Title: [191515] branches/safari-601-branch

Diff

Modified: branches/safari-601-branch/Source/_javascript_Core/ChangeLog (191514 => 191515)


--- branches/safari-601-branch/Source/_javascript_Core/ChangeLog	2015-10-23 22:45:34 UTC (rev 191514)
+++ branches/safari-601-branch/Source/_javascript_Core/ChangeLog	2015-10-23 23:03:15 UTC (rev 191515)
@@ -1,5 +1,24 @@
 2015-10-22  Matthew Hanson  <[email protected]>
 
+        Rollout r191395. rdar://problem/22847057
+
+    2015-10-21  Filip Pizlo  <[email protected]>
+
+            Failures in PutStackSinkingPhase should be less severe
+            https://bugs.webkit.org/show_bug.cgi?id=150400
+
+            Reviewed by Geoffrey Garen.
+
+            Make the PutStackSinkingPhase abort instead of asserting. To test that it's OK to not have
+            PutStackSinkingPhase run, this adds a test mode where we run without PutStackSinkingPhase.
+
+            * dfg/DFGPlan.cpp: Make it possible to not run PutStackSinkingPhase for tests.
+            (JSC::DFG::Plan::compileInThreadImpl):
+            * dfg/DFGPutStackSinkingPhase.cpp: PutStackSinkingPhase should abort instead of asserting, except when validation is enabled.
+            * runtime/Options.h: Add an option for disabling PutStackSinkingPhase.
+
+2015-10-22  Matthew Hanson  <[email protected]>
+
         Merge r191395. rdar://problem/22847057
 
     2015-10-21  Filip Pizlo  <[email protected]>
@@ -319,7 +338,7 @@
             (main):
             (CommandLine::parseArguments):
             (jscmain):
-            * runtime/Options.cpp: 
+            * runtime/Options.cpp:
             (JSC::OptionRange::dump): I needed a dump method on Options members when debugging this.
             (JSC::Options::initialize): This can now be called more than once.
             * runtime/Options.h:

Modified: branches/safari-601-branch/Source/_javascript_Core/dfg/DFGPlan.cpp (191514 => 191515)


--- branches/safari-601-branch/Source/_javascript_Core/dfg/DFGPlan.cpp	2015-10-23 22:45:34 UTC (rev 191514)
+++ branches/safari-601-branch/Source/_javascript_Core/dfg/DFGPlan.cpp	2015-10-23 23:03:15 UTC (rev 191515)
@@ -354,8 +354,7 @@
         
         // Ideally, these would be run to fixpoint with the object allocation sinking phase.
         performArgumentsElimination(dfg);
-        if (Options::usePutStackSinking())
-            performPutStackSinking(dfg);
+        performPutStackSinking(dfg);
         
         performConstantHoisting(dfg);
         performGlobalCSE(dfg);

Modified: branches/safari-601-branch/Source/_javascript_Core/dfg/DFGPutStackSinkingPhase.cpp (191514 => 191515)


--- branches/safari-601-branch/Source/_javascript_Core/dfg/DFGPutStackSinkingPhase.cpp	2015-10-23 22:45:34 UTC (rev 191514)
+++ branches/safari-601-branch/Source/_javascript_Core/dfg/DFGPutStackSinkingPhase.cpp	2015-10-23 23:03:15 UTC (rev 191515)
@@ -230,39 +230,6 @@
                         dataLog("Deferred at ", node, ":", deferred, "\n");
                     
                     if (node->op() == GetStack) {
-                        // Handle the case that the input doesn't match our requirements. This is
-                        // really a bug, but it's a benign one if we simply don't run this phase.
-                        // It usually arises because of patterns like:
-                        //
-                        // if (thing)
-                        //     PutStack()
-                        // ...
-                        // if (thing)
-                        //     GetStack()
-                        //
-                        // Or:
-                        //
-                        // if (never happens)
-                        //     GetStack()
-                        //
-                        // Because this phase runs early in SSA, it should be sensible to enforce
-                        // that no such code pattern has arisen yet. So, when validation is
-                        // enabled, we assert that we aren't seeing this. But with validation
-                        // disabled we silently let this fly and we just abort this phase.
-                        // FIXME: Get rid of all remaining cases of conflicting GetStacks.
-                        // https://bugs.webkit.org/show_bug.cgi?id=150398
-
-                        bool isConflicting =
-                            deferred.operand(node->stackAccessData()->local) == ConflictingFlush;
-                        
-                        if (validationEnabled())
-                            DFG_ASSERT(m_graph, node, !isConflicting);
-
-                        if (isConflicting) {
-                            // Oh noes! Abort!!
-                            return false;
-                        }
-
                         // A GetStack doesn't affect anything, since we know which local we are reading
                         // from.
                         continue;

