ELF: Add "fast link" mode.

[ruiu]

Summary

I'm not 100% sure if this is the right thing to do, but at least
it seems to be worth discussing. This patch adds a new command line
flag, --fast, to the linker. That flag disables costly but optional
features so that the linker produces semantically correct (but larger)
output quickly. Currently it only disables section merging.

This flag is not intended to be used for final production linking.
It is intended to be used in compile-link-test cycle.

Time to link clang with debug info is about 2x faster with the flag.

Head:
13.24 seconds
Output size: 1227189664 bytes

With this patch:
7.41 seconds
Output size: 2490281784 bytes

[davide]

Can't -O be used for that?

[ruiu]

Maybe we can, but this should be one below the default optimization level. So maybe we should define -O1 as default and -O0 as this? Then I think "O" makes sense.

[davide]

That was exactly what we had in mind. The set of optimizations performed by a linker is relatively small compared to the optimizations done by a compiler so I don't think we'll have hard time deciding what we want to put at any given level.

Cheers

[davide]

In D19705#416367, @davide wrote:

That was exactly what we had in mind. The set of optimizations performed by a linker is relatively small compared to the optimizations done by a compiler so I don't think we'll have hard time deciding what we want to put at any given level.

Cheers

*I had in mind.

[ruiu]

• Update as per Davide's comment.

[davide]

I think this is good, and I like that. Hopefully other people will share this feeling.

[ruiu]

Yeah, this is better than I originally thought. I like it.

[• rafael]

LGTM with a testcaste :-)

lld/trunk/ELF/Driver.cpp

void LinkerDriver::readConfigs(opt::InputArgList &Args) {
   Config->Entry = getString(Args, OPT_entry);
   Config->Fini = getString(Args, OPT_fini, "_fini");
   Config->Init = getString(Args, OPT_init, "_init");
   Config->OutputFile = getString(Args, OPT_o);
   Config->Reproduce = getString(Args, OPT_reproduce);
   Config->SoName = getString(Args, OPT_soname);
   Config->Sysroot = getString(Args, OPT_sysroot);
 
-  Config->Optimize = getInteger(Args, OPT_O, 0);
+  Config->Optimize = getInteger(Args, OPT_O, 1);
   Config->LtoO = getInteger(Args, OPT_lto_O, 2);
   if (Config->LtoO > 3)
     error("invalid optimization level for LTO: " + getString(Args, OPT_lto_O));
   Config->LtoJobs = getInteger(Args, OPT_lto_jobs, 1);
   if (Config->LtoJobs == 0)
     error("number of threads must be > 0");
 
   Config->ZExecStack = hasZOption(Args, "execstack");

lld/trunk/ELF/InputFiles.cpp

ArrayRef<Elf_Word> Entries =
       check(Obj.template getSectionContentsAsArray<Elf_Word>(&Sec));
   if (Entries.empty() || Entries[0] != GRP_COMDAT)
     fatal("unsupported SHT_GROUP format");
   return Entries.slice(1);
 }
 
 template <class ELFT> static bool shouldMerge(const typename ELFT::Shdr &Sec) {
   typedef typename ELFT::uint uintX_t;
+
+  // We don't merge sections if -O0 (default is -O1). This makes sometimes
+  // the linker significantly faster, although the output will be bigger.
+  if (Config->Optimize == 0)
+    return false;
+
   uintX_t Flags = Sec.sh_flags;
   if (!(Flags & SHF_MERGE))
     return false;
   if (Flags & SHF_WRITE)
     fatal("writable SHF_MERGE sections are not supported");
   uintX_t EntSize = Sec.sh_entsize;
   if (!EntSize || Sec.sh_size % EntSize)
     fatal("SHF_MERGE section size must be a multiple of sh_entsize");

lld/trunk/test/ELF/merge-string.s

 // REQUIRES: x86
 // RUN: llvm-mc -filetype=obj -triple=x86_64-pc-linux %s -o %t.o
 // RUN: ld.lld -O2 %t.o -o %t.so -shared
 // RUN: llvm-readobj -s -section-data -t %t.so | FileCheck %s
 // RUN: ld.lld -O1 %t.o -o %t.so -shared
 // RUN: llvm-readobj -s -section-data -t %t.so | FileCheck --check-prefix=NOTAIL %s
+// RUN: ld.lld -O0 %t.o -o %t.so -shared
+// RUN: llvm-readobj -s -section-data -t %t.so | FileCheck --check-prefix=NOMERGE %s
 
         .section  .rodata.str1.1,"aMS",@progbits,1
   .asciz  "abc"
 foo:
   .ascii  "a"
 bar:
         .asciz  "bc"
         .asciz  "bc"

 // NOTAIL-NEXT: Link: 0
 // NOTAIL-NEXT: Info: 0
 // NOTAIL-NEXT: AddressAlignment: 1
 // NOTAIL-NEXT: EntrySize: 1
 // NOTAIL-NEXT: SectionData (
 // NOTAIL-NEXT:   0000: 61626300 626300                     |abc.bc.|
 // NOTAIL-NEXT: )
 
+// NOMERGE:      Name:    .rodata
+// NOMERGE-NEXT: Type:    SHT_PROGBITS
+// NOMERGE-NEXT: Flags [
+// NOMERGE-NEXT:   SHF_ALLOC
+// NOMERGE-NEXT:   SHF_MERGE
+// NOMERGE-NEXT:   SHF_STRINGS
+// NOMERGE-NEXT: ]
+// NOMERGE-NEXT: Address:         0x1C8
+// NOMERGE-NEXT: Offset:  0x1C8
+// NOMERGE-NEXT: Size:    16
+// NOMERGE-NEXT: Link: 0
+// NOMERGE-NEXT: Info: 0
+// NOMERGE-NEXT: AddressAlignment: 2
+// NOMERGE-NEXT: EntrySize: 0
+// NOMERGE-NEXT: SectionData (
+// NOMERGE-NEXT:   0000: 61626300 61626300 62630000 14000000  |abc.abc.bc......|
+// NOMERGE-NEXT: )
+
 // CHECK:      Name: .rodata
 // CHECK-NEXT: Type: SHT_PROGBITS
 // CHECK-NEXT: Flags [
 // CHECK-NEXT:   SHF_ALLOC
 // CHECK-NEXT:   SHF_MERGE
 // CHECK-NEXT:   SHF_STRINGS
 // CHECK-NEXT: ]
 // CHECK-NEXT: Address: 0x1CC