<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.1//EN"
"http://www.w3.org/TR/xhtml11/DTD/xhtml11.dtd">
<html xmlns="http://www.w3.org/1999/xhtml">
<head><meta http-equiv="content-type" content="text/html; charset=utf-8" />
<title>[193987] trunk/Source/JavaScriptCore</title>
</head>
<body>
<style type="text/css"><!--
#msg dl.meta { border: 1px #006 solid; background: #369; padding: 6px; color: #fff; }
#msg dl.meta dt { float: left; width: 6em; font-weight: bold; }
#msg dt:after { content:':';}
#msg dl, #msg dt, #msg ul, #msg li, #header, #footer, #logmsg { font-family: verdana,arial,helvetica,sans-serif; font-size: 10pt; }
#msg dl a { font-weight: bold}
#msg dl a:link { color:#fc3; }
#msg dl a:active { color:#ff0; }
#msg dl a:visited { color:#cc6; }
h3 { font-family: verdana,arial,helvetica,sans-serif; font-size: 10pt; font-weight: bold; }
#msg pre { overflow: auto; background: #ffc; border: 1px #fa0 solid; padding: 6px; }
#logmsg { background: #ffc; border: 1px #fa0 solid; padding: 1em 1em 0 1em; }
#logmsg p, #logmsg pre, #logmsg blockquote { margin: 0 0 1em 0; }
#logmsg p, #logmsg li, #logmsg dt, #logmsg dd { line-height: 14pt; }
#logmsg h1, #logmsg h2, #logmsg h3, #logmsg h4, #logmsg h5, #logmsg h6 { margin: .5em 0; }
#logmsg h1:first-child, #logmsg h2:first-child, #logmsg h3:first-child, #logmsg h4:first-child, #logmsg h5:first-child, #logmsg h6:first-child { margin-top: 0; }
#logmsg ul, #logmsg ol { padding: 0; list-style-position: inside; margin: 0 0 0 1em; }
#logmsg ul { text-indent: -1em; padding-left: 1em; }#logmsg ol { text-indent: -1.5em; padding-left: 1.5em; }
#logmsg > ul, #logmsg > ol { margin: 0 0 1em 0; }
#logmsg pre { background: #eee; padding: 1em; }
#logmsg blockquote { border: 1px solid #fa0; border-left-width: 10px; padding: 1em 1em 0 1em; background: white;}
#logmsg dl { margin: 0; }
#logmsg dt { font-weight: bold; }
#logmsg dd { margin: 0; padding: 0 0 0.5em 0; }
#logmsg dd:before { content:'\00bb';}
#logmsg table { border-spacing: 0px; border-collapse: collapse; border-top: 4px solid #fa0; border-bottom: 1px solid #fa0; background: #fff; }
#logmsg table th { text-align: left; font-weight: normal; padding: 0.2em 0.5em; border-top: 1px dotted #fa0; }
#logmsg table td { text-align: right; border-top: 1px dotted #fa0; padding: 0.2em 0.5em; }
#logmsg table thead th { text-align: center; border-bottom: 1px solid #fa0; }
#logmsg table th.Corner { text-align: left; }
#logmsg hr { border: none 0; border-top: 2px dashed #fa0; height: 1px; }
#header, #footer { color: #fff; background: #636; border: 1px #300 solid; padding: 6px; }
#patch { width: 100%; }
#patch h4 {font-family: verdana,arial,helvetica,sans-serif;font-size:10pt;padding:8px;background:#369;color:#fff;margin:0;}
#patch .propset h4, #patch .binary h4 {margin:0;}
#patch pre {padding:0;line-height:1.2em;margin:0;}
#patch .diff {width:100%;background:#eee;padding: 0 0 10px 0;overflow:auto;}
#patch .propset .diff, #patch .binary .diff {padding:10px 0;}
#patch span {display:block;padding:0 10px;}
#patch .modfile, #patch .addfile, #patch .delfile, #patch .propset, #patch .binary, #patch .copfile {border:1px solid #ccc;margin:10px 0;}
#patch ins {background:#dfd;text-decoration:none;display:block;padding:0 10px;}
#patch del {background:#fdd;text-decoration:none;display:block;padding:0 10px;}
#patch .lines, .info {color:#888;background:#fff;}
--></style>
<div id="msg">
<dl class="meta">
<dt>Revision</dt> <dd><a href="http://trac.webkit.org/projects/webkit/changeset/193987">193987</a></dd>
<dt>Author</dt> <dd>fpizlo@apple.com</dd>
<dt>Date</dt> <dd>2015-12-11 15:20:20 -0800 (Fri, 11 Dec 2015)</dd>
</dl>
<h3>Log Message</h3>
<pre>B3 should have CSE
https://bugs.webkit.org/show_bug.cgi?id=150961
Reviewed by Benjamin Poulain.
