Thomas Harte
e1914b4f16
Attempts to add a proper intermediate buffer for sprites to allow the split of collection and output.
2017-12-08 22:12:39 -05:00
Thomas Harte
8653eb8b55
Corrects various latent errors in optimised TMS video collection.
2017-12-06 20:24:29 -05:00
Thomas Harte
a4f0a260fd
Reformulates the TMS graphics mode fetch loop to try to eliminate heavy conditionality. Temporarily introduces some sprite selection issues.
2017-12-05 22:39:03 -05:00
Thomas Harte
d4a53e82bb
Replaces manual retread of memcpy with standard memcpy.
2017-12-05 18:21:34 -05:00
Thomas Harte
6eedc99286
Makes substantial optimisations to text mode.
...
Character optimisations to come.
2017-12-04 22:18:51 -05:00
Thomas Harte
a473338abe
Makes minor type conversion fixes.
2017-12-03 22:24:48 -05:00
Thomas Harte
ad3df36c20
Corrects sprite information collection to cover all four.
2017-12-03 14:51:55 -05:00
Thomas Harte
38b11893e8
Takes first steps towards sprite display on the TMS.
2017-12-02 22:13:43 -05:00
Thomas Harte
e4534775b0
Cleans up and zooms in on the TMS slightly.
2017-12-02 17:48:31 -05:00
Thomas Harte
fe0cdc8d69
Corrects colour fetching in TMS Graphics II to be a function of row.
2017-12-02 16:10:29 -05:00
Thomas Harte
ca26ce8400
Slightly corrects style errors in the Cartridge hierarchy, and introduces mapping of .ROM to the MSX when appropriate.
2017-12-02 16:01:30 -05:00
Thomas Harte
d3dd8f3f2a
Implements screen 2 addressing.
2017-12-02 14:05:52 -05:00
Thomas Harte
3c8d2d579d
Resolves remaining sources of text mode instability.
2017-11-30 22:48:07 -05:00
Thomas Harte
edcbb3dfed
Tidies code a little and thereby uncovers and corrects one cause of output instability.
2017-11-30 22:19:53 -05:00
Thomas Harte
9c8158753e
Makes a first attempt at displaying text mode.
2017-11-30 21:35:26 -05:00
Thomas Harte
ee84f33ab5
Ensures that the 9918 admits that it is the source of interrupts.
2017-11-29 21:33:43 -05:00
Thomas Harte
aa4eef41d8
Seeks to introduce MSX interrupts.
2017-11-29 20:31:55 -05:00
Thomas Harte
ecd7d4731b
Advances emulation to showing what looks like appropriate text on screen.
2017-11-28 21:27:15 -05:00
Thomas Harte
563aa051e4
Simplifies code a little and gives something on screen.
2017-11-28 21:19:28 -05:00
Thomas Harte
642bb8333f
Introduces something of a first attempt at graphics collection and display. An unsuccessful attempt.
2017-11-28 21:10:30 -05:00
Thomas Harte
c558e86e03
Adds border colour output.
2017-11-27 22:05:40 -05:00
Thomas Harte
dbb14ea2e2
Corrects counting deficiencies that could produce an unstable display.
2017-11-27 21:36:12 -05:00
Thomas Harte
173e16b107
Corrects the 9918 so that it terminates.
2017-11-27 19:48:04 -05:00
Thomas Harte
7d2adad67e
Adds the absolute most basic version of in-frame time keeping, to display a white square.
2017-11-27 19:43:33 -05:00
Thomas Harte
d33612def5
Ensures the MSX provides a clock to the VDP.
2017-11-26 20:07:30 -05:00
Thomas Harte
9cb6ca3440
Adds elementary decoding of VDP accesses.
2017-11-26 20:01:11 -05:00
Thomas Harte
0eb5dd9688
Introduces the fundamentals of bus routing for the MSX.
2017-11-26 16:47:59 -05:00
Thomas Harte
a14b53a9ab
Adds a TMS9918 skeleton plus enough in the MSX to get to a blank screen in SDL/kiosk mode.
2017-11-26 13:28:26 -05:00
Thomas Harte
c827d14d97
Corrects various GCC warnings across the 6560, CPC, TIA, Oric video and elsewhere.
2017-11-12 17:17:27 -05:00
Thomas Harte
5408efe9b5
Flags obvious default options within the 6560, Vic-20 and DynamicMachine.
2017-11-12 16:41:09 -05:00
Thomas Harte
6d80856f02
Attempts to eliminate warnings around a meaningless value and an unused label in the 8272.
2017-11-12 16:34:51 -05:00
Thomas Harte
4778616fd7
Eliminates unused result and unused label.
2017-11-12 16:30:23 -05:00
Thomas Harte
5aef81cf24
Commutes cross-platform #pragma marks to //MARK:s.
2017-11-12 15:59:11 -05:00
Thomas Harte
2e15fab651
Doubles down on <cX> over <X.h> for C includes, and usage of the namespace for those types and functions.
2017-11-11 15:28:40 -05:00
Thomas Harte
5b6ea35d96
Corrects initialisation ordering for the ZX80/81, C1540 and AY-3-8910.
2017-11-10 22:31:27 -05:00
Thomas Harte
4cbc87a17d
Corrects out-of-order initialisations for the 1770, Atari 2600 joystick, Pitfall II bus extender, Microdisc and 6502.
2017-11-10 22:20:44 -05:00
Thomas Harte
ff7ba526fb
Corrects improper initialisation order on the 6560.
2017-11-10 22:05:35 -05:00
Thomas Harte
cb015c83e1
Eliminated C99-style struct initialisations.
