[PATCH] D88823: [CodeGen][TailDuplicator] Don't duplicate blocks with INLINEASM_BR

[Bill Wendling via Phabricator via llvm-commits]

void added a comment.

One thing, I believe we're trying to turn "asm goto" into something it was never meant to be. If you look at gcc's output for this code:

  int foo(int a, int b) {
  	int x;
  
  	if (a < 42) {
  		x = 0;
  		asm goto("# 1\n\t.data %l1" : : "r"(b) : : indirect);
  	}
  	if (b > 42) {
  		x = 37;
  		asm goto("# 2\n\t.data %l1" : : "r"(a) : : indirect);
  	}
  indirect:
  	x = a + b - x;
  ret:
  	return x;
  }

GCC doesn't combine the indirect destination:

  	.file	"asm-goto-phi.c"
  	.text
  	.p2align 4
  	.globl	foo
  	.type	foo, @function
  foo:
  .LFB0:
  	.cfi_startproc
  	movl	%esi, %eax
  	cmpl	$41, %edi
  	jg	.L2
  #APP
  # 6 "asm-goto-phi.c" 1
  	# 1
  	.data .L5
  # 0 "" 2
  #NO_APP
  .L2:
  	xorl	%edx, %edx
  	cmpl	$42, %eax
  	jle	.L3
  #APP
  # 10 "asm-goto-phi.c" 1
  	# 2
  	.data .L4
  # 0 "" 2
  #NO_APP
  .L4:
  	movl	$37, %edx
  .L3:
  	addl	%edi, %eax
  	subl	%edx, %eax
  	ret
  	.p2align 4,,10
  	.p2align 3
  .L5:
  	xorl	%edx, %edx
  	addl	%edi, %eax
  	subl	%edx, %eax
  	ret
  	.cfi_endproc
  .LFE0:
  	.size	foo, .-foo
  	.ident	"GCC: (Debian 9.3.0-14) 9.3.0"
  	.section	.note.GNU-stack,"", at progbits

This side steps the issue of resolving PHI nodes.

But whatever. You guys come up with something.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D88823/new/

https://reviews.llvm.org/D88823