<div dir="ltr"><div dir="ltr"><br></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Fri, Sep 11, 2020 at 9:06 AM Karel Gardas <<a href="mailto:karel.gardas@centrum.cz">karel.gardas@centrum.cz</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">On 9/11/20 3:13 PM, Joel Sherrill wrote:<br>
> FWIW i386 is an 80s CPU introduced in 1985. i486 was introduced<br>
> in 1989. The Pentium was 1993. Remember It's All About the Pentium!<br>
> Pentium II was 1997 and first with SMP but maybe not setting the baseline<br>
> we want.<br>
<br>
What about to support in master what man can currently purchase? If so,<br>
then based on my limited research it looks like the cpu family<br>
(orderable) goes back to 2016/2015 which support architecture of more or<br>
less NetBurst uarch isn set (both intel and amd).<br>
<br>
One exception found is Vertex86 cpus which still sells and boards are<br>
available and which are compatible only with i586/p5 from as you noted<br>
1993...<br></blockquote><div><br>Karel: Typo correction: Vortex86: <a href="https://en.wikipedia.org/wiki/Vortex86" rel="noreferrer" target="_blank">https://en.wikipedia.org/wiki/Vortex86</a><br></div><div><br></div><div>If I am reading that right, we may be ok dumping i386 and i486 completely</div><div>and assuming around a Pentium II as default. Rather than a blanket drop </div><div>32-bit support which kills a LOT of usable functionality, I would rather find</div><div>a new floor for the CPU model. </div><div><br></div><div>My recollection from SMP testing i386 on qemu was that the lowest CPU </div><div>model on Qemu that worked was core2duo which pc586-sse matches for</div><div>GCC arguments.</div><div><br></div><div>CPU_CFLAGS = -mtune=pentium -march=pentium -msse2<br></div><div><br></div><div>Is that lower than the broken code in GCC? What's the lowest multilib</div><div>where the atomics are right? Assuming that it isn't just a bug below that</div><div>level.</div><div><br></div><div>I would like to pick a floor based on some reasonable rationale </div><div>like GCC needs feature X for atomics which appears in CPU model Y.</div><div><br></div><div>I don't think Qemu provides a rationale for a floor. Availability of SoCs</div><div>for embedded systems seems like a short list and the wikipedia page</div><div>doesn't encourage me that they are viable options as most look dead</div><div>or barely breathing (parrot is dead):<br><br><a href="https://en.wikipedia.org/wiki/List_of_x86_manufacturers">https://en.wikipedia.org/wiki/List_of_x86_manufacturers</a><br></div><div><br></div><div>That leaves us with what's the lowest multilib in GCC where atomics</div><div>work (or can be fixed, assuming this is a bug).</div><div><br></div><div>--joel</div><div><br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
<br>
Karel<br>
</blockquote></div></div>