<div dir="ltr">FWIW, <a href="https://lld.llvm.org/windows_support.html">lld seems pretty well supported on Windows</a>. We have some evidence that lld is faster than gold and ld.bfd on Linux.<div><br></div><div>Thanks</div><div>Greg</div></div><br><div class="gmail_quote"><div dir="ltr">On Sat, Dec 22, 2018 at 7:44 AM Ben Gamari <<a href="mailto:ben@well-typed.com">ben@well-typed.com</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">"Greg Steuck (Sh-toy-k)" <<a href="mailto:gnezdo@google.com" target="_blank">gnezdo@google.com</a>> writes:<br>
<br>
> I checked and we can go 10x what we have been using up to now. We'll have a<br>
> spending cap of $4500/month set up.<br>
><br>
> If you can get away from Windows, the licensing fees will stop eating into<br>
> this budget.<br>
><br>
Indeed. Currently Windows is eating a disproportionate amount of our<br>
budget due to a perfect storm of considerations:<br>
<br>
a. Windows licensing fees, meaning Windows instances cost two-to-three<br>
times what we pay per-core for Linux instances<br>
<br>
b. the Windows linker is slow as mollasses, meaning we need lots of<br>
cores to keep up with our build volume<br>
<br>
Build times are currently over three hours on the current four-core<br>
machine we current have on GCE with over half of this being spent in the<br>
testsuite where almost all time is spent in ld. I have made some<br>
inquiries to see whether there is anything we can do about this.<br>
<br>
In the short term I intend on increasing the size of our instance to try<br>
to keep up with demand a bit better.<br>
<br>
Thanks again for your help!<br>
<br>
Cheers,<br>
<br>
- Ben<br>
</blockquote></div>