Merged main:8d203100e898 into amd-gfx:a01f70e2edd5

Local branch amd-gfx a01f70e Merged main:f88f090a2ef1 into amd-gfx:4e956125f560
Remote branch main 8d20310 Revert "[mlir][memref] Fix offset update in emulating narrow type for strided memref (llvm#67714)"

Author: SC llvm team

clang/lib/CodeGen/CGRecordLayoutBuilder.cpp

@@ -658,9 +658,9 @@ void CGRecordLowering::computeVolatileBitfields() {
 
 void CGRecordLowering::accumulateVPtrs() {
   if (Layout.hasOwnVFPtr())
-    Members.push_back(MemberInfo(CharUnits::Zero(), MemberInfo::VFPtr,
-        llvm::FunctionType::get(getIntNType(32), /*isVarArg=*/true)->
-            getPointerTo()->getPointerTo()));
+    Members.push_back(
+        MemberInfo(CharUnits::Zero(), MemberInfo::VFPtr,
+                   llvm::PointerType::getUnqual(Types.getLLVMContext())));
   if (Layout.hasOwnVBPtr())
     Members.push_back(
         MemberInfo(Layout.getVBPtrOffset(), MemberInfo::VBPtr,

clang/lib/CodeGen/MicrosoftCXXABI.cpp

@@ -1262,8 +1262,6 @@ void MicrosoftCXXABI::initializeHiddenVirtualInheritanceMembers(
         Builder.CreateInBoundsGEP(CGF.Int8Ty, Int8This, VBaseOffset);
     // vtorDisp is always the 32-bits before the vbase in the class layout.
     VtorDispPtr = Builder.CreateConstGEP1_32(CGF.Int8Ty, VtorDispPtr, -4);
-    VtorDispPtr = Builder.CreateBitCast(
-        VtorDispPtr, CGF.Int32Ty->getPointerTo(AS), "vtordisp.ptr");
 
     Builder.CreateAlignedStore(VtorDispValue, VtorDispPtr,
                                CharUnits::fromQuantity(4));
@@ -1568,14 +1566,9 @@ void MicrosoftCXXABI::EmitInstanceFunctionProlog(CodeGenFunction &CGF) {
   if (!CGF.CurFuncIsThunk && MD->isVirtual()) {
     CharUnits Adjustment = getVirtualFunctionPrologueThisAdjustment(CGF.CurGD);
     if (!Adjustment.isZero()) {
-      unsigned AS = cast<llvm::PointerType>(This->getType())->getAddressSpace();
-      llvm::Type *charPtrTy = CGF.Int8Ty->getPointerTo(AS),
-                 *thisTy = This->getType();
-      This = CGF.Builder.CreateBitCast(This, charPtrTy);
       assert(Adjustment.isPositive());
       This = CGF.Builder.CreateConstInBoundsGEP1_32(CGF.Int8Ty, This,
                                                     -Adjustment.getQuantity());
-      This = CGF.Builder.CreateBitCast(This, thisTy, "this.adjusted");
     }
   }
   setCXXABIThisValue(CGF, This);
@@ -2511,9 +2504,6 @@ LValue MicrosoftCXXABI::EmitThreadLocalVarDeclLValue(CodeGenFunction &CGF,
   llvm::Value *V = CGF.CGM.GetAddrOfGlobalVar(VD);
   llvm::Type *RealVarTy = CGF.getTypes().ConvertTypeForMem(VD->getType());
 
-  unsigned AS = cast<llvm::PointerType>(V->getType())->getAddressSpace();
-  V = CGF.Builder.CreateBitCast(V, RealVarTy->getPointerTo(AS));
-
   CharUnits Alignment = CGF.getContext().getDeclAlign(VD);
   Address Addr(V, RealVarTy, Alignment);
 
