[PowerPC] Add vector pair calling convention for AIX

This is AIX part of update after https://reviews.llvm.org/D117225

Fixed the issue that AIX64 with vector pair enabled saw redundant
spill/reload of callee saved vector registers.

Based on original patch by: Kai Luo

Reviewed By: lkail

Differential Revision: https://reviews.llvm.org/D133466

4 files changed, 49 insertions, 77 deletions

diff --git a/llvm/lib/Target/PowerPC/PPCCallingConv.td b/llvm/lib/Target/PowerPC/PPCCallingConv.td
index 8f9dc98..9df1b1d 100644
--- a/llvm/lib/Target/PowerPC/PPCCallingConv.td
+++ b/llvm/lib/Target/PowerPC/PPCCallingConv.td
@@ -386,3 +386,9 @@ def CSR_SVR64_ColdCC_R2_VSRP : CalleeSavedRegs<(add CSR_SVR64_ColdCC_VSRP, X2)>;
 
 def CSR_64_AllRegs_VSRP :
   CalleeSavedRegs<(add CSR_64_AllRegs_VSX, CSR_ALL_VSRP)>;
+
+def CSR_AIX64_VSRP : CalleeSavedRegs<(add CSR_PPC64_Altivec, CSR_VSRP)>;
+
+def CSR_AIX64_R2_VSRP : CalleeSavedRegs<(add CSR_AIX64_VSRP, X2)>;
+
+def CSR_AIX32_VSRP : CalleeSavedRegs<(add CSR_AIX32_Altivec, CSR_VSRP)>;
diff --git a/llvm/lib/Target/PowerPC/PPCRegisterInfo.cpp b/llvm/lib/Target/PowerPC/PPCRegisterInfo.cpp
index a60b993..d39934b 100644
--- a/llvm/lib/Target/PowerPC/PPCRegisterInfo.cpp
+++ b/llvm/lib/Target/PowerPC/PPCRegisterInfo.cpp
@@ -237,8 +237,14 @@ PPCRegisterInfo::getCalleeSavedRegs(const MachineFunction *MF) const {
   }
   // Standard calling convention CSRs.
   if (TM.isPPC64()) {
-    if (Subtarget.pairedVectorMemops())
+    if (Subtarget.pairedVectorMemops()) {
+      if (Subtarget.isAIXABI()) {
+        if (!TM.getAIXExtendedAltivecABI())
+          return SaveR2 ? CSR_PPC64_R2_SaveList : CSR_PPC64_SaveList;
+        return SaveR2 ? CSR_AIX64_R2_VSRP_SaveList : CSR_AIX64_VSRP_SaveList;
+      }
       return SaveR2 ? CSR_SVR464_R2_VSRP_SaveList : CSR_SVR464_VSRP_SaveList;
+    }
     if (Subtarget.hasAltivec() &&
         (!Subtarget.isAIXABI() || TM.getAIXExtendedAltivecABI())) {
       return SaveR2 ? CSR_PPC64_R2_Altivec_SaveList
@@ -248,6 +254,9 @@ PPCRegisterInfo::getCalleeSavedRegs(const MachineFunction *MF) const {
   }
   // 32-bit targets.
   if (Subtarget.isAIXABI()) {
+    if (Subtarget.pairedVectorMemops())
+      return TM.getAIXExtendedAltivecABI() ? CSR_AIX32_VSRP_SaveList
+                                           : CSR_AIX32_SaveList;
     if (Subtarget.hasAltivec())
       return TM.getAIXExtendedAltivecABI() ? CSR_AIX32_Altivec_SaveList
                                            : CSR_AIX32_SaveList;
@@ -286,6 +295,11 @@ PPCRegisterInfo::getCallPreservedMask(const MachineFunction &MF,
   }
 
   if (Subtarget.isAIXABI()) {
+    if (Subtarget.pairedVectorMemops()) {
+      if (!TM.getAIXExtendedAltivecABI())
+        return TM.isPPC64() ? CSR_PPC64_RegMask : CSR_AIX32_RegMask;
+      return TM.isPPC64() ? CSR_AIX64_VSRP_RegMask : CSR_AIX32_VSRP_RegMask;
+    }
     return TM.isPPC64()
                ? ((Subtarget.hasAltivec() && TM.getAIXExtendedAltivecABI())
                       ? CSR_PPC64_Altivec_RegMask
diff --git a/llvm/test/CodeGen/PowerPC/aix64-vector-pair-cc-spills.ll b/llvm/test/CodeGen/PowerPC/aix64-vector-pair-cc-spills.ll
index c532a43..34db2e3 100644
--- a/llvm/test/CodeGen/PowerPC/aix64-vector-pair-cc-spills.ll
+++ b/llvm/test/CodeGen/PowerPC/aix64-vector-pair-cc-spills.ll
@@ -3,32 +3,20 @@
 ; RUN: llc -O0 -mtriple=powerpc64-ibm-aix-xcoff -mcpu=pwr10 -vec-extabi -stop-after=prologepilog -verify-machineinstrs < %s | \
 ; RUN: FileCheck --check-prefix=CHECK-VEXT %s
 
