Re: [RFC] Improve memset

From: Rasmus Villemoes
Date: Fri Sep 13 2019 - 05:18:05 EST

On 13/09/2019 11.00, Linus Torvalds wrote:
> On Fri, Sep 13, 2019 at 8:22 AM Borislav Petkov <bp@xxxxxxxxx> wrote:
>> since the merge window is closing in and y'all are on a conference, I
>> thought I should take another stab at it. It being something which Ingo,
>> Linus and Peter have suggested in the past at least once.
>> Instead of calling memset:
>> ffffffff8100cd8d: e8 0e 15 7a 00 callq ffffffff817ae2a0 <__memset>
>> and having a JMP inside it depending on the feature supported, let's simply
>> have the REP; STOSB directly in the code:
> That's probably fine for when the memset *is* a call, but:
>> The result is this:
>> static __always_inline void *memset(void *dest, int c, size_t n)
>> {
>> void *ret, *dummy;
>> asm volatile(ALTERNATIVE_2_REVERSE("rep; stosb",
> Forcing this code means that if you do
> struct { long hi, low; } a;
> memset(&a, 0, sizeof(a));
> you force that "rep stosb". Which is HORRID.
> The compiler should turn it into just one single 8-byte store. But
> because you took over all of memset(), now that doesn't happen.

OK, that answers my question.

> So we do need to have gcc do the __builtin_memset() for the simple cases..

Something like

if (__builtin_constant_p(c) && __builtin_constant_p(n) && n <= 32)
return __builtin_memset(dest, c, n);

might be enough? Of course it would be sad if 32 was so high that this
turned into a memset() call, but there's -mmemset-strategy= if one wants
complete control. Though that's of course build-time, so can't consider
differences between cpu models.