Add an option for ICFing data

[• rafael]

Summary

An internal linker has support for merging identical data and in some cases it can be a significant win.

This is behind an off by default flag so it has to be requested explicitly.

[ruiu]

We used to merge data for COFF but not anymore because it causes nasty problems that are hard to find and debug. E.g. https://bugs.chromium.org/p/chromium/issues/detail?id=682773#c24

Are you sure you don't have that kind of issue?

[ruiu]

Got it. If you are using this option from the beginning, it should be fine.

It's not a big deal, but --icf-data sounds a bit odd, because ICF is short for Identical *Code* Folding. I have no better idea though.

ELF/Options.td

146

Please rewrite the description so that it doesn't sound odd even if it's not displayed after --icf.

[• rafael]

Rebased and updated description.

[ruiu]

LGTM

--icf-data is not technically a correct name but that's probably the best name.

[• espindola]

r320448

ELF/Config.h

struct Configuration {
   bool ExportDynamic;
   bool FixCortexA53Errata843419;
   bool GcSections;
   bool GdbIndex;
   bool GnuHash = false;
   bool HasDynamicList = false;
   bool HasDynSymTab;
   bool ICF;
+  bool ICFData;
   bool MipsN32Abi = false;
   bool NoGnuUnique;
   bool NoUndefinedVersion;
   bool NoinhibitExec;
   bool Nostdlib;
   bool OFormatBinary;
   bool Omagic;
   bool OptRemarksWithHotness;

ELF/Driver.cpp

void LinkerDriver::readConfigs(opt::InputArgList &Args) {
   errorHandler().FatalWarnings =
       Args.hasFlag(OPT_fatal_warnings, OPT_no_fatal_warnings, false);
   Config->FilterList = args::getStrings(Args, OPT_filter);
   Config->Fini = Args.getLastArgValue(OPT_fini, "_fini");
   Config->FixCortexA53Errata843419 = Args.hasArg(OPT_fix_cortex_a53_843419);
   Config->GcSections = Args.hasFlag(OPT_gc_sections, OPT_no_gc_sections, false);
   Config->GdbIndex = Args.hasFlag(OPT_gdb_index, OPT_no_gdb_index, false);
   Config->ICF = Args.hasFlag(OPT_icf_all, OPT_icf_none, false);
+  Config->ICFData = Args.hasArg(OPT_icf_data);
   Config->Init = Args.getLastArgValue(OPT_init, "_init");
   Config->LTOAAPipeline = Args.getLastArgValue(OPT_lto_aa_pipeline);
   Config->LTONewPmPasses = Args.getLastArgValue(OPT_lto_newpm_passes);
   Config->LTOO = args::getInteger(Args, OPT_lto_O, 2);
   Config->LTOPartitions = args::getInteger(Args, OPT_lto_partitions, 1);
   Config->MapFile = Args.getLastArgValue(OPT_Map);
   Config->NoGnuUnique = Args.hasArg(OPT_no_gnu_unique);
   Config->NoUndefinedVersion = Args.hasArg(OPT_no_undefined_version);

ELF/ICF.cpp

 // Returns a hash value for S. Note that the information about
 // relocation targets is not included in the hash value.
 template <class ELFT> static uint32_t getHash(InputSection *S) {
   return hash_combine(S->Flags, S->getSize(), S->NumRelocations, S->Data);
 }
 
 // Returns true if section S is subject of ICF.
 static bool isEligible(InputSection *S) {
+  // Don't merge read only data sections unless --icf-data was passed.
+  if (!(S->Flags & SHF_EXECINSTR) && !Config->ICFData)
+    return false;
+
   // .init and .fini contains instructions that must be executed to
   // initialize and finalize the process. They cannot and should not
   // be merged.
-  return S->Live && (S->Flags & SHF_ALLOC) && (S->Flags & SHF_EXECINSTR) &&
-         !(S->Flags & SHF_WRITE) && S->Name != ".init" && S->Name != ".fini";
+  return S->Live && (S->Flags & SHF_ALLOC) && !(S->Flags & SHF_WRITE) &&
+         S->Name != ".init" && S->Name != ".fini";
 }
 
 // Split an equivalence class into smaller classes.
 template <class ELFT>
 void ICF<ELFT>::segregate(size_t Begin, size_t End, bool Constant) {
   // This loop rearranges sections in [Begin, End) so that all sections
   // that are equal in terms of equals{Constant,Variable} are contiguous
   // in [Begin, End).

ELF/Options.td

 
 defm hash_style: Eq<"hash-style">,
   HelpText<"Specify hash style (sysv, gnu or both)">;
 
 def help: F<"help">, HelpText<"Print option help">;
 
 def icf_all: F<"icf=all">, HelpText<"Enable identical code folding">;
 
+def icf_data: F<"icf-data">,
+  HelpText<"Enable ICF to also fold identical read only data">;
+
 def icf_none: F<"icf=none">, HelpText<"Disable identical code folding">;
 
 defm image_base : Eq<"image-base">, HelpText<"Set the base address">;
 
 defm init: Eq<"init">, HelpText<"Specify an initializer function">,
   MetaVarName<"<symbol>">;
 
 defm library: Eq<"library">, HelpText<"Root name of library to use">,

test/ELF/icf9.s

 # REQUIRES: x86
 
 ### Make sure that we do not merge data.
 # RUN: llvm-mc -filetype=obj -triple=x86_64-unknown-linux %s -o %t
 # RUN: ld.lld %t -o %t2 --icf=all --verbose 2>&1 | FileCheck %s
 # RUN: llvm-readelf -S -W %t2 | FileCheck --check-prefix=SEC %s
 
 # SEC:  .rodata      PROGBITS  0000000000200120 000120 000002 00 A 0 0 1
 
 # CHECK-NOT: selected .rodata.d1
 # CHECK-NOT: selected .rodata.d2
 
+# We do merge rodata if passed --icf-data
+# RUN: ld.lld %t -o %t2 --icf=all --verbose --icf-data 2>&1 | FileCheck --check-prefix=DATA %s
+# RUN: llvm-readelf -S -W %t2 | FileCheck --check-prefix=DATA-SEC %s
+
+# DATA: selected .rodata.d1
+# DATA: removed .rodata.d2
+
+# DATA-SEC:  .rodata      PROGBITS  0000000000200120 000120 000001 00 A 0 0 1
+
 .globl _start, d1, d2
 _start:
   ret
 
-.section .rodata.f1, "a"
+.section .rodata.d1, "a"
 d1:
   .byte 1
 
-.section .rodata.f2, "a"
+.section .rodata.d2, "a"
 d2:
   .byte 1