-; Error pattern will be fixed in https://reviews.llvm.org/D133466
 ; CHECK-LABEL: name: foo
-; CHECK: spill-slot
-; CHECK-NEXT: callee-saved-register: '$v31'
-; CHECK: spill-slot
-; CHECK-NEXT: callee-saved-register: '$v30'
-; CHECK: spill-slot
-; CHECK-NEXT: callee-saved-register: '$v29'
-; CHECK: spill-slot
-; CHECK-NEXT: callee-saved-register: '$v28'
-; CHECK: spill-slot
-; CHECK-NEXT: callee-saved-register: '$v27'
-; CHECK: spill-slot
-; CHECK-NEXT: callee-saved-register: '$v26'
-; CHECK: spill-slot
-; CHECK-NEXT: callee-saved-register: '$v25'
-; CHECK: spill-slot
-; CHECK-NEXT: callee-saved-register: '$v24'
-; CHECK: spill-slot
-; CHECK-NEXT: callee-saved-register: '$v23'
-; CHECK: spill-slot
-; CHECK-NEXT: callee-saved-register: '$v22'
-; CHECK: spill-slot
-; CHECK-NEXT: callee-saved-register: '$v21'
-; CHECK: spill-slot
-; CHECK-NEXT: callee-saved-register: '$v20'
+; CHECK-NOT: spill-slot
+; CHECK-NOT: callee-saved-register: '$v31'
+; CHECK-NOT: callee-saved-register: '$v30'
+; CHECK-NOT: callee-saved-register: '$v29'
+; CHECK-NOT: callee-saved-register: '$v28'
+; CHECK-NOT: callee-saved-register: '$v27'
+; CHECK-NOT: callee-saved-register: '$v26'
+; CHECK-NOT: callee-saved-register: '$v25'
+; CHECK-NOT: callee-saved-register: '$v24'
+; CHECK-NOT: callee-saved-register: '$v23'
+; CHECK-NOT: callee-saved-register: '$v22'
+; CHECK-NOT: callee-saved-register: '$v21'
+; CHECK-NOT: callee-saved-register: '$v20'
 
 ; CHECK-VEXT-LABEL: name: foo
 ; CHECK-VEXT-NOT: spill-slot
@@ -50,32 +38,20 @@ entry:
   ret void
 }
 
-; Error pattern will be fixed in https://reviews.llvm.org/D133466
 ; CHECK-LABEL: name: spill
-; CHECK: spill-slot
-; CHECK-NEXT: callee-saved-register: '$v31'
-; CHECK: spill-slot
-; CHECK-NEXT: callee-saved-register: '$v30'
-; CHECK: spill-slot
-; CHECK-NEXT: callee-saved-register: '$v29'
-; CHECK: spill-slot
-; CHECK-NEXT: callee-saved-register: '$v28'
-; CHECK: spill-slot
-; CHECK-NEXT: callee-saved-register: '$v27'
-; CHECK: spill-slot
-; CHECK-NEXT: callee-saved-register: '$v26'
-; CHECK: spill-slot
-; CHECK-NEXT: callee-saved-register: '$v25'
-; CHECK: spill-slot
-; CHECK-NEXT: callee-saved-register: '$v24'
-; CHECK: spill-slot
-; CHECK-NEXT: callee-saved-register: '$v23'
-; CHECK: spill-slot
-; CHECK-NEXT: callee-saved-register: '$v22'
-; CHECK: spill-slot
-; CHECK-NEXT: callee-saved-register: '$v21'
-; CHECK: spill-slot
-; CHECK-NEXT: callee-saved-register: '$v20'
+; CHECK-NOT: spill-slot
+; CHECK-NOT: callee-saved-register: '$v31'
+; CHECK-NOT: callee-saved-register: '$v30'
+; CHECK-NOT: callee-saved-register: '$v29'
+; CHECK-NOT: callee-saved-register: '$v28'
+; CHECK-NOT: callee-saved-register: '$v27'
+; CHECK-NOT: callee-saved-register: '$v26'
+; CHECK-NOT: callee-saved-register: '$v25'
+; CHECK-NOT: callee-saved-register: '$v24'
+; CHECK-NOT: callee-saved-register: '$v23'
+; CHECK-NOT: callee-saved-register: '$v22'
+; CHECK-NOT: callee-saved-register: '$v21'
+; CHECK-NOT: callee-saved-register: '$v20'
 
