Many hyperlinks are disabled.
Use anonymous login
to enable hyperlinks.
216 check-ins using file cores/sia/Makefile version d8600c5a16
|
2020-02-08
| ||
| 07:08 | first steps for CLI wrapper Leaf check-in: e264c3cff1 user: kc5tja tags: makerom | |
|
2020-02-02
| ||
| 06:09 | I believe this completes the program loader for building a ROM image. There are a few error conditions that are not checked for; for example, segment IDs are not bounds checked. Similarly, it's not checked to see if a BSS segment has a HUNK_TEXT hunk with a non-zero length. However, these should be fairly easy to add later on, should these become problematic inputs to the MakeROM tool. check-in: 675b1e2afa user: kc5tja tags: makerom | |
| 05:31 | merge in (:remove) directive check-in: 051e950121 user: kc5tja tags: makerom | |
| 05:30 | Introduce the (:remove ...) directive Leaf check-in: 3a47f2f9d6 user: kc5tja tags: trunk | |
| 05:28 | Introduce (:remove ...) directive Closed-Leaf check-in: 36a562f18f user: kc5tja tags: brickie-remove-directive | |
| 05:28 | Create new branch named "brickie-remove-directive" check-in: 1b58c53c39 user: kc5tja tags: brickie-remove-directive | |
| 05:27 | Steps towards better support for BSS segments (and, now that I think about it, page protection as well). check-in: c3a23a7120 user: kc5tja tags: makerom | |
| 03:33 | HUNK_RELOC64 support added to load interpreter. Still need to implement support for distinguishing between BSS and CODE/DATA ahead of time, though; otherwise, we'd need two passes through the input file, which I'm not a fan of. check-in: 74b1dabdba user: kc5tja tags: makerom | |
| 01:59 | Import (:delete ...) function in Brickie check-in: 1bc0382449 user: kc5tja tags: makerom | |
| 01:58 | Introduce (:delete ...) command to Brickie. check-in: 33dcc7213e user: kc5tja tags: trunk | |
| 01:57 | Forgot the remove utility word Closed-Leaf check-in: 921e48fb0a user: kc5tja tags: brickie-delete-directive | |
| 01:53 | Add (:delete ...) directive check-in: f12588c011 user: kc5tja tags: brickie-delete-directive | |
| 01:52 | Create new branch named "brickie-delete-directive" check-in: a0542294ab user: kc5tja tags: brickie-delete-directive | |
| 01:49 | Lay down relocation hunks for processing. check-in: 439adf1ad6 user: kc5tja tags: makerom | |
|
2020-01-27
| ||
| 07:47 | Documentation fixes check-in: bb9ada72cc user: kc5tja tags: makerom | |
| 07:36 | Designed very basic load file interpreter. Correctly places segment contents. Does not yet understand relocations. check-in: 4844f0ae51 user: kc5tja tags: makerom | |
|
2020-01-26
| ||
| 23:03 | Choose better name for section check-in: 34a0e063d2 user: kc5tja tags: makerom | |
| 20:40 | Inserted a lot of explanation behind why I chose the magic constants that I did. They probably don't belong here. They probably need to reside in a separate reference document. But, at least I have them under revision control somewhere. check-in: 1f7133f0d8 user: kc5tja tags: makerom | |
| 18:50 | Update brickie check-in: da8e336246 user: kc5tja tags: makerom | |
| 18:49 | Introduce doc and code directives to make "weaving" easier. check-in: a28155cb66 user: kc5tja tags: trunk | |
| 18:45 | Rename (:code) to (:bottom), bringing symmetry with (:top), and also allowing me to use (:code) as a tag to identify where code picks up versus documentation. (:code) (which is also invoked implicitly in some other directives) enables outlist generation, while (:doc) disables it. This allows for bulk documentation in, e.g., raw Markdown or HTML to comingle with executable program code. Closed-Leaf check-in: b8c915baf6 user: kc5tja tags: doc-directive | |
| 18:10 | Create new branch named "doc-directive" check-in: f4190db675 user: kc5tja tags: doc-directive | |
| 18:09 | Starting work on makerom utility check-in: a06942ac8e user: kc5tja tags: makerom | |
| 16:25 | merge feature check-in: 55a465d93c user: kc5tja tags: makerom | |
| 16:25 | Place definition of lines in appropriate context check-in: 515b6aefb4 user: kc5tja tags: trunk | |
| 16:24 | Create new branch named "makerom" check-in: f754c53ba9 user: kc5tja tags: makerom | |
