<div dir="ltr"><div class="gmail_extra"><div class="gmail_quote">On 6 March 2018 at 13:38, Bob Tennent <span dir="ltr"><<a href="mailto:rdt@cs.queensu.ca" target="_blank">rdt@cs.queensu.ca</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Is anybody else using tl2018 finding it slower (particularly<br>
lua scripts)?<br>
<br>
Bob T.<br>
</blockquote></div><div class="gmail_extra"><br></div><div class="gmail_extra">As a general rule, software gets slower and CPU's faster over time.</div><div class="gmail_extra">This year, many will find CPU's running more slowly.</div><br>I/O intensive tasks have been reported to run as much as 30% slower </div><div class="gmail_extra">on current systems with meltdown and spectre mitigations. At work</div><div class="gmail_extra">we have seen very significant (deadline breaking) slowdowns of I/O </div><div class="gmail_extra">intensive processing, but we don't have good timing from "before </div><div class="gmail_extra">mitigation" and the input data changed in a way that may take longer</div><div class="gmail_extra">to process. I suspect those antique CPU's that are no longer being </div><div class="gmail_extra">made would suffer more, but not sure they get the same set (or any)</div><div class="gmail_extra">of mitigations available for current CPU's. The mitigations are </div><div class="gmail_extra">complex and in a state of flux. It is hard to predict whether changes to</div><div class="gmail_extra">compilers and lagnuage implementations can claw back some of the </div><div class="gmail_extra">lost performance. I think we can be sure Intel will be pushing new</div><div class="gmail_extra">CPU designs.</div><div class="gmail_extra"><div><br></div>-- <br><div class="gmail_signature" data-smartmail="gmail_signature"><div dir="ltr"><div>George N. White III<br><br></div></div></div>
</div></div>