Merged main:beffc821e829 into amd-gfx:010f535d2674

Local branch amd-gfx 010f535 Merged main:6eee238975e4 into amd-gfx:4d4ec95d174b
Remote branch main beffc82 [CodeLayout] CDSortImpl: remove HotChains and remove linear-time erase_value from mergeChains (llvm#69276)

Author: SC llvm team

compiler-rt/lib/sanitizer_common/sanitizer_platform.h

@@ -303,7 +303,15 @@
 #    define SANITIZER_MMAP_RANGE_SIZE FIRST_32_SECOND_64(1ULL << 32, 1ULL << 40)
 #  endif
 #elif SANITIZER_RISCV64
-#  define SANITIZER_MMAP_RANGE_SIZE FIRST_32_SECOND_64(1ULL << 32, 1ULL << 47)
+// FIXME: Rather than hardcoding the VMA here, we should rely on
+// GetMaxUserVirtualAddress(). This will require some refactoring though since
+// many places either hardcode some value or SANITIZER_MMAP_RANGE_SIZE is
+// assumed to be some constant integer.
+#  if SANITIZER_FUCHSIA
+#    define SANITIZER_MMAP_RANGE_SIZE (1ULL << 38)
+#  else
+#    define SANITIZER_MMAP_RANGE_SIZE FIRST_32_SECOND_64(1ULL << 32, 1ULL << 47)
+#  endif
 #elif defined(__aarch64__)
 #  if SANITIZER_APPLE
 #    if SANITIZER_OSX || SANITIZER_IOSSIM

lldb/examples/synthetic/gnu_libstdcpp.py

@@ -914,6 +914,11 @@ def get_variant_npos_value(index_byte_size):
     if index == npos_value:
         return " No Value"
 
+    # Invalid index can happen when the variant is not initialized yet.
+    template_arg_count = data_obj.GetType().GetNumberOfTemplateArguments()
+    if index >= template_arg_count:
+        return " <Invalid>"
+
     active_type = data_obj.GetType().GetTemplateArgumentType(index)
     return f" Active Type = {active_type.GetDisplayTypeName()} "
 

lldb/source/Plugins/TypeSystem/Clang/TypeSystemClang.cpp

@@ -7182,6 +7182,9 @@ GetNthTemplateArgument(const clang::ClassTemplateSpecializationDecl *decl,
   // Note that 'idx' counts from the beginning of all template arguments
   // (including the ones preceding the parameter pack).
   const auto &pack = args[last_idx];
+  if (idx >= pack.pack_size())
+    return nullptr;
+
   const size_t pack_idx = idx - last_idx;
   assert(pack_idx < pack.pack_size() && "parameter pack index out-of-bounds");
   return &pack.pack_elements()[pack_idx];

lldb/test/API/functionalities/data-formatter/data-formatter-stl/libstdcpp/variant/TestDataFormatterLibStdcxxVariant.py

@@ -71,3 +71,29 @@ def test_with_run_command(self):
             substrs=["v_many_types_no_value =  No Value"],
         )
         """
+
+    @add_test_categories(["libstdcxx"])
+    def test_invalid_variant_index(self):
+        """Test LibStdC++ data formatter for std::variant with invalid index."""
+        self.build()
+
+        (self.target, self.process, thread, bkpt) = lldbutil.run_to_source_breakpoint(
+            self, "// break here", lldb.SBFileSpec("main.cpp", False)
+        )
+
+        lldbutil.continue_to_breakpoint(self.process, bkpt)
+
+        self.expect(
+            "frame variable v1",
+            substrs=["v1 =  Active Type = int  {", "Value = 12", "}"],
+        )
+
+        var_v1 = thread.frames[0].FindVariable("v1")
+        var_v1_raw_obj = var_v1.GetNonSyntheticValue()
+        index_obj = var_v1_raw_obj.GetChildMemberWithName("_M_index")
+        self.assertTrue(index_obj and index_obj.IsValid())
+
+        INVALID_INDEX = "100"
+        index_obj.SetValueFromCString(INVALID_INDEX)
+
+        self.expect("frame variable v1", substrs=["v1 =  <Invalid>"])

llvm/docs/LangRef.rst

@@ -309,7 +309,7 @@ added in the future:
     prototype of all callees to exactly match the prototype of the
     function definition. Furthermore the inliner doesn't consider such function
     calls for inlining.
-"``cc 10``" - GHC convention
+"``ghccc``" - GHC convention
     This calling convention has been implemented specifically for use by
     the `Glasgow Haskell Compiler (GHC) <http://www.haskell.org/ghc>`_.
     It passes everything in registers, going to extremes to achieve this

llvm/docs/TestSuiteGuide.md

@@ -177,7 +177,7 @@ benchmarks. CMake can print a list of them:
 - `TEST_SUITE_FORTRAN`
 
