Title: [237553] trunk
- Revision
- 237553
- Author
- [email protected]
- Date
- 2018-10-29 08:46:14 -0700 (Mon, 29 Oct 2018)
Log Message
Add support for GetStack FlushedDouble
https://bugs.webkit.org/show_bug.cgi?id=191012
<rdar://problem/45265141>
Reviewed by Saam Barati.
JSTests:
* stress/get-stack-double.js: Added.
(bar):
(noInline):
Source/_javascript_Core:
LowerDFGToB3::compileGetStack assumed that we would not emit GetStack
for doubles, but it turns out it may arise from the PutStack sinking
phase: if we sink a PutStack into a successor block, other predecessors
will emit a GetStack followed by a Upsilon.
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileGetStack):
Modified Paths
Added Paths
Diff
Modified: trunk/JSTests/ChangeLog (237552 => 237553)
--- trunk/JSTests/ChangeLog 2018-10-29 14:53:14 UTC (rev 237552)
+++ trunk/JSTests/ChangeLog 2018-10-29 15:46:14 UTC (rev 237553)
@@ -1,5 +1,17 @@
2018-10-29 Tadeu Zagallo <[email protected]>
+ Add support for GetStack FlushedDouble
+ https://bugs.webkit.org/show_bug.cgi?id=191012
+ <rdar://problem/45265141>
+
+ Reviewed by Saam Barati.
+
+ * stress/get-stack-double.js: Added.
+ (bar):
+ (noInline):
+
+2018-10-29 Tadeu Zagallo <[email protected]>
+
New bytecode format for JSC
https://bugs.webkit.org/show_bug.cgi?id=187373
<rdar://problem/44186758>
Added: trunk/JSTests/stress/get-stack-double.js (0 => 237553)
--- trunk/JSTests/stress/get-stack-double.js (rev 0)
+++ trunk/JSTests/stress/get-stack-double.js 2018-10-29 15:46:14 UTC (rev 237553)
@@ -0,0 +1,11 @@
+function bar(n)
+{
+ for (p = 0; p < 30; p++)
+ if (p + 0.1)
+ n -= 0.2
+}
+
+for (var i = 0; i < 100000; ++i)
+ bar(0);
+
+function noInline() { }
Modified: trunk/Source/_javascript_Core/ChangeLog (237552 => 237553)
--- trunk/Source/_javascript_Core/ChangeLog 2018-10-29 14:53:14 UTC (rev 237552)
+++ trunk/Source/_javascript_Core/ChangeLog 2018-10-29 15:46:14 UTC (rev 237553)
@@ -1,5 +1,21 @@
2018-10-29 Tadeu Zagallo <[email protected]>
+ Add support for GetStack FlushedDouble
+ https://bugs.webkit.org/show_bug.cgi?id=191012
+ <rdar://problem/45265141>
+
+ Reviewed by Saam Barati.
+
+ LowerDFGToB3::compileGetStack assumed that we would not emit GetStack
+ for doubles, but it turns out it may arise from the PutStack sinking
+ phase: if we sink a PutStack into a successor block, other predecessors
+ will emit a GetStack followed by a Upsilon.
+
+ * ftl/FTLLowerDFGToB3.cpp:
+ (JSC::FTL::DFG::LowerDFGToB3::compileGetStack):
+
+2018-10-29 Tadeu Zagallo <[email protected]>
+
New bytecode format for JSC
https://bugs.webkit.org/show_bug.cgi?id=187373
<rdar://problem/44186758>
Modified: trunk/Source/_javascript_Core/ftl/FTLLowerDFGToB3.cpp (237552 => 237553)
--- trunk/Source/_javascript_Core/ftl/FTLLowerDFGToB3.cpp 2018-10-29 14:53:14 UTC (rev 237552)
+++ trunk/Source/_javascript_Core/ftl/FTLLowerDFGToB3.cpp 2018-10-29 15:46:14 UTC (rev 237553)
@@ -1737,9 +1737,10 @@
AbstractValue& value = m_state.operand(data->local);
DFG_ASSERT(m_graph, m_node, isConcrete(data->format), data->format);
- DFG_ASSERT(m_graph, m_node, data->format != FlushedDouble, data->format); // This just happens to not arise for GetStacks, right now. It would be trivial to support.
- if (isInt32Speculation(value.m_type))
+ if (data->format == FlushedDouble)
+ setDouble(m_out.loadDouble(addressFor(data->machineLocal)));
+ else if (isInt32Speculation(value.m_type))
setInt32(m_out.load32(payloadFor(data->machineLocal)));
else
setJSValue(m_out.load64(addressFor(data->machineLocal)));
_______________________________________________
webkit-changes mailing list
[email protected]
https://lists.webkit.org/mailman/listinfo/webkit-changes
