org" target="_blank">cfe-dev@lists.llvm.org</a>
<a class="m_6348209695862640419gmail-m_-7615144704090997817moz-txt-link-freetext" href="http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev" target="_blank">http://lists.llvm.org/cgi-bin/<wbr>mailman/listinfo/cfe-dev</a>
</pre>
</blockquote>
<span class="m_6348209695862640419gmail-"> <br>
<pre class="m_6348209695862640419gmail-m_-7615144704090997817moz-signature" cols="72">--
Hal Finkel
Lead, Compiler Technology and Programming Languages
Leadership Computing Facility
Argonne National Laboratory</pre>
</span></div>
<br>
______________________________<wbr>_________________<br>
cfe-dev mailing list<br>
<a href="mailto:cfe-dev@lists.llvm.org" target="_blank">cfe-dev@lists.llvm.org</a><br>
<a href="http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev" rel="noreferrer" target="_blank">http://lists.llvm.org/cgi-bin/<wbr>mailman/listinfo/cfe-dev</a><br>
<br>
</blockquote>
</div>
<br>
</div>
</div>
</blockquote>
<br>
<pre class="m_6348209695862640419moz-signature" cols="72">--
Hal Finkel
Lead, Compiler Technology and Programming Languages
Leadership Computing Facility
Argonne National Laboratory</pre>
</span></div>
<br>______________________________<wbr>_________________<br>
cfe-dev mailing list<br>
<a href="mailto:cfe-dev@lists.llvm.org">cfe-dev@lists.llvm.org</a><br>
<a href="http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev" rel="noreferrer" target="_blank">http://lists.llvm.org/cgi-bin/<wbr>mailman/listinfo/cfe-dev</a><br>
<br></blockquote></div><br></div></div>