<!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>[200897] 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/200897">200897</a></dd>
<dt>Author</dt> <dd>benjamin@webkit.org</dd>
<dt>Date</dt> <dd>2016-05-13 16:45:15 -0700 (Fri, 13 May 2016)</dd>
</dl>
<h3>Log Message</h3>
<pre>[JSC] Move the CheckTierUp function calls out of the main path
https://bugs.webkit.org/show_bug.cgi?id=157668
Patch by Benjamin Poulain <bpoulain@apple.com> on 2016-05-13
Reviewed by Mark Lam.
If you have a tiny tiny loop (for example, Sunspider's bits-in-byte),
the size of CheckTierUp is a problem.
On multi-issue CPUs, the node is so big that we do not
get to run anything from the loop in the instruction fetch.
On x86, having a bigger loop also pushes us out of the LSD.
This is a 6% improvement on bits-in-byte. Other Sunspider tests
only improves marginally.
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::addSlowPathGenerator):
(JSC::DFG::SpeculativeJIT::runSlowPathGenerators):
* dfg/DFGSpeculativeJIT.h:
(JSC::DFG::SpeculativeJIT::silentSpill):
(JSC::DFG::SpeculativeJIT::silentFill):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::compile):</pre>
<h3>Modified Paths</h3>
<ul>
<li><a href="#trunkSourceJavaScriptCoreChangeLog">trunk/Source/JavaScriptCore/ChangeLog</a></li>
<li><a href="#trunkSourceJavaScriptCoredfgDFGSpeculativeJITcpp">trunk/Source/JavaScriptCore/dfg/DFGSpeculativeJIT.cpp</a></li>
<li><a href="#trunkSourceJavaScriptCoredfgDFGSpeculativeJITh">trunk/Source/JavaScriptCore/dfg/DFGSpeculativeJIT.h</a></li>
<li><a href="#trunkSourceJavaScriptCoredfgDFGSpeculativeJIT64cpp">trunk/Source/JavaScriptCore/dfg/DFGSpeculativeJIT64.cpp</a></li>
</ul>
</div>
<div id="patch">
<h3>Diff</h3>
<a id="trunkSourceJavaScriptCoreChangeLog"></a>
<div class="modfile"><h4>Modified: trunk/Source/JavaScriptCore/ChangeLog (200896 => 200897)</h4>
<pre class="diff"><span>
<span class="info">--- trunk/Source/JavaScriptCore/ChangeLog 2016-05-13 23:44:31 UTC (rev 200896)
+++ trunk/Source/JavaScriptCore/ChangeLog 2016-05-13 23:45:15 UTC (rev 200897)
</span><span class="lines">@@ -1,5 +1,32 @@
</span><span class="cx"> 2016-05-13 Benjamin Poulain <bpoulain@apple.com>
</span><span class="cx">
</span><ins>+ [JSC] Move the CheckTierUp function calls out of the main path
+ https://bugs.webkit.org/show_bug.cgi?id=157668
+
+ Reviewed by Mark Lam.
+
+ If you have a tiny tiny loop (for example, Sunspider's bits-in-byte),
+ the size of CheckTierUp is a problem.
+
+ On multi-issue CPUs, the node is so big that we do not
+ get to run anything from the loop in the instruction fetch.
+
+ On x86, having a bigger loop also pushes us out of the LSD.
+
+ This is a 6% improvement on bits-in-byte. Other Sunspider tests
+ only improves marginally.
