openpowerbot | [slack] <Benjamin Herrenschmidt> wht's going on with ghdl these days ? | 02:46 |
---|---|---|
openpowerbot | [slack] <Benjamin Herrenschmidt> ```microwatt$ make | 02:46 |
openpowerbot | [slack] <Benjamin Herrenschmidt> cc -O3 -Wall -c -o sim_vhpi_c.o sim_vhpi_c.c | 02:46 |
openpowerbot | [slack] <Benjamin Herrenschmidt> cc -O3 -Wall -c -o sim_bram_helpers_c.o sim_bram_helpers_c.c | 02:46 |
openpowerbot | [slack] <Benjamin Herrenschmidt> cc -O3 -Wall -c -o sim_console_c.o sim_console_c.c | 02:46 |
openpowerbot | [slack] <Benjamin Herrenschmidt> cc -O3 -Wall -c -o sim_jtag_socket_c.o sim_jtag_socket_c.c | 02:46 |
openpowerbot | [slack] <Benjamin Herrenschmidt> ghdl -i --std=08 --work=unisim --workdir=sim-unisim sim-unisim/BSCANE2.vhdl sim-unisim/BUFG.vhdl sim-unisim/unisim_vcomponents.vhdl | 02:47 |
openpowerbot | [slack] <Benjamin Herrenschmidt> ghdl -c --std=08 -Psim-unisim -Wl,sim_vhpi_c.o -Wl,sim_bram_helpers_c.o -Wl,sim_console_c.o -Wl,sim_jtag_socket_c.o decode_types.vhdl common.vhdl wishbone_types.vhdl fetch1.vhdl utils.vhdl plru.vhdl cache_ram.vhdl icache.vhdl predecode.vhdl decode1.vhdl helpers.vhdl insn_helpers.vhdl control.vhdl decode2.vhdl register_file.vhdl cr_file.vhdl crhelpers.vhdl ppc_fx_insns.vhdl rotator.vhdl logical | 02:47 |
openpowerbot | [slack] <Benjamin Herrenschmidt> ghdl:error: unknown warning identifier: l,sim_vhpi_c.o``` | 02:47 |
openpowerbot | [slack] <Benjamin Herrenschmidt> it doesn't support -Wl to pass linker options ? | 02:47 |
openpowerbot | [slack] <Benjamin Herrenschmidt> what is that vunit thing ? I can't figure out how that's supposed to work... | 02:51 |
openpowerbot | [slack] <joel> What version are you running? | 03:14 |
openpowerbot | [slack] <joel> ```$ ghdl --version | 03:15 |
openpowerbot | [slack] <joel> GHDL 1.0.0 (Debian 1.0.0+dfsg-8+b1) [Dunoon edition] | 03:15 |
openpowerbot | [slack] <joel> Compiled with GNAT Version: 10.3.0 | 03:15 |
openpowerbot | [slack] <joel> llvm code generator``` | 03:15 |
openpowerbot | [slack] <Benjamin Herrenschmidt> ```$ ghdl --version | 04:17 |
openpowerbot | [slack] <Benjamin Herrenschmidt> GHDL 3.0.0-dev (2.0.0.r652.g6961b3f82) [Dunoon edition] | 04:17 |
openpowerbot | [slack] <Benjamin Herrenschmidt> Compiled with GNAT Version: 10.3.0 | 04:17 |
openpowerbot | [slack] <Benjamin Herrenschmidt> mcode code generator | 04:17 |
openpowerbot | [slack] <Benjamin Herrenschmidt> Written by Tristan Gingold.``` | 04:17 |
openpowerbot | [slack] <Benjamin Herrenschmidt> but I had the same problem with `1.0.0+dfsg-6` (which I think was installing ghdl-mcode which could be the problem) | 04:18 |
openpowerbot | [slack] <joel> I tried it with today's ghdl git and it was ok. I can't remember if it was GCC or llvm built | 04:19 |
openpowerbot | [slack] <Matt Johnston> I'm seeing that same error here, using nightly oss-cad-suite tools | 04:19 |
openpowerbot | [slack] <Benjamin Herrenschmidt> yeah it's an mcode build, I'll rebuild with llvm | 04:19 |
openpowerbot | [slack] <Benjamin Herrenschmidt> it looks like ubuntu ships an mcode variant by default | 04:19 |
openpowerbot | [slack] <Matt Johnston> I have a vague recollection of sorting out that problem one other time | 04:19 |
openpowerbot | [slack] <Benjamin Herrenschmidt> ```$ apt-cache search ghdl | 04:19 |
openpowerbot | [slack] <Benjamin Herrenschmidt> ghdl - VHDL compiler/simulator | 04:19 |
openpowerbot | [slack] <Benjamin Herrenschmidt> ghdl-common - VHDL compiler/simulator (common files) | 04:19 |
openpowerbot | [slack] <Benjamin Herrenschmidt> ghdl-gcc - VHDL compiler/simulator (GCC backend) | 04:19 |
openpowerbot | [slack] <Benjamin Herrenschmidt> ghdl-llvm - VHDL compiler/simulator (LLVM backend) | 04:19 |
