 |
Post AuxgwzLC0EMu93gmxs by [email protected] |
 |
More posts by [email protected] |
|
Post #AuxgwyVR6ey5YXTSIC by [email protected] |
|
0 likes, 0 repeats |
|
|
|
on my blog!performance of random floatsin which i examine the relative cost of … |
|
|
|
Post #AuxgwycWgIdhuWn7LM by [email protected] |
|
0 likes, 0 repeats |
|
|
|
oh good grief i have discovered that a better compiler significantly changes th… |
|
|
|
Post #AuxgwyovwAZ2X0b1gO by [email protected] |
|
0 likes, 0 repeats |
|
|
|
also bloody amd64 is running my baseline slower than my functions that do workh… |
|
|
|
Post #Auxgwyw1VoEeszugjY by [email protected] |
|
0 likes, 0 repeats |
|
|
|
hmm no, `cpuid` is not a good replacement for `isb sy` |
|
|
|
Post #Auxgwz3p2oTRHBYutE by [email protected] |
|
0 likes, 0 repeats |
|
|
|
aha!i think i have succeeded by passing arguments and return values via pointer… |
|
|
|
Post #AuxgwzBcZoiDfND92u by [email protected] |
|
0 likes, 0 repeats |
|
|
|
RIGHTi added a baseline to my benchmark which revealed my amd64 numbers were no… |
|
|
|
Post #AuxgwzLC0EMu93gmxs by [email protected] |
|
0 likes, 1 repeats |
|
|
|
@fanf Still believe that baseline and Vaseline SHOULD rhyme. |
|
|
|
Post #Auzke1YYKV6JeB4RgO by [email protected] |
|
0 likes, 0 repeats |
|
|
|
@fanf does it spit out the bit hacking version? |
|
|
|
Post #Auzke1eZy5vBwrtG4m by [email protected] |
|
0 likes, 0 repeats |
|
|
|
@dysfun no that's the same (it uses a very neat bfxil instruction)the new t… |
|
|
|
Post #Auzke1lJZ3JEHl2dZg by [email protected] |
|
0 likes, 0 repeats |
|
|
|
@fanf @dysfun Oh! That's a nice one. |
|
|
|
Post #Auzke1rhBKPgbY1jWK by [email protected] |
|
0 likes, 0 repeats |
|
|
|
@mbr @dysfun isn’t it! arm says that variety of ucvtf is intended for fixed-p… |
|
|
|
Post #AuzkeROAHkExlUPW6K by [email protected] |
|
0 likes, 0 repeats |
|
|
|
@fanf isn't LFENCE enough? (since it does double duty as speculation barrie… |
|
|
|
Post #AuzkeRTTxyUg1ytlOC by [email protected] |
|
0 likes, 0 repeats |
|
|
|
@harold oh, lfence might be what i was looking for!feedback is a good idea, tha… |
|
|
|
Post #AuzkeRcLR1aCTT2qCe by [email protected] |
|
0 likes, 0 repeats |
|
|
|
@harold so i tried lfence and it sort-of workswith one lfence in the loop i get… |
|
|
|
Post #AuzkekheTW7LfQfqc4 by [email protected] |
|
0 likes, 0 repeats |
|
|
|
@fanf I'd love to see some ministat output (assuming the timings are approx… |
|
|
|
Post #Auzkekok39my1PzVfE by [email protected] |
|
0 likes, 0 repeats |
|
|
|
@wollman several samples examined with a mark one eyeball is enough to see that… |
|
|
|
Post #AuzkfKFYJT1zuhnuoS by [email protected] |
|
0 likes, 0 repeats |
|
|
|
@fanf ah, the lovely "helping hand" of the compiler optimizer 🙃 |
|
|
|
Post #AuzkfKMHuQQ2FaxIJM by [email protected] |
|
0 likes, 0 repeats |
|
|
|
@jelu i have defeated the compiler by using separate translation units; it'… |
|
|
|
Post #AuzkfMo0oJmfq9wSno by [email protected] |
|
0 likes, 0 repeats |
|
|
|
@fanf yeah, layers of re-compilers and optimizers 😒 |