2017-11-10 19:14:19 -05:00
Thomas Harte
c0055a5a5f
Further builds up SConstruct, correcting many missed imports and a couple of improper uses of C99 in C++ code.
2017-11-09 22:04:49 -05:00
Thomas Harte
f95515ae81
Eliminates a large number of instance of end-of-line tabs.
2017-11-07 22:51:06 -05:00
Thomas Harte
ad9df4bb90
Commutes uint8_t *, uint16_t *, uint32_t *, size_t, off_t and long to functional-style casts.
2017-10-21 22:30:15 -04:00
Thomas Harte
ec999446e8
Commutes int and unsigned casts to the functional style.
2017-10-21 21:00:40 -04:00
Thomas Harte
5e3e91373a
Switches all unsigned int and double casts to functional style.
2017-10-21 19:49:04 -04:00
Thomas Harte
91b867a7b3
Ensures full 8272 instance state initialisation.
2017-10-17 22:11:01 -04:00
Thomas Harte
3944e734d3
Ensures full 6845 instance state initialisation and uses an unsigned shifter.
2017-10-17 22:10:28 -04:00
Thomas Harte
97a2be71e3
Introduces flush_tracks to Drive, while switching its interface to using Track::Address and adjusting associated integer types.
2017-10-06 21:45:12 -04:00
Thomas Harte
edb9fd301c
Begins this project's conversion to functional-style casts.
2017-10-03 22:04:15 -04:00
Thomas Harte
c7f27b2db4
Renames MFM.[c/h]pp as per its new remit: encoding only.
2017-09-24 21:40:43 -04:00
Thomas Harte
2a08bd9ecc
Factors shifting plus stateful [M]FM token recognition out of the MFMDiskController.
...
Given the proliferation of MFM-related classes, establishes a subdirectory for them.
2017-09-24 20:07:56 -04:00
Thomas Harte
698e4fe550
Tidies the Disk file hierarchy.
2017-09-22 22:39:23 -04:00
Thomas Harte
d6a5f9a29e
Revokes unnecessary change.
2017-09-16 18:24:13 -04:00
Thomas Harte
0d84b4b9dd
Removes some redundant end_writing calls.
2017-09-16 17:09:17 -04:00
Thomas Harte
98751e6ac8
Ensures that all result phases are exactly the intended length by replacing accumulation with assignment.
...
Also attempts a different version of control mark behaviour. Experiments.
2017-09-15 22:59:26 -04:00
Thomas Harte
35fe4d50d4
Adds command termination upon drive becoming unready, and copies head and drive selection into ST0.
2017-09-15 20:26:41 -04:00
Thomas Harte
4d4a0cf1d2
Puts the disk controller back into the loop with knowledge about reading mode, and uses that knowledge to cut off the PLL.
2017-09-14 22:30:40 -04:00
Thomas Harte
b62f3e726a
Adds a start-of-execution-phase get-out for drives that aren't ready.
2017-09-12 20:43:53 -04:00
Thomas Harte
2f13517f38
Adjusts the 1770 not to talk directly to the drive about motor status.
2017-09-11 22:10:56 -04:00
Thomas Harte
d3c385b471
Separates the 8272's drive selection signalling from actual drive ownership.
...
Thereby returns working motor control to the CPC.
2017-09-11 21:25:26 -04:00
Thomas Harte
96bf133924
Withdraws requirement for DiskController users to specify a PLL multiplier or to provide rotation speed.
...
In the latter case because it's no longer of any interest to the controller, and in the former because I'd rather it be picked automatically.
2017-09-10 22:56:05 -04:00
Thomas Harte
0622187ddf
Strips Controller of all capabilities now housed on the Drive.
2017-09-10 19:23:23 -04:00
Thomas Harte
90c7056d12
Started devolving timed event loop logic down to the drives, moving them closer to modelling real life.
2017-09-10 14:43:20 -04:00
Thomas Harte
ff510f3b84
Explicitly disallows copying of VIAs, and marks the constructor as noexcept.
2017-09-05 21:21:23 -04:00
Thomas Harte
7fd6699e0b
Corrects comment indentation.
2017-09-05 21:15:15 -04:00
Thomas Harte
450712f39c
Improves and corrects 6522 header documentation.
2017-09-04 14:32:34 -04:00
Thomas Harte
24b3faa427
Deconstitutes the 6522 into component parts, templated and non-templated.
...
Adjusts the Oric, Vic-20 and C-1540 accordingly, albeit with the quickest possible solutions.
2017-09-04 14:26:04 -04:00
Thomas Harte
b30bb2a234
Adds an initial implementation of display skew, as a completely live property.
2017-08-29 22:16:40 -04:00
Thomas Harte
334afbc710
Removes const from get_status and get_register, as both may now logically mutate the object.
2017-08-27 18:13:55 -04:00
Thomas Harte
17c13624e5
Improved comments.
2017-08-27 18:11:40 -04:00
Thomas Harte
113349d272
Started making some formal admissions that different CRTC models exist. Plenty yet to do.
2017-08-27 18:10:07 -04:00
Thomas Harte
bdda701207
Reverts previous unevidenced change.
2017-08-26 22:58:16 -04:00
Thomas Harte
487fe83dca
Ensures that vertical sync and end-of-visible-lines conditions potentially trigger whenever line_counter_ changes, not only when it increments.
2017-08-26 17:54:54 -04:00
Thomas Harte
6c5a03187b
Relocates the HSYNC start test, in order to pass Arnold's cpctest HSYNC start position conformance test.
