<!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>[184110] trunk</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/184110">184110</a></dd>
<dt>Author</dt> <dd>fpizlo@apple.com</dd>
<dt>Date</dt> <dd>2015-05-11 11:18:10 -0700 (Mon, 11 May 2015)</dd>
</dl>
<h3>Log Message</h3>
<pre>We shouldn't promote LoadVarargs to a sequence of GetStacks and PutStacks if doing so would exceed the LoadVarargs' limit
https://bugs.webkit.org/show_bug.cgi?id=144851
Reviewed by Michael Saboff.
Source/JavaScriptCore:
LoadVarargs loads arguments from some object and puts them on the stack. The region of
stack is controlled by a bunch of meta-data, including InlineCallFrame. InlineCallFrame
shouldn't really be edited after ByteCodeParser, so we cannot convert LoadVarargs to
something that uses more stack than the LoadVarargs wanted to.
This check was missing in the ArgumentsEliminationPhase's LoadVarargs->GetStack+PutStack
promoter. This is an important promotion rule for performance, and in cases where we are
compiling truly hot code, the LoadVarargs limit will be at least as big as the length of
the phantom arguments array that this phase sees. The LoadVarargs limit is based on
profiling and the phantom arguments array is a proof; in most cases the profiling is more
conservative.
But, you could write some crazy code where the statically obvious arguments array value is
bigger than what the profiling would have told you. When this happens, this promotion
effectively removes a bounds check. This either results in us clobbering a bunch of stack,
or it means that we never initialize a region of the stack that a later operation will read
(the uninitialization happens because PutStackSinkingPhase removes PutStacks that appear
unnecessary, and a GetMyArgumentByVal will claim not to use the region of the stack outside
the original LoadVarargs limit).
* dfg/DFGArgumentsEliminationPhase.cpp:
* tests/stress/load-varargs-elimination-bounds-check-barely.js: Added.
(foo):
(bar):
(baz):
* tests/stress/load-varargs-elimination-bounds-check.js: Added.
(foo):
(bar):
(baz):
LayoutTests:
* js/regress/load-varargs-elimination-expected.txt: Added.
* js/regress/load-varargs-elimination.html: Added.
* js/regress/script-tests/load-varargs-elimination.js: Added.
(foo):
(bar):
(baz):
* js/regress/sink-huge-activation-expected.txt: Added.
* js/regress/sink-huge-activation.html: Added.</pre>
<h3>Modified Paths</h3>
<ul>
<li><a href="#trunkLayoutTestsChangeLog">trunk/LayoutTests/ChangeLog</a></li>
<li><a href="#trunkSourceJavaScriptCoreChangeLog">trunk/Source/JavaScriptCore/ChangeLog</a></li>
<li><a href="#trunkSourceJavaScriptCoredfgDFGArgumentsEliminationPhasecpp">trunk/Source/JavaScriptCore/dfg/DFGArgumentsEliminationPhase.cpp</a></li>
</ul>
<h3>Added Paths</h3>
<ul>
<li><a href="#trunkLayoutTestsjsregressloadvarargseliminationexpectedtxt">trunk/LayoutTests/js/regress/load-varargs-elimination-expected.txt</a></li>
<li><a href="#trunkLayoutTestsjsregressloadvarargseliminationhtml">trunk/LayoutTests/js/regress/load-varargs-elimination.html</a></li>
<li><a href="#trunkLayoutTestsjsregressscripttestsloadvarargseliminationjs">trunk/LayoutTests/js/regress/script-tests/load-varargs-elimination.js</a></li>
<li><a href="#trunkLayoutTestsjsregresssinkhugeactivationexpectedtxt">trunk/LayoutTests/js/regress/sink-huge-activation-expected.txt</a></li>
<li><a href="#trunkLayoutTestsjsregresssinkhugeactivationhtml">trunk/LayoutTests/js/regress/sink-huge-activation.html</a></li>
<li><a href="#trunkSourceJavaScriptCoretestsstressloadvarargseliminationboundscheckbarelyjs">trunk/Source/JavaScriptCore/tests/stress/load-varargs-elimination-bounds-check-barely.js</a></li>
<li><a href="#trunkSourceJavaScriptCoretestsstressloadvarargseliminationboundscheckjs">trunk/Source/JavaScriptCore/tests/stress/load-varargs-elimination-bounds-check.js</a></li>
</ul>
</div>
<div id="patch">
<h3>Diff</h3>
<a id="trunkLayoutTestsChangeLog"></a>
<div class="modfile"><h4>Modified: trunk/LayoutTests/ChangeLog (184109 => 184110)</h4>
<pre class="diff"><span>
<span class="info">--- trunk/LayoutTests/ChangeLog 2015-05-11 18:02:57 UTC (rev 184109)
+++ trunk/LayoutTests/ChangeLog 2015-05-11 18:18:10 UTC (rev 184110)
</span><span class="lines">@@ -1,3 +1,19 @@
</span><ins>+2015-05-10 Filip Pizlo <fpizlo@apple.com>
+
+ We shouldn't promote LoadVarargs to a sequence of GetStacks and PutStacks if doing so would exceed the LoadVarargs' limit
+ https://bugs.webkit.org/show_bug.cgi?id=144851
+
+ Reviewed by Michael Saboff.