+
+ * dfg/DFGSpeculativeJIT.cpp:
+ (JSC::DFG::SpeculativeJIT::addSlowPathGenerator):
+ (JSC::DFG::SpeculativeJIT::runSlowPathGenerators):
+ * dfg/DFGSpeculativeJIT.h:
+ (JSC::DFG::SpeculativeJIT::silentSpill):
+ (JSC::DFG::SpeculativeJIT::silentFill):
+ * dfg/DFGSpeculativeJIT64.cpp:
+ (JSC::DFG::SpeculativeJIT::compile):
+
+2016-05-13 Benjamin Poulain <bpoulain@apple.com>
+
</ins><span class="cx"> [JSC] Emit the loads of emitLoadWithStructureCheck() in the order they are used
</span><span class="cx"> https://bugs.webkit.org/show_bug.cgi?id=157671
</span><span class="cx">
</span></span></pre></div>
<a id="trunkSourceJavaScriptCoredfgDFGSpeculativeJITcpp"></a>
<div class="modfile"><h4>Modified: trunk/Source/JavaScriptCore/dfg/DFGSpeculativeJIT.cpp (200896 => 200897)</h4>
<pre class="diff"><span>
<span class="info">--- trunk/Source/JavaScriptCore/dfg/DFGSpeculativeJIT.cpp 2016-05-13 23:44:31 UTC (rev 200896)
+++ trunk/Source/JavaScriptCore/dfg/DFGSpeculativeJIT.cpp 2016-05-13 23:45:15 UTC (rev 200897)
</span><span class="lines">@@ -354,12 +354,21 @@
</span><span class="cx"> m_slowPathGenerators.append(WTFMove(slowPathGenerator));
</span><span class="cx"> }
</span><span class="cx">
</span><ins>+void SpeculativeJIT::addSlowPathGenerator(std::function<void()> lambda)
+{
+ m_slowPathLambdas.append(std::make_pair(lambda, m_origin.semantic));
+}
+
</ins><span class="cx"> void SpeculativeJIT::runSlowPathGenerators(PCToCodeOriginMapBuilder& pcToCodeOriginMapBuilder)
</span><span class="cx"> {
</span><del>- for (unsigned i = 0; i < m_slowPathGenerators.size(); ++i) {
- pcToCodeOriginMapBuilder.appendItem(m_jit.label(), m_slowPathGenerators[i]->origin().semantic);
- m_slowPathGenerators[i]->generate(this);
</del><ins>+ for (auto& slowPathGenerator : m_slowPathGenerators) {
+ pcToCodeOriginMapBuilder.appendItem(m_jit.label(), slowPathGenerator->origin().semantic);
+ slowPathGenerator->generate(this);
</ins><span class="cx"> }
</span><ins>+ for (auto& generatorPair : m_slowPathLambdas) {
+ pcToCodeOriginMapBuilder.appendItem(m_jit.label(), generatorPair.second);
+ generatorPair.first();
+ }
</ins><span class="cx"> }
</span><span class="cx">
</span><span class="cx"> void SpeculativeJIT::clearGenerationInfo()
</span></span></pre></div>
<a id="trunkSourceJavaScriptCoredfgDFGSpeculativeJITh"></a>
<div class="modfile"><h4>Modified: trunk/Source/JavaScriptCore/dfg/DFGSpeculativeJIT.h (200896 => 200897)</h4>
<pre class="diff"><span>
<span class="info">--- trunk/Source/JavaScriptCore/dfg/DFGSpeculativeJIT.h 2016-05-13 23:44:31 UTC (rev 200896)
+++ trunk/Source/JavaScriptCore/dfg/DFGSpeculativeJIT.h 2016-05-13 23:45:15 UTC (rev 200897)
</span><span class="lines">@@ -319,6 +319,7 @@
</span><span class="cx"> GeneratedOperandType checkGeneratedTypeForToInt32(Node*);
</span><span class="cx">
</span><span class="cx"> void addSlowPathGenerator(std::unique_ptr<SlowPathGenerator>);
</span><ins>+ void addSlowPathGenerator(std::function<void()>);
</ins><span class="cx"> void runSlowPathGenerators(PCToCodeOriginMapBuilder&);
</span><span class="cx">
</span><span class="cx"> void compile(Node*);
</span><span class="lines">@@ -339,8 +340,23 @@
