AMDGPU: Switch over reg class size instead of checking all super classes

[oota-llvm.git] / lib / Target / AMDGPU / SIRegisterInfo.cpp

diff --git a/lib/Target/AMDGPU/SIRegisterInfo.cpp b/lib/Target/AMDGPU/SIRegisterInfo.cpp
index c89948bb3c2dc7e2f09c05d81adc0e068dd92436..a3976f5e59182b6cd74aaea22ab5989ad8eb7baa 100644 (file)
--- a/lib/Target/AMDGPU/SIRegisterInfo.cpp
+++ b/lib/Target/AMDGPU/SIRegisterInfo.cpp
@@ -351,31 +351,45 @@ const TargetRegisterClass *SIRegisterInfo::getPhysRegClass(unsigned Reg) const {
   return nullptr;
 }
 
+// TODO: It might be helpful to have some target specific flags in
+// TargetRegisterClass to mark which classes are VGPRs to make this trivial.
 bool SIRegisterInfo::hasVGPRs(const TargetRegisterClass *RC) const {
-  return getCommonSubClass(&AMDGPU::VGPR_32RegClass, RC) ||
-         getCommonSubClass(&AMDGPU::VReg_64RegClass, RC) ||
-         getCommonSubClass(&AMDGPU::VReg_96RegClass, RC) ||
-         getCommonSubClass(&AMDGPU::VReg_128RegClass, RC) ||
-         getCommonSubClass(&AMDGPU::VReg_256RegClass, RC) ||
-         getCommonSubClass(&AMDGPU::VReg_512RegClass, RC);
+  switch (RC->getSize()) {
+  case 4:
+    return getCommonSubClass(&AMDGPU::VGPR_32RegClass, RC) != nullptr;
+  case 8:
+    return getCommonSubClass(&AMDGPU::VReg_64RegClass, RC) != nullptr;
+  case 12:
+    return getCommonSubClass(&AMDGPU::VReg_96RegClass, RC) != nullptr;
+  case 16:
+    return getCommonSubClass(&AMDGPU::VReg_128RegClass, RC) != nullptr;
+  case 32:
+    return getCommonSubClass(&AMDGPU::VReg_256RegClass, RC) != nullptr;
+  case 64:
+    return getCommonSubClass(&AMDGPU::VReg_512RegClass, RC) != nullptr;
+  default:
+    llvm_unreachable("Invalid register class size");
+  }
 }
 
 const TargetRegisterClass *SIRegisterInfo::getEquivalentVGPRClass(
                                          const TargetRegisterClass *SRC) const {
-    if (hasVGPRs(SRC)) {
-      return SRC;
-    } else if (getCommonSubClass(SRC, &AMDGPU::SGPR_32RegClass)) {
-      return &AMDGPU::VGPR_32RegClass;
-    } else if (getCommonSubClass(SRC, &AMDGPU::SGPR_64RegClass)) {
-      return &AMDGPU::VReg_64RegClass;
-    } else if (getCommonSubClass(SRC, &AMDGPU::SReg_128RegClass)) {
-      return &AMDGPU::VReg_128RegClass;
-    } else if (getCommonSubClass(SRC, &AMDGPU::SReg_256RegClass)) {
-      return &AMDGPU::VReg_256RegClass;
-    } else if (getCommonSubClass(SRC, &AMDGPU::SReg_512RegClass)) {
-      return &AMDGPU::VReg_512RegClass;
-    }
-    return nullptr;
+  switch (SRC->getSize()) {
+  case 4:
+    return &AMDGPU::VGPR_32RegClass;
+  case 8:
+    return &AMDGPU::VReg_64RegClass;
+  case 12:
+    return &AMDGPU::VReg_96RegClass;
+  case 16:
+    return &AMDGPU::VReg_128RegClass;
+  case 32:
+    return &AMDGPU::VReg_256RegClass;
+  case 64:
+    return &AMDGPU::VReg_512RegClass;
+  default:
+    llvm_unreachable("Invalid register class size");
+  }
 }
 
 const TargetRegisterClass *SIRegisterInfo::getSubRegClass(
@@ -392,6 +406,30 @@ const TargetRegisterClass *SIRegisterInfo::getSubRegClass(
   }
 }
 
+bool SIRegisterInfo::shouldRewriteCopySrc(
+  const TargetRegisterClass *DefRC,
+  unsigned DefSubReg,
+  const TargetRegisterClass *SrcRC,
+  unsigned SrcSubReg) const {
+  // We want to prefer the smallest register class possible, so we don't want to
+  // stop and rewrite on anything that looks like a subregister
+  // extract. Operations mostly don't care about the super register class, so we
+  // only want to stop on the most basic of copies between the smae register
+  // class.
+  //
+  // e.g. if we have something like
+  // vreg0 = ...
+  // vreg1 = ...
+  // vreg2 = REG_SEQUENCE vreg0, sub0, vreg1, sub1, vreg2, sub2
+  // vreg3 = COPY vreg2, sub0
+  //
+  // We want to look through the COPY to find:
+  //  => vreg3 = COPY vreg0
+
+  // Plain copy.
+  return getCommonSubClass(DefRC, SrcRC) != nullptr;
+}
+
 unsigned SIRegisterInfo::getPhysRegSubReg(unsigned Reg,
                                           const TargetRegisterClass *SubRC,
                                           unsigned Channel) const {