   Activate that Fortran tests. This is a work in progress. More information can be
-  found in the [Flang documentation](https://flang.llvm.org/docs/html/FortranLLVMTestSuite.html)
+  found in the [Flang documentation](https://flang.llvm.org/docs/FortranLLVMTestSuite.html)
 
 - `TEST_SUITE_RUN_UNDER`
 

llvm/docs/WritingAnLLVMPass.rst

@@ -1020,8 +1020,7 @@ series of passes:
    instead of traversing the entire program.  It reduces the memory consumption
    of compiler, because, for example, only one `DominatorSet
    <https://llvm.org/doxygen/classllvm_1_1DominatorSet.html>`_ needs to be
-   calculated at a time.  This also makes it possible to implement some
-   :ref:`interesting enhancements <writing-an-llvm-pass-SMP>` in the future.
+   calculated at a time.
 
 The effectiveness of the ``PassManager`` is influenced directly by how much
 information it has about the behaviors of the passes it is scheduling.  For
@@ -1350,29 +1349,3 @@ some with solutions, some without.
 Hopefully these tips will help with common case debugging situations.  If you'd
 like to contribute some tips of your own, just contact `Chris
 <mailto:[email protected]>`_.
-
-Future extensions planned
--------------------------
-
-Although the LLVM Pass Infrastructure is very capable as it stands, and does
-some nifty stuff, there are things we'd like to add in the future.  Here is
-where we are going:
-
-.. _writing-an-llvm-pass-SMP:
-
-Multithreaded LLVM
-^^^^^^^^^^^^^^^^^^
-
-Multiple CPU machines are becoming more common and compilation can never be
-fast enough: obviously we should allow for a multithreaded compiler.  Because
-of the semantics defined for passes above (specifically they cannot maintain
-state across invocations of their ``run*`` methods), a nice clean way to
-implement a multithreaded compiler would be for the ``PassManager`` class to
-create multiple instances of each pass object, and allow the separate instances
-to be hacking on different parts of the program at the same time.
-
-This implementation would prevent each of the passes from having to implement
-multithreaded constructs, requiring only the LLVM core to have locking in a few
-places (for global resources).  Although this is a simple extension, we simply
-haven't had time (or multiprocessor machines, thus a reason) to implement this.
-Despite that, we have kept the LLVM passes SMP ready, and you should too.

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 477827
+#define LLVM_MAIN_REVISION 477835
 
 /* Define if LLVM_ENABLE_DUMP is enabled */
 #cmakedefine LLVM_ENABLE_DUMP

llvm/lib/Target/M68k/MCTargetDesc/M68kMCCodeEmitter.cpp

@@ -204,7 +204,7 @@ void M68kMCCodeEmitter::encodeInstruction(const MCInst &MI,
                                           SmallVectorImpl<MCFixup> &Fixups,
                                           const MCSubtargetInfo &STI) const {
   LLVM_DEBUG(dbgs() << "EncodeInstruction: " << MCII.getName(MI.getOpcode())
-                    << "(" << Opcode << ")\n");
+                    << "(" << MI.getOpcode() << ")\n");
   (void)MCII;
 
   // Try using the new method first.

