llvm dev - Mar 2010 - [LLVMdev] Which floating-point comparison?

I notice llvm provides both ordered and unordered variants of
floating-point comparison. Which of these is the right one to use by
default? I suppose the two criteria would be, in order of importance:

1. Which is more efficient (more directly maps to typical hardware)?

2. Which is more familiar (more like the way C and Fortran do it)?

On Sun, Mar 28, 2010 at 7:45 AM, Russell Wallace
<russell.wallace at gmail.com> wrote:
> I notice llvm provides both ordered and unordered variants of
> floating-point comparison. Which of these is the right one to use by
> default? I suppose the two criteria would be, in order of importance:
>
> 1. Which is more efficient (more directly maps to typical hardware)?
You can figure this out by looking at the output of llc:

$ cat test.ll
define i1 @less(double %x, double %y) nounwind readnone {
entry:
  %0 = fcmp ult double %x, %y                     ; <i1> [#uses=1]
  ret i1 %0
}
$ Debug/bin/llc  <test.ll
	.section	__TEXT,__text,regular,pure_instructions
	.globl	_less
	.align	4, 0x90
_less:                                  ## @less
## BB#0:                                ## %entry
	movsd	4(%esp), %xmm0
	ucomisd	12(%esp), %xmm0
	sbbb	%al, %al
	andb	$1, %al
	ret
> 2. Which is more familiar (more like the way C and Fortran do it)?
You can use http://llvm.org/demo/ to figure that out.

That works, thanks! It turns out that x86/SSE at least handles both
equally well, but the ordered version is what C uses.

On Sun, Mar 28, 2010 at 6:38 PM, Jeffrey Yasskin <jyasskin at google.com>
wrote:
> On Sun, Mar 28, 2010 at 7:45 AM, Russell Wallace
> <russell.wallace at gmail.com> wrote:
>> I notice llvm provides both ordered and unordered variants of
>> floating-point comparison. Which of these is the right one to use by
>> default? I suppose the two criteria would be, in order of importance:
>>
>> 1. Which is more efficient (more directly maps to typical hardware)?
>
> You can figure this out by looking at the output of llc:
>
> $ cat test.ll
> define i1 @less(double %x, double %y) nounwind readnone {
> entry:
>  %0 = fcmp ult double %x, %y                     ; <i1> [#uses=1]
>  ret i1 %0
> }
> $ Debug/bin/llc  <test.ll
>        .section        __TEXT,__text,regular,pure_instructions
>        .globl  _less
>        .align  4, 0x90
> _less:                                  ## @less
> ## BB#0:                                ## %entry
>        movsd   4(%esp), %xmm0
>        ucomisd 12(%esp), %xmm0
>        sbbb    %al, %al
>        andb    $1, %al
>        ret
>
>> 2. Which is more familiar (more like the way C and Fortran do it)?
>
> You can use http://llvm.org/demo/ to figure that out.
>