[llvm] r265469 - [RegisterBankInfo] Make addRegBankCoverage more capable to ease

[Quentin Colombet via llvm-commits]

Author: qcolombet
Date: Tue Apr  5 16:20:12 2016
New Revision: 265469

URL: http://llvm.org/viewvc/llvm-project?rev=265469&view=rev
Log:
[RegisterBankInfo] Make addRegBankCoverage more capable to ease
targeting jobs.
Now, addRegBankCoverage also adds the subreg-classes not just the
sub-classes of the given register class.

Modified:
    llvm/trunk/include/llvm/CodeGen/GlobalISel/RegisterBankInfo.h
    llvm/trunk/lib/CodeGen/GlobalISel/RegisterBankInfo.cpp

Modified: llvm/trunk/include/llvm/CodeGen/GlobalISel/RegisterBankInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/GlobalISel/RegisterBankInfo.h?rev=265469&r1=265468&r2=265469&view=diff
==============================================================================
--- llvm/trunk/include/llvm/CodeGen/GlobalISel/RegisterBankInfo.h (original)
+++ llvm/trunk/include/llvm/CodeGen/GlobalISel/RegisterBankInfo.h Tue Apr  5 16:20:12 2016
@@ -47,8 +47,8 @@ protected:
 
   /// Add \p RC to the set of register class that the register bank
   /// identified \p ID covers.
-  /// This method transitively adds all the sub classes of \p RC
-  /// to the set of covered register classes.
+  /// This method transitively adds all the sub classes and the subreg-classes
+  /// of \p RC to the set of covered register classes.
   /// It also adjusts the size of the register bank to reflect the maximal
   /// size of a value that can be hold into that register bank.
   ///
@@ -56,6 +56,9 @@ protected:
   /// The rationale is if \p ID covers the registers of \p RC, that
   /// does not necessarily mean that \p ID covers the set of registers
   /// of RC's superclasses.
+  /// This method does *not* add the superreg classes as well for consistents.
+  /// The expected use is to add the coverage top-down with respect to the
+  /// register hierarchy.
   ///
   /// \todo TableGen should just generate the BitSet vector for us.
   void addRegBankCoverage(unsigned ID, const TargetRegisterClass &RC,

Modified: llvm/trunk/lib/CodeGen/GlobalISel/RegisterBankInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/GlobalISel/RegisterBankInfo.cpp?rev=265469&r1=265468&r2=265469&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/GlobalISel/RegisterBankInfo.cpp (original)
+++ llvm/trunk/lib/CodeGen/GlobalISel/RegisterBankInfo.cpp Tue Apr  5 16:20:12 2016
@@ -76,15 +76,23 @@ void RegisterBankInfo::addRegBankCoverag
     const uint32_t *SubClassMask = CurRC.getSubClassMask();
     // The subclasses mask is broken down into chunks of uint32_t, but it still
     // represents all register classes.
+    bool First = true;
     for (unsigned Base = 0; Base < NbOfRegClasses; Base += 32) {
       unsigned Idx = Base;
       for (uint32_t Mask = *SubClassMask++; Mask; Mask >>= 1, ++Idx) {
         unsigned Offset = countTrailingZeros(Mask);
         unsigned SubRCId = Idx + Offset;
-        if (!Covered.test(SubRCId))
+        if (!Covered.test(SubRCId)) {
+          if (First)
+            DEBUG(dbgs() << "  Enqueue sub-class: ");
+          DEBUG(dbgs() << TRI.getRegClassName(TRI.getRegClass(SubRCId))
+                       << ", ");
           WorkList.push_back(SubRCId);
-        // Remember that we saw the sub class.
-        Covered.set(SubRCId);
+          // Remember that we saw the sub class.
+          Covered.set(SubRCId);
+          First = false;
+        }
+
         // Move the cursor to the next sub class.
         // I.e., eat up the zeros then move to the next bit.
         // This last part is done as part of the loop increment.
@@ -96,6 +104,61 @@ void RegisterBankInfo::addRegBankCoverag
         Idx += Offset;
       }
     }
+    if (!First)
+      DEBUG(dbgs() << '\n');
+
+    // Push also all the register classes that can be accessed via a
+    // subreg index, i.e., its subreg-class (which is different than
+    // its subclass).
+    //
+    // Note: It would probably be faster to go the other way around
+    // and have this method add only super classes, since this
+    // information is available in a more efficient way. However, it
+    // feels less natural for the client of this APIs plus we will
+    // TableGen the whole bitset at some point, so compile time for
+    // the initialization is not very important.
+    First = true;
+    for (unsigned SubRCId = 0; SubRCId < NbOfRegClasses; ++SubRCId) {
+      if (Covered.test(SubRCId))
+        continue;
+      bool Pushed = false;
+      const TargetRegisterClass *SubRC = TRI.getRegClass(SubRCId);
+      for (SuperRegClassIterator SuperRCIt(SubRC, &TRI); SuperRCIt.isValid();
+           ++SuperRCIt) {
+        if (Pushed)
+          break;
+        const uint32_t *SuperRCMask = SuperRCIt.getMask();
+        for (unsigned Base = 0; Base < NbOfRegClasses; Base += 32) {
+          unsigned Idx = Base;
+          for (uint32_t Mask = *SuperRCMask++; Mask; Mask >>= 1, ++Idx) {
+            unsigned Offset = countTrailingZeros(Mask);
+            unsigned SuperRCId = Idx + Offset;
+            if (SuperRCId == RCId) {
+              if (First)
+                DEBUG(dbgs() << "  Enqueue subreg-class: ");
+              DEBUG(dbgs() << TRI.getRegClassName(SubRC) << ", ");
+              WorkList.push_back(SubRCId);
+              // Remember that we saw the sub class.
+              Covered.set(SubRCId);
+              Pushed = true;
+              First = false;
+              break;
+            }
+
+            // Move the cursor to the next sub class.
+            // I.e., eat up the zeros then move to the next bit.
+            // This last part is done as part of the loop increment.
 
+            // By construction, Offset must be less than 32.
+            // Otherwise, than means Mask was zero. I.e., no UB.
+            Mask >>= Offset;
+            // Remember that we shifted the base offset.
+            Idx += Offset;
+          }
+        }
+      }
+    }
+    if (!First)
+      DEBUG(dbgs() << '\n');
   } while (!WorkList.empty());
 }