Merged main:08136d822c73 into amd-gfx:5c4b01290188

Local branch amd-gfx 5c4b012 Merged main:3ff7d51eb823 into amd-gfx:fb1d7733ee1c
Remote branch main 08136d8 [DWARFLinkerParallel] Add support of accelerator tables to DWARFLinkerParallel.

libcxx/test/CMakeLists.txt

@@ -54,22 +54,22 @@ if (MSVC)
   set(dbg_include "")
 
   if (NOT CMAKE_MSVC_RUNTIME_LIBRARY OR CMAKE_MSVC_RUNTIME_LIBRARY MATCHES "DLL$")
-    set(crt_lib "msvcrt")
+    set(fms_runtime_lib "dll")
     set(cxx_lib "msvcprt")
   else()
-    set(crt_lib "libcmt")
+    set(fms_runtime_lib "static")
     set(cxx_lib "libcpmt")
   endif()
 
   if ((NOT CMAKE_MSVC_RUNTIME_LIBRARY AND uppercase_CMAKE_BUILD_TYPE STREQUAL "DEBUG")
       OR (CMAKE_MSVC_RUNTIME_LIBRARY MATCHES "Debug"))
-    set(dbg_include " -D_DEBUG -include set_windows_crt_report_mode.h")
-    set(crt_lib "${crt_lib}d")
+    set(dbg_include " -include set_windows_crt_report_mode.h")
+    set(fms_runtime_lib "${fms_runtime_lib}_dbg")
     set(cxx_lib "${cxx_lib}d")
   endif()
 
   serialize_lit_param(dbg_include "\"${dbg_include}\"")
-  serialize_lit_param(crt_lib "\"${crt_lib}\"")
+  serialize_lit_param(fms_runtime_lib "\"${fms_runtime_lib}\"")
   serialize_lit_param(cxx_lib "\"${cxx_lib}\"")
 endif()
 

libcxx/test/configs/llvm-libc++-shared-clangcl.cfg.in

@@ -5,10 +5,10 @@ lit_config.load_config(config, '@CMAKE_CURRENT_BINARY_DIR@/cmake-bridge.cfg')
 
 config.substitutions.append(('%{flags}', '--driver-mode=g++'))
 config.substitutions.append(('%{compile_flags}',
-    '-nostdinc++ -I %{include} -I %{target-include} -I %{libcxx}/test/support -D_CRT_SECURE_NO_WARNINGS -D_CRT_NONSTDC_NO_WARNINGS -D_CRT_STDIO_ISO_WIDE_SPECIFIERS -DNOMINMAX' + config.dbg_include
+    '-fms-runtime-lib=' + config.fms_runtime_lib + ' -nostdinc++ -I %{include} -I %{target-include} -I %{libcxx}/test/support -D_CRT_SECURE_NO_WARNINGS -D_CRT_NONSTDC_NO_WARNINGS -D_CRT_STDIO_ISO_WIDE_SPECIFIERS -DNOMINMAX' + config.dbg_include
 ))
 config.substitutions.append(('%{link_flags}',
-    '-nostdlib -L %%{lib} -lc++ -l%s -l%s -loldnames' % (config.crt_lib, config.cxx_lib)
+    '-nostdlib -L %{lib} -lc++ -l' + config.cxx_lib
 ))
 config.substitutions.append(('%{exec}',
     '%{executor} --execdir %T --prepend_env PATH=%{lib} -- '

libcxx/test/configs/llvm-libc++-shared-no-vcruntime-clangcl.cfg.in

@@ -6,10 +6,10 @@ lit_config.load_config(config, '@CMAKE_CURRENT_BINARY_DIR@/cmake-bridge.cfg')
 
