07:52 <joseng> Hi all,

07:52 <joseng> can anybody help me to understand the Streams in litex? I have a gateware design with classic FIFOs and now I wanted to try to move over to Streams between my modules and therefore use the stream.AsyncFIFO.

07:52 <joseng> https://gist.github.com/Wardstein/4d3004c2a91f8fbb9594f49d8f3aea57

07:52 <joseng> This is what I currently have. But I get never any data at the output side of the FIFO. Did I get the sources/sinks the wrong way round? Can't I pass an endpoint as a parameter? Or did I connect the FIFO wrong?

09:53 <acathla> _florent_, good, removing the line 937 of soc.py fixed the reboot (I still have to understand how's everything is connected)

09:54 <acathla> and flash works now!

09:57 <acathla> with vexriscv or serv!

10:27 <acathla> well, uploading worked, flashing...not yet

10:34 <zyp> joseng, I think you've got lines 27-28 the wrong way around, IIRC it should be sink.connect(source), not source.connect(sink)

10:47 <_florent_> Hi, I just tried to write a new LiteX's Wiki guide to prepare the Acorn CLE 215+ and use it with LiteX: https://github.com/enjoy-digital/litex/wiki/Use-LiteX-on-the-Acorn-CLE-215

10:49 <_florent_> Hoping it will be useful for those who want to use it as a FPGA development board

10:50 <_florent_> acathla: thanks for the feedback, I'll try to fix the reset today

11:11 <trabucayre> _florent_: I've received my acorn. I've just a question about power supply: 3.3V is enough or an external one (PCIe) is required

11:25 <zyp> trabucayre, which acorn? the cle-215? it gets all the power it needs from the m.2 slot

11:26 <trabucayre> zyp: yep cle-215+

11:29 <trabucayre> argg :-/

11:32 <trabucayre> power supply for m.2 m key is only 3.3V no?

11:32 <zyp> yes

11:35 <trabucayre> in fact I've received the cle-215 but forgot to order m.2 <-> PCIe so if it's possible to provide 3.3V through Pico-ZEMate it's allow me to start before receiving adapter

11:36 <zyp> ah, it probably is but I can check the schematic

11:38 <zyp> yeah, the pico-ezmates are connected directly to the same 3.3V rail as the m.2 edge

11:38 <trabucayre> yes!!!

11:39 <trabucayre> thanks

14:32 <st-gourichon-fid> Quick survey (re regarding https://github.com/enjoy-digital/litex/pull/725 I guess you _florent_ replied ): we chose "extract" and "inject" rather than "get" and "set" that seemed "generic"/"vague".

14:33 <st-gourichon-fid> Can anyone provide kind of "prior art" where a convenience function would extract an integer value from a multi-field words? How is it typically called?

14:34 <st-gourichon-fid> Same question for a convenience function that takes a word and an integer, and returns the word with some bit range changed from the integer value. How it this typically called?

14:35 <daveshah> it's called extract in yosys: https://github.com/YosysHQ/yosys/blob/master/kernel/rtlil.h#L660-L666

14:35 <daveshah> I'm not sure about a good term for the other way round though

14:42 <st-gourichon-fid> thanks daveshah, interesting. We are open to using any wording, it's best if we follow some existing common practice.

14:52 <st-gourichon-fid> In book "The Art of Assembly Language" Randall Hyde uses "extract" and "insert" https://flylib.com/books/en/1.330.1.30/1/ . https://www.plantation-productions.com/Webster/www.artofasm.com/Windows/HTML/BitManipulationa2.html

14:52 <tpb> Title: 3.7 Packing and Unpacking Data | The Art of Assembly Language (at flylib.com)

14:53 <st-gourichon-fid> I guess "get" and "set" suggested by _florent_ may feel familiar, like setting and getting variables in an object, something lightweight?

14:53 <_florent_> st-gourichon-fid, daveshah: thanks, I was also going to suggest extract/insert

14:55 <st-gourichon-fid> "Insert" did not come to mind because it usually suggests moving existing things (e.g. insert in list), not changing in-place.

14:56 <_florent_> st-gourichon-fid: I think extract was fine on my side but can't explain why I was not satisfied by inject. set/get are maybe indeed too generic and vague

14:59 <st-gourichon-fid> To be fair, the meaning of "inject" also kind-of suggests displacement. It just avoids "insert" which already has a specific meaning of moving existing thing to make room for the new thing.

15:00 <st-gourichon-fid> How about "extract"/"set"?

