summary refs log tree commit diff
path: root/quine.asm
AgeCommit message (Collapse)Author
34 hoursall the non-64-bit addressing modes for movIrene Knapp
or at least, all the ones we need yay it's theoretically possible these work! it was a lot of details to get right, and they're not tested at all Force-Push: config Change-Id: I4f1bef007982af7901aa263af8e618d19e262f12
36 hoursdid all the pending instructions that use indexed addressing modesIrene Knapp
we're starting to feel vaguely proficient in stack juggling, but it will still be quite a thing if these don't need fixing later Force-Push: yes Change-Id: Ic1285d01cc42c6e8a30d36fc53e67135d8d9e175
40 hoursrename all the modrm_* words to addressing_*Irene Knapp
also re-order "sib" to be before them we were about to add one for indexed modes, and realized that really, these words are responsible for both the ModRM and SIB bytes, it's just that some modes only need one of them. this makes their responsibilities clearer, and as a bonus it draws a stronger distinction between these high-level words and the low-level "modrm", which omits consistency checking. having it called something else reduces the likelihood of inappropriately using the low-level one when a high-level one would be more suitable. their names are a bit long now, but that's probably also good, it will incentivize building higher-level abstractions that define multiple variants of instructions rather than writing each one by hand. someday. Force-Push: yes Change-Id: Ia8f8fcc0e092f94e66f17a3fa55e8bf3e927fdff
2 daysimplement a bunch of assembly-in-ForthIrene Knapp
Force-Push: yes Change-Id: Ia9844da8f6e5d59ed77348b85200c28a7113bd1f
2 dayslisted out all the assembly instructions that need Forth versionsIrene Knapp
Force-Push: yes Change-Id: Ib9efd1cd1c8b362547bef8b15829f9da6a28b4d1
2 daysyay now it has "exit" on the heap, tooIrene Knapp
Force-Push: yes Change-Id: I19215825a8891bbc8b36bac03a0f0ca440721cb3
2 daysthere is now a full implementation of docol on the heapIrene Knapp
yay that was hard. lots of bug-fixing. for some reason the condition code for not-equal was being output as 0x6 instead of 0x4. there was also a lot of stack manipulation to trace through very carefully. Force-Push: yes Change-Id: I8f2c19d5be96ba27a3e8c56c460c761a3b16666f
4 dayssome notes to figure out the topological sort and so onIrene Knapp
also a minor renaming for consistency Force-Push: yes Change-Id: I80f7a091272fa9a73cedd7caba495a45d5866d16
4 daysthe variable words are fully defined on the heap nowIrene Knapp
not used for anything yet, but hey verifying this required a lot of time in the debugger Force-Push: yes Change-Id: I38a9536ce8efde236b1ed81ca94e0cf7e101413d
4 daysmove the output helpers and assembly-in-forth stuff earlier in the fileIrene Knapp
this better reflects the logical dependencies, per the new plan Force-Push: yes Change-Id: I2caa0e6bb1c722a7839f6695c48b05d0b2cfad25
4 daysadd the beginnings of some word-defining wordsIrene Knapp
Force-Push: yes Change-Id: I7e4c678c27a87b79a9f2c2aad12bd65d7d5600cc
5 daysdowncase _all_ the thingsIrene Knapp
this will get more painful the longer we put it off Force-Push: yes Change-Id: I82430303d9d9d8a317eb8f020316e2145020a5c8
5 daysthose ;codeword comments were unnecessarily verboseIrene Knapp
and the 0x in front of the 8 was silly Force-Push: yes Change-Id: I23a8ffd6aeec1d78ee7362c1f97d99e94947e5b6
5 daysokay some words-as-variables are defined nowIrene Knapp
and the heap bootstrapping strategy is explained, or part of it Force-Push: yes Change-Id: I4c48cf081709a7200a70041e321ace4d6713be94
6 daysoutput a whole bunch of assembly from ForthIrene Knapp
and begin to grow the Forth-styled assembly language it needs runtime word definition next, in our view, so we're checkpointing it here so as to work on that Force-Push: yes Change-Id: I08927581f6b6fdc69f6543936e46fe9ff83c8d48
9 daysclean up the first two instruction macros ever writtenIrene Knapp
they had some minor confusing stuff going on; now they don't Force-Push: yes Change-Id: Ie3756cf18222cb8b3f4794f57d91cb5229be715e
9 daysget rid of the redundant copy of the label valueIrene Knapp
Force-Push: yes Change-Id: I5709b96b401362564cc0f1dec3953dad13c319b7
9 daysget it down to a single call to write()Irene Knapp
this also adds a block-copy routine, further breaking up the handful of remaining monolithic things Force-Push: yes Change-Id: Ida960d83bab6176d016168cdf25e0763aa6050ef
10 daysthe two-pass magick works!!!!!Irene Knapp
yessssssss that was a lot of debugging, wow. there were two bugs in UNROLL: the source and destination for the block move were set to the high end instead of the low end; and the encoding of an lea variant fell into an unsupported case but was emitted anyway due to an incorrect guard clause that was seriously a lot of debugging. also all the stack logic for grabbing and stashing the label stuff was wrong several times, but it works now the original "old code" is now fully eliminated, wow! Force-Push: yeah Change-Id: I85d951eabc7de6ba502e2a1aa0f5998f9b399765
12 daysROLL, UNROLL, and a whole bunch of addressing modesIrene Knapp
also a couple more gdb tips Force-Push: yes Change-Id: I31038334449ea45238c811b4e97e2d87833d8ea6