</span><span class="cx"> SilentRegisterSavePlan silentSavePlanForFPR(VirtualRegister spillMe, FPRReg source);
</span><span class="cx"> void silentSpill(const SilentRegisterSavePlan&);
</span><span class="cx"> void silentFill(const SilentRegisterSavePlan&, GPRReg canTrample);
</span><del>-
</del><ins>+
</ins><span class="cx"> template<typename CollectionType>
</span><ins>+ void silentSpill(const CollectionType& savePlans)
+ {
+ for (unsigned i = 0; i < savePlans.size(); ++i)
+ silentSpill(savePlans[i]);
+ }
+
+ template<typename CollectionType>
+ void silentFill(const CollectionType& savePlans, GPRReg exclude = InvalidGPRReg)
+ {
+ GPRReg canTrample = SpeculativeJIT::pickCanTrample(exclude);
+ for (unsigned i = savePlans.size(); i--;)
+ silentFill(savePlans[i], canTrample);
+ }
+
+ template<typename CollectionType>
</ins><span class="cx"> void silentSpillAllRegistersImpl(bool doSpill, CollectionType& plans, GPRReg exclude, GPRReg exclude2 = InvalidGPRReg, FPRReg fprExclude = InvalidFPRReg)
</span><span class="cx"> {
</span><span class="cx"> ASSERT(plans.isEmpty());
</span><span class="lines">@@ -2812,6 +2828,7 @@
</span><span class="cx"> MinifiedGraph* m_minifiedGraph;
</span><span class="cx">
</span><span class="cx"> Vector<std::unique_ptr<SlowPathGenerator>, 8> m_slowPathGenerators;
</span><ins>+ Vector<std::pair<std::function<void()>, CodeOrigin>, 8> m_slowPathLambdas;
</ins><span class="cx"> Vector<SilentRegisterSavePlan> m_plans;
</span><span class="cx"> unsigned m_outOfLineStreamIndex { UINT_MAX };
</span><span class="cx"> };
</span></span></pre></div>
<a id="trunkSourceJavaScriptCoredfgDFGSpeculativeJIT64cpp"></a>
<div class="modfile"><h4>Modified: trunk/Source/JavaScriptCore/dfg/DFGSpeculativeJIT64.cpp (200896 => 200897)</h4>
<pre class="diff"><span>
<span class="info">--- trunk/Source/JavaScriptCore/dfg/DFGSpeculativeJIT64.cpp 2016-05-13 23:44:31 UTC (rev 200896)
+++ trunk/Source/JavaScriptCore/dfg/DFGSpeculativeJIT64.cpp 2016-05-13 23:45:15 UTC (rev 200897)
</span><span class="lines">@@ -5166,18 +5166,27 @@
</span><span class="cx">
</span><span class="cx"> #if ENABLE(FTL_JIT)
</span><span class="cx"> case CheckTierUpInLoop: {
</span><del>- MacroAssembler::Jump done = m_jit.branchAdd32(
- MacroAssembler::Signed,
</del><ins>+ MacroAssembler::Jump callTierUp = m_jit.branchAdd32(
+ MacroAssembler::PositiveOrZero,
</ins><span class="cx"> TrustedImm32(Options::ftlTierUpCounterIncrementForLoop()),
</span><span class="cx"> MacroAssembler::AbsoluteAddress(&m_jit.jitCode()->tierUpCounter.m_counter));
</span><del>-
- silentSpillAllRegisters(InvalidGPRReg);
- m_jit.setupArgumentsWithExecState(
- TrustedImm32(node->origin.semantic.bytecodeIndex));
- appendCall(triggerTierUpNowInLoop);
- silentFillAllRegisters(InvalidGPRReg);
-
- done.link(&m_jit);
</del><ins>+
+ MacroAssembler::Label toNextOperation = m_jit.label();
+
+ Vector<SilentRegisterSavePlan> savePlans;
+ silentSpillAllRegistersImpl(false, savePlans, InvalidGPRReg);
+ unsigned bytecodeIndex = node->origin.semantic.bytecodeIndex;
+
+ addSlowPathGenerator([=]() {
+ callTierUp.link(&m_jit);
+
+ silentSpill(savePlans);