2017-08-26 17:22:48 -04:00
Thomas Harte
7d7aa2f5d5
Eliminates repetition of the unpacking of register 3 into a horizontal sync count.
2017-08-26 14:37:03 -04:00
Thomas Harte
28550c0227
Breaks the 6845 bus cycle into a phase 1 and a phase 2 per the belief that sync line changes, which are observable, happen at the end of the first phase rather than at the beginning of the next. This may have interrupt timing effects, as machines often derive an interrupt from sync.
2017-08-26 13:56:23 -04:00
Thomas Harte
6e99169348
Permits the 6845's bus state to be examined by an owner, eliminating the need to buffer it in the bus handler. But more than that it allows the CRTC to decide when it adjusts the various outputs respective to the main phase. So a net effect of the change is that the CPC now sees vsync a cycle earlier, because my current reading of the 6845 datasheet is that it is set at the end of phase 1, not the beginning of the next phase 1.
2017-08-26 12:59:59 -04:00
Thomas Harte
3caa4705ca
Limits sync counter size.
2017-08-26 12:31:19 -04:00
Thomas Harte
039aed1bd1
Switches the two sync counters to upward-going rather than downward, as a more likely match to the way the rest of the 6845 implementation.
2017-08-25 21:26:01 -04:00
Thomas Harte
a914eadc85
Ensured that register 6 is checked on every loop.
2017-08-22 22:17:45 -04:00
Thomas Harte
e956740c56
Refactors the 6845 more clearly to break out the acts of ending a line and ending a frame, changing the way the memory address is altered — the end-of-line value is provisionally stored and then used if necessary — in order to do so.
2017-08-22 21:54:48 -04:00
Thomas Harte
e88a51e75e
Worked logic all the way down to the CPC. If the 8272 announces that it is asleep, it is now no longer clocked. Also very slightly cut down on IRQ line chatter to the Z80.
2017-08-20 12:05:00 -04:00
Thomas Harte
2d9efccc98
Introduced a master 'is sleeping' flag. I'm starting to think there's a pattern forming here.
2017-08-20 10:43:53 -04:00
Thomas Harte
8ce46b6e49
Having spotted that I was using my single-character loop counter names incorrectly (quelle surprise!), got a bit more explicit. Also flattened into a single loop so that I can break rather than returning.
2017-08-20 10:32:09 -04:00
Thomas Harte
669e0caff5
Ensured the head_unload_delay values are properly seeded, and generalised the quick escape.
2017-08-19 22:06:56 -04:00
Thomas Harte
e208f03636
Corrects the US colour palette, effectively undoing what was a mistaken adjustment for the time when Oric-centric phase alignment was built into the CRT based on a false calculation that it wouldn't affect the machines that generate chrominance functionally.
2017-08-16 09:58:34 -04:00
Thomas Harte
cc9d23f23b
Inverted meaning of register_masks, as it's a bit weird that the mask is inverted immediately upon usage. It's a left-over from thinking the unused bits should be 1s; unit tests reveal they should be 0s. Comment updated appropriately.
2017-08-16 09:29:48 -04:00
Thomas Harte
1a831bcf9b
Quick fix: supply the port being written to correctly.
2017-08-16 09:15:57 -04:00
Thomas Harte
82367a2246
Added documentation.
2017-08-16 09:14:56 -04:00
Thomas Harte
3947347d88
Introduces active input handling for the AY and uses it in the CPC to give proper, active keyboard input, rather than push-on-select, which was only ever a temporary hack. Also maps a few more keys for the Amstrad.
2017-08-15 22:47:17 -04:00
Thomas Harte
aefbafa18d
Centralised the 8272's actions of setting the non-DMA execution flag, picking the drive and head and loading the head for accessing commands, and switched error flag if read ID doesn't find anything.
2017-08-15 21:49:10 -04:00
Thomas Harte
d12c834f9c
Increased amount reported.
2017-08-15 20:35:33 -04:00
Thomas Harte
56de5cb2b3
Removed one further logging event.
2017-08-15 20:18:32 -04:00
Thomas Harte
709257a0c5
Quick fix: also treat reception of sync as a reason not to stop looking for a data address mark.
2017-08-15 20:16:56 -04:00
Thomas Harte
75a9d2bb33
Started withdrawing logging, reduced index hole count back to the correct number and tried to increase read_data rigour.
2017-08-15 20:12:01 -04:00
Thomas Harte
73080d6c36
Added an easy way for disk controllers to clamp termination of written data exactly to the index hole.
...
This commit also temporarily provides a whole load of extra logging and minor logic improvements from the 8272. I'm mid-flow on finding a particularly vicious error in its handling of writing; wait for the pull request. But, at least: now waits for the first part of a post-ID gap before writing data, and attempts partially to handle appearance of the index hole during writing a track. More work to do on that though.
2017-08-15 16:05:10 -04:00
Thomas Harte
9d77f33611
Dealt with another source of repeating magic constants: the command numbers.
2017-08-15 11:06:10 -04:00
Thomas Harte
7d132f81f7
Increased logging by quite a distance and made an attempt once again to allow the processor some time to supply the first byte when writing before declaring overrun.
2017-08-15 10:50:28 -04:00
Thomas Harte
6553bf05b4
Corrected multi-sector reads: ensured the incremented sector number isn't replaced by the original, and that the controller returns to scanning mode.
2017-08-14 22:27:31 -04:00
Thomas Harte
0816d3f5a9
Corrected 'read deleted data' command. It's 0xc, not 0xb.
2017-08-14 21:41:20 -04:00
Thomas Harte
55055c7847
Minor: ensured immediate line comparison works. But I think my problem might be trying to do this as straight line logic?
