Add TRI::getCommonSuperRegClass().

This function is a generalization of getMatchingSuperRegClass() to the
symmetric case where both sides are using a sub-register index. It will
find a super-register class and sub-register indexes that make this
diagram commute:

                                   PreA
                       SuperRC  ---------->  RCA

                          |                   |
                          |                   |
                     PreB |                   | SubA
                          |                   |
                          |                   |
                          V                   V

                         RCB    ----------> SubRC
                                   SubB

This can be used to coalesce copies like:

  %vreg1:sub16 = COPY %vreg2:sub16; GR64:%vreg1, GR32: %vreg2

llvm-svn: 156317

include/llvm/Target/TargetRegisterInfo.h

@@ -458,6 +458,34 @@ public:
     return b;
   }
 
+  /// getCommonSuperRegClass - Find a common super-register class if it exists.
+  ///
+  /// Find a register class, SuperRC and two sub-register indices, PreA and
+  /// PreB, such that:
+  ///
+  ///   1. PreA + SubA == PreB + SubB  (using composeSubRegIndices()), and
+  ///
+  ///   2. For all Reg in SuperRC: Reg:PreA in RCA and Reg:PreB in RCB, and
+  ///
+  ///   3. SuperRC->getSize() >= max(RCA->getSize(), RCB->getSize()).
+  ///
+  /// SuperRC will be chosen such that no super-class of SuperRC satisfies the
+  /// requirements, and there is no register class with a smaller spill size
+  /// that satisfies the requirements.
+  ///
+  /// SubA and SubB must not be 0. Use getMatchingSuperRegClass() instead.
+  ///
+  /// Either of the PreA and PreB sub-register indices may be returned as 0. In
+  /// that case, the returned register class will be a sub-class of the
+  /// corresponding argument register class.
+  ///
+  /// The function returns NULL if no register class can be found.
+  ///
+  const TargetRegisterClass*
+  getCommonSuperRegClass(const TargetRegisterClass *RCA, unsigned SubA,
+                         const TargetRegisterClass *RCB, unsigned SubB,
+                         unsigned &PreA, unsigned &PreB) const;
+
   //===--------------------------------------------------------------------===//
   // Register Class Information
   //

lib/Target/TargetRegisterInfo.cpp

@@ -122,6 +122,16 @@ BitVector TargetRegisterInfo::getAllocatableSet(const MachineFunction &MF,
   return Allocatable;
 }
 
+static inline
+const TargetRegisterClass *firstCommonClass(const uint32_t *A,
+                                            const uint32_t *B,
+                                            const TargetRegisterInfo *TRI) {
+  for (unsigned I = 0, E = TRI->getNumRegClasses(); I < E; I += 32)
+    if (unsigned Common = *A++ & *B++)
+      return TRI->getRegClass(I + CountTrailingZeros_32(Common));
+  return 0;
+}
+
 const TargetRegisterClass *
 TargetRegisterInfo::getCommonSubClass(const TargetRegisterClass *A,
                                       const TargetRegisterClass *B) const {
@@ -173,3 +183,64 @@ TargetRegisterInfo::getMatchingSuperRegClass(const TargetRegisterClass *A,
       return getRegClass(Base + CountTrailingZeros_32(Common));
   return 0;
 }
+
+const TargetRegisterClass *TargetRegisterInfo::
+getCommonSuperRegClass(const TargetRegisterClass *RCA, unsigned SubA,
+                       const TargetRegisterClass *RCB, unsigned SubB,
+                       unsigned &PreA, unsigned &PreB) const {
+  assert(RCA && SubA && RCB && SubB && "Invalid arguments");
+
+  // Search all pairs of sub-register indices that project into RCA and RCB
+  // respectively. This is quadratic, but usually the sets are very small. On
+  // most targets like X86, there will only be a single sub-register index
+  // (e.g., sub_16bit projecting into GR16).
+  //
+  // The worst case is a register class like DPR on ARM.
+  // We have indices dsub_0..dsub_7 projecting into that class.
+  //
+  // It is very common that one register class is a sub-register of the other.
+  // Arrange for RCA to be the larger register so the answer will be found in
+  // the first iteration. This makes the search linear for the most common
+  // case.
+  const TargetRegisterClass *BestRC = 0;
+  unsigned *BestPreA = &PreA;
+  unsigned *BestPreB = &PreB;
+  if (RCA->getSize() < RCB->getSize()) {
+    std::swap(RCA, RCB);
+    std::swap(BestPreA, BestPreB);
+  }
+
+  // Also terminate the search one we have found a register class as small as
+  // RCA.
+  unsigned MinSize = RCA->getSize();
+
+  for (SuperRegClassIterator IA(RCA, this, true); IA.isValid(); ++IA) {
+    unsigned FinalA = composeSubRegIndices(IA.getSubReg(), SubA);
+    for (SuperRegClassIterator IB(RCB, this, true); IB.isValid(); ++IB) {
+      // Check if a common super-register class exists for this index pair.
+      const TargetRegisterClass *RC =
+        firstCommonClass(IA.getMask(), IB.getMask(), this);
+      if (!RC || RC->getSize() < MinSize)
+        continue;
+
+      // The indexes must compose identically: PreA+SubA == PreB+SubB.
+      unsigned FinalB = composeSubRegIndices(IB.getSubReg(), SubB);
+      if (FinalA != FinalB)
+        continue;
+
+      // Is RC a better candidate than BestRC?
+      if (BestRC && RC->getSize() >= BestRC->getSize())
+        continue;
+
+      // Yes, RC is the smallest super-register seen so far.
+      BestRC = RC;
+      *BestPreA = IA.getSubReg();
+      *BestPreB = IB.getSubReg();
+
+      // Bail early if we reached MinSize. We won't find a better candidate.
+      if (BestRC->getSize() == MinSize)
+        return BestRC;
+    }
+  }
+  return BestRC;
+}