Another epic optimiser failure

Another epic optimiser failure

[Stefan Kanthak]

--- .c ---
int ispowerof2(unsigned long long argument) {
    return __builtin_popcountll(argument) == 1;
}
--- EOF ---

GCC 13.3    gcc -m32 -march=alderlake -O3
            gcc -m32 -march=sapphirerapids -O3
            gcc -m32 -mpopcnt -mtune=sapphirerapids -O3

https://gcc.godbolt.org/z/cToYrrYPq
ispowerof2(unsigned long long):
        xor     eax, eax        # superfluous
        xor     edx, edx        # superfluous
        popcnt  eax, [esp+4]
        popcnt  edx, [esp+8]
        add     eax, edx
        cmp     eax, 1      ->    dec  eax
        sete    al
        movzx   eax, al         # superfluous
        ret

9 instructions in 28 bytes      # 6 instructions in 20 bytes

OUCH: popcnt writes the WHOLE result register, there is ABSOLUTELY
      no need to clear it beforehand nor to clear the higher 24 bits
      afterwards!

JFTR: before GCC zealots write nonsense: see -march= or -mtune=

GCC 13.3    gcc -mpopcnt -mtune=barcelona -O3

https://gcc.godbolt.org/z/3Ks8vh7a6
ispowerof2(unsigned long long):
        popcnt  rdi, rdi    ->    popcnt  rax, rdi
        xor     eax, eax        # superfluous!
        dec     edi         ->    dec     eax
        sete    al          ->    setz    al
        ret

GCC 13.3    gcc -m32 -mpopcnt -mtune=barcelona -O3

https://gcc.godbolt.org/z/s5s5KTGnv
ispowerof2(unsigned long long):
        popcnt  eax, [esp+4]
        popcnt  edx, [esp+8]
        add     eax, edx
        dec     eax
        sete    al
        movzx   eax, al        # superfluous!
        ret

Will GCC eventually generate properly optimised code instead of bloat?

Stefan

Re: Another epic optimiser failure

[Jakub Jelinek]

On Sat, May 27, 2023 at 11:04:11PM +0200, Stefan Kanthak wrote:
> OUCH: popcnt writes the WHOLE result register, there is ABSOLUTELY
>       no need to clear it beforehand nor to clear the higher 24 bits
>       afterwards!

Except that there is.  See https://gcc.gnu.org/PR62011 for details.

	Jakub

Re: Another epic optimiser failure

[Stefan Kanthak]

"Jakub Jelinek" <jakub@redhat.com> wrote, completely clueless:

> On Sat, May 27, 2023 at 11:04:11PM +0200, Stefan Kanthak wrote:
>> OUCH: popcnt writes the WHOLE result register, there is ABSOLUTELY
>>       no need to clear it beforehand nor to clear the higher 24 bits
>>       afterwards!
> 
> Except that there is.  See https://gcc.gnu.org/PR62011 for details.

GUESS WHY I EXPLICITLY WROTE

| JFTR: before GCC zealots write nonsense: see -march= or -mtune=

NOT AMUSED ABOUT YOUR CRYING INCOMPETENCE!
Stefan

Re: Another epic optimiser failure

[Andrew Pinski]

On Sat, May 27, 2023 at 2:38 PM Stefan Kanthak <stefan.kanthak@nexgo.de> wrote:
>
> "Jakub Jelinek" <jakub@redhat.com> wrote, completely clueless:
>
> > On Sat, May 27, 2023 at 11:04:11PM +0200, Stefan Kanthak wrote:
> >> OUCH: popcnt writes the WHOLE result register, there is ABSOLUTELY
> >>       no need to clear it beforehand nor to clear the higher 24 bits
> >>       afterwards!
> >
> > Except that there is.  See https://gcc.gnu.org/PR62011 for details.
>
> GUESS WHY I EXPLICITLY WROTE
>
> | JFTR: before GCC zealots write nonsense: see -march= or -mtune=
>
> NOT AMUSED ABOUT YOUR CRYING INCOMPETENCE!