2017-08-14 19:08:20 -04:00
Thomas Harte
cddcd0fb79
Put my money where my mouth is and switched the superclass of WD1770 to MFMController, eliminating duplicated (/factored out) code.
2017-08-14 16:32:53 -04:00
Thomas Harte
a366298022
Factored out the standard [M]FM gap and mark groups, to increase 8272 readability and because it's pretty-much certain I'll need them again if ever I try to tackle e.g. the 8271.
2017-08-14 16:03:35 -04:00
Thomas Harte
4df9307d25
Factored out the dull and repetitious stuff of writing n bytes of the same value.
2017-08-14 15:50:36 -04:00
Thomas Harte
9038ba622e
Added a quick version of read track.
2017-08-14 14:34:56 -04:00
Thomas Harte
7b8bb0297a
Implemented single density version of format track.
2017-08-14 13:03:17 -04:00
Thomas Harte
0da02d3902
Added read/write escape clauses if faced with a read-only disk.
2017-08-14 12:53:18 -04:00
Thomas Harte
2e5ad19fe1
Minor tidying.
2017-08-14 12:42:48 -04:00
Thomas Harte
a10389a22c
Factored out the stuff of stuffing the bus.
2017-08-14 12:42:22 -04:00
Thomas Harte
cefec7a19f
Sought more robustly (i.e. less repetitively) to handle dispatch, including cancelling seeks where appropriate.
2017-08-14 10:37:39 -04:00
Thomas Harte
7264fbb3d2
read_id now clears status. I probably need to find a way to generalise this.
2017-08-14 09:58:55 -04:00
Thomas Harte
8a7b23dc9e
Ensured data-accessing commands cancel seeks on their drives. Also introduced a count of drives currently seeking in order to make for a slightly better broad-phase test in run_for.
2017-08-14 09:45:39 -04:00
Thomas Harte
b7065575f3
Added (empty) call-ins for DMA usage; switched to having the 'is seeking' bit in the status register stay high until sense interrupt status, but now it goes high even for seeks that don't actually go anywhere, and corrected interpretation of the specify command, with a positive result: the received step rate time, now that it's being interpreted correctly, is much shorter.
2017-08-14 09:04:22 -04:00
Thomas Harte
7ea703f150
Started making provisions for a DMA-compatible implementation. Re: the CPC, it sounds like DMA acknowledge might be permanently wired, causing DMA mode seemingly to work from the 8272's point of view.
2017-08-14 08:38:00 -04:00
Thomas Harte
1011143dbe
Sought to correct my interpretation of 'gap 3'.
2017-08-13 21:52:48 -04:00
Thomas Harte
9ace6e1f71
Applied minimum constraints for specified parameters.
2017-08-13 19:25:57 -04:00
Thomas Harte
5221837be8
Fixed Non-DMA flag for the format track execution phase. The emulated machine now provides sector details.
2017-08-13 18:51:06 -04:00
Thomas Harte
e1e9a06712
Made an attempt at format a track.
2017-08-13 18:05:19 -04:00
Thomas Harte
b0a7208cc7
Strung together a very basic version of 8272 write [/deleted] data. Lots of cases as-yet unhandled.
2017-08-13 12:50:07 -04:00
Thomas Harte
eec42aa7ae
Entrusted further status to drives; also adjusted them to report read only if diskless, which I now believe to be correct.
2017-08-13 11:50:49 -04:00
Thomas Harte
5f42022c1d
Added a tester for the control mark.
2017-08-12 17:35:14 -04:00
Thomas Harte
11d0c37506
Attempted to find a more expressive way for maintaining state — macros for all conditions, to bind both values and destinations.
2017-08-12 17:33:52 -04:00
Thomas Harte
27d1dc5c37
Removed some old printfs.
2017-08-12 16:49:20 -04:00
Thomas Harte
7135259cc1
Sought to flesh out error conditions.
2017-08-12 16:36:37 -04:00
Thomas Harte
4909325e79
Implemented read deleted data.
2017-08-12 13:01:17 -04:00
Thomas Harte
a4ee697ed1
Quickie: head unload is scheduled only if the head is presently loaded.
2017-08-12 12:53:45 -04:00
Thomas Harte
0f15a2f97f
Relented: it actually looks like status bytes aren't per-drive. But each drive may fail at seeking individually. So that piece of state accumulates at the 8272 drive.
2017-08-12 12:52:36 -04:00
Thomas Harte
89ace671a4
Corrected unload time. Was 8000 times too short.
2017-08-12 09:44:01 -04:00
Thomas Harte
e7db2a2f6d
Sought to introduce head loading and unloading delays.
2017-08-12 09:36:21 -04:00
Thomas Harte
ffb1a14ace
Minor: clear status registers before a read data.
2017-08-11 18:56:33 -04:00
Thomas Harte
e4f04d0977
Merge branch 'master' into AYFidelity
2017-08-11 14:41:08 -04:00
Thomas Harte
a5593bec79
Threw in support for the light-pen trigger.
2017-08-10 15:00:14 -04:00
Thomas Harte
a1e2646301
Imposed counter size limits.
2017-08-10 14:58:24 -04:00
Thomas Harte
6a6e5ae79c
Forced users of the 6845 to be explicit about which type. So far with no effect.
2017-08-10 12:28:57 -04:00
Thomas Harte
02d792c003
Simplified logic slightly, avoiding repetition.
2017-08-10 11:48:37 -04:00
Thomas Harte
be8e7a4144
Eliminated false register aliasing, restricted register sizes and locked out reading and writing where appropriate.
