1
0
mirror of https://github.com/TomHarte/CLK.git synced 2024-11-30 04:50:08 +00:00
Commit Graph

342 Commits

Author SHA1 Message Date
Thomas Harte
fe6e2eb0a1 Generalise CBW. 2023-10-09 15:00:04 -04:00
Thomas Harte
08aed3bac5 Implement CWD. 2023-10-09 14:54:14 -04:00
Thomas Harte
6bbd896c34 Add DAS with a manageable number of failures. 2023-10-09 14:47:39 -04:00
Thomas Harte
0bf2099a70 Improve DAA. 2023-10-09 14:42:32 -04:00
Thomas Harte
1b9e6e8c8e Add DAA, which doesn't yet pass all tests. 2023-10-09 14:27:02 -04:00
Thomas Harte
59521f9d38 Implement CBW, CLC, CLD, CLI, CMC. 2023-10-09 11:59:38 -04:00
Thomas Harte
769aed10ea Reduce repetition. 2023-10-09 11:49:38 -04:00
Thomas Harte
5a77f0c93c Implement CALL. 2023-10-09 11:46:59 -04:00
Thomas Harte
4f14210ee0 Remove ideas discarded. 2023-10-08 22:27:01 -04:00
Thomas Harte
f618ca6046 Implement, test AND. 2023-10-08 22:18:40 -04:00
Thomas Harte
e3b18708c7 Handle segment-boundary word accesses.
With all ADDs and ADCs enabled, no remaining failures.
2023-10-08 22:11:05 -04:00
Thomas Harte
bd0b62232f Consider that displacements may always be signed.
Down to 1 failure.
2023-10-08 21:41:36 -04:00
Thomas Harte
dbfaef632a Fix DataPointer reference.
Down from 4521 to 1248 failures within 00.json.gz
2023-10-08 15:59:30 -04:00
Thomas Harte
0d2af80f7f Avoid access issues if there's no index. 2023-10-08 13:50:36 -04:00
Thomas Harte
6f768d9a3d Start climbing towards address resolution. 2023-10-08 13:47:43 -04:00
Thomas Harte
dd3cc1f510 Fix ADD and ADC sign flags. 2023-10-08 13:39:46 -04:00
Thomas Harte
a4b1d2b00a Float out data resolution. 2023-10-08 13:34:28 -04:00
Thomas Harte
5c62606154 Simplify parity logic. 2023-10-07 13:38:36 -04:00
Thomas Harte
16bf7c6f26 Fix include guard. 2023-10-07 13:31:35 -04:00
Thomas Harte
cf4603cb33 Attempt to check defined flags only. 2023-10-06 16:32:35 -04:00
Thomas Harte
b6d000ac5e Add enough wiring to consolidate failure on lazy handling of flags. 2023-10-06 13:22:35 -04:00
Thomas Harte
82f0cd790f Find first failing execution, note reason. 2023-10-06 11:43:18 -04:00
Thomas Harte
2d17d9d316 Execute some tests at some facile level. 2023-10-06 11:31:45 -04:00
Thomas Harte
a0ca0bb3c0 Mark non-templates as inline. 2023-10-06 11:11:29 -04:00
Thomas Harte
c6b311b84a Explain source of comments. 2023-10-06 11:10:54 -04:00
Thomas Harte
28c7d27cac Establish some proportion of state, ready to execute _something_. 2023-10-06 11:07:33 -04:00
Thomas Harte
90a8999b4b Fix typo. 2023-10-05 22:29:15 -04:00
Thomas Harte
6d392852d2 Hack on through to something that builds. 2023-10-05 22:27:52 -04:00
Thomas Harte
f411a961a3 Create a central location for avoiding segment conditionality. 2023-10-05 17:12:38 -04:00
Thomas Harte
ada411c0d8 It's differing mildly from DataPointResolver, but segue towards a world of real data. 2023-10-05 17:06:00 -04:00
Thomas Harte
eb100e3b29 Start reforming; data size plus register aren't independent in finding a source. 2023-10-05 16:49:02 -04:00
Thomas Harte
15acb1fc7c Add ADC and ADD. 2023-10-05 15:49:07 -04:00
Thomas Harte
09b2cfad8a Add AAM and AAS. 2023-10-05 14:52:24 -04:00
Thomas Harte
059f300500 Start fleshing out x86 performance. 2023-10-05 14:37:58 -04:00