|
2020-01-16
| ||
| 20:10 | Introducing Brickie, a literate programming tool written for GForth, but it should be portable to DX-Forth with modest effort. It is based on Akkartik's "tangle" tool, and has similar syntax for markup. Brickie is optimized for my project's needs, however, and will evolve independently of Akkartik's work. Long story short, you write software as layers of related functionality: in a sense, you explicitly name and compose patches which ultimately leads to the finished product. This allows for code and commentary to co-reside, even when chunks of related functionality are widely separated in the source code. See http://akkartik.name/post/wart-layers for more behind the theory of this approach to literate programming. check-in: 63fa3ba241 user: kc5tja tags: trunk | |
| 20:00 | Clean up working file if tangle succeeds Closed-Leaf check-in: 9ac2855656 user: kc5tja tags: x-brickie | |
| 19:57 | Added makefile; makes tangle a binary check-in: 3f4f6d01a1 user: kc5tja tags: x-brickie | |
| 19:03 | Add tangle script check-in: ca2352c8b1 user: kc5tja tags: x-brickie | |
| 08:41 | Introduces :before and :after directives. Precise syntax isn't *quite* the same as Akkartik's version, but pretty close. This basically completes Brickie's inner guts. Next up, wrapping this up into a convenient shell-invokable script of some kind. check-in: 16be7efd90 user: kc5tja tags: x-brickie | |
| 07:27 | Evaluate directives as Forth command strings. check-in: b53ea0892d user: kc5tja tags: x-brickie | |
| 06:44 | Introduced an "insertion point" (IP) variable. This tracks where the next line of input will be placed in the output list. Future directives will adjust the insertion point; for example (:before XXX) will look for XXX in the outlist, and if found, will set IP to the immediately preceding line. (:after YYY) will do the same, but sets IP directly to the line containing YYY, so that new text will appear after it. (No facility exists for deletion of code yet; I'm not sure how that works. Need to study akkartik's code to learn more about that.) check-in: 86bd346a6c user: kc5tja tags: x-brickie | |
| 06:12 | refactor check-in: 2589910629 user: kc5tja tags: x-brickie | |
| 06:07 | refactor check-in: bece2ded3d user: kc5tja tags: x-brickie | |
| 05:51 | Forth instructions to read in a file, line by line. Accumulate the input; directives not yet implemented. Writes output file. check-in: 3192ac7b76 user: kc5tja tags: x-brickie | |
| 00:16 | Create new branch named "x-brickie" check-in: f2b941471e user: kc5tja tags: x-brickie | |
|
2020-01-11
| ||
| 04:56 | forgot about m4 check-in: 02c01ec8d4 user: kc5tja tags: trunk | |
| 04:50 | fix formatting check-in: 77617ea1f5 user: kc5tja tags: trunk | |
| 04:42 | Contribute official E3 user documentation check-in: b205b9737b user: kc5tja tags: trunk | |
| 03:54 | Try to fix build bug on Ubuntu 19.04 cloud box. (This might go on for a while...) check-in: 0154dbff6a user: kc5tja tags: trunk | |
|
2020-01-09
| ||
| 06:11 | Feature parity with E2, but much easier to add new virtual hardware. Also, much easier to support interrupts in the future as well. check-in: 7b40e899bd user: kc5tja tags: trunk | |
| 05:55 | BUG FIX: Distribute sub-dword-sized data across the data bus, as hardware would. Closed-Leaf check-in: 51649cc385 user: kc5tja tags: e3-cpu | |
|
2020-01-06
| ||
| 16:59 | Add KEY? and DRAINKEY to deal with SIA1 garbage from emulation check-in: f84f6c3126 user: kc5tja tags: e3-cpu | |
| 16:51 | s/buffer/character/ in SIA code, for greater communication of intent. check-in: 32387908da user: kc5tja tags: e3-cpu | |
| 02:30 | You can now leave minicom to terminate the emulator. check-in: 7a457a320b user: kc5tja tags: e3-cpu | |
| 02:25 | Forgot to remove debugging code check-in: 687e009e23 user: kc5tja tags: e3-cpu | |
