<!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>[193993] 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/193993">193993</a></dd>
<dt>Author</dt> <dd>fpizlo@apple.com</dd>
<dt>Date</dt> <dd>2015-12-11 16:17:24 -0800 (Fri, 11 Dec 2015)</dd>
</dl>
<h3>Log Message</h3>
<pre>B3::reduceStrength should remove redundant Phi's
https://bugs.webkit.org/show_bug.cgi?id=152184
Reviewed by Benjamin Poulain.
This adds redundant Phi removal using Aycock and Horspools SSA simplification algorithm. This
is needed because even in simple asm.js code, we see a lot of CFG simplification that leaves
behind totally useless Phi's.
* b3/B3PhiChildren.cpp:
(JSC::B3::PhiChildren::PhiChildren):
* b3/B3PhiChildren.h:
(JSC::B3::PhiChildren::at):
(JSC::B3::PhiChildren::operator[]):
(JSC::B3::PhiChildren::phis):
* b3/B3ReduceStrength.cpp:</pre>
<h3>Modified Paths</h3>
<ul>
<li><a href="#trunkSourceJavaScriptCoreChangeLog">trunk/Source/JavaScriptCore/ChangeLog</a></li>
<li><a href="#trunkSourceJavaScriptCoreb3B3PhiChildrencpp">trunk/Source/JavaScriptCore/b3/B3PhiChildren.cpp</a></li>
<li><a href="#trunkSourceJavaScriptCoreb3B3PhiChildrenh">trunk/Source/JavaScriptCore/b3/B3PhiChildren.h</a></li>
<li><a href="#trunkSourceJavaScriptCoreb3B3ReduceStrengthcpp">trunk/Source/JavaScriptCore/b3/B3ReduceStrength.cpp</a></li>
</ul>
</div>
<div id="patch">
<h3>Diff</h3>
<a id="trunkSourceJavaScriptCoreChangeLog"></a>
<div class="modfile"><h4>Modified: trunk/Source/JavaScriptCore/ChangeLog (193992 => 193993)</h4>
<pre class="diff"><span>
<span class="info">--- trunk/Source/JavaScriptCore/ChangeLog 2015-12-12 00:11:40 UTC (rev 193992)
+++ trunk/Source/JavaScriptCore/ChangeLog 2015-12-12 00:17:24 UTC (rev 193993)
</span><span class="lines">@@ -1,3 +1,22 @@
</span><ins>+2015-12-11 Filip Pizlo <fpizlo@apple.com>
+
+ B3::reduceStrength should remove redundant Phi's
+ https://bugs.webkit.org/show_bug.cgi?id=152184
+
+ Reviewed by Benjamin Poulain.
+
+ This adds redundant Phi removal using Aycock and Horspools SSA simplification algorithm. This
+ is needed because even in simple asm.js code, we see a lot of CFG simplification that leaves
+ behind totally useless Phi's.
+
+ * b3/B3PhiChildren.cpp:
+ (JSC::B3::PhiChildren::PhiChildren):
+ * b3/B3PhiChildren.h:
+ (JSC::B3::PhiChildren::at):
+ (JSC::B3::PhiChildren::operator[]):
+ (JSC::B3::PhiChildren::phis):
+ * b3/B3ReduceStrength.cpp:
+
</ins><span class="cx"> 2015-12-11 Benjamin Poulain <benjamin@webkit.org>
</span><span class="cx">
</span><span class="cx"> [JSC] Add an implementation of pow() taking an integer exponent to B3
</span></span></pre></div>
<a id="trunkSourceJavaScriptCoreb3B3PhiChildrencpp"></a>
<div class="modfile"><h4>Modified: trunk/Source/JavaScriptCore/b3/B3PhiChildren.cpp (193992 => 193993)</h4>
<pre class="diff"><span>
<span class="info">--- trunk/Source/JavaScriptCore/b3/B3PhiChildren.cpp 2015-12-12 00:11:40 UTC (rev 193992)
+++ trunk/Source/JavaScriptCore/b3/B3PhiChildren.cpp 2015-12-12 00:17:24 UTC (rev 193993)
</span><span class="lines">@@ -36,8 +36,13 @@
</span><span class="cx"> : m_upsilons(proc.values().size())
</span><span class="cx"> {
</span><span class="cx"> for (Value* value : proc.values()) {
</span><del>- if (UpsilonValue* upsilon = value->as<UpsilonValue>())
- m_upsilons[upsilon->phi()].append(upsilon);
</del><ins>+ if (UpsilonValue* upsilon = value->as<UpsilonValue>()) {
+ Value* phi = upsilon->phi();
+ Vector<UpsilonValue*>& vector = m_upsilons[phi];
+ if (vector.isEmpty())
+ m_phis.append(phi);
+ vector.append(upsilon);
+ }
</ins><span class="cx"> }
</span><span class="cx"> }
</span><span class="cx">
</span></span></pre></div>
<a id="trunkSourceJavaScriptCoreb3B3PhiChildrenh"></a>
<div class="modfile"><h4>Modified: trunk/Source/JavaScriptCore/b3/B3PhiChildren.h (193992 => 193993)</h4>
<pre class="diff"><span>
<span class="info">--- trunk/Source/JavaScriptCore/b3/B3PhiChildren.h 2015-12-12 00:11:40 UTC (rev 193992)
+++ trunk/Source/JavaScriptCore/b3/B3PhiChildren.h 2015-12-12 00:17:24 UTC (rev 193993)
</span><span class="lines">@@ -166,8 +166,11 @@
</span><span class="cx"> UpsilonCollection at(Value* value) { return UpsilonCollection(this, value, &m_upsilons[value]); }
</span><span class="cx"> UpsilonCollection operator[](Value* value) { return at(value); }
</span><span class="cx">