Thomas Harte
524e4ae65c Tidy up just slightly more. 2023-10-05 11:26:52 -04:00
Thomas Harte
488fceb42b Clean up, add a TODO. 2023-10-05 11:23:58 -04:00
Thomas Harte
01851874ea I guess this is what a perform looks like. 2023-10-05 11:23:41 -04:00
Thomas Harte
7f6e3cf8b7 Define the available flags. 2023-10-05 10:51:55 -04:00
Thomas Harte
2d20175472 Explain absence. 2023-10-05 09:27:02 -04:00
Thomas Harte
6597283c34 Simplify roll/shift case. 2023-10-05 09:26:12 -04:00
Thomas Harte
f6fd49d950 Relocate all text wrangling; this isn't really test-specific. 2023-10-04 22:35:52 -04:00
Thomas Harte
40af162214 Be overt about what's here to aid with printing only. 2023-10-04 22:15:13 -04:00
Thomas Harte
92c46faf84 Add SETMO and SETMOC. 2023-09-29 22:28:23 -04:00
Thomas Harte
ff9237be9f Decode SALC. 2023-09-29 22:06:42 -04:00
Thomas Harte
6cbb434482 Deal with all dangling aliases.
Leaves just five undocumented instructions.
2023-09-29 15:36:34 -04:00
Thomas Harte
103f42f0b0 Introduce FF.7 alias. 2023-09-29 15:26:25 -04:00
Thomas Harte
f2732962d0 Add 6x 8086 aliases. 2023-09-29 15:22:05 -04:00
Thomas Harte
ef5ee8cf94 Include missing context on JMP/CALL far.
Zero failing tests amongst official opcodes.
2023-09-29 14:57:08 -04:00
Thomas Harte
1a6c8a2aed Add outputters for IN and OUT.
2 failures remaining.
2023-09-29 09:39:51 -04:00
Thomas Harte
b76899f2bc Undo broken extension-word DS assumption.
8 failures.
2023-09-28 22:17:14 -04:00
Thomas Harte
245919e67d Resolve REPNE and whitespace issues. 2023-09-28 22:01:12 -04:00
Thomas Harte
ae4a588de3 Adjust semantics to avoid culling end of relevant RETs. 2023-09-28 15:24:15 -04:00
Thomas Harte
960cca163e Make better guess at CALL/JMP size; apply same sizing-logic as offset for disassembly matching.
13 failures.
2023-09-28 14:52:42 -04:00
Thomas Harte
249da884a7 Trim trailing space. 2023-09-28 13:59:41 -04:00
Thomas Harte
035a1265f6 Map invalid reg numbers properly for the 8086.
17 failures.
2023-09-28 13:11:15 -04:00
Thomas Harte
ff4d79e77e Add test synonym, fix operand size.
19 failures.
2023-09-28 09:43:26 -04:00
Thomas Harte
78cb39ad67 Also fix AddrReg.
24 failures.
2023-09-27 22:47:14 -04:00
Thomas Harte
9207de4164 Fix RegAddr macro.
26 failures.
2023-09-27 22:44:10 -04:00
Thomas Harte
c20e7ed9b6 Fix TEST.
28 failures.
2023-09-27 22:30:40 -04:00
Thomas Harte
2d882d2153 Switch shift/roll semantics to reduce extension words and for sanity generally.
37 failures.
2023-09-27 16:40:46 -04:00
Thomas Harte
b59eae3676 Adopt normative ESC decoding.
55 failures.
2023-09-27 10:32:22 -04:00
Thomas Harte
5368f789f6 Shuffle list slightly. 2023-09-26 17:30:27 -04:00
Thomas Harte
b03b408984 Give the decoder responsibility for sanity-checking repetitions.
This may avoid some spurious extension words.
2023-09-26 17:29:20 -04:00
Thomas Harte
cd072e1b57 LEA implies a word. Otherwise add TODOs.
So that's now 69 failures.
2023-09-26 15:41:51 -04:00
Thomas Harte
f16ac603f2 Deal with printing segment:offset.
70 failing files remaining.
2023-09-26 15:28:51 -04:00
Thomas Harte
0a0051eb59 I've just been inconsistent with POP. Stop being so.
71 failures from 288 tests.