2017-08-10 11:22:30 -04:00
Thomas Harte
b1dbd7833a
Merge branch 'master' into 6845Address
2017-08-10 11:15:08 -04:00
Thomas Harte
a4c910f1de
This appears to be a more accurate take on 6845 address advancement — it is necessary that character output has finished for the line address to be updated.
2017-08-10 11:12:53 -04:00
Thomas Harte
2eed24e859
Made an initial attempt at [a subset of] multi-sector reads.
2017-08-10 11:11:26 -04:00
Thomas Harte
b11d142cff
Switched to descriptive names.
2017-08-08 20:35:41 -04:00
Thomas Harte
021ff8674e
Added something for sense drive status.
2017-08-08 20:30:54 -04:00
Thomas Harte
e3d1f4fe1e
Subjectively, this might be more correct. It definitely prevents intermediate frequencies. More research required.
2017-08-08 17:58:35 -04:00
Thomas Harte
3bdedfd749
Improved comments.
2017-08-08 07:44:46 -04:00
Thomas Harte
46278ff297
Experimental: is this meant to be a compare-before-increment?
2017-08-07 23:02:29 -04:00
Thomas Harte
390ecec3d9
Added: now declines to pass on output if in input mode for ports A and B.
2017-08-07 19:56:22 -04:00
Thomas Harte
41a30c147d
Adjusted: invalid register selection simply deselects all registers.
2017-08-07 19:51:36 -04:00
Thomas Harte
4709ae80cb
Added port direction tests.
2017-08-07 19:36:55 -04:00
Thomas Harte
7fbb455836
Per the CPC test I'm checking, 0s should be returned for non-retained bits, not 1s.
2017-08-07 19:07:12 -04:00
Thomas Harte
745afd217f
The port input/output flags are now honoured; reading a port that is set as an output returns the current output value.
2017-08-07 19:01:18 -04:00
Thomas Harte
47732ffb98
Prevented the 8272 from overreading ID fields (and, by doing so, overrunning its internal buffer). Exposed the MFMController's CRC generator for inspection.
2017-08-07 12:37:22 -04:00
Thomas Harte
d07f3216ab
Added a broad phase on whether seeking is ongoing.
2017-08-07 12:12:59 -04:00
Thomas Harte
68c73184b1
Had failed to spot that by taking control of stepping at this level, the appropriate invalidate_tracks were not being sent.
2017-08-07 10:36:53 -04:00
Thomas Harte
7f824d6494
Ensured seeks and recalibrates end immediately if no seeking is required.
2017-08-07 10:31:32 -04:00
Thomas Harte
3219212f03
A closer inspection of the data sheet seems to suggest that invalid command sequences will post ST0.
2017-08-07 07:35:41 -04:00
Thomas Harte
d90e35e5bd
Added a bunch of comments, and ensured that the data request bit remains set for the entire period that command bytes are accepted.
2017-08-07 07:27:00 -04:00
Thomas Harte
73f8488150
Reaching the end of the usable part of my day, decided to tidy up a little before bed with indentation that reflects a distinction between top-level entry points and mere loops.
2017-08-06 22:14:18 -04:00
Thomas Harte
3853966a1e
Removed formal storage of ST3, as it just seems to be composed live. This may turn out also to be the best way to deal with ST0–2, time will tell. Also took a stab at the error in responding properly to the ROM's intended use of seek might be accepting new commands as replacements for old ones rather than rejecting them. That didn't seem to do the trick.
2017-08-06 22:10:12 -04:00
Thomas Harte
d63893a437
Collapsed implementations of recalibrate and seek, and decided to intend to go for an upward count on steps taken rather than a downward one. But seek continues presently to fail.
2017-08-06 21:52:52 -04:00
Thomas Harte
90c74043f5
Remembered to toggle off RQM between bytes. CAT now works.
2017-08-06 21:21:59 -04:00
Thomas Harte
600445d90a
Made a first attempt to return sector contents.
2017-08-06 20:40:29 -04:00
Thomas Harte
e4b405fd3d
With the ROM now using a read ID to set its expectations, implemented that and fixed FIND/READ_HEADER macros for multiple use. Execution now reaches the unimplemented section of read data.
2017-08-06 20:32:46 -04:00
Thomas Harte
3b7ecbdf0d
Renamed result_ to result_stack_ to emphasis the fact that it goes backwards. Switched meaning of CB so that it is set for the entire command, execution and result phases.
2017-08-06 20:17:12 -04:00
Thomas Harte
01efb645cb
Took a reasonable gamble that the CHRN reported is from internal registers, not from the last-found header.
2017-08-06 19:57:34 -04:00
Thomas Harte
b5ec1f42d5
Started resetting 'busy' when entering the result phase. AMSDOS now complains of a missing disk after failing to find sector 01. My belief is that it should end up asking for C1. So this is not even getting through a failure to find a sector correctly yet.
2017-08-06 19:48:17 -04:00
Thomas Harte
e9972aa0dd
Added respect for the index-hole limit on reading, and an error phase.
2017-08-06 19:25:44 -04:00
Thomas Harte
1c9a744b01
Made an effort to start inspecting ID fields, at least. Discovered that my emulation has somehow stopped proceeding beyond sense interrupt status though. Fix one in that area: adjust ST0 just in time for the sense interrupt status response, as that'll need to specify the drive number properly.
2017-08-06 18:06:20 -04:00
Thomas Harte
e6d4bb29d8
Discovered correct sense interrupt status result if nobody is in the completed seeking state, and switched to it. It's a single 0x80 rather than two bytes.