This implements a very simple CSE for pure values. I need this as a prerequisite for other
optimizations that I'm implementing. For now, this is neutral on imaging-gaussian-blur but a
slow-down on asm.js code. I suspect that the asm.js slow-down is because of other things that are
still going wrong, and anyway, I need CSE to be able to do even the most basic asm.js strength
reductions.
* b3/B3ReduceStrength.cpp:
* b3/B3ReduceStrength.h:
* b3/B3Value.cpp:
(JSC::B3::Value::replaceWithIdentity):
(JSC::B3::Value::key):</pre>
<h3>Modified Paths</h3>
<ul>
<li><a href="#trunkSourceJavaScriptCoreChangeLog">trunk/Source/JavaScriptCore/ChangeLog</a></li>
<li><a href="#trunkSourceJavaScriptCoreb3B3ReduceStrengthcpp">trunk/Source/JavaScriptCore/b3/B3ReduceStrength.cpp</a></li>
<li><a href="#trunkSourceJavaScriptCoreb3B3ReduceStrengthh">trunk/Source/JavaScriptCore/b3/B3ReduceStrength.h</a></li>
<li><a href="#trunkSourceJavaScriptCoreb3B3Valuecpp">trunk/Source/JavaScriptCore/b3/B3Value.cpp</a></li>
</ul>
</div>
<div id="patch">
<h3>Diff</h3>
<a id="trunkSourceJavaScriptCoreChangeLog"></a>
<div class="modfile"><h4>Modified: trunk/Source/JavaScriptCore/ChangeLog (193986 => 193987)</h4>
<pre class="diff"><span>
<span class="info">--- trunk/Source/JavaScriptCore/ChangeLog        2015-12-11 23:09:35 UTC (rev 193986)
+++ trunk/Source/JavaScriptCore/ChangeLog        2015-12-11 23:20:20 UTC (rev 193987)
</span><span class="lines">@@ -1,3 +1,22 @@
</span><ins>+2015-12-11 Filip Pizlo <fpizlo@apple.com>
+
+ B3 should have CSE
+ https://bugs.webkit.org/show_bug.cgi?id=150961
+
+ Reviewed by Benjamin Poulain.
+
+ This implements a very simple CSE for pure values. I need this as a prerequisite for other
+ optimizations that I'm implementing. For now, this is neutral on imaging-gaussian-blur but a
+ slow-down on asm.js code. I suspect that the asm.js slow-down is because of other things that are
+ still going wrong, and anyway, I need CSE to be able to do even the most basic asm.js strength
+ reductions.
+
+ * b3/B3ReduceStrength.cpp:
+ * b3/B3ReduceStrength.h:
+ * b3/B3Value.cpp:
+ (JSC::B3::Value::replaceWithIdentity):
+ (JSC::B3::Value::key):
+
</ins><span class="cx"> 2015-12-11 Mark Lam <mark.lam@apple.com>
</span><span class="cx">
</span><span class="cx"> Refactoring to reduce potential cut-paste errors with the FTL ICs.