So you want to complain about GCC knowing about an Intel performance errata????
If you read that bug report, you would have learned why the zeroing is
done. I am sorry you hate GCC for actually following advice from the
processor maker after all.

> Stefan

Re: Another epic optimiser failure

[Stefan Kanthak]

"Andrew Pinski" <pinskia@gmail.com> wrote:

> On Sat, May 27, 2023 at 2:38 PM Stefan Kanthak <stefan.kanthak@nexgo.de> wrote:
>>
>> "Jakub Jelinek" <jakub@redhat.com> wrote, completely clueless:
>>
>>> On Sat, May 27, 2023 at 11:04:11PM +0200, Stefan Kanthak wrote:
>>>> OUCH: popcnt writes the WHOLE result register, there is ABSOLUTELY
>>>>       no need to clear it beforehand nor to clear the higher 24 bits
>>>>       afterwards!
>>>
>>> Except that there is.  See https://gcc.gnu.org/PR62011 for details.
>>
>> GUESS WHY I EXPLICITLY WROTE
>>
>> | JFTR: before GCC zealots write nonsense: see -march= or -mtune=
>>
>> NOT AMUSED ABOUT YOUR CRYING INCOMPETENCE!
>
> So you want to complain about GCC knowing about an Intel performance errata????

Ever heard about cargo cult?

Read the options I used CAREFULLY

| gcc -m32 -march=alderlake -O3
| gcc -m32 -march=sapphirerapids -O3
| gcc -m32 -mpopcnt -mtune=sapphirerapids -O3

> If you read that bug report, you would have learned why the zeroing is
> done. I am sorry you hate GCC for actually following advice from the
> processor maker after all.

HSD146 does NOT apply to Alder Lake or Sapphire Rapids!

NOT AMUSED ABOUT YOUR INCOMPETENCE TOO!
Stefan

Re: Another epic optimiser failure

[Jonathan Wakely]

On Sat, 27 May 2023 at 23:03, Stefan Kanthak <stefan.kanthak@nexgo.de> wrote:
>
> "Andrew Pinski" <pinskia@gmail.com> wrote:
>
> > On Sat, May 27, 2023 at 2:38 PM Stefan Kanthak <stefan.kanthak@nexgo.de> wrote:
> >>
> >> "Jakub Jelinek" <jakub@redhat.com> wrote, completely clueless:
> >>
> >>> On Sat, May 27, 2023 at 11:04:11PM +0200, Stefan Kanthak wrote:
> >>>> OUCH: popcnt writes the WHOLE result register, there is ABSOLUTELY
> >>>>       no need to clear it beforehand nor to clear the higher 24 bits
> >>>>       afterwards!
> >>>
> >>> Except that there is.  See https://gcc.gnu.org/PR62011 for details.
> >>
> >> GUESS WHY I EXPLICITLY WROTE
> >>
> >> | JFTR: before GCC zealots write nonsense: see -march= or -mtune=
> >>
> >> NOT AMUSED ABOUT YOUR CRYING INCOMPETENCE!
> >
> > So you want to complain about GCC knowing about an Intel performance errata????
>
> Ever heard about cargo cult?
>
> Read the options I used CAREFULLY
>
> | gcc -m32 -march=alderlake -O3
> | gcc -m32 -march=sapphirerapids -O3
> | gcc -m32 -mpopcnt -mtune=sapphirerapids -O3
>
> > If you read that bug report, you would have learned why the zeroing is
> > done. I am sorry you hate GCC for actually following advice from the
> > processor maker after all.
>
> HSD146 does NOT apply to Alder Lake or Sapphire Rapids!
>
> NOT AMUSED ABOUT YOUR INCOMPETENCE TOO!
> Stefan

File a proper bug report then, instead of subjecting this mailing list
to your clown show.

