[llvm] r358522 - Time profiler: optimize json output time

Anton Afanasyev via llvm-commits llvm-commits at lists.llvm.org
Tue Apr 16 13:36:56 PDT 2019


Author: anton-afanasyev
Date: Tue Apr 16 13:36:56 2019
New Revision: 358522

URL: http://llvm.org/viewvc/llvm-project?rev=358522&view=rev
Log:
Time profiler: optimize json output time

Summary:
Use llvm::json::Array.reserve() to optimize json output time. Here is motivation:
https://reviews.llvm.org/D60609#1468941. In short: for the json array
with ~32K entries, pushing back each entry takes ~4% of whole time compared
to the method of preliminary memory reservation: (3995-3845)/3995 = 3.75%.

Reviewers: lebedev.ri

Subscribers: hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D60792

Modified:
    llvm/trunk/lib/Support/TimeProfiler.cpp

Modified: llvm/trunk/lib/Support/TimeProfiler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Support/TimeProfiler.cpp?rev=358522&r1=358521&r2=358522&view=diff
==============================================================================
--- llvm/trunk/lib/Support/TimeProfiler.cpp (original)
+++ llvm/trunk/lib/Support/TimeProfiler.cpp Tue Apr 16 13:36:56 2019
@@ -89,6 +89,9 @@ struct TimeTraceProfiler {
            "All profiler sections should be ended when calling Write");
 
     json::Array Events;
+    const size_t ExpectedEntryCount =
+        Entries.size() + CountAndTotalPerName.size() + 1;
+    Events.reserve(ExpectedEntryCount);
 
     // Emit all events for the main flame graph.
     for (const auto &E : Entries) {
@@ -149,6 +152,8 @@ struct TimeTraceProfiler {
         {"args", json::Object{{"name", "clang"}}},
     });
 
+    assert(Events.size() == ExpectedEntryCount && "Size prediction failed!");
+
     OS << formatv("{0:2}", json::Value(json::Object(
                                {{"traceEvents", std::move(Events)}})));
   }




More information about the llvm-commits mailing list