@@ -3226,9 +3216,6 @@ llvm::Value *MicrosoftCXXABI::EmitMemberDataPointerAddress(
     CodeGenFunction &CGF, const Expr *E, Address Base, llvm::Value *MemPtr,
     const MemberPointerType *MPT) {
   assert(MPT->isMemberDataPointer());
-  unsigned AS = Base.getAddressSpace();
-  llvm::Type *PType =
-      CGF.ConvertTypeForMem(MPT->getPointeeType())->getPointerTo(AS);
   CGBuilderTy &Builder = CGF.Builder;
   const CXXRecordDecl *RD = MPT->getMostRecentCXXRecordDecl();
   MSInheritanceModel Inheritance = RD->getMSInheritanceModel();
@@ -3256,16 +3243,9 @@ llvm::Value *MicrosoftCXXABI::EmitMemberDataPointerAddress(
     Addr = Base.getPointer();
   }
 
-  // Cast to char*.
-  Addr = Builder.CreateBitCast(Addr, CGF.Int8Ty->getPointerTo(AS));
-
   // Apply the offset, which we assume is non-null.
-  Addr = Builder.CreateInBoundsGEP(CGF.Int8Ty, Addr, FieldOffset,
+  return Builder.CreateInBoundsGEP(CGF.Int8Ty, Addr, FieldOffset,
                                    "memptr.offset");
-
-  // Cast the address to the appropriate pointer type, adopting the address
-  // space of the base pointer.
-  return Builder.CreateBitCast(Addr, PType);
 }
 
 llvm::Value *

llvm/include/llvm/CodeGen/MachineBlockFrequencyInfo.h

@@ -65,9 +65,10 @@ class MachineBlockFrequencyInfo : public MachineFunctionPass {
 
   /// Compute the frequency of the block, relative to the entry block.
   /// This API assumes getEntryFreq() is non-zero.
-  float getBlockFreqRelativeToEntryBlock(const MachineBasicBlock *MBB) const {
+  double getBlockFreqRelativeToEntryBlock(const MachineBasicBlock *MBB) const {
     assert(getEntryFreq() != 0 && "getEntryFreq() should not return 0 here!");
-    return getBlockFreq(MBB).getFrequency() * (1.0f / getEntryFreq());
+    return static_cast<double>(getBlockFreq(MBB).getFrequency()) /
+           static_cast<double>(getEntryFreq());
   }
 
   std::optional<uint64_t>

llvm/include/llvm/Config/llvm-config.h.cmake

@@ -16,7 +16,7 @@
 
 /* Indicate that this is LLVM compiled from the amd-gfx branch. */
 #define LLVM_HAVE_BRANCH_AMD_GFX
-#define LLVM_MAIN_REVISION 476431
+#define LLVM_MAIN_REVISION 476438
 
 /* Define if LLVM_ENABLE_DUMP is enabled */
 #cmakedefine LLVM_ENABLE_DUMP

llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp

@@ -1929,18 +1929,32 @@ void AsmPrinter::emitFunctionBody() {
 
   // Output MBB ids, function names, and frequencies if the flag to dump
   // MBB profile information has been set
-  if (MBBProfileDumpFileOutput) {
+  if (MBBProfileDumpFileOutput && !MF->empty() &&
+      MF->getFunction().getEntryCount()) {
     if (!MF->hasBBLabels())
       MF->getContext().reportError(
           SMLoc(),
           "Unable to find BB labels for MBB profile dump. -mbb-profile-dump "
           "must be called with -basic-block-sections=labels");
     MachineBlockFrequencyInfo &MBFI =
         getAnalysis<LazyMachineBlockFrequencyInfoPass>().getBFI();
+    // The entry count and the entry basic block frequency aren't the same. We
+    // want to capture "absolute" frequencies, i.e. the frequency with which a
+    // MBB is executed when the program is executed. From there, we can derive
+    // Function-relative frequencies (divide by the value for the first MBB).
+    // We also have the information about frequency with which functions
+    // were called. This helps, for example, in a type of integration tests
+    // where we want to cross-validate the compiler's profile with a real
+    // profile.
+    // Using double precision because uint64 values used to encode mbb
+    // "frequencies" may be quite large.
+    const double EntryCount =
+        static_cast<double>(MF->getFunction().getEntryCount()->getCount());
     for (const auto &MBB : *MF) {
+      const double MBBRelFreq = MBFI.getBlockFreqRelativeToEntryBlock(&MBB);
+      const double AbsMBBFreq = MBBRelFreq * EntryCount;
       *MBBProfileDumpFileOutput.get()
-          << MF->getName() << "," << MBB.getBBID() << ","
-          << MBFI.getBlockFreq(&MBB).getFrequency() << "\n";
+          << MF->getName() << "," << MBB.getBBID() << "," << AbsMBBFreq << "\n";
     }
   }
 }