| 02:25 | NEW: Implemented SIA read functionality. And, it successfully is running DX-Forth without causing the emulator to crash. Holy crap, I never thought I'd get this far. BUT: there's problems. When DX-Forth starts up, the banner comes right back as input to the interpreter. This is highly undesirable, and I really wish I could figure out a way to prevent this from happening. I don't think it's causing the interpreter distress; but, still, it's damn annoying. There are other problems too: for some reason, numeric output is not working. AGAIN. Also, Forth refuses to recognize input longer than a single character. Neither issue is clear to me why it happens. I don't yet have a plan to debug this yet either. Baffling. check-in: 52ff954d87 user: kc5tja tags: e3-cpu | |
| 01:32 | NEW: Instruction fetch, load, and store misalignment traps. CPU bus interface properly models Wishbone byte lanes. (Turns out this was easier all along.) External bus errors now produce instruction fetch, load, and store access fault traps. SIA now modeled in software. ROM and RAM are now modeled in software. FIX: loads from memory properly assert their transfer sizes. Instruction fetch now acquires its instruction from the correct byte lanes. Memory writes no longer deadlock the CPU. check-in: 212569624f user: kc5tja tags: e3-cpu | |
|
2020-01-03
| ||
| 07:03 | In theory, we have a KCP53000B emulated without support for external interrupts. In theory. Next steps is to get RAM, ROM, and SIA core access support implemented, and with *any* luck at all, we can see the DX-Forth 1.2 prompt come up. (I suspect this will not be easy.) check-in: 502b06662b user: kc5tja tags: e3-cpu | |
| 07:00 | Bedtime soon; committing changes made so far. check-in: 8f62974eab user: kc5tja tags: e3-cpu | |
| 01:17 | CPU state lifecycle check-in: 119f5aaaaa user: kc5tja tags: e3-cpu | |
| 00:47 | Create new branch named "e3-cpu" check-in: 700336c879 user: kc5tja tags: e3-cpu | |
|
2020-01-02
| ||
| 23:59 | Restore trunk back to pre-repkg state. check-in: 8571b9ca73 user: kc5tja tags: trunk | |
| 23:48 | Goddammit, this is getting annoying. Leaf check-in: 4bd5a1a55d user: kc5tja tags: replace-nrmf-with-repkg | |
| 23:45 | Remove vestiges of NRMF. check-in: db7a0e50d6 user: kc5tja tags: trunk | |
| 23:42 | Replace NRMF with another non-recursive make system, purpose built. check-in: 1da36e75d2 user: kc5tja tags: trunk | |
| 23:41 | Create new branch named "replace-nrmf-with-repkg" check-in: 38cb48ead1 user: kc5tja tags: replace-nrmf-with-repkg | |
| 05:07 | Remove config.od; no longer needed. check-in: d7b42491d9 user: kc5tja tags: trunk | |
| 03:28 | e3 builds, processes some basic arguments. Need to work on CPU and SIA emulation next. check-in: 456f3f082f user: kc5tja tags: trunk | |
| 00:50 | Initial import of initial main.c of initial code for E3. check-in: 2d4b6522a0 user: kc5tja tags: trunk | |
|
2020-01-01
| ||
| 03:00 | Divorce DX-Forth from processor card-specific ROM image check-in: efcd68a84e user: kc5tja tags: trunk | |
| 03:00 | Oops, I forgot to link Kestrel ROM access into the address decode logic of the emulator. check-in: 5551f78da0 user: kc5tja tags: trunk | |
| 02:22 | Rename rom.asm to cpurom.asm, and clean up compiler warnings check-in: 2db30aba6b user: kc5tja tags: trunk | |
| 02:05 | Move origin to $40000 to prep for icoBoard Gamma FPGA bitstream check-in: 62f70e562e user: kc5tja tags: trunk | |
| 01:51 | ...by which I mean split apart the CPU-card functionality from the Kestrel ROM image contents. check-in: df7de75d56 user: kc5tja tags: trunk | |
| 01:50 | Emulator E2 now conforms fully to E3 memory map semantics. Just need to update the DX-Forth image to exploit this. check-in: e1402a7070 user: kc5tja tags: trunk | |
| 01:40 | Rename unified ROM to the "CPU ROM", the ROM which is intended to reside on the active processor card. Right now, this ROM behaves exactly as it did in the E2 emulator. However, eventually, this ROM will hold the processor's machine-mode bootstrap code and software shims for missing hardware features. check-in: 2433110035 user: kc5tja tags: trunk | |
|
2019-12-31
| ||