Re: Another epic optimiser failure

[Nicholas Vinson]

On 5/27/23 17:04, Stefan Kanthak wrote:
> --- .c ---
> int ispowerof2(unsigned long long argument) {
>      return __builtin_popcountll(argument) == 1;
> }
> --- EOF ---
>
> GCC 13.3    gcc -m32 -march=alderlake -O3
>              gcc -m32 -march=sapphirerapids -O3
>              gcc -m32 -mpopcnt -mtune=sapphirerapids -O3
>
> https://gcc.godbolt.org/z/cToYrrYPq
> ispowerof2(unsigned long long):
>          xor     eax, eax        # superfluous
>          xor     edx, edx        # superfluous
>          popcnt  eax, [esp+4]
>          popcnt  edx, [esp+8]
>          add     eax, edx
>          cmp     eax, 1      ->    dec  eax
>          sete    al
>          movzx   eax, al         # superfluous
>          ret
>
> 9 instructions in 28 bytes      # 6 instructions in 20 bytes

I agree this can be done using 6 instructions, but you cannot do it 
using the dec instruction. If you use the dec instruction, "movzx eax, 
al" becomes a required instruction (consider the case when the input is 
0) resulting in 7 instructions and 22 bytes.

Re: Another epic optimiser failure

[Julian Waters]

[-- Attachment #1: Type: text/plain, Size: 5497 bytes --]

"There's your first mistake. Hint: people who are able to hand deconstruct
the output of a compiler's code generator and point out exactly how
instructions are wasted are never correctly referred to as an "idiot", in
the context of computer programming at least."

gdb -batch -ex 'set disassembly-flavor intel' -ex 'file /bin/ls' -ex
'disassemble method'


Being able to run a disassembler does not make you intelligent or hot shit ;)

"Strange reasoning you've used here. Is this sort of like how if I'm
against Donald Trump, then I must be for Hillary Clinton, or vice
versa?


That's called a "false dichotomy" FYI."


Literally a few sentences later:


"Are the GCC developers *trying* to subtly push everyone toward Clang,
by slowly degrading GCC over time in hopes that people will eventually
give up and leave in frustration? Serious question."


"You mean the ones which are unclear and uncertain, because the GCC
documentation is inaccurate or simply lies?"


I will concede that gcc's documentation is pretty horrible, but that's
the exact reason I asked the gcc developers

and maintainers to list the options out here instead. Maybe you need
to brush up on your comprehension skills?

"Do you have any rebuttals of his argument to present yourself? Or do
you prefer to just sit back and wait on "y'all" to do the heavy
lifting?"

I would, if he hadn't already been absolutely schooled in more recent
replies pointing out why gcc produces certain code sequences.

At least he had the maturity to apologize and make his leave once he
found out he had made several mistakes, unlike

a certain someone I'm speaking to right now. And the few corner cases
he mentions that are valid he didn't even bother

filing a bug report, but instead resorted to endlessly screaming on a
mailing list without letting anyone talk to

him properly.


"What version of GCC can we expect to generate efficient and correct
code for this brand new, just-released "x86" instruction set? Maybe
GCC 97 will finally get it right...which at the current rate of major
version number increase, should be some time next year I guess.

Or rather more accurately, when will GCC's code generator stop
regressing as it seemingly has done for many versions now, and finally
Make Compiling Great Again?"

Again, you may want to look at the more recent code snippets posted
after the main argument got sidetracked, I don't

even have to do anything, hilariously enough.


It's also amazing how you managed to overpoliticize the rest of your
post after using that criticism against me

initially, how ironic.


"Ever heard the saying "if you can't run with the big dogs, stay under
the porch"?"

You think you're one of the big dogs? Pffft, that's cute.


I have no interest in getting into a compiler pissing match past this
point, so be on your way. Go annoy other people instead with

your garbage.