llvm/lib/CodeGen/GlobalISel/CallLowering.cpp

@@ -847,7 +847,8 @@ void CallLowering::insertSRetLoads(MachineIRBuilder &MIRBuilder, Type *RetTy,
 
   unsigned NumValues = SplitVTs.size();
   Align BaseAlign = DL.getPrefTypeAlign(RetTy);
-  Type *RetPtrTy = RetTy->getPointerTo(DL.getAllocaAddrSpace());
+  Type *RetPtrTy =
+      PointerType::get(RetTy->getContext(), DL.getAllocaAddrSpace());
   LLT OffsetLLTy = getLLTForType(*DL.getIndexType(RetPtrTy), DL);
 
   MachinePointerInfo PtrInfo = MachinePointerInfo::getFixedStack(MF, FI);

llvm/test/CodeGen/Generic/bb-profile-dump.ll

@@ -7,26 +7,48 @@
 
 ; Check that given a simple case, we can return the default MBFI
 
-define i64 @f2(i64 %a, i64 %b) {
+define i64 @f2(i64 %a, i64 %b) !prof !1{
     %sum = add i64 %a, %b
     ret i64 %sum
 }
 
-; CHECK: f2,0,8
+; CHECK: f2,0,1.000000e+03
 
-define i64 @f1() {
+define i64 @f1() !prof !2{
     %sum = call i64 @f2(i64 2, i64 2)
     %isEqual = icmp eq i64 %sum, 4
-    br i1 %isEqual, label %ifEqual, label %ifNotEqual
+    br i1 %isEqual, label %ifEqual, label %ifNotEqual, !prof !3
 ifEqual:
     ret i64 0
 ifNotEqual:
     ret i64 %sum
 }
 
-; CHECK-NEXT: f1,0,16
-; CHECK-NEXT: f1,1,8
-; CHECK-NEXT: f1,2,16
+; CHECK-NEXT: f1,0,1.000000e+01
+; CHECK-NEXT: f1,2,6.000000e+00
+; CHECK-NEXT: f1,1,4.000000e+00
+
+define void @f3(i32 %iter) !prof !4 {
+entry:
+    br label %loop
+loop:
+    %i = phi i32 [0, %entry], [%i_next, %loop]
+    %i_next = add i32 %i, 1
+    %exit_cond = icmp slt i32 %i_next, %iter
+    br i1 %exit_cond, label %loop, label %exit, !prof !5
+exit:
+    ret void
+}
+
+; CHECK-NEXT: f3,0,2.000000e+00
+; CHECK-NEXT: f3,1,2.002000e+03
+; CHECK-NEXT: f3,2,2.000000e+00
+
+!1 = !{!"function_entry_count", i64 1000}
+!2 = !{!"function_entry_count", i64 10}
+!3 = !{!"branch_weights", i32 2, i32 3}
+!4 = !{!"function_entry_count", i64 2}
+!5 = !{!"branch_weights", i32 1000, i32 1}
 
 ; Check that if we pass -mbb-profile-dump but don't set -basic-block-sections,
 ; we get an appropriate error message