[AutoBump] Merge with fixes of c1d01b2f (Jan 08) (30) [Only tested MLIR]

flang/docs/Extensions.md

@@ -160,7 +160,11 @@ end
 * `<>` as synonym for `.NE.` and `/=`
 * `$` and `@` as legal characters in names
 * Initialization in type declaration statements using `/values/`
-* Saved variables without explicit or default initializers are zero initialized.
+* Saved variables without explicit or default initializers are zero initialized,
+  except for scalar variables from the main program that are not explicitly
+  initialized or marked with an explicit SAVE attribute (these variables may be
+  placed on the stack by flang and not zero initialized). It is not advised to
+  rely on this extension in new code.
 * In a saved entity of a type with a default initializer, components without default
   values are zero initialized.
 * Kind specification with `*`, e.g. `REAL*4`

flang/lib/Optimizer/Transforms/StackArrays.cpp

@@ -330,6 +330,18 @@ std::optional<AllocationState> LatticePoint::get(mlir::Value val) const {
   return it->second;
 }
 
+static mlir::Value lookThroughDeclaresAndConverts(mlir::Value value) {
+  while (mlir::Operation *op = value.getDefiningOp()) {
+    if (auto declareOp = llvm::dyn_cast<fir::DeclareOp>(op))
+      value = declareOp.getMemref();
+    else if (auto convertOp = llvm::dyn_cast<fir::ConvertOp>(op))
+      value = convertOp->getOperand(0);
+    else
+      return value;
+  }
+  return value;
+}
+
 mlir::LogicalResult AllocationAnalysis::visitOperation(
     mlir::Operation *op, const LatticePoint &before, LatticePoint *after) {
   LLVM_DEBUG(llvm::dbgs() << "StackArrays: Visiting operation: " << *op
@@ -363,10 +375,10 @@ mlir::LogicalResult AllocationAnalysis::visitOperation(
     mlir::Value operand = op->getOperand(0);
 
     // Note: StackArrays is scheduled in the pass pipeline after lowering hlfir
-    // to fir. Therefore, we only need to handle `fir::DeclareOp`s.
-    if (auto declareOp =
-            llvm::dyn_cast_if_present<fir::DeclareOp>(operand.getDefiningOp()))
-      operand = declareOp.getMemref();
+    // to fir. Therefore, we only need to handle `fir::DeclareOp`s. Also look
+    // past converts in case the pointer was changed between different pointer
+    // types.
+    operand = lookThroughDeclaresAndConverts(operand);
 
     std::optional<AllocationState> operandState = before.get(operand);
     if (operandState && *operandState == AllocationState::Allocated) {
@@ -535,17 +547,12 @@ AllocMemConversion::matchAndRewrite(fir::AllocMemOp allocmem,
 
   // remove freemem operations
   llvm::SmallVector<mlir::Operation *> erases;
-  for (mlir::Operation *user : allocmem.getOperation()->getUsers()) {
-    if (auto declareOp = mlir::dyn_cast_if_present<fir::DeclareOp>(user)) {
-      for (mlir::Operation *user : declareOp->getUsers()) {
-        if (mlir::isa<fir::FreeMemOp>(user))
-          erases.push_back(user);
-      }
-    }
-
-    if (mlir::isa<fir::FreeMemOp>(user))
-      erases.push_back(user);
-  }
+  mlir::Operation *parent = allocmem->getParentOp();
+  // TODO: this shouldn't need to be re-calculated for every allocmem
+  parent->walk([&](fir::FreeMemOp freeOp) {
+    if (lookThroughDeclaresAndConverts(freeOp->getOperand(0)) == allocmem)
+      erases.push_back(freeOp);
+  });
 
   // now we are done iterating the users, it is safe to mutate them
   for (mlir::Operation *erase : erases)