| 22:56 | Upgrade to DX-Forth 1.2, which now supports the Kestrel-3/E3 memory map standards (even though the emulator doesn't quite get the firmware address decoding right yet). Fixes a number of signed vs. unsigned bugs found within DX-Forth 1.1, thanks to RAM addresses now being located in negative address space (on 32-bit machines). KNOWN BUG IN DX-FORTH 1.2: there is no way to enter an unsigned 32-bit number or unsigned 16-bit number in source form. So, HEX C000000 or HEX C000 will result in negative numbers. (This bug also existed in DX-Forth 1.1, but I wasn't aware of it.) This is an artifact of LITERAL attempting to save space in the dictionary by using the most compact numeric representation. Three workarounds exist: (1) create a parsing word which parses the next number and manually compiles a 64-bit number; (2) Wait for DX-Forth 1.3 to roll out some kind of fix for this issue; OR, (3) follow such numbers (perhaps computed from elsewhere) with explicit masks, like -65536 AND, etc. Of course, the third option is to replace DX-Forth 1.x with DX-Forth 2.x and go with native-code compilation, but this will take much longer to get working. check-in: f1c79c7f80 user: kc5tja tags: trunk | |
|
2019-12-30
| ||
| 17:34 | Switch from old Redo-based builds back to using GNU Make (specifically using my NRMF package). check-in: da0ce0827c user: kc5tja tags: trunk | |
| 17:28 | Include DX-Forth ROM image in build steps in new Makefile Closed-Leaf check-in: 5a48d1c21f user: kc5tja tags: switch-to-nrmf | |
|
2019-12-28
| ||
| 02:35 | Progress on putting DXForth ROM image into Makefile ecosystem check-in: 4050eb5042 user: kc5tja tags: switch-to-nrmf | |
| 01:41 | Repair compiler warnings check-in: 8c13288d87 user: kc5tja tags: switch-to-nrmf | |
|
2019-12-27
| ||
| 23:11 | Added e2 emulator to set of build targets. check-in: f7b2890ea2 user: kc5tja tags: switch-to-nrmf | |
| 19:31 | test check-in: 29370814b5 user: kc5tja tags: switch-to-nrmf | |
| 19:31 | why was trunk set as the current branch? check-in: 7f1232f979 user: kc5tja tags: switch-to-nrmf | |
| 19:27 | Integrated the `a` assembler into the NRMF configuration. check-in: 9f4e4c255d user: kc5tja tags: trunk | |
| 17:37 | Import NRMF files from Github repo check-in: 803c2a6dc9 user: kc5tja tags: trunk | |
| 17:37 | Create new branch named "switch-to-nrmf" check-in: 6514ad3dc5 user: kc5tja tags: switch-to-nrmf | |
|
2019-12-26
| ||
| 23:07 | restoring order now? check-in: 11b423ba53 user: kc5tja tags: trunk | |
| 23:06 | Closing private branch Closed-Leaf check-in: 86c35572ce user: kc5tja tags: private | |
| 23:05 | Restoring order, I hope check-in: d193aa34a1 user: kc5tja tags: private | |
| 23:02 | testing more check-in: d941dd928c user: kc5tja tags: private | |
| 23:00 | Remembering previous configuration; testing concluded. check-in: 3d9da246a8 user: kc5tja tags: private | |
| 22:57 | More testing check-in: 7d864db8c9 user: kc5tja tags: private | |
| 22:56 | More testing check-in: 8182ca29ca user: kc5tja tags: trunk | |
| 22:52 | testing, please disregard check-in: 8c2b8498c9 user: kc5tja tags: trunk | |
|
2019-12-21
| ||
| 16:36 | Testing new install of fossil to see if this will preserve user credentials. check-in: aafe97f23b user: kc5tja tags: trunk | |
|
2019-11-12
| ||
| 05:30 | Bringing REPORT up to date. Still not complete, but not entirely happy with the complexity of the firmware design either. I reserve the right to scale things way down in the next version. We shall see where this leads. check-in: 5b58b60f8a user: kc5tja tags: trunk | |
|
2019-10-27
| ||
| 00:52 | Re-read doc; identified one code section and one whole sub-section which needs updating to the latest I/O channel model. check-in: cb2c29bfa3 user: kc5tja tags: trunk | |
|
2019-09-15
| ||
| 23:57 | Refined some more SSP procedures. But, I am finding that complexity is growing beyond my ability to keep it in my head, and am running into limitations when it comes to distinguishing commands intended for the controller or for the units. check-in: c6e384b2d2 user: kc5tja tags: trunk | |
| 06:24 | refining some of the storage protocol requirements before my laptop loses power check-in: a6c0c403cb user: kc5tja tags: trunk | |
|
2019-09-11
| ||
| 03:25 | Simplify forth auto-load block logic check-in: 46379ac38b user: kc5tja tags: trunk | |
|
2019-09-10
| ||