+ m_jit.setupArgumentsWithExecState(TrustedImm32(bytecodeIndex));
+ appendCall(triggerTierUpNowInLoop);
+ silentFill(savePlans);
+
+ m_jit.jump().linkTo(toNextOperation, &m_jit);
+ });
</ins><span class="cx"> break;
</span><span class="cx"> }
</span><span class="cx">
</span><span class="lines">@@ -5198,7 +5207,7 @@
</span><span class="cx">
</span><span class="cx"> case CheckTierUpAndOSREnter: {
</span><span class="cx"> ASSERT(!node->origin.semantic.inlineCallFrame);
</span><del>-
</del><ins>+
</ins><span class="cx"> GPRTemporary temp(this);
</span><span class="cx"> GPRReg tempGPR = temp.gpr();
</span><span class="cx">
</span><span class="lines">@@ -5206,26 +5215,39 @@
</span><span class="cx"> auto triggerIterator = m_jit.jitCode()->tierUpEntryTriggers.find(bytecodeIndex);
</span><span class="cx"> RELEASE_ASSERT(triggerIterator != m_jit.jitCode()->tierUpEntryTriggers.end());
</span><span class="cx"> uint8_t* forceEntryTrigger = &(m_jit.jitCode()->tierUpEntryTriggers.find(bytecodeIndex)->value);
</span><ins>+
</ins><span class="cx"> MacroAssembler::Jump forceOSREntry = m_jit.branchTest8(MacroAssembler::NonZero, MacroAssembler::AbsoluteAddress(forceEntryTrigger));
</span><del>-
- MacroAssembler::Jump done = m_jit.branchAdd32(
- MacroAssembler::Signed,
</del><ins>+ MacroAssembler::Jump overflowedCounter = m_jit.branchAdd32(
+ MacroAssembler::PositiveOrZero,
</ins><span class="cx"> TrustedImm32(Options::ftlTierUpCounterIncrementForLoop()),
</span><span class="cx"> MacroAssembler::AbsoluteAddress(&m_jit.jitCode()->tierUpCounter.m_counter));
</span><ins>+ MacroAssembler::Label toNextOperation = m_jit.label();
</ins><span class="cx">
</span><del>- forceOSREntry.link(&m_jit);
- silentSpillAllRegisters(tempGPR);
</del><ins>+ Vector<SilentRegisterSavePlan> savePlans;
+ silentSpillAllRegistersImpl(false, savePlans, tempGPR);
+
</ins><span class="cx"> unsigned streamIndex = m_stream->size();
</span><span class="cx"> m_jit.jitCode()->bytecodeIndexToStreamIndex.add(bytecodeIndex, streamIndex);
</span><del>- m_jit.setupArgumentsWithExecState(TrustedImm32(bytecodeIndex));
- appendCallSetResult(triggerOSREntryNow, tempGPR);
- MacroAssembler::Jump dontEnter = m_jit.branchTestPtr(MacroAssembler::Zero, tempGPR);
- m_jit.emitRestoreCalleeSaves();
- m_jit.jump(tempGPR);
- dontEnter.link(&m_jit);
- silentFillAllRegisters(tempGPR);
-
- done.link(&m_jit);
</del><ins>+
+ addSlowPathGenerator([=]() {
+ forceOSREntry.link(&m_jit);
+ overflowedCounter.link(&m_jit);
+
+ silentSpill(savePlans);
+ m_jit.setupArgumentsWithExecState(TrustedImm32(bytecodeIndex));
+ appendCallSetResult(triggerOSREntryNow, tempGPR);
+
+ if (savePlans.isEmpty())
+ m_jit.branchTestPtr(MacroAssembler::Zero, tempGPR).linkTo(toNextOperation, &m_jit);
+ else {
+ MacroAssembler::Jump osrEnter = m_jit.branchTestPtr(MacroAssembler::NonZero, tempGPR);
+ silentFill(savePlans);
+ m_jit.jump().linkTo(toNextOperation, &m_jit);
+ osrEnter.link(&m_jit);
+ }
+ m_jit.emitRestoreCalleeSaves();
+ m_jit.jump(tempGPR);
+ });
</ins><span class="cx"> break;
</span><span class="cx"> }
</span><span class="cx"> #else // ENABLE(FTL_JIT)
</span></span></pre>
</div>
</div>
</body>
</html>