*** tpb <[email protected]> has joined #litex | 00:00 | |
MoeIcenowy | somlo: so it looks like a proper H increases the size at ~50% ? | 02:01 |
---|---|---|
MoeIcenowy | (not properly calculated | 02:01 |
MoeIcenowy | somlo: chipyard is https://github.com/ucb-bar/chipyard | 02:01 |
cr1901 | proper H? | 02:01 |
MoeIcenowy | cr1901: previously we didn't properly enable H because of some weirdness of Rocket | 02:02 |
cr1901 | I didn't even know "H" was an extension | 02:03 |
* cr1901 lost count | 02:03 | |
MoeIcenowy | Hypervisor | 02:03 |
cr1901 | ahhhhh | 02:03 |
cr1901 | RISCV IMAFDCOMGWTFBBQ | 02:04 |
MoeIcenowy | cr1901: you are out of date | 02:05 |
cr1901 | Oh? | 02:05 |
MoeIcenowy | now it's RV64IMAFDCVSUZicsrZifenceiSs1p12 | 02:05 |
MoeIcenowy | like this | 02:05 |
cr1901 | It's missing the L spec | 02:05 |
cr1901 | :P | 02:05 |
cr1901 | (Also, My God we've run of out alphabet, haven't we?) | 02:06 |
MoeIcenowy | new single-letter extension now rarely appear | 02:06 |
MoeIcenowy | cr1901: yes | 02:06 |
MoeIcenowy | and it tend to divide single letter ones to multiple multi letter ones | 02:06 |
MoeIcenowy | e.g. original B is now ZbaZbbZbs ... | 02:06 |
cr1901 | Well, L spec is Decimal Floating Point, which I'm actually curious about just because it would be interesting. But there hasn't been movement on it AFAIK | 02:06 |
cr1901 | ahhhh | 02:06 |
cr1901 | Is the single letter "B" still used if the entirity of "ZbaZbbZb..." is implemented? | 02:07 |
MoeIcenowy | well I dont know... | 02:08 |
MoeIcenowy | Waterman says there's now no B extension | 02:09 |
MoeIcenowy | SiFive has some B option | 02:09 |
cr1901 | ahhh, I remember when Claire Wolf (yosys dev) was working on the B extension. Idk if she still is, but you're right that I'm out of date. | 02:09 |
MoeIcenowy | but it does not even contain Zbs | 02:09 |
MoeIcenowy | it's only ZbaZbb | 02:09 |
MoeIcenowy | cr1901: well in RV context maybe you should better call her picorv32 dev ;-) | 02:10 |
cr1901 | Hehehe fair enough :D! | 02:10 |
MoeIcenowy | I do admire picorv32 and it inspired me when I was doing my homework in college on Computer Orgnization | 02:11 |
cr1901 | There was no RV when I took Comp Arch, so we made our own bare minimum RISC CPU in Verilog w/ custom insns. | 02:12 |
MoeIcenowy | (well I did a very rubbish rv32i (even without Zicsr) core then, and did run something produced by GCC | 02:12 |
cr1901 | I have a RV32 core that's "mostly" ready, but I need to write the microcode for it (size optimized!). I'll get it merged into LiteX after that | 02:12 |
MoeIcenowy | oh I want to see it ;-) | 02:13 |
MoeIcenowy | I am always curious about microcoding, but I haven't really understood it | 02:13 |
MoeIcenowy | _florent_: could main ram be used as DMA destination? | 02:14 |
MoeIcenowy | GW2A-18 is not very rich of BRAM | 02:14 |
MoeIcenowy | but I got the 8MB SDRAM on GW2AR-18C running | 02:14 |
cr1901 | MoeIcenowy: https://mastodon.social/@cr1901/109847259530300392 Re: microcode, "Bit-Slice Microprocessor Design" by Mick and Brick | 02:19 |
tpb | Title: William D. Jones: "For those who have no idea how microcoded CPUs wo…" - Mastodon (at mastodon.social) | 02:19 |
MoeIcenowy | somlo: btw Zenithal says Zk* Zb* are heavy of size | 02:19 |
MoeIcenowy | ah thanks | 02:19 |
cr1901 | Like pico/femtorv32, my RV core isn't fast. It's meant to be small. Microcoding it is intended to abuse block RAMs as a way to use less FPGA logic | 02:22 |
MoeIcenowy | cr1901: or maybe it can use LUTRAM? | 02:27 |
cr1901 | Yes, that's also an option for non-ice40 FPGAs | 02:29 |
cr1901 | My scripts for testing size were meant for the worst case of ice40 (no LUTRAM) | 02:29 |
*** bl0x_ <[email protected]> has joined #litex | 02:51 | |
*** bl0x <[email protected]> has quit IRC (Ping timeout: 255 seconds) | 02:52 | |
MoeIcenowy | _florent_: GW2A-18 seems to be very short on BRAM, when I use standard vexriscv + sdcard, I got `The number(49) of BSRAM in the design exceeds the resource limit(46) of current device.` | 03:06 |
MoeIcenowy | maybe I should look at GowinSynthesis result | 03:09 |
somlo | All I wish for is yosys/nextpnr for whatever FPGA is on the Alveo boards ;P | 03:38 |
somlo | if only I didn't have to work for a living, I could dedicate the next 10 years to making it happen :D | 03:39 |
*** Degi_ <[email protected]> has joined #litex | 03:43 | |
*** Degi <[email protected]> has quit IRC (Ping timeout: 255 seconds) | 03:44 | |
*** Degi_ is now known as Degi | 03:44 | |
*** so-offish <so-offish!~so-offish@2610:148:610:2b11::17> has quit IRC (Ping timeout: 260 seconds) | 05:03 | |
*** GNUmoon <GNUmoon!~GNUmoon@gateway/tor-sasl/gnumoon> has quit IRC (Remote host closed the connection) | 06:54 | |
*** GNUmoon <GNUmoon!~GNUmoon@gateway/tor-sasl/gnumoon> has joined #litex | 06:54 | |
MoeIcenowy | well litesdcard seems to use 18 BSRAMs | 07:13 |
*** cr1901 <cr1901!~cr1901@2601:8d:8600:911:49ba:3ff7:be12:607f> has quit IRC (Read error: Connection reset by peer) | 08:12 | |
*** cr1901 <cr1901!~cr1901@2601:8d:8600:911:c093:ef31:d5fc:cfb> has joined #litex | 09:42 | |
*** shorne <[email protected]> has quit IRC (Read error: Connection reset by peer) | 11:30 | |
*** shorne <[email protected]> has joined #litex | 11:30 | |
*** SpaceCoaster <SpaceCoaster!~derek@user/spacecoaster> has joined #litex | 13:35 | |
*** so-offish <[email protected]> has joined #litex | 13:41 | |
*** so-offishul <so-offishul!~so-offish@2610:148:610:2b11::b> has joined #litex | 13:42 | |
*** so-offish <[email protected]> has quit IRC (Ping timeout: 276 seconds) | 13:46 | |
*** so-offishul <so-offishul!~so-offish@2610:148:610:2b11::b> has quit IRC (Quit: Leaving) | 18:23 | |
*** so-offish <so-offish!~so-offish@2610:148:610:2b11::b> has joined #litex | 18:23 | |
*** hrberg <[email protected]> has quit IRC (Ping timeout: 255 seconds) | 21:47 | |
*** somlo <[email protected]> has quit IRC (Ping timeout: 250 seconds) | 23:09 | |
*** somlo <[email protected]> has joined #litex | 23:12 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!