Modified: branches/safari-601-branch/Source/_javascript_Core/runtime/Options.h (191514 => 191515)


--- branches/safari-601-branch/Source/_javascript_Core/runtime/Options.h	2015-10-23 22:45:34 UTC (rev 191514)
+++ branches/safari-601-branch/Source/_javascript_Core/runtime/Options.h	2015-10-23 23:03:15 UTC (rev 191515)
@@ -190,7 +190,6 @@
     v(bool, optimizeNativeCalls, false, nullptr) \
     v(bool, enableMovHintRemoval, true, nullptr) \
     v(bool, enableObjectAllocationSinking, true, nullptr) \
-    v(bool, usePutStackSinking, true, nullptr) \
     \
     v(bool, enableConcurrentJIT, true, "allows the DFG / FTL compilation in threads other than the executing JS thread") \
     v(unsigned, numberOfDFGCompilerThreads, computeNumberOfWorkerThreads(2, 2) - 1, nullptr) \

Modified: branches/safari-601-branch/Tools/ChangeLog (191514 => 191515)


--- branches/safari-601-branch/Tools/ChangeLog	2015-10-23 22:45:34 UTC (rev 191514)
+++ branches/safari-601-branch/Tools/ChangeLog	2015-10-23 23:03:15 UTC (rev 191515)
@@ -1,5 +1,20 @@
 2015-10-22  Matthew Hanson  <[email protected]>
 
+        Rollout r191395. rdar://problem/22847057
+
+    2015-10-21  Filip Pizlo  <[email protected]>
+
+            Failures in PutStackSinkingPhase should be less severe
+            https://bugs.webkit.org/show_bug.cgi?id=150400
+
+            Reviewed by Geoffrey Garen.
+
+            Add a test mode for no PutStackSinkingPhase.
+
+            * Scripts/run-jsc-stress-tests:
+
+2015-10-22  Matthew Hanson  <[email protected]>
+
         Merge r191395. rdar://problem/22847057
 
     2015-10-21  Filip Pizlo  <[email protected]>
@@ -137,7 +152,7 @@
             rdar://problem/22044000
 
             Reviewed by Dan Bernstein.
-            
+
             Added an API test.
 
             * TestWebKitAPI/Configurations/WebProcessPlugIn.xcconfig: Linked libWebCoreTestSupport in order to use MockContentFilter.

Modified: branches/safari-601-branch/Tools/Scripts/run-jsc-stress-tests (191514 => 191515)


--- branches/safari-601-branch/Tools/Scripts/run-jsc-stress-tests	2015-10-23 22:45:34 UTC (rev 191514)
+++ branches/safari-601-branch/Tools/Scripts/run-jsc-stress-tests	2015-10-23 23:03:15 UTC (rev 191515)
@@ -710,10 +710,6 @@
     run("ftl-no-cjit-validate", "--validateGraph=true", *(FTL_OPTIONS + NO_CJIT_OPTIONS)) if $enableFTL
 end
 
-def runFTLNoCJITNoPutStackValidate
-    run("ftl-no-cjit-no-put-stack-validate", "--validateGraph=true", "--usePutStackSinking=false", *(FTL_OPTIONS + NO_CJIT_OPTIONS)) if $enableFTL
-end
-
 def runFTLNoCJITNoInlineValidate
     run("ftl-no-cjit-no-inline-validate", "--validateGraph=true", "--maximumInliningDepth=1", *(FTL_OPTIONS + NO_CJIT_OPTIONS)) if $enableFTL
 end
@@ -777,7 +773,6 @@
         runDefaultFTL
         runFTLNoCJITValidate
         runFTLNoCJITNoInlineValidate
-        runFTLNoCJITNoPutStackValidate
         runFTLEager
         runFTLEagerNoCJITValidate
         runFTLNoCJITSmallPool
_______________________________________________
webkit-changes mailing list
[email protected]
https://lists.webkit.org/mailman/listinfo/webkit-changes

Reply via email to