| 05:48 | I think I have completed the refinement of the boot menu logic. I feel confident I can proceed with the boot source discovery. check-in: bfb02a2af0 user: kc5tja tags: trunk | |
| 04:30 | Switch methods of tracking which steps need refinement. This seems to be an easier approach, and I can track progress in an automated manner with clever use of grep. check-in: 01899c8062 user: kc5tja tags: trunk | |
|
2019-09-09
| ||
| 05:06 | More refinements; done for this evening though. check-in: 985c171204 user: kc5tja tags: trunk | |
| 04:35 | More stepwise refinements to flesh out more boot loader behavioral details, menu item structure, etc. check-in: a903a4c39f user: kc5tja tags: trunk | |
| 00:34 | Missing word radically alters intended meaning of a sentence. check-in: 2c2434a794 user: kc5tja tags: trunk | |
|
2019-09-08
| ||
| 23:48 | Bring in the project report org-mode document. Tracks stepwise refinement of system firmware, hardware design choices, etc. check-in: e485309137 user: kc5tja tags: trunk | |
|
2019-09-05
| ||
| 03:41 | .skip and .zero support check-in: 447c21378e user: kc5tja tags: trunk | |
| 03:10 | Support .asciz and fix unit test breakage in scanner check-in: f23325b630 user: kc5tja tags: trunk | |
| 02:51 | .ascii support check-in: 29d317a6a2 user: kc5tja tags: trunk | |
|
2019-09-03
| ||
| 04:41 | .8byte, .4byte, and .2byte implemented check-in: 8a46823df5 user: kc5tja tags: trunk | |
| 04:31 | .half and .byte supported check-in: 4f12102c36 user: kc5tja tags: trunk | |
| 04:23 | Support .word check-in: c0b13ae1bb user: kc5tja tags: trunk | |
| 02:12 | Remove debugging code check-in: 233d4785af user: kc5tja tags: trunk | |
| 02:07 | String expression support check-in: 88de4ec636 user: kc5tja tags: trunk | |
| 01:29 | Basic .dword primitive check-in: 0dda2d09e7 user: kc5tja tags: trunk | |
|
2019-09-02
| ||
| 23:11 | First parser integration test check-in: 9688a01a13 user: kc5tja tags: trunk | |
| 22:27 | Relocate tokens beyond ASCII/UTF-8 boundaries check-in: 82629340d1 user: kc5tja tags: trunk | |
| 22:25 | Unrecognized punctuation are their own symbols. check-in: 09e88c1248 user: kc5tja tags: trunk | |
| 18:24 | Dead code removal check-in: f01e8709f2 user: kc5tja tags: trunk | |
| 18:19 | tests for octal numbers check-in: ab025af620 user: kc5tja tags: trunk | |
| 18:15 | support binary constants as well check-in: 590014a9f8 user: kc5tja tags: trunk | |
| 17:59 | Numeric constants (unsigned only; parser needed for signed constants) check-in: feb73894e7 user: kc5tja tags: trunk | |
| 16:41 | Remove dead code from manifest check-in: 0c54627c0d user: kc5tja tags: trunk | |
| 16:40 | Remove dead code from manifest check-in: 824187d941 user: kc5tja tags: trunk | |
| 16:35 | keyword check (naive implementation) check-in: e2b6e66be3 user: kc5tja tags: trunk | |
| 06:18 | Track line numbers check-in: e28f0d9a4e user: kc5tja tags: trunk | |
| 05:54 | Scanner grabs identifiers check-in: 1b3c87c3b9 user: kc5tja tags: trunk | |
| 04:13 | Removing parser again; switching to Wirth-style recursive descent implementation check-in: d3ab39fb86 user: kc5tja tags: trunk | |
| 04:12 | First scanner tests check-in: fcf8bd4e65 user: kc5tja tags: trunk | |
|
2019-08-26
| ||
| 06:49 | Introduce failing parser test. Tried shunting yard; does not work for my needs. Must revert to recursive descent with Wirth emission. check-in: 6c6a725a22 user: kc5tja tags: trunk | |
|
2019-08-25
| ||
| 23:17 | Forgot to include buffer migration check check-in: 651073af86 user: kc5tja tags: trunk | |
| 23:08 | Make sure vector contents is preserved after expansion check-in: 0449beaa7a user: kc5tja tags: trunk | |
| 22:55 | Dynamic vectors completed check-in: a52343650b user: kc5tja tags: trunk | |
| 19:16 | Start on dynamic vector type. Major refactoring into smaller modules. Inconvenience from lack of shared libraries is showing. check-in: af4b7000f7 user: kc5tja tags: trunk | |
|
2019-08-24
| ||
| 21:28 | forgot runtests check-in: de3957748f user: kc5tja tags: trunk | |
| 21:27 | Remove parser; started on wrong foot. Starting over check-in: de174e1877 user: kc5tja tags: trunk | |
|
2019-08-23
| ||
