[www] r233526 - Add keynote abstracts
Tobias Grosser
tobias at grosser.es
Mon Mar 30 01:31:29 PDT 2015
Author: grosser
Date: Mon Mar 30 03:31:29 2015
New Revision: 233526
URL: http://llvm.org/viewvc/llvm-project?rev=233526&view=rev
Log:
Add keynote abstracts
Modified:
www/trunk/devmtg/2015-04/index.html
Modified: www/trunk/devmtg/2015-04/index.html
URL: http://llvm.org/viewvc/llvm-project/www/trunk/devmtg/2015-04/index.html?rev=233526&r1=233525&r2=233526&view=diff
==============================================================================
--- www/trunk/devmtg/2015-04/index.html (original)
+++ www/trunk/devmtg/2015-04/index.html Mon Mar 30 03:31:29 2015
@@ -355,7 +355,11 @@ soon.
.s0{text-align:left;font-weight:bold;color:#000000;background-color:#ffffff;font-size:10pt;vertical-align:bottom;white-space:nowrap;direction:ltr;padding:2px
3px 2px 3px;}.ritz .waffle
.s1{text-align:left;color:#000000;background-color:#ffffff;font-size:10pt;vertical-align:bottom;white-space:nowrap;direction:ltr;padding:2px
-3px 2px 3px;}.ritz .waffle .s2{border-right:
+3px 2px 3px;}
+.abstract{text-align:left;color:#000000;background-color:#ffffff;font-size:10pt;vertical-align:bottom;direction:ltr;padding:2px
+3px 2px 3px;}
+
+.ritz .waffle .s2{border-right:
none;text-align:left;color:#000000;background-color:#ffffff;font-size:11pt;vertical-align:bottom;white-space:nowrap;direction:ltr; padding:2px
3px 2px 3px;font-weight:bold}.ritz .waffle
.s3{text-align:left;color:#ff0000;background-color:#ffffff;font-size:10pt;vertical-align:bottom;white-space:nowrap;direction:ltr;padding:2px;font-weight: bold
@@ -376,6 +380,43 @@ none;text-align:left;color:#000000;backg
</td>
</tr>
<tr style='height:21px;'>
+ <td class="abstract" dir="ltr" colspan="2">
+<br>
+<div style="width: 80em">
+The C and C++ languages were originally designed without
+concurrency support. The recent revision of the C and C++
+standards introduced an intricate but precise semantics for
+threads; todayâs C and C++ compilers, whose optimisers were
+initially developed in absence of any well-defined concurrency
+memory model, are being extended to support this new
+concurrency standard.<br/>
+In this talk, after recalling the C and C++ memory models, we
+will explore the correctness of compiler optimisations and
+present simple necessary conditions that can be used as a
+reference by compiler implementers. As an application, we
+will show how this theory enables building an automatic
+compiler tester that hunts âconcurrency compiler bugsâ: subtle
+wrong code generation bugs which are observable only when the
+miscompiled functions interact with concurrent contexts.<br/>
+Perhaps surprisingly, we will also show that by leveraging the
+semantics of low-level relaxed atomic accesses (which allows
+programmers to take full advantage of weakly-ordered memory
+operations), it is possible to build counterexamples to
+several common source-to-source program transformations (such
+as expression linearisation and âroach motelâ reorderings)
+that modern compilers perform and that are deemed to be
+correct. This implies that this memory model cannot be used
+to give semantics to compiler intermediate languages, as the
+LLVM IR documentation suggests. We will evaluate a number of
+possible local fixes, some strengthening and some weakening
+the model, and discuss their implications for compiler
+writers.
+<br>
+<br>
+</div>
+</td>
+ </tr>
+ <tr style='height:21px;'>
<td class="s1" dir="ltr">Ivan Godard</td>
<td class="s2 softmerge">
<div class="softmerge-inner"> LLVM meets the truly alien:
@@ -383,6 +424,32 @@ none;text-align:left;color:#000000;backg
</td>
</tr>
<tr style='height:21px;'>
+ <td class="abstract" dir="ltr" colspan="2">
+<br>
+<div style="width: 80em">
+<br>
+The Mill is a new commercial CPU family reflecting a ground-up rethink of single-thread architecture
+
+<ul>
+<li>Mill has no general registers</li>
+<li>Mill is SSA - in hardware</li>
+<li>Mill code has two instruction streams and two program counters, one of which runs backwards</li>
+<li>Mill instructions look (kinda) like six VLIWs, side-by-side</li>
+<li>Mill operations are polymorphic, both for scalar width and vector length</li>
+<li>x = a < 0x12345678901234567 ? F(G(b+c), e-f) : g + h; is (less than) one instruction</li>
+<li>and much more</li>
+</ul>
+
+ As a result, the Mill tool chain is - different. Except that it is LLVM. The retarget is not yet done, and has been - interesting. In the Chinese sense of interesting.
+
+ The talk will explain how the Mill tool chain works, and why it is done that way, and where LLVM fits into it.
+
+ Or doesn't. Ranting and raving will ensue.
+<br>
+</div>
+</td>
+ </tr>
+ <tr style='height:21px;'>
<td></td>
<td></td>
</tr>
More information about the llvm-commits
mailing list