flang/test/Transforms/stack-arrays.fir

@@ -379,7 +379,8 @@ func.func @placement_loop_declare() {
     %3 = arith.addi %c1, %c2 : index
     // operand is now available
     %4 = fir.allocmem !fir.array<?xi32>, %3
-    %5 = fir.declare %4 {uniq_name = "temp"} : (!fir.heap<!fir.array<?xi32>>) -> !fir.heap<!fir.array<?xi32>>
+    %shape = fir.shape %3 : (index) -> !fir.shape<1>
+    %5 = fir.declare %4(%shape) {uniq_name = "temp"} : (!fir.heap<!fir.array<?xi32>>, !fir.shape<1>) -> !fir.heap<!fir.array<?xi32>>
     // ...
     fir.freemem %5 : !fir.heap<!fir.array<?xi32>>
     fir.result %3, %c1_i32 : index, i32
@@ -400,3 +401,20 @@ func.func @placement_loop_declare() {
 // CHECK-NEXT:   }
 // CHECK-NEXT:   return
 // CHECK-NEXT: }
+
+// Can we look through fir.convert and fir.declare?
+func.func @lookthrough() {
+  %0 = fir.allocmem !fir.array<42xi32>
+  %c42 = arith.constant 42 : index
+  %shape = fir.shape %c42 : (index) -> !fir.shape<1>
+  %1 = fir.declare %0(%shape) {uniq_name = "name"} : (!fir.heap<!fir.array<42xi32>>, !fir.shape<1>) -> !fir.heap<!fir.array<42xi32>>
+  %2 = fir.convert %1 : (!fir.heap<!fir.array<42xi32>>) -> !fir.ref<!fir.array<42xi32>>
+  // use the ref so the converts aren't folded
+  %3 = fir.load %2 : !fir.ref<!fir.array<42xi32>>
+  %4 = fir.convert %2 : (!fir.ref<!fir.array<42xi32>>) -> !fir.heap<!fir.array<42xi32>>
+  fir.freemem %4 : !fir.heap<!fir.array<42xi32>>
+  return
+}
+// CHECK: func.func @lookthrough() {
+// CHECK:     fir.alloca !fir.array<42xi32>
+// CHECK-NOT: fir.freemem

llvm/Maintainers.md

@@ -45,8 +45,12 @@ Matthew.Arsenault@amd.com, arsenm2@gmail.com (email), [arsenm](https://github.co
 
 #### Inlining
 
-Chandler Carruth \
-chandlerc@gmail.com, chandlerc@google.com (email), [chandlerc](https://github.com/chandlerc) (GitHub)
+Arthur Eubanks \
+aeubanks@google.com (email), [aeubanks](https://github.com/aeubanks) (GitHub) \
+Mircea Trofin (esp. ML inliner) \
+mtrofin@google.com (email), [mtrofin](https://github.com/mtrofin) (GitHub) \
+Kazu Hirata (esp. module inliner and inline order) \
+kazu@google.com (email), [kazutakahirata](https://github.com/kazutakahirata) (GitHub)
 
 #### InstCombine, InstSimplify, ValueTracking, ConstantFold
 
@@ -65,6 +69,11 @@ mail@justinbogner.com (email), [bogner](https://github.com/bogner) (GitHub)
 Diego Novillo \
 dnovillo@google.com (email), [dnovillo](https://github.com/dnovillo) (GitHub)
 
+#### New pass manager, CGSCC, LazyCallGraph
+
+Arthur Eubanks \
+aeubanks@google.com (email), [aeubanks](https://github.com/aeubanks) (GitHub)
+
 #### LoopStrengthReduce
 
 Quentin Colombet \
@@ -462,7 +471,7 @@ sabre@nondot.org (email), [lattner](https://github.com/lattner) (GitHub), clattn
 
 Paul C. Anagnostopoulos (paul@windfall.com, [Paul-C-Anagnostopoulos](https://github.com/Paul-C-Anagnostopoulos)) -- TableGen \
 Justin Bogner (mail@justinbogner.com, [bogner](https://github.com/bogner)) -- SelectionDAG \
-Chandler Carruth (chandlerc@gmail.com, chandlerc@google.com, [chandlerc](https://github.com/chandlerc)) -- ADT, Support \
+Chandler Carruth (chandlerc@gmail.com, chandlerc@google.com, [chandlerc](https://github.com/chandlerc)) -- ADT, Support, Inlining \
 Peter Collingbourne (peter@pcc.me.uk, [pcc](https://github.com/pcc)) -- LTO \
 Evan Cheng (evan.cheng@apple.com) -- Parts of code generator not covered by someone else \
 Jake Ehrlich (jakehehrlich@google.com, [jakehehrlich](https://github.com/jakehehrlich)) -- llvm-objcopy and ObjCopy library \

llvm/include/llvm/ADT/STLFunctionalExtras.h

@@ -36,8 +36,8 @@ namespace llvm {
 /// a function_ref.
 template<typename Fn> class function_ref;
 