On Tue, May 30, 2023 at 2:59 AM Dave Blanchard <dave@killthe.net> wrote:

> On Sun, 28 May 2023 15:50:41 +0800
> Julian Waters via Gcc <gcc@gcc.gnu.org> wrote:
>
> > Man, these clang fanboys sure are getting out of hand
>
> Strange reasoning you've used here. Is this sort of like how if I'm
> against Donald Trump, then I must be for Hillary Clinton, or vice versa?
>
> That's called a "false dichotomy" FYI.
>
> > I feel like all this garbage can be easily resolved by y'all showing this
> > idiot
>
> There's your first mistake. Hint: people who are able to hand deconstruct
> the output of a compiler's code generator and point out exactly how
> instructions are wasted are never correctly referred to as an "idiot", in
> the context of computer programming at least.
>
> He's certainly got a few things wrong from time to time in his zeal, but
> his overall point seems to stand. Do you have any rebuttals of his argument
> to present yourself? Or do you prefer to just sit back and wait on "y'all"
> to do the heavy lifting?
>
> > the exact proper options required
>
> You mean the ones which are unclear and uncertain, because the GCC
> documentation is inaccurate or simply lies?
>
> > and attaching the resulting compiled assembly exactly as he wants it
>
> And what if GCC is unable to produce anything like that, because the code
> generator is at the very least questionable, as his postings seems to prove?
>
> > or if gcc doesn't compile the exact assembly he wants, explaining why
> gcc chose a different
> > route than the quote on quote "Perfect assembly" that he expects it to
> spit
> > out
>
> What version of GCC can we expect to generate efficient and correct code
> for this brand new, just-released "x86" instruction set? Maybe GCC 97 will
> finally get it right...which at the current rate of major version number
> increase, should be some time next year I guess.
>
> Or rather more accurately, when will GCC's code generator stop regressing
> as it seemingly has done for many versions now, and finally Make Compiling
> Great Again?
>
> > And Stefan? Ever heard of the saying that "the loudest man in the room is
> > always the weakest"?
>
> Ever heard the saying "if you can't run with the big dogs, stay under the
> porch"?
>
> Are the GCC developers *trying* to subtly push everyone toward Clang, by
> slowly degrading GCC over time in hopes that people will eventually give up
> and leave in frustration? Serious question.
>
> Dave
>
>

Re: Another epic optimiser failure

[Nicholas Vinson]

[-- Attachment #1: Type: text/plain, Size: 3887 bytes --]

On 5/29/23 15:01, Dave Blanchard wrote:

> He's certainly got a few things wrong from time to time in his zeal, but his overall point seems to stand. Do you have any rebuttals of his argument to present yourself? Or do you prefer to just sit back and wait on "y'all" to do the heavy lifting?

He's gotten many details wrong including the proper flags to set for gcc 
(and the "bad documentation" does not justify all the errors he's made), 
his hand-generated assembly (I've personally pointed out logic errors in 
his assembly on more than on occasion), and has failed to provide 
evidence that his solutions are better.

In almost all of his examples, he uses -O3 which is basically the "speed 
above all else" optimization level. I pointed this out before; I also 
pointed out that the smallest code (in bytes) with the fewest 
instructions is not always the fastest. He has not provided any data 
showing that his solutions result in faster executing code than what gcc 
produces. He has also raised questions that show a distinct lack of 
understanding when it comes to storage hierarchy; something I feel one 
would need to know to properly write fast assembly. Finally, I will 
admit some of the examples of gcc produced code are a bit suspicious, 
and probably should be reviewed.

In short Stefan is not being taken seriously because he is not 
presenting himself, or his arguments, in a manner that would convince 
people to take him seriously. As long as Stefan continues to communicate 
in such a manner, we're going to see similar such responses from (some 
of) the gcc devs (unfortunately).

The best next steps for Stefan, would be to review the constructive 
criticism, expand on his examples by providing explanation and proof as 
to why they're better, and then present these updated findings in the 
proper manner.

Using his first example as my own, take the C code:

	int ispowerof2(unsigned long long argument)
	{
      		return (argument & argument - 1) == 0;
	}

when compiled produces:

% gcc -m32 -O3 -c ispowerof2.c && objdump -d -Mintel ispowerof2.o

ispowerof2.o:     file format elf32-i386

Disassembly of section .text:

