[llvm-mc] - Produce R_X86_64_PLT32 relocation for branches with JCC opcodes too.

[grimar]

Summary

The idea is to produce R_X86_64_PLT32 instead of
R_X86_64_PC32 for branches.

It fixes https://bugs.llvm.org/show_bug.cgi?id=44397.

This patch teaches MC to do that for JCC (jump if condition is met)
instructions. The new behavior matches modern GNU as.
It is similar to D43383, which did the same for "call/jmp foo",
but missed JCC cases.

[MaskRay]

R_X86_64_PLT32 is a branch relocation type which means the exact address is insignificant.

• In an executable, it can prevent the creation of a canonical PLT entry.
• In a shared object, an R_X86_64_PC32 to a preemptible symbol is not allowed.

This can be favorable since it can enable us to use JCC_4 for tail calls.

[smeenai]

CC @hans ... there was discussion in https://bugs.llvm.org/show_bug.cgi?id=44397 about whether this is worth picking to 10.0.

[MaskRay]

In D72831#1839293, @smeenai wrote:

CC @hans ... there was discussion in https://bugs.llvm.org/show_bug.cgi?id=44397 about whether this is worth picking to 10.0.

This is related to CodeGen/BranchFolding.cpp (D29856).

I think certain -Os and -Oz code with static relocation model could create such bne foo with R_X86_64_PC32. If the target is defined in a shared object, it will create a canonical PLT.
It does not hurt to include it in release/10.x but I also don't think there is a demanding need. (Non -Os/-Oz optimization levels do not trigger. -fpie-/-fpic does not trigger.)

define void @foo(i32* %a) minsize {
entry:
  %0 = load i32, i32* %a
  %cmp = icmp eq i32 %0, 0
  br i1 %cmp, label %if.then, label %if.else
if.then:
  tail call void @bar()
  br label %if.else
if.else:
  ret void
}

declare void @bar()

llvm/lib/Target/X86/MCTargetDesc/X86MCCodeEmitter.cpp

if (Expr->getKind() == MCExpr::SymbolRef) {
     return Ref->getKind() == MCSymbolRefExpr::VK_SECREL;
   }
   return false;
 }
 
 static bool isPCRel32Branch(const MCInst &MI, const MCInstrInfo &MCII) {
   unsigned Opcode = MI.getOpcode();
   const MCInstrDesc &Desc = MCII.get(Opcode);
-  if ((Opcode != X86::CALL64pcrel32 && Opcode != X86::JMP_4) ||
+  if ((Opcode != X86::CALL64pcrel32 && Opcode != X86::JMP_4 &&
+       Opcode != X86::JCC_4) ||
       getImmFixupKind(Desc.TSFlags) != FK_PCRel_4)
     return false;
 
   unsigned CurOp = X86II::getOperandBias(Desc);
   const MCOperand &Op = MI.getOperand(CurOp);
   if (!Op.isExpr())
     return false;
 

llvm/test/MC/ELF/basic-elf-64.s

 // RUN: llvm-mc -filetype=obj -triple x86_64-pc-linux-gnu %s -o - | llvm-readobj -h -S -r --symbols | FileCheck %s
 
         .text
   .globl  main
   .align  16, 0x90
   .type main,@function
 main:                                   # @main
 # %bb.0:
   subq  $8, %rsp
   movl  $.L.str1, %edi
   callq puts
   movl  $.L.str2, %edi
   callq puts
   xorl  %eax, %eax
   addq  $8, %rsp
     call foo@GOTPCREL
+    ja foo
+    jae foo
+    jb foo
+    jbe foo
+    jc foo
+    je foo
+    jz foo
+    jg foo
+    jge foo
+    jl foo
+    jle foo
+    jna foo
+    jnae foo
+    jnb foo
+    jnbe foo
+    jnc foo
+    jne foo
+    jng foo
+    jnge foo
+    jnl foo
+    jnle foo
+    jno foo
+    jnp foo
+    jns foo
+    jnz foo
+    jo foo
+    jp foo
+    jpe foo
+    jpo foo
+    js foo
+    jz foo
   ret
 .Ltmp0:
   .size main, .Ltmp0-main
 
   .type .L.str1,@object         # @.str1
   .section  .rodata.str1.1,"aMS",@progbits,1
 .L.str1:
   .asciz   "Hello"

 
 // CHECK:      Relocations [
 // CHECK:        Section {{.*}} .rela.text {
 // CHECK-NEXT:     0x5  R_X86_64_32   .rodata.str1.1 0x0
 // CHECK-NEXT:     0xA  R_X86_64_PLT32 puts           0xFFFFFFFFFFFFFFFC
 // CHECK-NEXT:     0xF  R_X86_64_32   .rodata.str1.1 0x6
 // CHECK-NEXT:     0x14 R_X86_64_PLT32 puts           0xFFFFFFFFFFFFFFFC
 // CHECK-NEXT:     0x1F R_X86_64_GOTPCREL foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0x25 R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0x2B R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0x31 R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0x37 R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0x3D R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0x43 R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0x49 R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0x4F R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0x55 R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0x5B R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0x61 R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0x67 R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0x6D R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0x73 R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0x79 R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0x7F R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0x85 R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0x8B R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0x91 R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0x97 R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0x9D R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0xA3 R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0xA9 R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0xAF R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0xB5 R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0xBB R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0xC1 R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0xC7 R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0xCD R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0xD3 R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
+// CHECK-NEXT:     0xD9 R_X86_64_PLT32 foo 0xFFFFFFFFFFFFFFFC
 // CHECK-NEXT:   }
 // CHECK-NEXT: ]
 
 // CHECK:   Symbol {
 // CHECK:     Binding: Local
 // CHECK:     Type: Section
 
 // CHECK:   Symbol {