| 06:54 | Try to refactor test engine into a separate module; however, it does not work nearly as well as I would like. check-in: 0be7136182 user: kc5tja tags: trunk | |
|
2019-08-21
| ||
| 22:27 | prepare to work on parser next check-in: ffc86e1698 user: kc5tja tags: trunk | |
|
2019-08-17
| ||
| 00:07 | Finished (for now at least) the section abstraction. check-in: 5594d490f9 user: kc5tja tags: trunk | |
|
2019-08-16
| ||
| 21:38 | WIP: you can emit bytes..dwords into a section now check-in: 0482ee37fa user: kc5tja tags: trunk | |
| 06:59 | Start work on a next-generation assembler for the Tripos/KOS environment. Uses official RISC-V assembly language directives as much as it makes sense to. Not intended to replace gas; but I need something usable in a BCPL and Tripos environment, and gas won't be useful until it either supports hunk format executables or runs under Tripos. check-in: 431ffa1b72 user: kc5tja tags: trunk | |
|
2019-08-05
| ||
| 07:53 | Bug fix: branch displacement was missing a bit check-in: 797fafc3c9 user: kc5tja tags: trunk | |
| 06:31 | writeback logic first pass implemented. check-in: 0708c27675 user: kc5tja tags: trunk | |
| 03:44 | SRLI and SRAI -- this completes OP-IMM check-in: b705e135a7 user: kc5tja tags: trunk | |
| 03:23 | AND, OR, XOR check-in: 83070c8d95 user: kc5tja tags: trunk | |
| 03:21 | Bug fix for SLTI: -4 < 4 would fail test check-in: 26f3c58373 user: kc5tja tags: trunk | |
| 03:02 | SLTI and SLTUI check-in: fe12c3b40e user: kc5tja tags: trunk | |
| 02:49 | SLLI implemented in the IXU check-in: d265210c64 user: kc5tja tags: trunk | |
| 02:30 | ANDI instruction works. check-in: b0c8e8b48a user: kc5tja tags: trunk | |
| 01:20 | Initial integration of ALU into IXU; no tests yet check-in: 3d81d8061e user: kc5tja tags: trunk | |
| 00:11 | Simplify the ALU code significantly check-in: 84a26d5484 user: kc5tja tags: trunk | |
|
2019-08-04
| ||
| 21:41 | Introduce (untested) ALU module, clone of KCP53000 ALU check-in: 9357ad3b2f user: kc5tja tags: trunk | |
| 19:38 | capture_operands implemented check-in: 313bd82cd1 user: kc5tja tags: trunk | |
| 07:49 | Preparing to switch everything over to Wishbone B4 Pipelined. check-in: d9bfda7c63 user: kc5tja tags: trunk | |
| 07:48 | New IXU design, going back to my original ideas of how to build the thing. check-in: 2556401d8a user: kc5tja tags: trunk | |
|
2019-07-16
| ||
| 05:45 | clarify comment check-in: 8568cc35ae user: kc5tja tags: trunk | |
| 05:18 | WIP: Implement misaligned address detection, and prevent misaligned writes from generating Tilelink transactions. It's up to the FU to terminate the instruction by committing a trap to the trunk bus in a subsequent cycle. check-in: 561d833304 user: kc5tja tags: trunk | |
|
2019-07-15
| ||
| 04:47 | WIP: Implement A-channel of TileLink interface. DOES NOT YET DEAL WITH UNALIGNED ADDRESSES; instead, just ignores low-order address bits depending on size of datum being written. check-in: a783307bfd user: kc5tja tags: trunk | |
|
2019-06-06
| ||
| 07:41 | update psp check-in: daf647810c user: kc5tja tags: trunk | |
| 07:39 | Complete trunk register file source indices logic. check-in: 7170d9c66c user: kc5tja tags: trunk | |
| 07:19 | Add SUM_REQ, but I renamed it to CU_READ_VALUES. When asserted, it tells the CU when to capture the values of the register file on the trunk's register data ports. check-in: c387b96556 user: kc5tja tags: trunk | |
| 03:13 | When GO_READ is asserted by the picker, drop out of read state. check-in: 2e36a268b3 user: kc5tja tags: trunk | |
|
2019-06-04
| ||
| 15:40 | WIP: FU logic recognizes when function units broadcast results on the trunk. Keeps track of operand available status. Drives READ_REQ signal when the FU feels all operands are now ready. check-in: 87d90f5c93 user: kc5tja tags: trunk | |
|
2019-06-02
| ||
| 22:38 | initial import for FU logic check-in: 2d67e26722 user: kc5tja tags: trunk | |
| 22:34 | Relabel some of the CU signals for easier recall of who drives them check-in: e3c8c82fe7 user: kc5tja tags: trunk | |
| 06:12 | psp update check-in: 3425f3c9f4 user: kc5tja tags: trunk | |