 config.substitutions.append(('%{flags}', '--driver-mode=g++'))
 config.substitutions.append(('%{compile_flags}',
-    '-nostdinc++ -I %{include} -I %{target-include} -I %{libcxx}/test/support -D_CRT_SECURE_NO_WARNINGS -D_CRT_NONSTDC_NO_WARNINGS -D_CRT_STDIO_ISO_WIDE_SPECIFIERS -DNOMINMAX -D_HAS_EXCEPTIONS=0' + config.dbg_include
+    '-fms-runtime-lib=' + config.fms_runtime_lib + ' -nostdinc++ -I %{include} -I %{target-include} -I %{libcxx}/test/support -D_CRT_SECURE_NO_WARNINGS -D_CRT_NONSTDC_NO_WARNINGS -D_CRT_STDIO_ISO_WIDE_SPECIFIERS -DNOMINMAX -D_HAS_EXCEPTIONS=0' + config.dbg_include
 ))
 config.substitutions.append(('%{link_flags}',
-    '-nostdlib -L %%{lib} -lc++ -l%s -l%s -loldnames' % (config.crt_lib, config.cxx_lib)
+    '-nostdlib -L %{lib} -lc++ -l' + config.cxx_lib
 ))
 config.substitutions.append(('%{exec}',
     '%{executor} --execdir %T --prepend_env PATH=%{lib} -- '

libcxx/test/configs/llvm-libc++-static-clangcl.cfg.in

@@ -5,10 +5,10 @@ lit_config.load_config(config, '@CMAKE_CURRENT_BINARY_DIR@/cmake-bridge.cfg')
 
 config.substitutions.append(('%{flags}', '--driver-mode=g++'))
 config.substitutions.append(('%{compile_flags}',
-    '-nostdinc++ -I %{include} -I %{target-include} -I %{libcxx}/test/support -D_CRT_SECURE_NO_WARNINGS -D_CRT_NONSTDC_NO_WARNINGS -D_CRT_STDIO_ISO_WIDE_SPECIFIERS -DNOMINMAX' + config.dbg_include
+    '-fms-runtime-lib=' + config.fms_runtime_lib + ' -nostdinc++ -I %{include} -I %{target-include} -I %{libcxx}/test/support -D_CRT_SECURE_NO_WARNINGS -D_CRT_NONSTDC_NO_WARNINGS -D_CRT_STDIO_ISO_WIDE_SPECIFIERS -DNOMINMAX' + config.dbg_include
 ))
 config.substitutions.append(('%{link_flags}',
-    '-nostdlib -L %%{lib} -llibc++ -l%s -l%s -loldnames' % (config.crt_lib, config.cxx_lib)
+    '-nostdlib -L %{lib} -llibc++ -l' + config.cxx_lib
 ))
 config.substitutions.append(('%{exec}',
     '%{executor} --execdir %T -- '

llvm/include/llvm/CodeGen/AccelTable.h

@@ -25,6 +25,7 @@
 #include "llvm/Support/DJB.h"
 #include "llvm/Support/Debug.h"
 #include <cstdint>
+#include <variant>
 #include <vector>
 
 /// \file
@@ -310,9 +311,13 @@ void emitDWARF5AccelTable(AsmPrinter *Asm,
                           const DwarfDebug &DD,
                           ArrayRef<std::unique_ptr<DwarfCompileUnit>> CUs);
 
+/// Emit a DWARFv5 Accelerator Table consisting of entries in the specified
+/// AccelTable. The \p CUs contains either symbols keeping offsets to the
+/// start of compilation unit, either offsets to the start of compilation
+/// unit themselves.
 void emitDWARF5AccelTable(
     AsmPrinter *Asm, AccelTable<DWARF5AccelTableStaticData> &Contents,
-    ArrayRef<MCSymbol *> CUs,
+    ArrayRef<std::variant<MCSymbol *, uint64_t>> CUs,
     llvm::function_ref<unsigned(const DWARF5AccelTableStaticData &)>
         getCUIndexForEntry);
 

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 476071
+#define LLVM_MAIN_REVISION 476085
 
 /* Define if LLVM_ENABLE_DUMP is enabled */
 #cmakedefine LLVM_ENABLE_DUMP

llvm/include/llvm/DWARFLinkerParallel/DWARFLinker.h

@@ -100,9 +100,6 @@ class ExtraDwarfEmitter {
   /// Emit section named SecName with data SecData.
   virtual void emitSectionContents(StringRef SecData, StringRef SecName) = 0;
 