13 daysadded a macro BEFORENEXT and a bunch of words like LITPACK64Irene Knapp
this shaves like 700 bytes off the file size, though it feels like more should be achievable Force-Push: yes Change-Id: Iaa51c462cc80cea1a16a67b428b5064e40d9b9f1
13 daysmove almost all the ELF stuff to code writen in ForthIrene Knapp
wow, this sure does balloon the file size. must be something we can do about that... Force-Push: yes Change-Id: I477c4fdc9b42c3e5d50654d2864e3933da65a64d
13 daysheap allocation now works! also fix some encoding issuesIrene Knapp
implement indirect adds and subtracts; it turns out they're convenient for heap management the operand order for lea.qreg.qreg.disp32 didn't match what the name suggested, so it's fixed now the semantics of push.dimm are quite surprising, for reasons that are now extensively documented also add notes on helpful ways to use gdb Force-Push: yes Change-Id: I717eb6427520e2c336586d4e8066ef172d97ce55
14 daysbreak up OLD_CODE into separate callsIrene Knapp
the goal is to make the assembly monolith, less of one. no individual piece of this is complicated, it was just tightly coupled... now it is less so. Force-Push: yes Change-Id: I4bf5711ee6eae18e68be9d4e1f5107933c45fbbd
14 daysmemory manipulationIrene Knapp
Force-Push: yes Change-Id: I5587783da20301abf11489caad6990531b13e3f1
2025-10-24add stack manip, arithmetic, bitwise stuff, comparisonsIrene Knapp
this is all untested, but I'm highly confident in my reading of the Intel manual at this point Force-Push: yes Change-Id: I3dbd59da02994db58f2a53d1890136d583803bc6
2025-10-23a convenience macro for word headersIrene Knapp
Force-Push: yes Change-Id: Ifda1a168808d541f8899a06d39bb9d13733dbadc
2025-10-23pick a header format, document itIrene Knapp
also all the existing stuff has word headers now. it doesn't need them to run, it just has them anyway. and EXIT was moved after DOCOL because it feels poetic that DOCOL is the first Forth word in the file Force-Push: yes Change-Id: I592445310712bfde42fde8cffd7bc672f16c4e6c
2025-10-20implement Forth EXITIrene Knapp
this is the final word called by interpreted words, to return Force-Push: yes Change-Id: I8301fa31009915e4beb069664c9bd16cdf33d879
2025-10-20woooooo successful debuggingIrene Knapp
it turns out things are real enough now that we can't just scribble wherever :) Change-Id: I26c71beaaefecae018ebb3ced4f1f579879a030a Force-Push: yes
2025-10-19it does threaded execution now!!!!!!!!Irene Knapp
this is the core Forth thing, or a large part of it Force-Push: yes Change-Id: Id15cf5289cb7b41e0ea824ef68791f663c2f90e6
2025-10-19much closer to being able to move to threaded executionIrene Knapp
not quite, but very very near Force-Push: yes Change-Id: I79ae9f2970e4d7263db5c511e3a5398c22c4771c
2025-10-19there was an important bug in mov.qreg.disp8.bimmIrene Knapp
it was using rax as an index register, when it should have been using no index register. this was working by coincidence, because rax was zero, but now it's not. yay fixed! there's also lots of decisions made and documented about the execution model, but they're only part-implemented and I wouldn't be checking them in right now if that were all that's new Force-Push: yes Change-Id: Ie8f64914484cd405272d640feeb1cf586ad915d7
2025-10-19rename a bunch of instructionsIrene Knapp
this is just cleanup. once they're all named in a uniform way, it will make sense to write higher-level wrappers around them... Force-Push: yes Change-Id: Id110111d7cd0bf936de8f101cf6886513e1cca84
2025-10-19add extensive documentation on the execution modelIrene Knapp
implement some of the earliest Forth loading stuff, following Jonesforth closely the Forth memory space isn't totally set up yet - we are going to use the data segment after all, though it won't be heap-ish really. everything up through _start is ready for Forthy-y use though, and it does `cld` besides. Force-Push: yeah Change-Id: Ibd12223f304aff71f4e78744f7a9da09b9072a45
2025-10-18Add documentation! Yay :)Irene Knapp
This also gets rid of the `load_origin` variable, it was redundant with flatassembler's `$$` feature. Force-Push: yes Change-Id: I0abc4cf8da65fcb443220e70b69267919ca43ba7
2025-10-11generalized mov.qreg.qregIrene Knapp
Force-Push: yes Change-Id: I668a2738ac19e98dd30c58964ba4c6706613ea7d
2025-10-11comment the code in preparation for restructuring itIrene Knapp
Force-Push: yes Change-Id: I561d5fbb35db97b9d236a159c6c888b0651248c3
2025-10-11clean up mnemonics a bitIrene Knapp
Force-Push: yes Change-Id: I71ad8876fa9c0bdae9c12f1661b084918a1cb61d
2025-10-11use qwordreg in another placeIrene Knapp
Force-Push: yes Change-Id: I0d328c24639b993f8140b18b7b3d4efa450c33c7
2025-10-10sorta outputs itself, but it reads too much of itself to do that, but very ↵Irene Knapp
good progress Force-Push: yeah Change-Id: I14fdd705643af7ae23ea75c778e9671d3787e0e6
2025-10-05hello works! yayIrene Knapp
Force-Push: yes Change-Id: I6c70991d16bd27da2b75ee0666203f043714a685
2025-10-05yay the earliest examples are in a solid stateIrene Knapp
Force-Push: yes Change-Id: I7111b9a9ed3cb5c5cf162af336b2278cf5850b76
generated by cgit-pink 1.4.1 (git 2.36.1) at 2025-11-08 14:26:47 +0000