llvm/lib/Target/RISCV/GISel/RISCVInstructionSelector.cpp

@@ -194,6 +194,14 @@ bool RISCVInstructionSelector::select(MachineInstr &MI) {
   }
   case TargetOpcode::G_SEXT_INREG:
     return selectSExtInreg(MI, MIB);
+  case TargetOpcode::G_FRAME_INDEX: {
+    // TODO: We may want to replace this code with the SelectionDAG patterns,
+    // which fail to get imported because it uses FrameAddrRegImm, which is a
+    // ComplexPattern
+    MI.setDesc(TII.get(RISCV::ADDI));
+    MI.addOperand(MachineOperand::CreateImm(0));
+    return constrainSelectedInstRegOperands(MI, TII, TRI, RBI);
+  }
   case TargetOpcode::G_SELECT:
     return selectSelect(MI, MIB, MRI);
   default:

llvm/lib/Target/RISCV/GISel/RISCVRegisterBankInfo.cpp

@@ -135,6 +135,7 @@ RISCVRegisterBankInfo::getInstrMapping(const MachineInstr &MI) const {
   case TargetOpcode::G_STORE:
     break;
   case TargetOpcode::G_CONSTANT:
+  case TargetOpcode::G_FRAME_INDEX:
   case TargetOpcode::G_GLOBAL_VALUE:
   case TargetOpcode::G_BRCOND:
     OperandsMapping = getOperandsMapping({GPRValueMapping, nullptr});

llvm/lib/Transforms/Utils/CodeLayout.cpp

@@ -1025,10 +1025,6 @@ class CDSortImpl {
     // Merge pairs of chains while improving the objective.
     mergeChainPairs();
 
-    LLVM_DEBUG(dbgs() << "Cache-directed function sorting reduced the number"
-                      << " of chains from " << NumNodes << " to "
-                      << HotChains.size() << "\n");
-
     // Collect nodes from all the chains.
     return concatChains();
   }
@@ -1074,16 +1070,13 @@ class CDSortImpl {
 
     // Initialize chains.
     AllChains.reserve(NumNodes);
-    HotChains.reserve(NumNodes);
     for (NodeT &Node : AllNodes) {
       // Adjust execution counts.
       Node.ExecutionCount = std::max(Node.ExecutionCount, Node.inCount());
       Node.ExecutionCount = std::max(Node.ExecutionCount, Node.outCount());
       // Create chain.
       AllChains.emplace_back(Node.Index, &Node);
       Node.CurChain = &AllChains.back();
-      if (Node.ExecutionCount > 0)
-        HotChains.push_back(&AllChains.back());
     }
 
     // Initialize chain edges.
@@ -1116,8 +1109,12 @@ class CDSortImpl {
     std::set<ChainEdge *, decltype(GainComparator)> Queue(GainComparator);
 
     // Insert the edges into the queue.
-    for (ChainT *ChainPred : HotChains) {
-      for (const auto &[_, Edge] : ChainPred->Edges) {
+    [[maybe_unused]] size_t NumActiveChains = 0;
+    for (NodeT &Node : AllNodes) {
+      if (Node.ExecutionCount == 0)
+        continue;
+      ++NumActiveChains;
+      for (const auto &[_, Edge] : Node.CurChain->Edges) {
         // Ignore self-edges.
         if (Edge->isSelfEdge())
           continue;
@@ -1152,6 +1149,7 @@ class CDSortImpl {
       MergeGainT BestGain = BestEdge->getMergeGain();
       mergeChains(BestSrcChain, BestDstChain, BestGain.mergeOffset(),
                   BestGain.mergeType());
+      --NumActiveChains;
 
       // Insert newly created edges into the queue.
       for (const auto &[_, Edge] : BestSrcChain->Edges) {
@@ -1167,6 +1165,10 @@ class CDSortImpl {
           Queue.insert(Edge);
       }
     }
+
+    LLVM_DEBUG(dbgs() << "Cache-directed function sorting reduced the number"
+                      << " of chains from " << NumNodes << " to "
+                      << NumActiveChains << "\n");
   }
 
   /// Compute the gain of merging two chains.
@@ -1301,9 +1303,6 @@ class CDSortImpl {
     // Merge the edges.
     Into->mergeEdges(From);
     From->clear();
-
-    // Remove the chain from the list of active chains.
-    llvm::erase_value(HotChains, From);
   }
 
   /// Concatenate all chains into the final order.
@@ -1370,9 +1369,6 @@ class CDSortImpl {
   /// All edges between the chains.
   std::vector<ChainEdge> AllEdges;
 