2023-09-26 15:16:41 -04:00
Thomas Harte
92c8e1ca93 Add missing #include. 2023-09-26 14:52:08 -04:00
Thomas Harte
87097c44b9 Curate list of known failures; apply easiest fixes.
Now at 157 failures of 288 applicable tests.
2023-09-25 11:39:12 -04:00
Thomas Harte
7fadf01e4e BP in isolation acts as a base. 2023-09-24 18:06:53 -04:00
Thomas Harte
d2b9c435e5 Allow for non-sign-extended offsets/displacements. 2023-09-24 15:00:16 -04:00
Thomas Harte
5fd98e9833 Add an ignore list.
Leaves 180 failures amongst the valid 306 instructions.
2023-09-22 22:56:33 -04:00
Thomas Harte
787e9e770e Retain baseless addresses correctly. 2023-09-22 17:27:27 -04:00
Thomas Harte
c8c0c3ca6d Default segment is ::DS if there was no base. 2023-09-22 17:03:40 -04:00
Thomas Harte
5a5f71e703 JMPs imply their size. 2023-09-22 17:00:10 -04:00
Thomas Harte
587ec81900 Improve string output, better to find actual errors.
Still at 194/324 failures, but a lot of them seem reasonable.
2023-09-22 11:24:33 -04:00
Thomas Harte
9f63db991c Capture default segments, fix base/index confusion. 2023-09-22 11:07:09 -04:00
Thomas Harte
7ebecd2f41 Add notes to self, finally figuring out segment issue. 2023-09-19 23:27:42 -04:00
Thomas Harte
6f5fcf23dc Add missing substitutions. 2023-09-19 14:00:27 -04:00
Thomas Harte
02fcaf0dbd JCXZ seems to be preferred over JPCX. 2023-09-19 13:56:48 -04:00
Thomas Harte
a7cf7d3183 Resolve LOOPNE, LOOPE, etc. 2023-09-19 13:55:19 -04:00
Thomas Harte
9b3199d97b Reduce failures to 205/324. 2023-09-19 13:45:19 -04:00
Thomas Harte
e5dfc882cb Agree that JZ/JNZ are clearer (for me) of the synonyms. 2023-09-19 13:38:08 -04:00
Thomas Harte
3582d2c9e3 Start to beef up operand count list. 2023-09-18 16:34:52 -04:00
Thomas Harte
da953fdf0d Complete 8086 operation list; standardise enum order. 2023-09-18 16:25:04 -04:00
Thomas Harte
710017ada2 Largely resolve the operation-name problem. 2023-09-18 15:57:26 -04:00
Thomas Harte
f8dc5b8ebc Attempt to get close on index + base addresses. 2023-09-17 17:05:19 -04:00
Thomas Harte
f039d44ee3 Fully handle rm = 6, mod = 0. 2023-09-15 22:08:20 -04:00
Thomas Harte
7ee5adc481 Forcing a displacement upon BP reduces to 29 failures.
(At the current limited fidelity of testing)
2023-09-15 16:09:04 -04:00
Thomas Harte
3e09afbb59
Remove errant square bracket. 2023-06-21 11:57:09 -04:00
Thomas Harte
8578dfbf22 Eliminate various other errant spaces. 2023-05-16 16:40:09 -04:00
Thomas Harte
28c79b2885 Eliminate redundant [space][tab] pairs. 2023-05-12 14:14:45 -04:00
Thomas Harte
2b56b7be0d Simplify namespace syntax. 2023-05-10 16:02:18 -05:00
Thomas Harte
5e355383df Correct SIB test. 2022-04-27 19:53:15 -04:00
Thomas Harte
1725894fe9 Eliminate redundant CMPSD, CDQ, CWDE.
Also removes IBTS for now, as I'm unclear where it should sit in the opcode map.
2022-03-12 12:24:44 -05:00
Thomas Harte
fd4f85eb19 Add SMSW. 2022-03-12 12:23:48 -05:00
Thomas Harte
f1c4864016 Eliminate INSD. 2022-03-12 11:37:21 -05:00
Thomas Harte
e6bd265729 Explain which BOUNDs operand is which. 2022-03-11 20:34:28 -05:00
Thomas Harte
c22e8112e7 Expand exposition. 2022-03-11 20:30:56 -05:00
Thomas Harte
44252984c2 Eliminate INT3 special case. 2022-03-11 14:03:46 -05:00
Thomas Harte
4b4f92780e Shuffle extension word order.