2017-08-06 15:34:33 -04:00
Thomas Harte
6c5b562d97
Made an attempt at some of the correct seek/recalibrate behaviour: it's now asynchronous from command processing and able to work on up to four drives at once. I just probably am not yet hitting all the status flags I need to hit.
2017-08-06 15:22:07 -04:00
Thomas Harte
a7103f9333
Disks are now communicated to the 8272. Which is able to handle four of them.
2017-08-06 13:24:14 -04:00
Thomas Harte
c12425e141
Added storage for the extended four status registers, and made an attempt at implementing the two most trivial result-phase commands. Am slightly paused momentarily trying to figure out whether seek activity is orthogonal to read/write activity.
2017-08-06 12:55:57 -04:00
Thomas Harte
89f6de1383
Started on a real ugly-implementation coroutine approach, and implemented specify as a fairly trivial first command: it has no result phase, and is the only thing called by AMSDOS as part of the initialisation sequence.
2017-08-06 12:36:18 -04:00
Thomas Harte
34eaf75352
Fixed WAIT_FOR_TIME macro.
2017-08-06 12:08:54 -04:00
Thomas Harte
29288b690e
Switched disk controllers to be instantiated explicitly in terms of cycles, created an Amstrad-specific subclass of the 8272 to record the direct programmatic availability of all disk motors bundled together, and otherwise adjusted to ensure the thing is clocked and that the motor is enabled and disabled appropriately. The 8272 is also now formally a subclass of the incoming MDM controller.
2017-08-06 09:45:16 -04:00
Thomas Harte
25fd3f7e50
Mildly increased work in here, still primarily oriented towards logging what I actually need to get done.
2017-08-05 22:26:59 -04:00
Thomas Harte
3e984e75b6
Strung up an empty shell that eventually should contain the 8272, and added appropriate IO decoding to the Amstrad.
2017-08-05 19:45:52 -04:00
Thomas Harte
26ce6cdab2
Permitted register 3 to dictate vertical sync length.
2017-08-04 08:56:36 -04:00
Thomas Harte
3ca9c38777
Attempted to move to more accurate bus reading — if control lines are set then all subsequent data inputs should act according to the current control lines; changes to port input should be reflected live upon readings, etc.
2017-08-02 19:45:58 -04:00
Thomas Harte
0267bc237f
Added the ability to set a port input, and relaxed bus state testing. I think my on-demand bus reactions here are inappropriate, so more work to do here probably.
2017-08-01 18:04:51 -04:00
Thomas Harte
e6854ff8db
Corrected typo: the input to an AY is BDIR, not BCDIR.
2017-08-01 17:06:57 -04:00
Thomas Harte
2d4e202be3
Completed dangling comment.
2017-08-01 17:01:36 -04:00
Thomas Harte
64da8e17d1
Fixed: of course this should take a reference to an existing port handler rather than hatching its own; otherwise additional communication with a port handler by an i8255 owner doesn't work as intended.
2017-08-01 17:01:20 -04:00
Thomas Harte
08ad35efd9
It's barely an implementation of the 8255, but ensured that data is bounced into the PortHandler, conveniently assuming the interaction mode used by the CPC.
2017-08-01 16:34:13 -04:00
Thomas Harte
58b98267fc
Formally transferred ownership of PIO accesses to an incoming template, and decided to start being explicit about how to specify the interfaces and provide fallbacks for optional behaviour for the new, clean generation of interfaces. A full-project sweep will inevitably occur but I'll try to tie off this branch first.
2017-08-01 16:15:19 -04:00
Thomas Harte
ace71280a0
Removed implementation file; this is only ever going to be a template.
2017-08-01 16:00:17 -04:00
Thomas Harte
1d99c116e7
Actually, this is probably more correct: increment and then compare, but increment the refresh address once more after the final character, to avoid repeating it.
2017-08-01 15:29:37 -04:00
Thomas Harte
ee27e16fb1
Switched to post-tests increment. Seems to give proper screen width, but also eliminates that 'compare to +1' step that felt unlikely.
2017-08-01 15:19:25 -04:00
Thomas Harte
3b1db14817
Made a quick attempt at properly updating the refresh address.
2017-08-01 07:36:03 -04:00
Thomas Harte
e3f677fa37
I was under-counting row lines. Adjusted comparison. The emulator now produces a solid white square of approximately correct proportions. I'm sure that filling in pixels will reveal the next set of bugs.
2017-07-31 22:21:46 -04:00
Thomas Harte
5c68b6cc21
Fixed display enable reset when there's no adjustment area. A practical lesson in failure to factor.
2017-07-31 22:16:08 -04:00
Thomas Harte
ffaa627820
Fixed frame restart when there is no adjustment period.
2017-07-31 22:13:45 -04:00
Thomas Harte
5a396f6787
Added an explicit cast.
2017-07-31 22:04:31 -04:00
Thomas Harte
cb0dc7b434
I'm sure it's not going to be this easy, but this is a genuine attempt at full horizontal and vertical timing.
2017-07-31 22:01:54 -04:00
Thomas Harte
e28829bd1b
Corrected CRTC timing, gave it someone to talk to and a means with which to talk.
2017-07-31 20:14:46 -04:00
Thomas Harte
68ceeab610
Created a 6845 class and started pushing data at it and clocking it. It doesn't currently have the concept of a bus but will do, hence the in-header implementation.
2017-07-31 19:56:59 -04:00
Thomas Harte
4abd62e62b
Standardises on const [Half]Cycles as the thing called and returned, rather than const [Half]Cycles & as it's explicitly defined to be only one int in size, so using a reference is overly weighty.
2017-07-27 22:05:29 -04:00
Thomas Harte
1da24d10fd
Corrected a couple of build errors.