-template<typename Ret, typename ...Params>
-class function_ref<Ret(Params...)> {
+template <typename Ret, typename... Params>
+class LLVM_GSL_POINTER function_ref<Ret(Params...)> {
   Ret (*callback)(intptr_t callable, Params ...params) = nullptr;
   intptr_t callable;
 

llvm/lib/Analysis/ScalarEvolution.cpp

@@ -226,7 +226,7 @@ static cl::opt<unsigned> RangeIterThreshold(
 
 static cl::opt<unsigned> MaxLoopGuardCollectionDepth(
     "scalar-evolution-max-loop-guard-collection-depth", cl::Hidden,
-    cl::desc("Maximum depth for recrusive loop guard collection"), cl::init(1));
+    cl::desc("Maximum depth for recursive loop guard collection"), cl::init(1));
 
 static cl::opt<bool>
 ClassifyExpressions("scalar-evolution-classify-expressions",

llvm/lib/CodeGen/CodeGenPrepare.cpp

@@ -152,7 +152,7 @@ static cl::opt<bool>
 
 static cl::opt<bool>
     EnableAndCmpSinking("enable-andcmp-sinking", cl::Hidden, cl::init(true),
-                        cl::desc("Enable sinkinig and/cmp into branches."));
+                        cl::desc("Enable sinking and/cmp into branches."));
 
 static cl::opt<bool> DisableStoreExtract(
     "disable-cgp-store-extract", cl::Hidden, cl::init(false),

llvm/lib/CodeGen/MIRSampleProfile.cpp

@@ -46,8 +46,9 @@ static cl::opt<bool> ShowFSBranchProb(
     cl::desc("Print setting flow sensitive branch probabilities"));
 static cl::opt<unsigned> FSProfileDebugProbDiffThreshold(
     "fs-profile-debug-prob-diff-threshold", cl::init(10),
-    cl::desc("Only show debug message if the branch probility is greater than "
-             "this value (in percentage)."));
+    cl::desc(
+        "Only show debug message if the branch probability is greater than "
+        "this value (in percentage)."));
 
 static cl::opt<unsigned> FSProfileDebugBWThreshold(
     "fs-profile-debug-bw-threshold", cl::init(10000),

llvm/lib/CodeGen/MachineBlockPlacement.cpp

@@ -149,7 +149,7 @@ static cl::opt<unsigned> JumpInstCost("jump-inst-cost",
 static cl::opt<bool>
     TailDupPlacement("tail-dup-placement",
                      cl::desc("Perform tail duplication during placement. "
-                              "Creates more fallthrough opportunites in "
+                              "Creates more fallthrough opportunities in "
                               "outline branches."),
                      cl::init(true), cl::Hidden);
 

llvm/lib/CodeGen/MachineBranchProbabilityInfo.cpp

@@ -29,7 +29,7 @@ namespace llvm {
 cl::opt<unsigned>
     StaticLikelyProb("static-likely-prob",
                      cl::desc("branch probability threshold in percentage"
-                              "to be considered very likely"),
+                              " to be considered very likely"),
                      cl::init(80), cl::Hidden);
 
 cl::opt<unsigned> ProfileLikelyProb(

llvm/lib/CodeGen/RegAllocGreedy.cpp

@@ -140,7 +140,7 @@ static cl::opt<bool> GreedyReverseLocalAssignment(
 static cl::opt<unsigned> SplitThresholdForRegWithHint(
     "split-threshold-for-reg-with-hint",
     cl::desc("The threshold for splitting a virtual register with a hint, in "
-             "percentate"),
+             "percentage"),
     cl::init(75), cl::Hidden);
 
 static RegisterRegAlloc greedyRegAlloc("greedy", "greedy register allocator",

llvm/lib/CodeGen/RegisterCoalescer.cpp

@@ -113,7 +113,7 @@ static cl::opt<unsigned> LargeIntervalSizeThreshold(
 
 static cl::opt<unsigned> LargeIntervalFreqThreshold(
     "large-interval-freq-threshold", cl::Hidden,
-    cl::desc("For a large interval, if it is coalesed with other live "
+    cl::desc("For a large interval, if it is coalesced with other live "
              "intervals many times more than the threshold, stop its "
              "coalescing to control the compile time. "),
     cl::init(256));

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

@@ -141,7 +141,7 @@ static cl::opt<bool> EnableReduceLoadOpStoreWidth(
 static cl::opt<bool> ReduceLoadOpStoreWidthForceNarrowingProfitable(
     "combiner-reduce-load-op-store-width-force-narrowing-profitable",
     cl::Hidden, cl::init(false),
-    cl::desc("DAG combiner force override the narrowing profitable check when"
+    cl::desc("DAG combiner force override the narrowing profitable check when "
              "reducing the width of load/op/store sequences"));
 
 static cl::opt<bool> EnableShrinkLoadReplaceStoreWithStore(

llvm/lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp

@@ -1777,6 +1777,31 @@ void VectorLegalizer::ExpandUINT_TO_FLOAT(SDNode *Node,
   assert((BW == 64 || BW == 32) &&
          "Elements in vector-UINT_TO_FP must be 32 or 64 bits wide");
 