-  /// Emit temporarily symbol named \p SymName inside section \p SecName.
-  virtual MCSymbol *emitTempSym(StringRef SecName, StringRef SymName) = 0;
-
   /// Emit the swift_ast section stored in \p Buffer.
   virtual void emitSwiftAST(StringRef Buffer) = 0;
 

llvm/include/llvm/Transforms/Utils/UnrollLoop.h

@@ -99,17 +99,6 @@ LoopUnrollResult UnrollAndJamLoop(Loop *L, unsigned Count, unsigned TripCount,
 bool isSafeToUnrollAndJam(Loop *L, ScalarEvolution &SE, DominatorTree &DT,
                           DependenceInfo &DI, LoopInfo &LI);
 
-bool computeUnrollCount(Loop *L, const TargetTransformInfo &TTI,
-                        DominatorTree &DT, LoopInfo *LI, AssumptionCache *AC,
-                        ScalarEvolution &SE,
-                        const SmallPtrSetImpl<const Value *> &EphValues,
-                        OptimizationRemarkEmitter *ORE, unsigned TripCount,
-                        unsigned MaxTripCount, bool MaxOrZero,
-                        unsigned TripMultiple, unsigned LoopSize,
-                        TargetTransformInfo::UnrollingPreferences &UP,
-                        TargetTransformInfo::PeelingPreferences &PP,
-                        bool &UseUpperBound);
-
 void simplifyLoopAfterUnroll(Loop *L, bool SimplifyIVs, LoopInfo *LI,
                              ScalarEvolution *SE, DominatorTree *DT,
                              AssumptionCache *AC,
@@ -126,9 +115,44 @@ TargetTransformInfo::UnrollingPreferences gatherUnrollingPreferences(
     std::optional<bool> UserUpperBound,
     std::optional<unsigned> UserFullUnrollMaxCount);
 
-InstructionCost ApproximateLoopSize(const Loop *L, unsigned &NumCalls,
-    bool &NotDuplicatable, bool &Convergent, const TargetTransformInfo &TTI,
-    const SmallPtrSetImpl<const Value *> &EphValues, unsigned BEInsns);
+/// Produce an estimate of the unrolled cost of the specified loop.  This
+/// is used to a) produce a cost estimate for partial unrolling and b) to
+/// cheaply estimate cost for full unrolling when we don't want to symbolically
+/// evaluate all iterations.
+class UnrollCostEstimator {
+  InstructionCost LoopSize;
+  bool NotDuplicatable;
+
+public:
+  unsigned NumInlineCandidates;
+  bool Convergent;
+
+  UnrollCostEstimator(const Loop *L, const TargetTransformInfo &TTI,
+                      const SmallPtrSetImpl<const Value *> &EphValues,
+                      unsigned BEInsns);
+
+  /// Whether it is legal to unroll this loop.
+  bool canUnroll() const { return LoopSize.isValid() && !NotDuplicatable; }
+
+  uint64_t getRolledLoopSize() const { return *LoopSize.getValue(); }
+
+  /// Returns loop size estimation for unrolled loop, given the unrolling
+  /// configuration specified by UP.
+  uint64_t
+  getUnrolledLoopSize(const TargetTransformInfo::UnrollingPreferences &UP,
+                      unsigned CountOverwrite = 0) const;
+};
+
+bool computeUnrollCount(Loop *L, const TargetTransformInfo &TTI,
+                        DominatorTree &DT, LoopInfo *LI, AssumptionCache *AC,
+                        ScalarEvolution &SE,
+                        const SmallPtrSetImpl<const Value *> &EphValues,
+                        OptimizationRemarkEmitter *ORE, unsigned TripCount,
+                        unsigned MaxTripCount, bool MaxOrZero,
+                        unsigned TripMultiple, const UnrollCostEstimator &UCE,
+                        TargetTransformInfo::UnrollingPreferences &UP,
+                        TargetTransformInfo::PeelingPreferences &PP,
+                        bool &UseUpperBound);
 
