Re: [PATCH] Builtins handling in IVOPT

[Wei Mi <wmi at google dot com>]

On Fri, Nov 22, 2013 at 9:21 AM, Wei Mi <wmi@google.com> wrote:
>> I think the problem can be showed by below example:
>> struct tag
>> {
>>   int a[10];
>>   int b;
>> };
>> struct tag s;
>> int foo(int len)
>> {
>>   int i = 0;
>>   int sum = 0;
>>   for (i = 0; i < len; i++)
>>     sum += barr (&s.a[i]);
>>
>>   return sum;
>> }
>> The dump before IVOPT is like:
>>
>>   <bb 4>:
>>   # i_16 = PHI <i_10(6), 0(3)>
>>   # sum_17 = PHI <sum_9(6), 0(3)>
>>   _6 = &s.a[i_16];
>>   _8 = barr (_6);
>>   sum_9 = _8 + sum_17;
>>   i_10 = i_16 + 1;
>>   if (len_5(D) > i_10)
>>     goto <bb 6>;
>>   else
>>     goto <bb 5>;
>>
>>   <bb 5>:
>>   # sum_11 = PHI <sum_9(4)>
>>   goto <bb 7>;
>>
>>   <bb 6>:
>>   goto <bb 4>;
>>
>> The iv use of _6 in bar(_6) is actually an memory address and it can
>> be computed efficiently for some targets.  For now IVOPT only honors
>> address type iv uses appeared in memory access, so this patch is
>> trying to handle this kind of address iv which is outside of memory
>> access just the same.  Please correct me if I am wrong.
>
> Yes, that is correct.
>

Sorry, to make a correction here. That is not my patch is doing. The
patch is not handling normal address exprs, but those exprs could be
expressed as mem accesses after builtins expand.

>>
>> After thought twice, I have two concerns about this:
>> 1) Why can't we just enhance the nolinear use logic to handle this
>> kind of iv use?  It's more complicated to handle it as a normal
>> address type iv, consider that IVOPT adds auto-increment candidates
>> according to them, you have to deal with that in this way
>> (auto-increment addressing mode can't apply to this kind of address iv
>> use).
>
> I think existing address iv use logic is enough to handle it. I am
> changing it and moving the gimple change from
> find_interesting_uses_stmt to rewrite_use_address in original patch.
>
>> 2) If I understand it right, this is an issue not only limited to
>> builtin functions, it stands for normal function calls too, right?
>>
>
> For builtin function, such as _mm_loadu_si128(b+4*i), it will be
> expanded to an insn: MOVDQU mem[b+4*i], $xmmreg, and the computation
> of b+4*i is for free. But for function call, the b+4*i will only be
> used as the value of an actual, and its computation cost cannot be
> avoided.
>
> Thanks,
> Wei.