+
+ * js/regress/load-varargs-elimination-expected.txt: Added.
+ * js/regress/load-varargs-elimination.html: Added.
+ * js/regress/script-tests/load-varargs-elimination.js: Added.
+ (foo):
+ (bar):
+ (baz):
+ * js/regress/sink-huge-activation-expected.txt: Added.
+ * js/regress/sink-huge-activation.html: Added.
+
</ins><span class="cx"> 2015-05-11 Marcos ChavarrÃa Teijeiro <chavarria1991@gmail.com>
</span><span class="cx">
</span><span class="cx"> [GTK] Gardening 11th May.
</span></span></pre></div>
<a id="trunkLayoutTestsjsregressloadvarargseliminationexpectedtxt"></a>
<div class="addfile"><h4>Added: trunk/LayoutTests/js/regress/load-varargs-elimination-expected.txt (0 => 184110)</h4>
<pre class="diff"><span>
<span class="info">--- trunk/LayoutTests/js/regress/load-varargs-elimination-expected.txt (rev 0)
+++ trunk/LayoutTests/js/regress/load-varargs-elimination-expected.txt 2015-05-11 18:18:10 UTC (rev 184110)
</span><span class="lines">@@ -0,0 +1,10 @@
</span><ins>+JSRegress/load-varargs-elimination
+
+On success, you will see a series of "PASS" messages, followed by "TEST COMPLETE".
+
+
+PASS no exception thrown
+PASS successfullyParsed is true
+
+TEST COMPLETE
+
</ins></span></pre></div>
<a id="trunkLayoutTestsjsregressloadvarargseliminationhtml"></a>
<div class="addfile"><h4>Added: trunk/LayoutTests/js/regress/load-varargs-elimination.html (0 => 184110)</h4>
<pre class="diff"><span>
<span class="info">--- trunk/LayoutTests/js/regress/load-varargs-elimination.html (rev 0)
+++ trunk/LayoutTests/js/regress/load-varargs-elimination.html 2015-05-11 18:18:10 UTC (rev 184110)
</span><span class="lines">@@ -0,0 +1,12 @@
</span><ins>+<!DOCTYPE HTML PUBLIC "-//IETF//DTD HTML//EN">
+<html>
+<head>
+<script src="../../resources/js-test-pre.js"></script>
+</head>
+<body>
+<script src="../../resources/regress-pre.js"></script>
+<script src="script-tests/load-varargs-elimination.js"></script>
+<script src="../../resources/regress-post.js"></script>
+<script src="../../resources/js-test-post.js"></script>
+</body>
+</html>
</ins></span></pre></div>
<a id="trunkLayoutTestsjsregressscripttestsloadvarargseliminationjs"></a>
<div class="addfile"><h4>Added: trunk/LayoutTests/js/regress/script-tests/load-varargs-elimination.js (0 => 184110)</h4>
<pre class="diff"><span>
<span class="info">--- trunk/LayoutTests/js/regress/script-tests/load-varargs-elimination.js (rev 0)
+++ trunk/LayoutTests/js/regress/script-tests/load-varargs-elimination.js 2015-05-11 18:18:10 UTC (rev 184110)
</span><span class="lines">@@ -0,0 +1,23 @@
</span><ins>+function foo() {
+ var result = 0;
+ for (var i = 0; i < arguments.length; ++i)
+ result += arguments[i];
+ return result;
+}
+
+function bar() {
+ return foo.apply(this, arguments);
+}
+
+function baz() {
+ return bar(1, 2, 3, 4);
+}
+
+noInline(baz);
+
+for (var i = 0; i < 1000000; ++i) {
+ var result = baz();
+ if (result != 10)
+ throw "Error: bad result: " + result;
+}
+
</ins></span></pre></div>
<a id="trunkLayoutTestsjsregresssinkhugeactivationexpectedtxt"></a>
<div class="addfile"><h4>Added: trunk/LayoutTests/js/regress/sink-huge-activation-expected.txt (0 => 184110)</h4>
<pre class="diff"><span>
<span class="info">--- trunk/LayoutTests/js/regress/sink-huge-activation-expected.txt (rev 0)
+++ trunk/LayoutTests/js/regress/sink-huge-activation-expected.txt 2015-05-11 18:18:10 UTC (rev 184110)
</span><span class="lines">@@ -0,0 +1,10 @@
</span><ins>+JSRegress/sink-huge-activation
+
+On success, you will see a series of "PASS" messages, followed by "TEST COMPLETE".
