| programmerjake | when you want a $5000 cable with 5% efficiency: https://xkcd.com/2642/ | 08:57 |
|---|---|---|
| lkcl | coool :) | 09:09 |
| lkcl | can anyone remember where it was that discussion of why you have LD/ST-with-update in an ISA? | 14:20 |
| lkcl | programmerjake, it might have been you who raised it. | 14:20 |
| lkcl | i remember a loop around a struct with three FP arguments | 14:20 |
| lkcl | and LD/ST-with-update very elegantly saved at least one instruction | 14:21 |
| programmerjake | icr, but i can come up with some code... | 18:59 |
| programmerjake | https://gcc.godbolt.org/z/zjYoP4KfK | 19:17 |
| programmerjake | the ldu saves 1 instruction: it doesn't need a separate addi to increment the pointer | 19:18 |
| lkcl | maaagic | 19:52 |
| lkcl | https://groups.google.com/g/comp.arch/c/PCyKnVva39A/m/iDnmFIh1AgAJ | 19:56 |
| ghostmansd[m] | Success story. I managed to compile this mp3 assembly so that it is the same between binutils and pysvp64asm. | 20:24 |
| ghostmansd[m] | The issue was not in remap algorithm, but in some logic around the sv_mode detection. | 20:25 |
| ghostmansd[m] | Took a lot of time to debug, but the results are encouraging. :-) | 20:25 |
| ghostmansd[m] | So yeah, we now can officially use binutils for that file. Especially since I already committed an updated version of this assembly into openpower-isa repository. | 20:26 |
| programmerjake | yay! | 20:37 |
| lkcl | ghostmansd[m], ikeee | 23:12 |
| lkcl | huh. can't type can't spell "oleee" | 23:12 |
| programmerjake | unintentional finger offset... | 23:13 |
Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!