The primary objective here is simplifying index calculation, but as per the note it does also potentially open up options with regard to packing in the future.
2022-03-11 13:24:45 -05:00
Thomas Harte
f694620087 Resolve TODO. 2022-03-11 13:10:44 -05:00
Thomas Harte
9b4048ec6e The address size modifier doesn't seem to affect far address sizes.
It's meant to affect only instructions with operands that reside in memory, I think. So probably only ::DirectAddress in my nomenclature. More research to do.
2022-03-11 12:46:07 -05:00
Thomas Harte
c744a97e3c Ensure no extensions for default constructed Instruction. 2022-03-11 11:55:26 -05:00
Thomas Harte
91d75d7704 Switch strategy on 8086 instruction lengths. 2022-03-11 09:48:26 -05:00
Thomas Harte
dc8cff364f Switch to common test. 2022-03-11 09:48:02 -05:00
Thomas Harte
572dc40e6b Allow assignments. 2022-03-11 09:47:23 -05:00
Thomas Harte
f92ffddb82 Add instruction length limits. 2022-03-10 20:47:56 -05:00
Thomas Harte
641e0c1afc Resolve default segment question. 2022-03-10 20:27:35 -05:00
Thomas Harte
bf7faa80c1 Add TODO. 2022-03-10 16:47:54 -05:00
Thomas Harte
a2ae3771eb Add test for switch to Source::IndirectNoBase. 2022-03-10 15:45:56 -05:00
Thomas Harte
673ffc50da Switch to intended compact version of Instruction. 2022-03-10 15:14:50 -05:00
Thomas Harte
6dc9973754 Incorporate length into Instruction. 2022-03-10 07:12:12 -05:00
Thomas Harte
cf6a910630 Handle no-base case directly in existing switch. 2022-03-09 20:20:32 -05:00
Thomas Harte
520baa6ec8 Formalise IndirectNoBase and permit a knowledgable caller to avoid conditionals. 2022-03-09 20:19:40 -05:00
Thomas Harte
bbf925a27e Clarify, unify and correct decoding and encoding of [CALL/RET/JMP][near/far/relative/absolute]. 2022-03-09 16:48:06 -05:00
Thomas Harte
381fd5dbe4 E8 is a relative call. 2022-03-09 16:37:07 -05:00
Thomas Harte
ead8b7437e Remove done TODO. 2022-03-09 15:26:20 -05:00
Thomas Harte
acd9df6745 Fix segment/offset sizes for far calls. 2022-03-09 15:23:43 -05:00
Thomas Harte
f96c051932 Record PUSH immediate operation size. 2022-03-09 14:24:57 -05:00
Thomas Harte
67b2e40fae Fixed: INs and OUTs remain single byte. 2022-03-09 10:51:16 -05:00
Thomas Harte
081a2acd61 Fix shift group operand size. 2022-03-09 09:33:25 -05:00
Thomas Harte
de79acc790 Fix RegAddr/AddrRegs and group 2 decoding. 2022-03-09 08:38:34 -05:00
Thomas Harte
21d4838322 Fix current implementation of data_segment.
As far as it goes.
2022-03-08 17:08:21 -05:00
Thomas Harte
926a373591 Extend SIB test, correct decoder. 2022-03-08 15:03:37 -05:00
Thomas Harte
a954f23642 Attempt 32-bit modregrm + SIB parsing. 2022-03-08 14:39:49 -05:00
Thomas Harte
41a104cc10 Adds special test/control/debug MOVs.
This'll do; it's not ideal but avoids bloating up the `Source` enum.