| 05:28 | WIP! This completes the add/subtract logic for the add/store computation unit. The precise needs of the "store" logic will be dictated by how the state machine in the function unit is implemented, so I'll cross that bridge when I get there. check-in: 0c5c321cf6 user: kc5tja tags: trunk | |
| 03:16 | WIP: First CDC 6600-inspired add/store computation unit. Experimental check-in: 46868584f7 user: kc5tja tags: trunk | |
|
2019-05-28
| ||
| 01:00 | Pass I port D channel error signal through to IXU for synchronous trap processing check-in: d7b385618d user: kc5tja tags: trunk | |
| 00:44 | Remember to double-check the lowest 2 address bits check-in: d23d900a8b user: kc5tja tags: trunk | |
| 00:40 | Relabel interfaces properly; expose iq_flush check-in: f37147ffb1 user: kc5tja tags: trunk | |
| 00:27 | Jump request and ack interface check-in: 16b97e0e8b user: kc5tja tags: trunk | |
|
2019-05-27
| ||
| 22:53 | Implement o_iq_valid handshake signal. Note that it is pulsed!! (See comments in code for more details.) check-in: 2dbe6f9e6d user: kc5tja tags: trunk | |
|
2019-05-26
| ||
| 01:37 | iq_address and iq_inst_addr implemented. Confirmed proper word routing from the fetched dword. I need to double-check proper handshaking next (e.g., iq_ready and iq_valid), and after that is done, work on the control flow functionality. check-in: 23a3a9c2af user: kc5tja tags: trunk | |
|
2019-05-06
| ||
| 01:41 | Finished IFU A-channel for instruction fetch port (TileLink). check-in: e91ccded29 user: kc5tja tags: trunk | |
|
2019-05-05
| ||
| 07:51 | Makefile update to remove old ilang files check-in: e17b606295 user: kc5tja tags: trunk | |
| 07:50 | Support the byte lane select mask check-in: 33a2716e57 user: kc5tja tags: trunk | |
| 07:34 | update psp check-in: 5c7b25760b user: kc5tja tags: trunk | |
| 07:20 | B/c the fetch_counter was treated as an anonymous input port to the module, allowing sby to tweak its value. Very unintuitive!! check-in: af1995818c user: kc5tja tags: trunk | |
| 07:17 | WIP: Why u no work? Fails induction for unknown reasons check-in: 3a96327a8f user: kc5tja tags: trunk | |
| 03:57 | update psp check-in: 866dcdadac user: kc5tja tags: trunk | |
| 03:56 | Initial IFU code check-in: 3018838aa9 user: kc5tja tags: trunk | |
|
2019-05-04
| ||
| 15:54 | PSP update and spec update check-in: 9f51a63046 user: kc5tja tags: trunk | |
|
2019-05-02
| ||
| 22:34 | Consolidate tests into per-module tests check-in: 51b29796b7 user: kc5tja tags: trunk | |
| 22:03 | Use nmigen version of tools.py (more maintainable) check-in: 13a75afbcd user: kc5tja tags: trunk | |
| 21:53 | Migrate tests for UpCounterCSR to nmigen check-in: c8c8464c5e user: kc5tja tags: trunk | |
| 18:30 | Migrate MemCSR tests to nmigen check-in: c09510ecfd user: kc5tja tags: trunk | |
| 17:52 | Augment clean target to account for nmigen's spec_xxx directories check-in: d38e951c0e user: kc5tja tags: trunk | |
| 17:50 | Migrate InputCSR tests to nmigen check-in: 09291e359f user: kc5tja tags: trunk | |
| 17:30 | Migrate tests for ConstantCSR to nmigen check-in: c216582c93 user: kc5tja tags: trunk | |
| 05:23 | Remove dead code check-in: 138c5505a1 user: kc5tja tags: trunk | |
| 05:22 | Convert formal testbench to nmigen for CSRSelect check-in: 0e9f6538b4 user: kc5tja tags: trunk | |
| 04:55 | Remove Verilog-based formal test bench for MStatus check-in: a5b58db554 user: kc5tja tags: trunk | |
| 04:44 | Remove from nmigen import * in favor of explicit imports check-in: 00a555b8ef user: kc5tja tags: trunk | |
| 03:21 | Thank you Luke Kenneth Casson Leighton for this contribution. This shows me both where I went wrong in my previous attempts to apply nmigen towards formal verification, but also a better interface to make formal tests run automatically than through a stack of mostly-but-not-exactly-identical Makefile rules. I will start evolving the codebase to use this new approach. check-in: c2e3819404 user: kc5tja tags: trunk | |
|
2019-04-29
| ||
| 06:52 | Updating spec and PSP check-in: d1cdbff14a user: kc5tja tags: trunk | |
|
2019-04-09
| ||
| 05:23 | Add (M)CYCLE, (M)INSTRET check-in: 6c559135a3 user: kc5tja tags: trunk | |
