1 |
Am 02.05.2015 um 14:38 schrieb Nikos Chantziaras: |
2 |
> On 02/05/15 15:10, Volker Armin Hemmann wrote: |
3 |
>> Am 02.05.2015 um 14:06 schrieb Nikos Chantziaras: |
4 |
>>> On 02/05/15 14:37, Volker Armin Hemmann wrote: |
5 |
>>>> Am 02.05.2015 um 13:25 schrieb Nikos Chantziaras: |
6 |
>>>>>>> |
7 |
>>>>>>> The kernel uses -O2 and several -march variants (e.g. |
8 |
>>>>>>> -march=core2). |
9 |
>>>>>>> Several other options are used to prevent GCC from generating |
10 |
>>>>>>> unsuitable code. |
11 |
>>>>>>> |
12 |
>>>>>>> Specifying another -march variant does not affect the optimizer |
13 |
>>>>>>> though. It only affects the code generator. If you don't modify the |
14 |
>>>>>>> other CFLAGS and only change -march, you will not get FP |
15 |
>>>>>>> instructions |
16 |
>>>>>>> unless you use FP in the code. |
17 |
>>>>>> |
18 |
>>>>>> http://www.agner.org/optimize/calling_conventions.pdf |
19 |
>>>>> |
20 |
>>>>> Not sure what you're trying to say. |
21 |
>>>>> |
22 |
>>>> |
23 |
>>>> that simd is not save in kernel if not carefully guarded. |
24 |
>>>> |
25 |
>>>> Really people, just don't fuck around with the cflags. |
26 |
>>> |
27 |
>>> I still fail to see the relevance. Unless you mean using a different |
28 |
>>> -O level. In that case, yes. You shouldn't. But I was talking about |
29 |
>>> -march. |
30 |
>>> |
31 |
>> |
32 |
>> you said this |
33 |
>> |
34 |
>>> |
35 |
>>> (note that SIMD is not FP and is perfectly fine in the kernel.) |
36 |
>> |
37 |
>> and I have shown you that you are wrong. |
38 |
> |
39 |
> Not sure why you think that. The kernel crypto routines are full of |
40 |
> SIMD code (like SSE and AVX.) Automatic vectorization wouldn't work. |
41 |
> But -march is not going to introduce that |
42 |
|
43 |
and never used in interrupt context and carefully guarded. You act like |
44 |
'oh, you can use simd instructions without any consideration' and that |
45 |
is just not true. |