2017-07-27 08:05:14 -04:00
Thomas Harte
8361756dc4
Switched definitively to the works-for-now approach of requiring an explicit opt-in where somebody wants to clock a whole-cycle receiver from a half-cycle clock.
2017-07-27 07:40:02 -04:00
Thomas Harte
1c2f68f129
Removed, as it's been relocated.
2017-07-25 20:43:05 -04:00
Thomas Harte
75d67ee770
Relocated ClockReceiver.hpp as it's a dependency for parts of the static analyser, and therefore needs to be distinct from the actual emulation parts.
2017-07-25 20:20:55 -04:00
Thomas Harte
a1e9a54765
Eliminated redundant uses of ClockReceiver and sought to ensure that proper run_fors are inherited all the way down.
2017-07-25 20:09:13 -04:00
Thomas Harte
545683df6f
Added some documentation, got explicit again about cycle/half-cycle intermingling, and added flush as what amounts to divide(1), for cleaner usage without a clock divider.
2017-07-25 19:50:40 -04:00
Thomas Harte
cfbd62a5dc
Attempted to fix implementation of divide, and marked everything as-yet unmarked as inline.
2017-07-25 07:43:39 -04:00
Thomas Harte
40339a12e1
Formalised the use of a cycles count with a divider, bringing a few additional plain-int users into the fold.
2017-07-25 07:15:31 -04:00
Thomas Harte
9be9bd9106
Neatened layout.
2017-07-24 22:52:35 -04:00
Thomas Harte
c1527cc9e2
Reduced back-and-forth between Cycles and ints within the Oric.
2017-07-24 22:46:31 -04:00
Thomas Harte
a1a3aab115
Fixed implicit sign conversion.
2017-07-24 22:40:15 -04:00
Thomas Harte
c77a83d86f
The 6560 is now a ClockReceiver. This reduces to zero the number of remaining instances of the text run_for_cycles in this codebase.
2017-07-24 22:38:35 -04:00
Thomas Harte
efdac2ce8c
The 6522 is now a ClockReceiver.
2017-07-24 22:29:09 -04:00
Thomas Harte
2912d7055b
The 6532 is now a ClockReceiver.
2017-07-24 21:57:24 -04:00
Thomas Harte
b7f88e8f61
Filter is now a ClockReciever, affecting all sound output devices.
2017-07-24 21:29:13 -04:00
Thomas Harte
8a2bdb8d22
Converted the TimedEventLoop and the things that sit atop it into ClockReceivers.
2017-07-24 21:19:05 -04:00
Thomas Harte
b82bef95f3
Decided to follow through on Cycles and HalfCycles as complete integer-alikes. Which means giving them the interesting range of operators. Also killed the implicit conversion to int as likely to lead to type confusion.
2017-07-24 20:10:05 -04:00
Thomas Harte
8a0b0cb3d7
Extended both classes to allow copy assignment, copy construction and implicit zero-length construction.
2017-07-23 22:13:41 -04:00
Thomas Harte
1ba3f262a2
Sketched out a template for clock-receiving components to allow them to be implemented in terms of either half or whole cycles.
2017-07-22 21:46:50 -04:00
Thomas Harte
8755824c64
Added some documentation.
2017-07-22 17:25:53 -04:00
Thomas Harte
64865b3f41
Signedness fixes.
2017-07-21 21:23:34 -04:00
Thomas Harte
53f0e1896b
Made delay_time_ unsigned for safe comparison.
2017-07-21 21:21:23 -04:00
Thomas Harte
aaa60dab12
Fixed signedness of index.
2017-07-21 21:21:01 -04:00
Thomas Harte
12f7e1b804
Enshrined a default colour burst amplitude. Which now everybody relies on. The 102 figure is derived from the burst apparently being 40 IRE.
2017-07-07 23:35:14 -04:00
Thomas Harte
eb8a2de5d6
Settled definitively on flush as more communicative than synchronise (and slightly more locale neutral); culled some more duplication from the Z80.
2017-05-15 07:38:59 -04:00
Thomas Harte
e270b726b3
Tweaked blue, increased saturation.
2017-05-13 22:01:02 -04:00
Thomas Harte
44ce7fa54c
Corrected luminances across the board, and PAL colours.
2017-05-13 21:50:09 -04:00
Thomas Harte
b0142cf050
Made an updated stab at NTSC colours.
2017-05-13 14:29:36 -04:00
Thomas Harte
a340331229
Introduced 1-bit of saturation, returning black and white as black and white.
2017-05-11 21:31:58 -04:00
Thomas Harte
15d17c12d5
Switched the 6560 to two bytes per pixel, since one isn't sufficient for precision and because mixing up the implementation might help me to figure out what's amiss.
2017-05-09 21:22:01 -04:00
Thomas Harte
5998123868
Added some consts, for a minor safety improvement.
2017-05-06 19:53:24 -04:00
Thomas Harte
e01f3f06c8
Completed curly bracket movement.
2017-03-26 14:34:47 -04:00
Thomas Harte
a4c5eebd1e
The latest Atari Age-discovered numbers suggest this starts up in 1024T mode.
2017-03-21 18:22:50 -04:00
Thomas Harte
c445eaec3e
Switched startup values, following a comment on AtariAge. May or may not be correct, the thread was speculative.
2017-03-19 17:38:26 -04:00
Thomas Harte
e0bca1e37b
Reinstated the 16 and 32 kb Atari pagers, and ensured the 6532 always starts in a valid state.
2017-03-18 17:34:34 -04:00
Thomas Harte
d3257c345a
Tested against public ROMs and corrected. Also moved the deferred adjustment into a more canonical place.
