All .xdata sections are eligble for ICF

[rnk]

Summary

Many small functions have identical unwind info because they push the
same sets of CSRs in the same order and have the same stack and prologue
size. The VC linker merges duplicate .xdata, and so should LLD.

This reduces the .xdata section size of clang.exe from 1.8MB to 94KB.

[ruiu]

LGTM

lld/COFF/ICF.cpp

83 ↗

(On Diff #123272)

I'd add bool Xdata for consistency.

[rnk]

lld/COFF/ICF.cpp

83 ↗

(On Diff #123272)

I want to short-circuit the section name string comparison, though. How about this?

[rnk]

• adjust isEligble conditions

lld/trunk/COFF/ICF.cpp

 
 // Returns true if section S is subject of ICF.
 //
 // Microsoft's documentation
 // (https://msdn.microsoft.com/en-us/library/bxwfs976.aspx; visited April
 // 2017) says that /opt:icf folds both functions and read-only data.
 // Despite that, the MSVC linker folds only functions. We found
 // a few instances of programs that are not safe for data merging.
-// Therefore, we merge only functions just like the MSVC tool.
+// Therefore, we merge only functions just like the MSVC tool. However, we merge
+// identical .xdata sections, because the address of unwind information is
+// insignificant to the user program and the Visual C++ linker does this.
 bool ICF::isEligible(SectionChunk *C) {
+  // Non-comdat chunks, dead chunks, and writable chunks are not elegible.
+  bool Writable = C->getPermissions() & llvm::COFF::IMAGE_SCN_MEM_WRITE;
+  if (!C->isCOMDAT() || !C->isLive() || Writable)
+    return false;
+
+  // Code sections with external symbols are eligible.
   bool Global = C->Sym && C->Sym->isExternal();
   bool Executable = C->getPermissions() & llvm::COFF::IMAGE_SCN_MEM_EXECUTE;
-  bool Writable = C->getPermissions() & llvm::COFF::IMAGE_SCN_MEM_WRITE;
-  return C->isCOMDAT() && C->isLive() && Global && Executable && !Writable;
+  if (Global && Executable)
+    return true;
+
+  // .xdata unwind info sections are eligble.
+  return C->getSectionName().split('$').first == ".xdata";
 }
 
 // Split an equivalence class into smaller classes.
 void ICF::segregate(size_t Begin, size_t End, bool Constant) {
   while (Begin < End) {
     // Divide [Begin, End) into two. Let Mid be the start index of the
     // second group.
     auto Bound = std::stable_partition(

lld/trunk/test/COFF/icf-xdata.s

+# RUN: llvm-mc %s -triple x86_64-windows-msvc -filetype=obj -o %t.obj
+# RUN: lld-link %t.obj -dll -noentry -out:%t.dll
+# RUN: llvm-readobj -sections %t.dll | FileCheck %s
+
+# There shouldn't be much xdata, because all three .pdata entries (12 bytes
+# each) should use the same .xdata unwind info.
+# CHECK:         Name: .pdata
+# CHECK-NEXT:    VirtualSize: 0x24
+# CHECK:         Name: .xdata
+# CHECK-NEXT:    VirtualSize: 0x8
+
+  .text
+callee:
+  ret
+
+  .def   xdata1;
+  .scl  2;
+  .type 32;
+  .endef
+  .section  .text,"xr",one_only,xdata1
+  .globl  xdata1                  # -- Begin function xdata1
+  .p2align  4, 0x90
+xdata1:                                 # @xdata1
+.seh_proc xdata1
+# BB#0:                                 # %entry
+  subq  $40, %rsp
+  .seh_stackalloc 40
+  .seh_endprologue
+  callq callee
+  nop
+  addq  $40, %rsp
+  jmp callee                  # TAILCALL
+  .seh_handlerdata
+  .section  .text,"xr",one_only,xdata1
+  .seh_endproc
+                                        # -- End function
+  .def   xdata2;
+  .scl  2;
+  .type 32;
+  .endef
+  .section  .text,"xr",one_only,xdata2
+  .globl  xdata2                  # -- Begin function xdata2
+  .p2align  4, 0x90
+xdata2:                                 # @xdata2
+.seh_proc xdata2
+# BB#0:                                 # %entry
+  subq  $40, %rsp
+  .seh_stackalloc 40
+  .seh_endprologue
+  callq callee
+  callq callee
+  nop
+  addq  $40, %rsp
+  jmp callee                  # TAILCALL
+  .seh_handlerdata
+  .section  .text,"xr",one_only,xdata2
+  .seh_endproc
+                                        # -- End function
+  .def   xdata3;
+  .scl  2;
+  .type 32;
+  .endef
+  .section  .text,"xr",one_only,xdata3
+  .globl  xdata3                  # -- Begin function xdata3
+  .p2align  4, 0x90
+xdata3:                                 # @xdata3
+.seh_proc xdata3
+# BB#0:                                 # %entry
+  subq  $40, %rsp
+  .seh_stackalloc 40
+  .seh_endprologue
+  callq callee
+  callq callee
+  callq callee
+  nop
+  addq  $40, %rsp
+  jmp callee                  # TAILCALL
+  .seh_handlerdata
+  .section  .text,"xr",one_only,xdata3
+  .seh_endproc
+                                        # -- End function
+  .section  .drectve,"yn"
+  .ascii  " -export:xdata1"
+  .ascii  " -export:xdata2"
+  .ascii  " -export:xdata3"
+