+  // If STRICT_/FMUL is not supported by the target (in case of f16) replace the
+  // UINT_TO_FP with a larger float and round to the smaller type
+  if ((!IsStrict && !TLI.isOperationLegalOrCustom(ISD::FMUL, DstVT)) ||
+      (IsStrict && !TLI.isOperationLegalOrCustom(ISD::STRICT_FMUL, DstVT))) {
+    EVT FPVT = BW == 32 ? MVT::f32 : MVT::f64;
+    SDValue UIToFP;
+    SDValue Result;
+    SDValue TargetZero = DAG.getIntPtrConstant(0, DL, /*isTarget=*/true);
+    EVT FloatVecVT = SrcVT.changeVectorElementType(FPVT);
+    if (IsStrict) {
+      UIToFP = DAG.getNode(ISD::STRICT_UINT_TO_FP, DL, {FloatVecVT, MVT::Other},
+                           {Node->getOperand(0), Src});
+      Result = DAG.getNode(ISD::STRICT_FP_ROUND, DL, {DstVT, MVT::Other},
+                           {Node->getOperand(0), UIToFP, TargetZero});
+      Results.push_back(Result);
+      Results.push_back(Result.getValue(1));
+    } else {
+      UIToFP = DAG.getNode(ISD::UINT_TO_FP, DL, FloatVecVT, Src);
+      Result = DAG.getNode(ISD::FP_ROUND, DL, DstVT, UIToFP, TargetZero);
+      Results.push_back(Result);
+    }
+
+    return;
+  }
+
   SDValue HalfWord = DAG.getConstant(BW / 2, DL, SrcVT);
 
   // Constants to clear the upper part of the word.

llvm/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp

@@ -125,9 +125,9 @@ static cl::opt<int> MaxReorderWindow(
   cl::desc("Number of instructions to allow ahead of the critical path "
            "in sched=list-ilp"));
 
-static cl::opt<unsigned> AvgIPC(
-  "sched-avg-ipc", cl::Hidden, cl::init(1),
-  cl::desc("Average inst/cycle whan no target itinerary exists."));
+static cl::opt<unsigned>
+    AvgIPC("sched-avg-ipc", cl::Hidden, cl::init(1),
+           cl::desc("Average inst/cycle when no target itinerary exists."));
 
 namespace {
 

llvm/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp

@@ -43,9 +43,9 @@ STATISTIC(LoadsClustered, "Number of loads clustered together");
 // without a target itinerary. The choice of number here has more to do with
 // balancing scheduler heuristics than with the actual machine latency.
 static cl::opt<int> HighLatencyCycles(
-  "sched-high-latency-cycles", cl::Hidden, cl::init(10),
-  cl::desc("Roughly estimate the number of cycles that 'long latency'"
-           "instructions take for targets with no itinerary"));
+    "sched-high-latency-cycles", cl::Hidden, cl::init(10),
+    cl::desc("Roughly estimate the number of cycles that 'long latency' "
+             "instructions take for targets with no itinerary"));
 
 ScheduleDAGSDNodes::ScheduleDAGSDNodes(MachineFunction &mf)
     : ScheduleDAG(mf), InstrItins(mf.getSubtarget().getInstrItineraryData()) {}

llvm/lib/Passes/PassBuilderPipelines.cpp

@@ -189,9 +189,9 @@ static cl::opt<bool> EnableGlobalAnalyses(
     "enable-global-analyses", cl::init(true), cl::Hidden,
     cl::desc("Enable inter-procedural analyses"));
 