</span></span></pre></div>
<a id="trunkSourceJavaScriptCoreb3B3ReduceStrengthcpp"></a>
<div class="modfile"><h4>Modified: trunk/Source/JavaScriptCore/b3/B3ReduceStrength.cpp (193986 => 193987)</h4>
<pre class="diff"><span>
<span class="info">--- trunk/Source/JavaScriptCore/b3/B3ReduceStrength.cpp        2015-12-11 23:09:35 UTC (rev 193986)
+++ trunk/Source/JavaScriptCore/b3/B3ReduceStrength.cpp        2015-12-11 23:20:20 UTC (rev 193987)
</span><span class="lines">@@ -30,6 +30,7 @@
</span><span class="cx">
</span><span class="cx"> #include "B3BasicBlockInlines.h"
</span><span class="cx"> #include "B3ControlValue.h"
</span><ins>+#include "B3Dominators.h"
</ins><span class="cx"> #include "B3IndexSet.h"
</span><span class="cx"> #include "B3InsertionSetInlines.h"
</span><span class="cx"> #include "B3MemoryValue.h"
</span><span class="lines">@@ -37,8 +38,10 @@
</span><span class="cx"> #include "B3ProcedureInlines.h"
</span><span class="cx"> #include "B3UpsilonValue.h"
</span><span class="cx"> #include "B3UseCounts.h"
</span><ins>+#include "B3ValueKey.h"
</ins><span class="cx"> #include "B3ValueInlines.h"
</span><span class="cx"> #include <wtf/GraphNodeWorklist.h>
</span><ins>+#include <wtf/HashMap.h>
</ins><span class="cx">
</span><span class="cx"> namespace JSC { namespace B3 {
</span><span class="cx">
</span><span class="lines">@@ -105,11 +108,20 @@
</span><span class="cx"> dataLog(m_proc);
</span><span class="cx"> }
</span><span class="cx">
</span><ins>+ m_proc.resetValueOwners();
+ m_dominators = &m_proc.dominators(); // Recompute if necessary.
+ m_pureValues.clear();
+
</ins><span class="cx"> for (BasicBlock* block : m_proc.blocksInPreOrder()) {
</span><span class="cx"> m_block = block;
</span><span class="cx">
</span><del>- for (m_index = 0; m_index < block->size(); ++m_index)
- process();
</del><ins>+ for (m_index = 0; m_index < block->size(); ++m_index) {
+ m_value = m_block->at(m_index);
+ m_value->performSubstitution();
+
+ reduceValueStrength();
+ replaceIfRedundant();
+ }
</ins><span class="cx"> m_insertionSet.execute(m_block);
</span><span class="cx"> }
</span><span class="cx">
</span><span class="lines">@@ -129,11 +141,8 @@
</span><span class="cx"> }
</span><span class="cx">
</span><span class="cx"> private:
</span><del>- void process()
</del><ins>+ void reduceValueStrength()
</ins><span class="cx"> {
</span><del>- m_value = m_block->at(m_index);
- m_value->performSubstitution();
-
</del><span class="cx"> switch (m_value->opcode()) {
</span><span class="cx"> case Add:
</span><span class="cx"> handleCommutativity();
</span><span class="lines">@@ -1014,6 +1023,35 @@
</span><span class="cx"> return false;
</span><span class="cx"> }
</span><span class="cx">
</span><ins>+ void replaceIfRedundant()
+ {
+ // This does a very simple pure dominator-based CSE. In the future we could add load elimination.
+ // Note that if we add load elimination, we should do it by directly matching load and store
+ // instructions instead of using the ValueKey functionality or doing DFG HeapLocation-like
+ // things.
+
+ // Don't bother with identities. We kill those anyway.
+ if (m_value->opcode() == Identity)
+ return;
+
+ ValueKey key = m_value->key();
+ if (!key)
+ return;
+
+ Vector<Value*, 1>& matches = m_pureValues.add(key, Vector<Value*, 1>()).iterator->value;
+
+ // Replace this value with whichever value dominates us.
