| --- Log opened Thu Nov 27 00:00:46 2014 | ||
| olofk | If I ever make a CPU bus, the first priority will be that all signal names should have equal length | 09:18 |
|---|---|---|
| olofk | wb_we, I'm looking at you! | 09:19 |
| olofk | hmmm... making a wishbone upsizer was messier than I had first thought | 09:54 |
| olofk | I'm getting ever more interested in adding a burst length signal | 09:55 |
| olofk | I guess that in practice most masters already use a fixed burst length anyway | 09:57 |
| olofk | sb0, ysionneau: Got any wishbone upsizers as well? | 10:08 |
| stekern | olofk: I on the other side would have the names by increasing number | 10:12 |
| stekern | to enforce the sorting | 10:13 |
| maxpaln | howdy all - a quick performance question. Looking around the IIRC logs and forums I can see a general DMIPS value of around 1.5 DMIPS/MHz and a coremark value of around 80-90 @ 50 MHz. Does anyone know if these are realistic/sensible values? I'm had a request for a processor with a DMIPS/MHz value of above 1.2. | 10:29 |
| stekern | maxpaln: those are realistic values | 10:30 |
| maxpaln | stekern: thanks. | 10:30 |
| maxpaln | FWIW, these are pretty impressive figures for an embedded uP - the NIOS II is much lower IIRC. | 10:31 |
| stekern | your lm32 has comparable numbers though ;) | 10:32 |
| stekern | but mor1kx has a cooler name, so that's of course the obvious choice | 10:33 |
| sb0 | maxpaln, lm32 uses much fewer LUTs | 10:33 |
| sb0 | and has nicer ABI and ISA. | 10:33 |
| maxpaln | :-) mor1kx is a MUCH cooler name. | 10:34 |
| maxpaln | Yes, our LM32 is also one of our considerations - but the need to run Linux makes the mor1kx a more appealing option. | 10:35 |
| maxpaln | we'll probably offer both and let the customer decide. | 10:35 |
| sb0 | more resources invested into developing software for an inferior processor. yay. | 10:39 |
| maxpaln | Whoop! The new BFM is proving its worth - I have cleared out several niggling corner case bugs and managed to reduce the latency in the process. I am now seeing consistent success in HW when there are no problems in simulation. This makes life a lot easier! :-) | 14:06 |
| olofk | maxpaln: Yes, those were great additions for finding bugs. Are you ready to implement byte masking and 64 bit now? I'm starting to need that :) | 15:07 |
| _franck__ | maxpaln: olofk : as I am curious I tested this new version of wb_bfm on my wb_altera_ddr_wrapper.... | 17:01 |
| _franck__ | as expected it found a bug :) | 17:02 |
| mor1kx | [mor1kx] bandvig pushed 1 new commit to withfpu: https://github.com/openrisc/mor1kx/commit/89c4fe5b72eddf1a955f130a52e86d159816c3ed | 19:31 |
| mor1kx | mor1kx/withfpu 89c4fe5 Andrey Bacherov: Rounding module is shared across FPU now. | 19:31 |
| --- Log closed Fri Nov 28 00:00:47 2014 | ||
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!