-static cl::opt<bool>
-    RunPartialInlining("enable-partial-inlining", cl::init(false), cl::Hidden,
-                       cl::desc("Run Partial inlinining pass"));
+static cl::opt<bool> RunPartialInlining("enable-partial-inlining",
+                                        cl::init(false), cl::Hidden,
+                                        cl::desc("Run Partial inlining pass"));
 
 static cl::opt<bool> ExtraVectorizerPasses(
     "extra-vectorizer-passes", cl::init(false), cl::Hidden,
@@ -264,7 +264,7 @@ static cl::opt<bool>
 static cl::opt<bool> FlattenedProfileUsed(
     "flattened-profile-used", cl::init(false), cl::Hidden,
     cl::desc("Indicate the sample profile being used is flattened, i.e., "
-             "no inline hierachy exists in the profile"));
+             "no inline hierarchy exists in the profile"));
 
 static cl::opt<bool> EnableOrderFileInstrumentation(
     "enable-order-file-instrumentation", cl::init(false), cl::Hidden,

llvm/lib/Target/AArch64/AArch64InstrFormats.td

@@ -1804,7 +1804,9 @@ class TMSystemException<bits<3> op1, string asm, list<dag> pattern>
 }
 
 class APASI : SimpleSystemI<0, (ins GPR64:$Xt), "apas", "\t$Xt">, Sched<[]> {
+  bits<5> Xt;
   let Inst{20-5} = 0b0111001110000000;
+  let Inst{4-0} = Xt;
   let DecoderNamespace = "APAS";
 }
 

llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp

@@ -4748,7 +4748,7 @@ InstructionCost AArch64TTIImpl::getShuffleCost(
             LT.second.getVectorNumElements() / 2) {
       if (Index == 0)
         return 0;
-      if (Index == LT.second.getVectorNumElements() / 2)
+      if (Index == (int)LT.second.getVectorNumElements() / 2)
         return 1;
     }
     Kind = TTI::SK_PermuteSingleSrc;

llvm/lib/Target/X86/X86ISelLowering.cpp

@@ -94,7 +94,7 @@ static cl::opt<int> BrMergingCcmpBias(
 
 static cl::opt<bool>
     WidenShift("x86-widen-shift", cl::init(true),
-               cl::desc("Replacte narrow shifts with wider shifts."),
+               cl::desc("Replace narrow shifts with wider shifts."),
                cl::Hidden);
 
 static cl::opt<int> BrMergingLikelyBias(

llvm/lib/Transforms/IPO/FunctionSpecialization.cpp

@@ -66,19 +66,19 @@ static cl::opt<unsigned> MaxCodeSizeGrowth(
     "Maximum codesize growth allowed per function"));
 
 static cl::opt<unsigned> MinCodeSizeSavings(
-    "funcspec-min-codesize-savings", cl::init(20), cl::Hidden, cl::desc(
-    "Reject specializations whose codesize savings are less than this"
-    "much percent of the original function size"));
+    "funcspec-min-codesize-savings", cl::init(20), cl::Hidden,
+    cl::desc("Reject specializations whose codesize savings are less than this "
+             "much percent of the original function size"));
 
 static cl::opt<unsigned> MinLatencySavings(
     "funcspec-min-latency-savings", cl::init(40), cl::Hidden,
-    cl::desc("Reject specializations whose latency savings are less than this"
+    cl::desc("Reject specializations whose latency savings are less than this "
              "much percent of the original function size"));
 
 static cl::opt<unsigned> MinInliningBonus(
-    "funcspec-min-inlining-bonus", cl::init(300), cl::Hidden, cl::desc(
-    "Reject specializations whose inlining bonus is less than this"
-    "much percent of the original function size"));
+    "funcspec-min-inlining-bonus", cl::init(300), cl::Hidden,
+    cl::desc("Reject specializations whose inlining bonus is less than this "
+             "much percent of the original function size"));
 
 static cl::opt<bool> SpecializeOnAddress(
     "funcspec-on-address", cl::init(false), cl::Hidden, cl::desc(

llvm/lib/Transforms/IPO/GlobalOpt.cpp

@@ -105,7 +105,7 @@ static cl::opt<int> ColdCCRelFreq(
     "coldcc-rel-freq", cl::Hidden, cl::init(2),
     cl::desc(
         "Maximum block frequency, expressed as a percentage of caller's "
-        "entry frequency, for a call site to be considered cold for enabling"
+        "entry frequency, for a call site to be considered cold for enabling "
         "coldcc"));
 
 /// Is this global variable possibly used by a leak checker as a root?  If so,

llvm/lib/Transforms/IPO/OpenMPOpt.cpp

@@ -129,7 +129,7 @@ static cl::opt<bool> PrintModuleBeforeOptimizations(
 
 static cl::opt<bool> AlwaysInlineDeviceFunctions(
     "openmp-opt-inline-device",
-    cl::desc("Inline all applicible functions on the device."), cl::Hidden,
+    cl::desc("Inline all applicable functions on the device."), cl::Hidden,
     cl::init(false));
 
 static cl::opt<bool>