15:04 <st-gourichon-fid> Or "field2int"/"int2field"? All seem kind of awkward. Perhaps "set" and "get" are food after all. Choose what you wish, I will adjust the PR according to your preference. ¯\_(ツ)_/¯

15:04 <st-gourichon-fid> s/food/good/

15:05 <_florent_> if we use set, then I have preference for get /set :)

15:05 <_florent_> in fact I think the context is a bit different than Yosys here

15:05 <st-gourichon-fid> https://www.wordreference.com/definition/inject When injecting something, it's inside but you don't see obvious displacement.

15:05 <tpb> Title: inject - WordReference.com Dictionary of English (at www.wordreference.com)

15:06 <_florent_> because we are creating C functions for each CSR registers and then here CSR fields

15:07 <_florent_> so it's a bit equivalent to the function we could write in Python for a class

15:07 <_florent_> reg.write(), reg.read()

15:08 <_florent_> field.set(), field.get()

15:11 <_florent_> in Yosys, with the extract function, the offset and len are argument, here the offset/len will not be arguments since will be provided as part of the function

15:52 <Melkhior> Not sure but I think this is similar to "2.5 Bit Extract/Deposit (bext, bdep)" in B ? https://raw.githubusercontent.com/riscv/riscv-bitmanip/master/bitmanip-draft.pdf

15:52 <Melkhior> so extract/deposit could be an appropriate terminology ?

16:16 <st-gourichon-fid> Thanks Melkhior for the reference.

16:16 <_florent_> Melkhior: thanks that's indeed interesting. set/get is simple but it's not clear if it's operating directly on the hardware or a value. Inject/Insert sounds like something cumulative (if you inject/insert twice you expect result to be different), so yes it could be extract/deposit.

16:16 <st-gourichon-fid> This is interesting. I read the C-equivalent code. Factually, "extract" is capable of doing exactly what we are talking about. But "deposit" only does the "bit shifting" part, not "inserting" the shifted bit in the existing word value. To say it another way, "deposit" produces a word where all bits outside of the mask are zero, which is not what "inject" does..

16:16 <st-gourichon-fid> We can live with that and call "deposit" what we do, still.

16:19 <st-gourichon-fid> https://en.wikipedia.org/wiki/Bit_manipulation_instruction_set#Parallel_bit_deposit_and_extract

16:19 <tpb> Title: Bit manipulation instruction set - Wikipedia (at en.wikipedia.org)

16:22 <st-gourichon-fid> What about "extract" and "replace" ? We "extract" an integer from a specific field in the word, and we "replace" the field in the word with the value of an integer.

16:25 <st-gourichon-fid> "Extract" exactly matches existing bit manipulation terminology, "deposit" is lower level.

16:25 <st-gourichon-fid> "replace" fixes the issue "if you inject/insert twice you expect result to be different". Indeed, if you replace something twice, the result is expected to be the same.

16:27 <_florent_> st-gourichon-fid: I like it, this is indeed representative to what we are doing here

16:28 <st-gourichon-fid> I like it too. I can adjust the PR now if it suits you.

16:30 <st-gourichon-fid> That's the exact terminology used in this question https://stackoverflow.com/questions/39684162/how-to-extract-values-in-bits-from-a-register-and-use-them-to-replace-bits-in-a (answer says "merge").

16:30 <tpb> Title: mips32 - How to extract values in bits from a register and use them to replace bits in a different register? (MIPs assembly language) - Stack Overflow (at stackoverflow.com)

16:31 <st-gourichon-fid> Same terminology in question and answer here: https://stackoverflow.com/questions/63967399/extract-certain-bits-from-a-and-replace-only-those-extracted-bits-in-b-at-a-cert

16:31 <tpb> Title: c - Extract certain bits from A and replace only those extracted bits in B at a certain position - Stack Overflow (at stackoverflow.com)

16:32 <st-gourichon-fid> Same here https://stackoverflow.com/questions/42755704/bits-manipulation-mips-assembly

16:32 <tpb> Title: Bits Manipulation MiPS Assembly - Stack Overflow (at stackoverflow.com)

16:36 <_florent_> st-gourichon-fid: if @daveshah @Melkhior are also fine with this, let's use extract/replace

16:36 <daveshah> sgtm

16:46 <Melkhior> works for me

16:46 <st-gourichon-fid> PR updated https://github.com/enjoy-digital/litex/pull/725

16:47 <_florent_> thanks all