+
+
+PASS no exception thrown
+PASS successfullyParsed is true
+
+TEST COMPLETE
+
</ins></span></pre></div>
<a id="trunkLayoutTestsjsregresssinkhugeactivationhtml"></a>
<div class="addfile"><h4>Added: trunk/LayoutTests/js/regress/sink-huge-activation.html (0 => 184110)</h4>
<pre class="diff"><span>
<span class="info">--- trunk/LayoutTests/js/regress/sink-huge-activation.html (rev 0)
+++ trunk/LayoutTests/js/regress/sink-huge-activation.html 2015-05-11 18:18:10 UTC (rev 184110)
</span><span class="lines">@@ -0,0 +1,12 @@
</span><ins>+<!DOCTYPE HTML PUBLIC "-//IETF//DTD HTML//EN">
+<html>
+<head>
+<script src="../../resources/js-test-pre.js"></script>
+</head>
+<body>
+<script src="../../resources/regress-pre.js"></script>
+<script src="script-tests/sink-huge-activation.js"></script>
+<script src="../../resources/regress-post.js"></script>
+<script src="../../resources/js-test-post.js"></script>
+</body>
+</html>
</ins></span></pre></div>
<a id="trunkSourceJavaScriptCoreChangeLog"></a>
<div class="modfile"><h4>Modified: trunk/Source/JavaScriptCore/ChangeLog (184109 => 184110)</h4>
<pre class="diff"><span>
<span class="info">--- trunk/Source/JavaScriptCore/ChangeLog 2015-05-11 18:02:57 UTC (rev 184109)
+++ trunk/Source/JavaScriptCore/ChangeLog 2015-05-11 18:18:10 UTC (rev 184110)
</span><span class="lines">@@ -1,3 +1,40 @@
</span><ins>+2015-05-10 Filip Pizlo <fpizlo@apple.com>
+
+ We shouldn't promote LoadVarargs to a sequence of GetStacks and PutStacks if doing so would exceed the LoadVarargs' limit
+ https://bugs.webkit.org/show_bug.cgi?id=144851
+
+ Reviewed by Michael Saboff.
+
+ LoadVarargs loads arguments from some object and puts them on the stack. The region of
+ stack is controlled by a bunch of meta-data, including InlineCallFrame. InlineCallFrame
+ shouldn't really be edited after ByteCodeParser, so we cannot convert LoadVarargs to
+ something that uses more stack than the LoadVarargs wanted to.
+
+ This check was missing in the ArgumentsEliminationPhase's LoadVarargs->GetStack+PutStack
+ promoter. This is an important promotion rule for performance, and in cases where we are
+ compiling truly hot code, the LoadVarargs limit will be at least as big as the length of
+ the phantom arguments array that this phase sees. The LoadVarargs limit is based on
+ profiling and the phantom arguments array is a proof; in most cases the profiling is more
+ conservative.
+
+ But, you could write some crazy code where the statically obvious arguments array value is
+ bigger than what the profiling would have told you. When this happens, this promotion
+ effectively removes a bounds check. This either results in us clobbering a bunch of stack,
+ or it means that we never initialize a region of the stack that a later operation will read
+ (the uninitialization happens because PutStackSinkingPhase removes PutStacks that appear
+ unnecessary, and a GetMyArgumentByVal will claim not to use the region of the stack outside
+ the original LoadVarargs limit).
+
+ * dfg/DFGArgumentsEliminationPhase.cpp:
+ * tests/stress/load-varargs-elimination-bounds-check-barely.js: Added.
+ (foo):
+ (bar):
+ (baz):
+ * tests/stress/load-varargs-elimination-bounds-check.js: Added.