	00000000 <ispowerof2>:
	   0:   f3 0f 7e 4c 24 04       movq   xmm1,QWORD PTR [esp+0x4]
	   6:   66 0f 76 c0             pcmpeqd xmm0,xmm0
	   a:   66 0f d4 c1             paddq  xmm0,xmm1
	   e:   66 0f db c1             pand   xmm0,xmm1
	  12:   66 0f 7e c2             movd   edx,xmm0
	  16:   66 0f 73 d0 20          psrlq  xmm0,0x20
	  1b:   66 0f 7e c0             movd   eax,xmm0
	  1f:   09 c2                   or     edx,eax
	  21:   0f 94 c0                sete   al
	  24:   0f b6 c0                movzx  eax,al
	  27:   c3                      ret

Whereas he claims the following is better:

	movq    xmm1, [esp+4]
	pcmpeqd xmm0, xmm0
	paddq   xmm0, xmm1
	pand    xmm0, xmm1
	pxor    xmm1, xmm1
	pcmpeqb xmm0, xmm1
	pmovmskb eax, xmm0
	cmp     al, 255
	sete    al
	ret

because it has 10 instructions and is 36 bytes long vs the 11 
instructions and 40 bytes. However, the rebuttals are 1. his code is 
wrong (can return values other than 0 or 1) and 2. -O3 doesn't optimize 
on instruction count or  byte size (as an aside: clang's output uses 14 
instructions but is only 32 bytes in size -- is it better or worse than 
gcc's?).

Therefore, while he's 1 instruction less and 4 bytes fewer (1 byte fewer 
if you add the needed correction), he presents no evidence that his 
solution is actually faster. What he would need to do instead is show 
proof that his solution is indeed faster than what gcc produces.

Afterwards, he would be in a position to represent this data in a proper 
manner.

Re: Another epic optimiser failure

[Dave Blanchard]

On Sun, 28 May 2023 15:50:41 +0800
Julian Waters via Gcc <gcc@gcc.gnu.org> wrote:

> Man, these clang fanboys sure are getting out of hand

Strange reasoning you've used here. Is this sort of like how if I'm against Donald Trump, then I must be for Hillary Clinton, or vice versa?

That's called a "false dichotomy" FYI.

> I feel like all this garbage can be easily resolved by y'all showing this
> idiot

There's your first mistake. Hint: people who are able to hand deconstruct the output of a compiler's code generator and point out exactly how instructions are wasted are never correctly referred to as an "idiot", in the context of computer programming at least.

He's certainly got a few things wrong from time to time in his zeal, but his overall point seems to stand. Do you have any rebuttals of his argument to present yourself? Or do you prefer to just sit back and wait on "y'all" to do the heavy lifting?

> the exact proper options required 

You mean the ones which are unclear and uncertain, because the GCC documentation is inaccurate or simply lies?

> and attaching the resulting compiled assembly exactly as he wants it

And what if GCC is unable to produce anything like that, because the code generator is at the very least questionable, as his postings seems to prove?

> or if gcc doesn't compile the exact assembly he wants, explaining why gcc chose a different
> route than the quote on quote "Perfect assembly" that he expects it to spit
> out

What version of GCC can we expect to generate efficient and correct code for this brand new, just-released "x86" instruction set? Maybe GCC 97 will finally get it right...which at the current rate of major version number increase, should be some time next year I guess.

Or rather more accurately, when will GCC's code generator stop regressing as it seemingly has done for many versions now, and finally Make Compiling Great Again?

> And Stefan? Ever heard of the saying that "the loudest man in the room is
> always the weakest"?

Ever heard the saying "if you can't run with the big dogs, stay under the porch"?

Are the GCC developers *trying* to subtly push everyone toward Clang, by slowly degrading GCC over time in hopes that people will eventually give up and leave in frustration? Serious question. 

Dave

Re: Another epic optimiser failure

[Julian Waters]

[-- Attachment #1: Type: text/plain, Size: 507 bytes --]

Man, these clang fanboys sure are getting out of hand

I feel like all this garbage can be easily resolved by y'all showing this
idiot the exact proper options required and attaching the resulting
compiled assembly exactly as he wants it, or if gcc doesn't compile the
exact assembly he wants, explaining why gcc chose a different
route than the quote on quote "Perfect assembly" that he expects it to spit
out

And Stefan? Ever heard of the saying that "the loudest man in the room is
always the weakest"?