2022-03-07 17:04:05 -05:00
Thomas Harte
f0b4971c7b Correct SHLD format. 2022-03-07 16:39:02 -05:00
Thomas Harte
8e669a32a3 Take a stab at group 8. 2022-03-07 16:34:56 -05:00
Thomas Harte
0e16e7935e Correct double reference to Group 6. 2022-03-07 16:26:17 -05:00
Thomas Harte
7ea84d9a4e Add MOVZX, MOVSX. 2022-03-07 16:25:44 -05:00
Thomas Harte
7313c89dec Add BT, BTS, BTR, BTC, BSF, BSR. 2022-03-07 16:23:25 -05:00
Thomas Harte
35a66c03c2 Add the SETs. 2022-03-07 10:32:34 -05:00
Thomas Harte
bbb3168bae Adds the missing shift group segues at c0 and c1. 2022-03-07 09:18:59 -05:00
Thomas Harte
1ea9d3faf8 Introduce additional forms of IMUL. 2022-03-07 09:05:22 -05:00
Thomas Harte
4479be4fd0 Add the two immediate PUSHes. 2022-03-06 14:28:41 -05:00
Thomas Harte
91a6bf671d Also 'easy': LSS, LFS, LGS.
Though perhaps I'm off on LES and LDS?
2022-03-06 09:28:43 -05:00
Thomas Harte
49b5889d9e 0x8c is available on the 8086. 2022-03-06 09:24:59 -05:00
Thomas Harte
ede61ae130 Flag up TODOs, for easier in-editor navigation. 2022-03-05 17:48:01 -05:00
Thomas Harte
7a79111767 Add the easiest 80386 extensions: PUSH/POP FS/GS and longer conditional jumps. 2022-03-05 17:32:21 -05:00
Thomas Harte
6432521b9d Correct two references to JP that should be JL. 2022-03-05 17:16:32 -05:00
Thomas Harte
65f578fe61 Add notes on all missing opcodes. 2022-03-05 17:16:13 -05:00
Thomas Harte
3a8eb4a4f0 Add 80386 segment overrides. 2022-03-05 17:03:46 -05:00
Thomas Harte
eb180656bb Fix $8e data size, add $8c. 2022-03-05 17:00:48 -05:00
Thomas Harte
1afcbba218 Clarify sign extension availability. 2022-03-05 16:44:26 -05:00
Thomas Harte
8a0902a83b Adapts existing opcodes for 32-bit parsing. 2022-03-05 13:52:07 -05:00
Thomas Harte
dfb312fee6 Make column and row meanings overt. 2022-03-05 11:56:08 -05:00
Thomas Harte
11bb594fa2 Sets up [ignored] memory and data size prefixes. 2022-03-02 20:23:35 -05:00
Thomas Harte
8e3ae2c78f Add opcode map as documentation. 2022-03-02 20:00:21 -05:00
Thomas Harte
4b4135e35a Correct #undef. 2022-03-01 18:23:24 -05:00
Thomas Harte
d1148c4cab Switch to constexpr function, for guaranteed semantics. 2022-03-01 17:30:41 -05:00
Thomas Harte
8ee62b4789 Simplify address size semantics.
Since it'll no longer be a mode-dependant toggle, but a fully-retained value.
2022-03-01 17:29:26 -05:00
Thomas Harte
5e7a142ff1 Fix is_write errors, update comment, add additional source for asserts. 2022-03-01 16:51:54 -05:00
Thomas Harte
2c816db45e Refactor: (i) to expose effective address calculation; and (ii) to include address size in Instruction. 2022-03-01 09:36:37 -05:00
Thomas Harte
b920507f34 Double down on AddressT, add an assert on memory_mask. 2022-02-28 10:03:58 -05:00
Thomas Harte
afbc57cc0c Incorporate displacement, switch macro flag. 2022-02-28 09:53:23 -05:00
Thomas Harte
9f12c009d6 Correct data size when accessing address registers. 2022-02-27 19:45:03 -05:00
Thomas Harte
84ac68a58b Fix indirect memory read/write 2022-02-27 18:43:00 -05:00
Thomas Harte
27d1df4699 Introduce enough of a DataPointerResolver test to build but fail. 2022-02-27 18:27:58 -05:00
Thomas Harte
0d7a7dc7c9 Introduce DataPointerResolver, to codify the meaning of DataPointer and validate that enough information is present. 2022-02-27 11:25:02 -05:00
Thomas Harte
b8bff0e7f5 Double up eSP, eBP, eSI, eDI and AH, CH, DH, BH enums, as per Intel's encoding. 2022-02-24 05:16:15 -05:00
Thomas Harte
60bf1ef7ea Rename SourceSIB to DataPointer, extend to allow for an absent base. 2022-02-23 08:28:20 -05:00