| 01:16 | Implement MIP register check-in: 615dbc6e78 user: kc5tja tags: trunk | |
|
2019-04-08
| ||
| 07:19 | Complete CSRU. No I/Os to outside world yet, though, but those can be added later. From CPU's POV, all CSRs are implemented and properly handle WLRL fields. check-in: e67e2d5951 user: kc5tja tags: trunk | |
| 00:30 | Remove commented old code. check-in: fcac6fca38 user: kc5tja tags: trunk | |
| 00:29 | Convert MSTATUS and MIE to use MemCSR check-in: 6074d7325c user: kc5tja tags: trunk | |
|
2019-04-07
| ||
| 22:57 | Generalize memory-only type CSRs. Some CSRs which have non-memory semantics can, in fact, use MemCSRs as a submodule, and just wrap CSR-specific logic around this. check-in: c0dba178f7 user: kc5tja tags: trunk | |
|
2019-03-25
| ||
| 08:15 | Begin proving mstatus check-in: 1629179d84 user: kc5tja tags: trunk | |
| 06:09 | Formally verify InputCSR check-in: f311559092 user: kc5tja tags: trunk | |
| 05:17 | Formally verify ConstantCSR check-in: 6facb8fbe2 user: kc5tja tags: trunk | |
| 04:32 | tighten CSRSelect properties check-in: d460e67f4a user: kc5tja tags: trunk | |
| 04:24 | Formally prove CSRSelect check-in: f96c5158b4 user: kc5tja tags: trunk | |
|
2019-03-24
| ||
| 19:56 | Data output logic. check-in: aa1efe0a7d user: kc5tja tags: trunk | |
|
2019-03-23
| ||
| 18:45 | Implement MSTATUS. check-in: f226214414 user: kc5tja tags: trunk | |
| 17:37 | Bringing repo up to date, as I'm giving a demo at SVFIG. check-in: e96a1d61c9 user: kc5tja tags: trunk | |
|
2019-03-18
| ||
| 08:25 | Begin implementation of the CSR Unit. Holy heck, using nMigen for this was totally the right choice. The CSRU module is **tiny** compared to its equivalent Verilog. check-in: 4c2460e8d1 user: kc5tja tags: trunk | |
|
2019-03-16
| ||
| 02:45 | Spec and project data updates. check-in: 2dc9dd15f4 user: kc5tja tags: trunk | |
|
2019-03-13
| ||
| 07:11 | Beginning specification work for the KCP53000B processor design. check-in: 04d68e52a6 user: kc5tja tags: trunk | |
|
2018-11-26
| ||
| 18:16 | SIA IS PROVEN! The core worked fine. The problem was a bug in my ROMA core instantiation. Celebration time!! check-in: 3989d347ab user: kc5tja tags: trunk | |
| 07:36 | Prove in hardware that the SIA core works. IT DOES. However, there is a bug when sending bytes with relatively few bits set on the low end of the byte (e.g., 0x05 becomes 0x0A on the receiver, while 0x51 remains 0x51 on the receiver). I genuinely have no idea how or why this is happening at this time. check-in: f70ab01418 user: kc5tja tags: trunk | |
| 07:35 | Parameterize the SIA core. check-in: 315211fe32 user: kc5tja tags: trunk | |
| 05:53 | Fucking reset circuitry gets me every time! check-in: 8d2bf519df user: kc5tja tags: trunk | |
| 01:34 | First cut at a hardware integration test for the SIA check-in: b06cdcafc5 user: kc5tja tags: trunk | |
|
2018-11-25
| ||
| 02:29 | Forgot to remove debugging change to default loopback settings check-in: 5511953cec user: kc5tja tags: trunk | |
| 02:03 | Integration test: SIA local loopback test passes! check-in: 2867a12bc1 user: kc5tja tags: trunk | |
| 02:03 | Bug fix: Draw the loopback control bits from the proper data lines. check-in: 8fe7daff6e user: kc5tja tags: trunk | |
| 02:01 | Bug fix: make sure loopback control bits are accounted for in formal assertions. check-in: f047e61a8d user: kc5tja tags: trunk | |
|
2018-11-01
| ||
| 17:58 | WIP: Renovate sia_tilelink.v and properties.vf to work better with Verilator. HOWEVER, Verilator now cannot locate "verilated.h" for some reason. Even if I manually include its path as a -I parameter to gcc, it refuses to find it. NOTE: I'll need to apply these kinds of renovations to all the other Verilog sources in the SIA core, and quite possibly throughout the Kestrel-3 repository as well. UUGH! I hate Verilog so much after this. The Verilog spec is total garbage if nobody is going to bother to follow it. Many thanks to ZipCPU (Dan G.) for isolating what could be one of the biggest bugs of the Kestrel-3, and a major source of what's driving me crazy with this project. check-in: 1dd9f6b5cf user: kc5tja tags: trunk | |