openpowerbot | [slack] <Benjamin Herrenschmidt> ghdl-mcode - VHDL compiler/simulator (mcode backend)``` | 04:19 |
openpowerbot | [slack] <Benjamin Herrenschmidt> have to install the right one 🙂 | 04:19 |
openpowerbot | [slack] <Benjamin Herrenschmidt> ghdl defaults to ghdl-mcode which is no good | 04:20 |
openpowerbot | [slack] <Benjamin Herrenschmidt> anyways I've sent a PR for the PLRU problem https://github.com/antonblanchard/microwatt/pull/399 | 04:20 |
openpowerbot | [slack] <Benjamin Herrenschmidt> hopefully improves performance a tiny bit | 04:20 |
openpowerbot | [slack] <Benjamin Herrenschmidt> the nature of the bug is such that bit 0 of LRU was correct, so we weren't hitting the worst case scenario of always replacing the entry we just used | 04:36 |
openpowerbot | [slack] <Benjamin Herrenschmidt> however other bits were wrong, so we would replace a "random" (sort-of) entry in the "other half" | 04:37 |
openpowerbot | [slack] <Benjamin Herrenschmidt> I wouldn't be surprised if the impact is negligible 🙂 | 04:37 |
openpowerbot | [slack] <Benjamin Herrenschmidt> @Paul Mackerras @Anton Blanchard @Michael Neuling Looking at the git history, you guys have been busy 🙂 | 04:41 |
openpowerbot | [slack] <Paul Mackerras> yep... | 04:45 |
openpowerbot | [slack] <Paul Mackerras> lately I have been hacking on the icache, so plru stuff is relevant | 04:45 |
openpowerbot | [slack] <Benjamin Herrenschmidt> @Paul Mackerras i like the 2 SPR RAMs, though I'm not fan of state in execute1 ... | 04:45 |
openpowerbot | [slack] <Benjamin Herrenschmidt> as long as we only update on complete instructions I suppose it's ok but ... how do you handle it bypass wise ? (ie back to back) | 04:46 |
openpowerbot | [slack] <Benjamin Herrenschmidt> hrm ... ignore my question | 04:47 |
openpowerbot | [slack] <Paul Mackerras> with a bypass path... 🙂 | 04:47 |
openpowerbot | [slack] <Benjamin Herrenschmidt> yeah... it's not too long though ? | 04:47 |
openpowerbot | [slack] <Benjamin Herrenschmidt> (the SRR1 path esp.) | 04:48 |
openpowerbot | [slack] <Benjamin Herrenschmidt> mind you it's no worse than the previous path via the regfile | 04:48 |
openpowerbot | [slack] <Benjamin Herrenschmidt> and should make synth life easier | 04:49 |
openpowerbot | [slack] <Benjamin Herrenschmidt> is there any value in getting a newer vivado btw ? | 04:53 |
openpowerbot | [slack] <Benjamin Herrenschmidt> I still have 2019.2 | 04:54 |
openpowerbot | [slack] <Benjamin Herrenschmidt> not that I have any time to play with this but ... 🙂 | 04:54 |
openpowerbot | [slack] <Benjamin Herrenschmidt> @Paul Mackerras ah I see you ended up adding an ex2 🙂 that was always going to happen I reckon 🙂 | 05:22 |
openpowerbot | [slack] <Paul Mackerras> yep | 05:23 |
openpowerbot | [slack] <Paul Mackerras> there's no extra latency for reading an SPR value that needs to be bypassed, and the need for the bypass is detected in decode2 | 05:32 |
openpowerbot | [slack] <Paul Mackerras> (mind you I don't provide for bypassing the SRR0/1 values written by an interrupt since we're redirecting and can't possibly be needing to read those values in the next cycle or two) | 05:33 |
openpowerbot | [slack] <Paul Mackerras> I haven't done extensive comparisons of vivado versions | 05:33 |
openpowerbot | [slack] <Paul Mackerras> I'm still using a 2019 version on my main machine | 05:33 |
openpowerbot | [slack] <Benjamin Herrenschmidt> ok | 05:58 |
*** fitzsim <fitzsim!~user@69-165-165-189.dsl.teksavvy.com> has quit IRC | 20:06 | |
*** fitzsim <fitzsim!~user@69-165-165-189.dsl.teksavvy.com> has joined #microwatt | 20:06 | |
*** ibot` <ibot`!~supybot@libre-soc.org> has joined #microwatt | 23:25 | |
*** ibot <ibot!~supybot@libre-soc.org> has quit IRC | 23:33 |
Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!