Thomas Harte
95976d8b58 Add missing #include. 2022-02-21 16:33:58 -05:00
Thomas Harte
ecb20cc29b Improve tabbing. 2022-02-21 16:09:03 -05:00
Thomas Harte
b6183e86eb Clarifies model tests by macro; adds the address size toggle. 2022-02-21 16:06:02 -05:00
Thomas Harte
229af0380c This is normatively called the address size. 2022-02-21 15:52:16 -05:00
Thomas Harte
b968a662d3 Dump notes on intended Instruction layout, add memory size flag. 2022-02-21 15:48:58 -05:00
Thomas Harte
159e869fe6 Justifies the templatisation. 2022-02-21 15:33:08 -05:00
Thomas Harte
76814588b8 Template Instruction on its content size. 2022-02-21 12:36:03 -05:00
Thomas Harte
1934c7faa2 Switch Decoder into a template. 2022-02-21 12:21:57 -05:00
Thomas Harte
9e9e160c43 Eliminate Ind[BXPlusSI/etc] in favour of specifying everything via a ScaleIndexBase. 2022-02-21 11:45:46 -05:00
Thomas Harte
546b4edbf1 Ensure ScaleIndexBase can be used constexpr; add note-to-self on indexing table. 2022-02-20 19:22:28 -05:00
Thomas Harte
63d8a88e2f Switch to holding the SIB as a typed ScaleIndexBase.
(and permit copy assignment)
2022-02-20 17:54:53 -05:00
Thomas Harte
75d2d64e7c Albeit that it requires nuanced shift/roll semantics, eliminates CL constant.
Shifts and rolls are already slightly semantically special for being undefined for values greater than 8/16/32 — i.e. in some implementations they don't even use the entirety of CL, just the low five bits. Which makes me feel a little better.

The upside of no ambiguity between eCX size 1 and CL justifies the trade.
2022-02-20 17:52:19 -05:00
Thomas Harte
a5113998e2 Accept that IN and OUT are going to have special semantics, thereby kill ::AX and ::DX. 2022-02-20 17:15:01 -05:00
Thomas Harte
4d2e8cd71d Adds a presently-unreachable step for SIB consumption. 2022-02-19 18:00:27 -05:00
Thomas Harte
30b355fd6f Chips away further at the legacy register names. 2022-02-18 18:37:47 -05:00
Thomas Harte
12df7112da Starts adjusting the concept of a Source. 2022-02-17 11:32:09 -05:00
Thomas Harte
cd5ca3f65b Attempts a full decoding of the 80286 instruction set. 2022-02-10 17:13:50 -05:00
Thomas Harte
0bd63cf00f Introduces the easy F page instructions. 2022-02-10 09:35:05 -05:00
Thomas Harte
7ceb3369eb Attempts decoding of the 80186 set. 2022-02-09 17:51:48 -05:00
Thomas Harte
ae21726287 Splits 80186 additions from 80286; fills in a touch more. 2022-02-01 20:38:10 -05:00
Thomas Harte
a4da1b6eb0 Begins enumerating the 80286 and 80386 instructions. 2022-01-31 09:11:06 -05:00
Thomas Harte
85bfd2eba3 Remove further errant 'Awaiting's. 2022-01-31 08:22:07 -05:00
Thomas Harte
2d543590dc Make a noun, for better consistency. 2022-01-31 08:14:33 -05:00
Thomas Harte
5f413a38df Switches all American-style dates.
I'd failed to configure my new computer appropriately, it seems.
2021-01-16 22:09:19 -05:00
Thomas Harte
2910faf963 Adds missing #include. 2021-01-15 22:33:14 -05:00
Thomas Harte
3c20e1f037 Adds files for the M50740 and corrects namespace errors elsewhere. 2021-01-15 21:30:30 -05:00
Thomas Harte
9c2c918760 Better sorts by function, corrects TEST description. 2021-01-15 21:07:02 -05:00
Thomas Harte
47d20699d8 Completes list, ensures POP acts as documented. 2021-01-15 20:48:31 -05:00
Thomas Harte
e8ce70dccb Chips further away at documentation. 2021-01-15 18:52:59 -05:00
Thomas Harte
fa4938f29c Establishes the reason I'm sort-of documenting these. 2021-01-15 18:27:55 -05:00
Thomas Harte
ddb4bb1421 Better plans project layout. 2021-01-15 18:16:01 -05:00