+ for (Value* match : matches) {
+ if (m_dominators->dominates(match->owner, m_value->owner)) {
+ m_value->replaceWithIdentity(match);
+ m_changed = true;
+ return;
+ }
+ }
+
+ matches.append(m_value);
+ }
+
</ins><span class="cx"> void simplifyCFG()
</span><span class="cx"> {
</span><span class="cx"> if (verbose) {
</span><span class="lines">@@ -1212,11 +1250,13 @@
</span><span class="cx">
</span><span class="cx"> Procedure& m_proc;
</span><span class="cx"> InsertionSet m_insertionSet;
</span><del>- BasicBlock* m_block;
- unsigned m_index;
- Value* m_value;
- bool m_changed;
- bool m_changedCFG;
</del><ins>+ BasicBlock* m_block { nullptr };
+ unsigned m_index { 0 };
+ Value* m_value { nullptr };
+ Dominators* m_dominators { nullptr };
+ HashMap<ValueKey, Vector<Value*, 1>> m_pureValues;
+ bool m_changed { false };
+ bool m_changedCFG { false };
</ins><span class="cx"> };
</span><span class="cx">
</span><span class="cx"> } // anonymous namespace
</span></span></pre></div>
<a id="trunkSourceJavaScriptCoreb3B3ReduceStrengthh"></a>
<div class="modfile"><h4>Modified: trunk/Source/JavaScriptCore/b3/B3ReduceStrength.h (193986 => 193987)</h4>
<pre class="diff"><span>
<span class="info">--- trunk/Source/JavaScriptCore/b3/B3ReduceStrength.h        2015-12-11 23:09:35 UTC (rev 193986)
+++ trunk/Source/JavaScriptCore/b3/B3ReduceStrength.h        2015-12-11 23:20:20 UTC (rev 193987)
</span><span class="lines">@@ -32,8 +32,11 @@
</span><span class="cx">
</span><span class="cx"> class Procedure;
</span><span class="cx">
</span><del>-// Does strength reduction, constant folding, canonicalization, CFG simplification, and DCE. In the
-// future we may also have it do CSE.
</del><ins>+// Does strength reduction, constant folding, canonicalization, CFG simplification, DCE, and CSE. This
+// phase runs those optimizations to fixpoint. The goal of the phase is to dramatically reduce the
+// complexity of the code. In the future, it's preferable to add optimizations to this phase rather than
+// creating new optimizations because then the optimizations can participate in the fixpoint. However,
+// this phase shouldn't become too expensive, so expensive optimizations should be separate.
</ins><span class="cx">
</span><span class="cx"> bool reduceStrength(Procedure&);
</span><span class="cx">
</span></span></pre></div>
<a id="trunkSourceJavaScriptCoreb3B3Valuecpp"></a>
<div class="modfile"><h4>Modified: trunk/Source/JavaScriptCore/b3/B3Value.cpp (193986 => 193987)</h4>
<pre class="diff"><span>
<span class="info">--- trunk/Source/JavaScriptCore/b3/B3Value.cpp        2015-12-11 23:09:35 UTC (rev 193986)
+++ trunk/Source/JavaScriptCore/b3/B3Value.cpp        2015-12-11 23:20:20 UTC (rev 193987)
</span><span class="lines">@@ -54,6 +54,13 @@
</span><span class="cx"> // a plain Identity Value. We first collect all of the information we need, then we destruct the
</span><span class="cx"> // previous value in place, and then we construct the Identity Value in place.
</span><span class="cx">
</span><ins>+ ASSERT(m_type == value->m_type);
+
+ if (m_type == Void) {
+ replaceWithNop();
+ return;
+ }
+
</ins><span class="cx"> unsigned index = m_index;
</span><span class="cx"> Type type = m_type;
</span><span class="cx"> Origin origin = m_origin;
</span><span class="lines">@@ -438,6 +445,7 @@
</span><span class="cx"> case FloatToDouble:
</span><span class="cx"> case DoubleToFloat:
</span><span class="cx"> case Check:
</span><ins>+ case BitwiseCast:
</ins><span class="cx"> return ValueKey(opcode(), type(), child(0));
</span><span class="cx"> case Add:
</span><span class="cx"> case Sub:
</span></span></pre>
</div>
</div>
</body>
</html>