[llvm] dee058c - [CSSPGO] Turn on ext-tsp by default for CSSPGO.

Hongtao Yu via llvm-commits llvm-commits at lists.llvm.org
Fri Feb 4 19:48:50 PST 2022


Author: Hongtao Yu
Date: 2022-02-04T19:46:44-08:00
New Revision: dee058c670593b999fec19c458dbbd882ad9de56

URL: https://github.com/llvm/llvm-project/commit/dee058c670593b999fec19c458dbbd882ad9de56
DIFF: https://github.com/llvm/llvm-project/commit/dee058c670593b999fec19c458dbbd882ad9de56.diff

LOG: [CSSPGO] Turn on ext-tsp by default for CSSPGO.

I'm seeing ext-tsp helps CSSPGO for our intern large benchmarks so I'm turning on it for CSSPGO. For non-CS AutoFDO, ext-tsp doesn't seem to help, probably because of lower profile counts quality.

Reviewed By: wenlei

Differential Revision: https://reviews.llvm.org/D119048

Added: 
    

Modified: 
    llvm/lib/CodeGen/MachineBlockPlacement.cpp
    llvm/lib/Transforms/IPO/SampleProfile.cpp
    llvm/lib/Transforms/Utils/CodeLayout.cpp

Removed: 
    


################################################################################
diff  --git a/llvm/lib/CodeGen/MachineBlockPlacement.cpp b/llvm/lib/CodeGen/MachineBlockPlacement.cpp
index c93ffaabf74c..5ad3d4d21ce2 100644
--- a/llvm/lib/CodeGen/MachineBlockPlacement.cpp
+++ b/llvm/lib/CodeGen/MachineBlockPlacement.cpp
@@ -200,10 +200,7 @@ static cl::opt<unsigned> TriangleChainCount(
     cl::init(2),
     cl::Hidden);
 
-static cl::opt<bool> EnableExtTspBlockPlacement(
-    "enable-ext-tsp-block-placement", cl::Hidden, cl::init(false),
-    cl::desc("Enable machine block placement based on the ext-tsp model, "
-             "optimizing I-cache utilization."));
+extern cl::opt<bool> EnableExtTspBlockPlacement;
 
 namespace llvm {
 extern cl::opt<unsigned> StaticLikelyProb;

diff  --git a/llvm/lib/Transforms/IPO/SampleProfile.cpp b/llvm/lib/Transforms/IPO/SampleProfile.cpp
index 0298f3c025db..38bbd506381b 100644
--- a/llvm/lib/Transforms/IPO/SampleProfile.cpp
+++ b/llvm/lib/Transforms/IPO/SampleProfile.cpp
@@ -295,6 +295,8 @@ static cl::opt<bool> OverwriteExistingWeights(
     "overwrite-existing-weights", cl::Hidden, cl::init(false),
     cl::desc("Ignore existing branch weights on IR and always overwrite."));
 
+extern cl::opt<bool> EnableExtTspBlockPlacement;
+
 namespace {
 
 using BlockWeightMap = DenseMap<const BasicBlock *, uint64_t>;
@@ -2000,6 +2002,10 @@ bool SampleProfileLoader::doInitialization(Module &M,
     if (!SampleProfileUseProfi.getNumOccurrences())
       SampleProfileUseProfi = true;
 
+    // Enable EXT-TSP block layout for CSSPGO.
+    if (!EnableExtTspBlockPlacement.getNumOccurrences())
+      EnableExtTspBlockPlacement = true;
+
     if (FunctionSamples::ProfileIsCSFlat) {
       // Tracker for profiles under 
diff erent context
       ContextTracker = std::make_unique<SampleContextTracker>(

diff  --git a/llvm/lib/Transforms/Utils/CodeLayout.cpp b/llvm/lib/Transforms/Utils/CodeLayout.cpp
index dfb9f608eab2..e46a939dcace 100644
--- a/llvm/lib/Transforms/Utils/CodeLayout.cpp
+++ b/llvm/lib/Transforms/Utils/CodeLayout.cpp
@@ -45,6 +45,11 @@
 using namespace llvm;
 #define DEBUG_TYPE "code-layout"
 
+cl::opt<bool> EnableExtTspBlockPlacement(
+    "enable-ext-tsp-block-placement", cl::Hidden, cl::init(false),
+    cl::desc("Enable machine block placement based on the ext-tsp model, "
+             "optimizing I-cache utilization."));
+
 // Algorithm-specific constants. The values are tuned for the best performance
 // of large-scale front-end bound binaries.
 static cl::opt<double>


        


More information about the llvm-commits mailing list