 } // end namespace llvm
 

llvm/lib/CodeGen/AsmPrinter/AccelTable.cpp

@@ -213,7 +213,7 @@ class Dwarf5AccelTableWriter : public AccelTableWriter {
 
   Header Header;
   DenseMap<uint32_t, SmallVector<AttributeEncoding, 2>> Abbreviations;
-  ArrayRef<MCSymbol *> CompUnits;
+  ArrayRef<std::variant<MCSymbol *, uint64_t>> CompUnits;
   llvm::function_ref<unsigned(const DataT &)> getCUIndexForEntry;
   MCSymbol *ContributionEnd = nullptr;
   MCSymbol *AbbrevStart = Asm->createTempSymbol("names_abbrev_start");
@@ -235,7 +235,7 @@ class Dwarf5AccelTableWriter : public AccelTableWriter {
 public:
   Dwarf5AccelTableWriter(
       AsmPrinter *Asm, const AccelTableBase &Contents,
-      ArrayRef<MCSymbol *> CompUnits,
+      ArrayRef<std::variant<MCSymbol *, uint64_t>> CompUnits,
       llvm::function_ref<unsigned(const DataT &)> GetCUIndexForEntry);
 
   void emit();
@@ -419,7 +419,10 @@ template <typename DataT>
 void Dwarf5AccelTableWriter<DataT>::emitCUList() const {
   for (const auto &CU : enumerate(CompUnits)) {
     Asm->OutStreamer->AddComment("Compilation unit " + Twine(CU.index()));
-    Asm->emitDwarfSymbolReference(CU.value());
+    if (std::holds_alternative<MCSymbol *>(CU.value()))
+      Asm->emitDwarfSymbolReference(std::get<MCSymbol *>(CU.value()));
+    else
+      Asm->emitDwarfLengthOrOffset(std::get<uint64_t>(CU.value()));
   }
 }
 
@@ -508,7 +511,7 @@ template <typename DataT> void Dwarf5AccelTableWriter<DataT>::emitData() const {
 template <typename DataT>
 Dwarf5AccelTableWriter<DataT>::Dwarf5AccelTableWriter(
     AsmPrinter *Asm, const AccelTableBase &Contents,
-    ArrayRef<MCSymbol *> CompUnits,
+    ArrayRef<std::variant<MCSymbol *, uint64_t>> CompUnits,
     llvm::function_ref<unsigned(const DataT &)> getCUIndexForEntry)
     : AccelTableWriter(Asm, Contents, false),
       Header(CompUnits.size(), Contents.getBucketCount(),
@@ -545,7 +548,7 @@ void llvm::emitAppleAccelTableImpl(AsmPrinter *Asm, AccelTableBase &Contents,
 void llvm::emitDWARF5AccelTable(
     AsmPrinter *Asm, AccelTable<DWARF5AccelTableData> &Contents,
     const DwarfDebug &DD, ArrayRef<std::unique_ptr<DwarfCompileUnit>> CUs) {
-  std::vector<MCSymbol *> CompUnits;
+  std::vector<std::variant<MCSymbol *, uint64_t>> CompUnits;
   SmallVector<unsigned, 1> CUIndex(CUs.size());
   int Count = 0;
   for (const auto &CU : enumerate(CUs)) {
@@ -581,7 +584,7 @@ void llvm::emitDWARF5AccelTable(
 
 void llvm::emitDWARF5AccelTable(
     AsmPrinter *Asm, AccelTable<DWARF5AccelTableStaticData> &Contents,
-    ArrayRef<MCSymbol *> CUs,
+    ArrayRef<std::variant<MCSymbol *, uint64_t>> CUs,
     llvm::function_ref<unsigned(const DWARF5AccelTableStaticData &)>
         getCUIndexForEntry) {
   Contents.finalize(Asm, "names");

llvm/lib/DWARFLinker/DWARFStreamer.cpp

@@ -297,7 +297,7 @@ void DwarfStreamer::emitDebugNames(
     return;
 
   // Build up data structures needed to emit this section.
-  std::vector<MCSymbol *> CompUnits;
+  std::vector<std::variant<MCSymbol *, uint64_t>> CompUnits;
   DenseMap<unsigned, size_t> UniqueIdToCuMap;
   unsigned Id = 0;
   for (auto &CU : EmittedUnits) {