</span><ins>+ const Vector<Value*, 8>& phis() const { return m_phis; }
+
</ins><span class="cx"> private:
</span><span class="cx"> IndexMap<Value, Vector<UpsilonValue*>> m_upsilons;
</span><ins>+ Vector<Value*, 8> m_phis;
</ins><span class="cx"> };
</span><span class="cx">
</span><span class="cx"> } } // namespace JSC::B3
</span></span></pre></div>
<a id="trunkSourceJavaScriptCoreb3B3ReduceStrengthcpp"></a>
<div class="modfile"><h4>Modified: trunk/Source/JavaScriptCore/b3/B3ReduceStrength.cpp (193992 => 193993)</h4>
<pre class="diff"><span>
<span class="info">--- trunk/Source/JavaScriptCore/b3/B3ReduceStrength.cpp 2015-12-12 00:11:40 UTC (rev 193992)
+++ trunk/Source/JavaScriptCore/b3/B3ReduceStrength.cpp 2015-12-12 00:17:24 UTC (rev 193993)
</span><span class="lines">@@ -35,6 +35,7 @@
</span><span class="cx"> #include "B3InsertionSetInlines.h"
</span><span class="cx"> #include "B3MemoryValue.h"
</span><span class="cx"> #include "B3PhaseScope.h"
</span><ins>+#include "B3PhiChildren.h"
</ins><span class="cx"> #include "B3ProcedureInlines.h"
</span><span class="cx"> #include "B3UpsilonValue.h"
</span><span class="cx"> #include "B3UseCounts.h"
</span><span class="lines">@@ -134,6 +135,7 @@
</span><span class="cx"> }
</span><span class="cx">
</span><span class="cx"> killDeadCode();
</span><ins>+ simplifySSA();
</ins><span class="cx">
</span><span class="cx"> result |= m_changed;
</span><span class="cx"> } while (m_changed);
</span><span class="lines">@@ -1248,6 +1250,67 @@
</span><span class="cx"> }
</span><span class="cx"> }
</span><span class="cx">
</span><ins>+ void simplifySSA()
+ {
+ // This runs Aycock and Horspool's algorithm on our Phi functions [1]. For most CFG patterns,
+ // this can take a suboptimal arrangement of Phi functions and make it optimal, as if you had
+ // run Cytron, Ferrante, Rosen, Wegman, and Zadeck. It's only suboptimal for irreducible
+ // CFGs. In practice, that doesn't matter, since we expect clients of B3 to run their own SSA
+ // conversion before lowering to B3, and in the case of the DFG, that conversion uses Cytron
+ // et al. In that context, this algorithm is intended to simplify Phi functions that were
+ // made redundant by prior CFG simplification. But according to Aycock and Horspool's paper,
+ // this algorithm is good enough that a B3 client could just give us maximal Phi's (i.e. Phi
+ // for each variable at each basic block) and we will make them optimal.
+ // [1] http://pages.cpsc.ucalgary.ca/~aycock/papers/ssa.ps
+
+ // Aycock and Horspool prescribe two rules that are to be run to fixpoint:
+ //
+ // 1) If all of the Phi's children are the same (i.e. it's one child referenced from one or
+ // more Upsilons), then replace all uses of the Phi with the one child.
+ //
+ // 2) If all of the Phi's children are either the Phi itself or exactly one other child, then
+ // replace all uses of the Phi with the one other child.
+ //
+ // Rule (2) subsumes rule (1), so we can just run (2). We only run one fixpoint iteration
+ // here. This premise is that in common cases, this will only find optimization opportunities
+ // as a result of CFG simplification and usually CFG simplification will only do one round
+ // of block merging per ReduceStrength fixpoint iteration, so it's OK for this to only do one
+ // round of Phi merging - since Phis are the value analogue of blocks.
+
+ PhiChildren phiChildren(m_proc);
+
+ for (Value* phi : phiChildren.phis()) {
+ Value* otherChild = nullptr;
+ bool ok = true;
+ for (Value* child : phiChildren[phi].values()) {
+ if (child == phi)
+ continue;
+ if (child == otherChild)
+ continue;
+ if (!otherChild) {
+ otherChild = child;
+ continue;
+ }
+ ok = false;
+ break;
+ }
+ if (!ok)
+ continue;
+ if (!otherChild) {
+ // Wow, this would be super weird. It probably won't happen, except that things could
+ // get weird as a consequence of stepwise simplifications in the strength reduction
+ // fixpoint.
+ continue;
+ }
+
+ // Turn the Phi into an Identity and turn the Upsilons into Nops.
+ m_changed = true;
+ for (Value* upsilon : phiChildren[phi])
+ upsilon->replaceWithNop();
+ phi->replaceWithIdentity(otherChild);
+ }
+ }
+
</ins><span class="cx"> Procedure& m_proc;
</span><span class="cx"> InsertionSet m_insertionSet;
</span><span class="cx"> BasicBlock* m_block { nullptr };
</span></span></pre>
</div>
</div>
</body>
</html>