-  /// Active chains. The vector gets updated at runtime when chains are merged.
-  std::vector<ChainT *> HotChains;
-
   /// The total number of samples in the graph.
   uint64_t TotalSamples{0};
 

llvm/test/CodeGen/RISCV/GlobalISel/instruction-select/frame-index-rv32.mir

@@ -0,0 +1,32 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -mtriple=riscv32 -run-pass=instruction-select %s -o - \
+# RUN: | FileCheck %s
+--- |
+  define ptr @frame_index() {
+  entry:
+    %x = alloca i32, align 4
+    ret ptr %x
+  }
+
+...
+---
+name:            frame_index
+legalized:       true
+regBankSelected: true
+registers:
+  - { id: 0, class: gprb, preferred-register: '' }
+stack:
+  - { id: 0, name: x, type: default, offset: 0, size: 4, alignment: 4,
+      stack-id: default, callee-saved-register: '', callee-saved-restored: true,
+      debug-info-variable: '', debug-info-expression: '', debug-info-location: '' }
+body:             |
+  bb.1.entry:
+    ; CHECK-LABEL: name: frame_index
+    ; CHECK: [[ADDI:%[0-9]+]]:gpr = ADDI %stack.0.x, 0
+    ; CHECK-NEXT: $x10 = COPY [[ADDI]]
+    ; CHECK-NEXT: PseudoRET implicit $x10
+    %0:gprb(p0) = G_FRAME_INDEX %stack.0.x
+    $x10 = COPY %0(p0)
+    PseudoRET implicit $x10
+
+...

llvm/test/CodeGen/RISCV/GlobalISel/instruction-select/frame-index-rv64.mir

@@ -0,0 +1,32 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -mtriple=riscv64 -run-pass=instruction-select %s -o - \
+# RUN: | FileCheck %s
+--- |
+  define ptr @frame_index() {
+  entry:
+    %x = alloca i32, align 4
+    ret ptr %x
+  }
+
+...
+---
+name:            frame_index
+legalized:       true
+regBankSelected: true
+registers:
+  - { id: 0, class: gprb, preferred-register: '' }
+stack:
+  - { id: 0, name: x, type: default, offset: 0, size: 4, alignment: 4,
+      stack-id: default, callee-saved-register: '', callee-saved-restored: true,
+      debug-info-variable: '', debug-info-expression: '', debug-info-location: '' }
+body:             |
+  bb.1.entry:
+    ; CHECK-LABEL: name: frame_index
+    ; CHECK: [[ADDI:%[0-9]+]]:gpr = ADDI %stack.0.x, 0
+    ; CHECK-NEXT: $x10 = COPY [[ADDI]]
+    ; CHECK-NEXT: PseudoRET implicit $x10
+    %0:gprb(p0) = G_FRAME_INDEX %stack.0.x
+    $x10 = COPY %0(p0)
+    PseudoRET implicit $x10
+
+...

llvm/test/CodeGen/RISCV/GlobalISel/regbankselect/frame-index.mir

@@ -0,0 +1,33 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -mtriple=riscv32 -run-pass=regbankselect %s -o - \
+# RUN: | FileCheck %s
+# RUN: llc -mtriple=riscv64 -run-pass=regbankselect %s -o - \
+# RUN: | FileCheck %s
+--- |
+  define ptr @frame_index() {
+  entry:
+    %x = alloca i32, align 4
+    ret ptr %x
+  }
+
+...
+---
+name:            frame_index
+legalized:       true
+registers:
+  - { id: 0, class: _, preferred-register: '' }
+stack:
+  - { id: 0, name: x, type: default, offset: 0, size: 4, alignment: 4,
+      stack-id: default, callee-saved-register: '', callee-saved-restored: true,
+      debug-info-variable: '', debug-info-expression: '', debug-info-location: '' }
+body:             |
+  bb.1.entry:
+    ; CHECK-LABEL: name: frame_index
+    ; CHECK: [[FRAME_INDEX:%[0-9]+]]:gprb(p0) = G_FRAME_INDEX %stack.0.x
+    ; CHECK-NEXT: $x10 = COPY [[FRAME_INDEX]](p0)
+    ; CHECK-NEXT: PseudoRET implicit $x10
+    %0:_(p0) = G_FRAME_INDEX %stack.0.x
+    $x10 = COPY %0(p0)
+    PseudoRET implicit $x10
+
+...