+ (foo):
+ (bar):
+ (baz):
+
</ins><span class="cx"> 2015-05-11 Andreas Kling <akling@apple.com>
</span><span class="cx">
</span><span class="cx"> JSON.stringify shouldn't use generic get() to access Array.length
</span></span></pre></div>
<a id="trunkSourceJavaScriptCoredfgDFGArgumentsEliminationPhasecpp"></a>
<div class="modfile"><h4>Modified: trunk/Source/JavaScriptCore/dfg/DFGArgumentsEliminationPhase.cpp (184109 => 184110)</h4>
<pre class="diff"><span>
<span class="info">--- trunk/Source/JavaScriptCore/dfg/DFGArgumentsEliminationPhase.cpp 2015-05-11 18:02:57 UTC (rev 184109)
+++ trunk/Source/JavaScriptCore/dfg/DFGArgumentsEliminationPhase.cpp 2015-05-11 18:18:10 UTC (rev 184110)
</span><span class="lines">@@ -450,7 +450,9 @@
</span><span class="cx">
</span><span class="cx"> LoadVarargsData* varargsData = node->loadVarargsData();
</span><span class="cx"> InlineCallFrame* inlineCallFrame = candidate->origin.semantic.inlineCallFrame;
</span><del>- if (inlineCallFrame && !inlineCallFrame->isVarargs()) {
</del><ins>+ if (inlineCallFrame
+ && !inlineCallFrame->isVarargs()
+ && inlineCallFrame->arguments.size() - varargsData->offset <= varargsData->limit) {
</ins><span class="cx"> Node* argumentCount = insertionSet.insertConstant(
</span><span class="cx"> nodeIndex, node->origin,
</span><span class="cx"> jsNumber(inlineCallFrame->arguments.size() - varargsData->offset));
</span></span></pre></div>
<a id="trunkSourceJavaScriptCoretestsstressloadvarargseliminationboundscheckbarelyjs"></a>
<div class="addfile"><h4>Added: trunk/Source/JavaScriptCore/tests/stress/load-varargs-elimination-bounds-check-barely.js (0 => 184110)</h4>
<pre class="diff"><span>
<span class="info">--- trunk/Source/JavaScriptCore/tests/stress/load-varargs-elimination-bounds-check-barely.js (rev 0)
+++ trunk/Source/JavaScriptCore/tests/stress/load-varargs-elimination-bounds-check-barely.js 2015-05-11 18:18:10 UTC (rev 184110)
</span><span class="lines">@@ -0,0 +1,35 @@
</span><ins>+function foo() {
+ var result = 0;
+ for (var i = 0; i < arguments.length; ++i)
+ result += arguments[i];
+ return result;
+}
+
+function bar() {
+ return foo.apply(this, arguments);
+}
+
+function baz(p) {
+ if (p)
+ return bar(1, 42);
+ return 0;
+}
+
+noInline(baz);
+
+// Execute baz() once with p set, so that the call has a valid prediction.
+baz(true);
+
+// Warm up profiling in bar and foo. Convince this profiling that bar()'s varargs call will tend to
+// pass a small number of arguments;
+for (var i = 0; i < 1000; ++i)
+ bar(1);
+
+// Now compile baz(), but don't run the bad code yet.
+for (var i = 0; i < 10000; ++i)
+ baz(false);
+
+// Finally, trigger the bug.
+var result = baz(true);
+if (result != 43)
+ throw "Error: bad result: " + result;
</ins></span></pre></div>
<a id="trunkSourceJavaScriptCoretestsstressloadvarargseliminationboundscheckjs"></a>
<div class="addfile"><h4>Added: trunk/Source/JavaScriptCore/tests/stress/load-varargs-elimination-bounds-check.js (0 => 184110)</h4>
<pre class="diff"><span>
<span class="info">--- trunk/Source/JavaScriptCore/tests/stress/load-varargs-elimination-bounds-check.js (rev 0)
+++ trunk/Source/JavaScriptCore/tests/stress/load-varargs-elimination-bounds-check.js 2015-05-11 18:18:10 UTC (rev 184110)
</span><span class="lines">@@ -0,0 +1,35 @@
</span><ins>+function foo() {
+ var result = 0;
+ for (var i = 0; i < arguments.length; ++i)
+ result += arguments[i];
+ return result;
+}
+
+function bar() {
+ return foo.apply(this, arguments);
+}
+
+function baz(p) {
+ if (p)
+ return bar(1, 2, 3, 4);
+ return 0;
+}
+
+noInline(baz);
+
+// Execute baz() once with p set, so that the call has a valid prediction.
+baz(true);
+
+// Warm up profiling in bar and foo. Convince this profiling that bar()'s varargs call will tend to
+// pass a small number of arguments;
+for (var i = 0; i < 1000; ++i)
+ bar(1);
+
+// Now compile baz(), but don't run the bad code yet.
+for (var i = 0; i < 10000; ++i)
+ baz(false);
+
+// Finally, trigger the bug.
+var result = baz(true);
+if (result != 10)
+ throw "Error: bad result: " + result;
</ins></span></pre>
</div>
</div>
</body>
</html>