 ; CHECK-VEXT-LABEL: name: spill
 ; CHECK-VEXT: spill-slot
diff --git a/llvm/test/CodeGen/PowerPC/aix64-virtual-call-no-spills.ll b/llvm/test/CodeGen/PowerPC/aix64-virtual-call-no-spills.ll
index fb4e5ae..1c89a5b 100644
--- a/llvm/test/CodeGen/PowerPC/aix64-virtual-call-no-spills.ll
+++ b/llvm/test/CodeGen/PowerPC/aix64-virtual-call-no-spills.ll
@@ -9,19 +9,7 @@ define dso_local noundef signext i32 @virtualCall(ptr noundef %b) #0 {
 ; CHECK:       # %bb.0: # %entry
 ; CHECK-NEXT:    mflr 0
 ; CHECK-NEXT:    std 0, 16(1)
-; CHECK-NEXT:    stdu 1, -320(1)
-; CHECK-NEXT:    stxv 52, 128(1) # 16-byte Folded Spill
-; CHECK-NEXT:    stxv 53, 144(1) # 16-byte Folded Spill
-; CHECK-NEXT:    stxv 54, 160(1) # 16-byte Folded Spill
-; CHECK-NEXT:    stxv 55, 176(1) # 16-byte Folded Spill
-; CHECK-NEXT:    stxv 56, 192(1) # 16-byte Folded Spill
-; CHECK-NEXT:    stxv 57, 208(1) # 16-byte Folded Spill
-; CHECK-NEXT:    stxv 58, 224(1) # 16-byte Folded Spill
-; CHECK-NEXT:    stxv 59, 240(1) # 16-byte Folded Spill
-; CHECK-NEXT:    stxv 60, 256(1) # 16-byte Folded Spill
-; CHECK-NEXT:    stxv 61, 272(1) # 16-byte Folded Spill
-; CHECK-NEXT:    stxv 62, 288(1) # 16-byte Folded Spill
-; CHECK-NEXT:    stxv 63, 304(1) # 16-byte Folded Spill
+; CHECK-NEXT:    stdu 1, -128(1)
 ; CHECK-NEXT:    std 3, 120(1)
 ; CHECK-NEXT:    ld 3, 120(1)
 ; CHECK-NEXT:    ld 4, 0(3)
@@ -37,19 +25,7 @@ define dso_local noundef signext i32 @virtualCall(ptr noundef %b) #0 {
 ; CHECK-NEXT:    ld 2, 40(1)
 ; CHECK-NEXT:    # kill: def $r3 killed $r3 killed $x3
 ; CHECK-NEXT:    extsw 3, 3
-; CHECK-NEXT:    lxv 63, 304(1) # 16-byte Folded Reload
-; CHECK-NEXT:    lxv 62, 288(1) # 16-byte Folded Reload
-; CHECK-NEXT:    lxv 61, 272(1) # 16-byte Folded Reload
-; CHECK-NEXT:    lxv 60, 256(1) # 16-byte Folded Reload
-; CHECK-NEXT:    lxv 59, 240(1) # 16-byte Folded Reload
-; CHECK-NEXT:    lxv 58, 224(1) # 16-byte Folded Reload
-; CHECK-NEXT:    lxv 57, 208(1) # 16-byte Folded Reload
-; CHECK-NEXT:    lxv 56, 192(1) # 16-byte Folded Reload
-; CHECK-NEXT:    lxv 55, 176(1) # 16-byte Folded Reload
-; CHECK-NEXT:    lxv 54, 160(1) # 16-byte Folded Reload
-; CHECK-NEXT:    lxv 53, 144(1) # 16-byte Folded Reload
-; CHECK-NEXT:    lxv 52, 128(1) # 16-byte Folded Reload
-; CHECK-NEXT:    addi 1, 1, 320
+; CHECK-NEXT:    addi 1, 1, 128
 ; CHECK-NEXT:    ld 0, 16(1)
 ; CHECK-NEXT:    mtlr 0
 ; CHECK-NEXT:    blr