1 |
On 02/05/15 15:10, Volker Armin Hemmann wrote: |
2 |
> Am 02.05.2015 um 14:06 schrieb Nikos Chantziaras: |
3 |
>> On 02/05/15 14:37, Volker Armin Hemmann wrote: |
4 |
>>> Am 02.05.2015 um 13:25 schrieb Nikos Chantziaras: |
5 |
>>>>>> |
6 |
>>>>>> The kernel uses -O2 and several -march variants (e.g. -march=core2). |
7 |
>>>>>> Several other options are used to prevent GCC from generating |
8 |
>>>>>> unsuitable code. |
9 |
>>>>>> |
10 |
>>>>>> Specifying another -march variant does not affect the optimizer |
11 |
>>>>>> though. It only affects the code generator. If you don't modify the |
12 |
>>>>>> other CFLAGS and only change -march, you will not get FP instructions |
13 |
>>>>>> unless you use FP in the code. |
14 |
>>>>> |
15 |
>>>>> http://www.agner.org/optimize/calling_conventions.pdf |
16 |
>>>> |
17 |
>>>> Not sure what you're trying to say. |
18 |
>>>> |
19 |
>>> |
20 |
>>> that simd is not save in kernel if not carefully guarded. |
21 |
>>> |
22 |
>>> Really people, just don't fuck around with the cflags. |
23 |
>> |
24 |
>> I still fail to see the relevance. Unless you mean using a different |
25 |
>> -O level. In that case, yes. You shouldn't. But I was talking about |
26 |
>> -march. |
27 |
>> |
28 |
> |
29 |
> you said this |
30 |
> |
31 |
>> |
32 |
>> (note that SIMD is not FP and is perfectly fine in the kernel.) |
33 |
> |
34 |
> and I have shown you that you are wrong. |
35 |
|
36 |
Not sure why you think that. The kernel crypto routines are full of SIMD |
37 |
code (like SSE and AVX.) Automatic vectorization wouldn't work. But |
38 |
-march is not going to introduce that. |