1 |
On Fri, Sep 15, 2006 at 04:47:14PM +0000, Duncan wrote: |
2 |
|
3 |
> I'm unclear as to what "vectorization" means as used here. My |
4 |
> understanding of "vector" is as a synonym for "line", thus implying loop |
5 |
> unrolling of some form or another, which will increase size. As I |
6 |
> explained in the grandparent, I believe such optimizations to be |
7 |
> counterproductive on modern processors due to the extreme cost of cache |
8 |
> misses as opposed to slight cycle inefficiencies. |
9 |
> |
10 |
> I am however aware that vectorization has a somewhat different meaning in |
11 |
> programming terms than the above, but am not sufficiently educated on the |
12 |
> topic to make an informed choice, so I've simply left gcc to go with its |
13 |
> default choice given my overall stated intention of -Os. |
14 |
> |
15 |
> If you can sufficiently explain the concept to me such that I |
16 |
> understand enough about it to feel comfortable going with other than the |
17 |
> default (which means I can explain why I chose it and why it won't |
18 |
> interfere with my overall strategy as outlined in the grandparent, or is |
19 |
> worth it even if it does), I'd be very grateful! =8^) |
20 |
|
21 |
Back in the day, vectorization was, I believe, a supercomputer SIMD |
22 |
(single instruction multiple data) concept, where instruction operands |
23 |
were pointers to data, so it would, for instance, add two arrays of |
24 |
numbers to produce a third array. Isn't this what the Altivec |
25 |
instructions do? |
26 |
|
27 |
I do not claim to know this is what gcc means, or even if the current |
28 |
concept of vectorization has any relation to Yee Olde Concepte. |
29 |
|
30 |
-- |
31 |
... _._. ._ ._. . _._. ._. ___ .__ ._. . .__. ._ .. ._. |
32 |
Felix Finch: scarecrow repairman & rocket surgeon / felix@×××××××.com |
33 |
GPG = E987 4493 C860 246C 3B1E 6477 7838 76E9 182E 8151 ITAR license #4933 |
34 |
I've found a solution to Fermat's Last Theorem but I see I've run out of room o |
35 |
-- |
36 |
gentoo-amd64@g.o mailing list |