2017-03-04 17:00:28 -05:00
Thomas Harte
e09b76bf32
Fixed 'same value, then immediate increment, then proper counting increments' behaviour and ensured it takes one cycle to commit a value. Adjusted tests to match.
2017-03-04 15:57:54 -05:00
Thomas Harte
ced644b103
It seems likely that an AY divides its clock by 8, not 16. I had conflated wave frequency and counter clock.
2017-01-11 22:03:01 -05:00
Thomas Harte
eca3995481
Added a CRC check for read address, ensured CRC, lost data and record not found are initially reset.
2017-01-01 21:00:25 -05:00
Thomas Harte
044c920a5b
Made it more explicit that there are no unhandled cases.
2017-01-01 20:56:52 -05:00
Thomas Harte
0df9ce5a76
Made an attempt at read address. So superficially that leaves only the force interrupts.
2017-01-01 20:55:09 -05:00
Thomas Harte
f94f34f053
Made an attempt at read track. Which means process_input_bit can't just swallow syncs any more; it now reports them as tokens of type ::Sync.
2017-01-01 20:39:19 -05:00
Thomas Harte
c994fa39f6
Ensured spin-up doesn't occur if there's no motor line.
2016-12-31 16:18:30 -05:00
Thomas Harte
1ea4f0d79d
Made an attempt to implement 'write track' and ensure that 'write sector' can't end without announcing that it has ended writing.
2016-12-31 16:01:44 -05:00
Thomas Harte
8eb21c6702
The "MFM...Byte"s aren't MFM-specific, they're relevant to both FM and MFM encoding. So renamed them. Also slimmed syntax within MFM.cpp mostly where emigration from the Acorn disk analyser had left a residue of lengthy namespace specification.
2016-12-31 15:25:11 -05:00
Thomas Harte
a8bc9d830e
Removed leftover very temporary debugging aid.
2016-12-28 23:03:05 -05:00
Thomas Harte
e4000bd060
Added some even more verbose logging; slightly simplified write loop logic, and decided it's definitely write_byte that's responsible for CRC generator feeding.
2016-12-28 21:24:19 -05:00
Thomas Harte
4adcb46665
Fixed FM-mode CRC generation.
2016-12-28 19:51:27 -05:00
Thomas Harte
1277a67f9a
Introduced data_mode_ to replace is_reading_data_, representing that there are now three possible modes. When writing, any input from the read head won't affect the CRC generator.
2016-12-28 19:26:21 -05:00
Thomas Harte
7a627b782d
Reintroduced writing of MFM sync marks when writing a sector.
2016-12-28 18:48:50 -05:00
Thomas Harte
a568172758
Made steps towards proper CRC generation. Am currently comparing against Oric disk images, as — amongst other things — they include precomputed CRCs.
2016-12-28 18:29:37 -05:00
Thomas Harte
9c0f622a2e
Started working CRC checking into the 1770. Discovered immediately that my generated CRC does not match that built into the Oric disk images. So mine is pretty-much certainly wrong. An opportunity for learning!
2016-12-26 16:46:26 -05:00
Thomas Harte
0490a47058
Worked on the all-around framework for decoding sectors back from tracks when closing down a file. Hit the wall that the parser is more observant of CRCs than the WD. No, really. So I guess I have to stop avoiding that whole issue.
2016-12-26 14:24:33 -05:00
Thomas Harte
83c433c142
Deviated from the data sheet, which seems likely to be correct. Hence removed a whole load of the temporary logging.
2016-12-26 12:48:49 -05:00
Thomas Harte
742c5df367
With lots of logging arising temporarily, fixed bug whereby conversion to a patched track would lead to holding a track with a distinct measure of time, leading to improperly-placed patches.
2016-12-25 22:00:39 -05:00
Thomas Harte
acc35885cd
Attempted to reduce track invalidations.
2016-12-25 20:38:25 -05:00
Thomas Harte
c0a1264ab0
Slightly improved legibility.
2016-12-25 20:19:47 -05:00
Thomas Harte
e2b829f68e
Made an attempt to write the proper address mark.
2016-12-25 20:15:07 -05:00
Thomas Harte
beaa868079
Factored the MFM parser out into encodings.
2016-12-25 20:00:57 -05:00
Thomas Harte
74e98fd097
Made an attempt to write actual data (albeit that CRC calculation is still missing).
2016-12-25 19:18:45 -05:00
Thomas Harte
98be6ede45
Shuffled a little to reduce risk of overflow, ensured writing is a loop, still seem to be writing too quickly for some reason.
2016-12-25 16:13:05 -05:00
Thomas Harte
d2ad2c756e
Added enough shovelling to write rubbish for an entire sector.
2016-12-25 15:46:49 -05:00
Thomas Harte
aceb7e3b6b
Started implementing write sector on the 1770, immediately deciding it would be useful to have a callback for end-of-queued-data-written from disk controller. So had a go at implementing that, naively. More investigation required.
2016-12-25 12:31:38 -05:00
Thomas Harte
901f19f89c
Added enough stuff that SSDs attached to a 1770 will now reach the entry point for writing.
2016-12-25 09:46:12 -05:00
Thomas Harte
c304db0f5a
Deintegrated the busy flag and the interrupt request line, as the latter is reset by status reads. Which also means I can start reporting the WD INTRQ line status directly from the Microdisc. That appears to be correct, rather than honouring the Microdisc IRQ select there.
2016-12-06 21:16:29 -05:00
Thomas Harte
ca50606e1d
Restored Vic audio.
2016-12-03 17:10:47 -05:00
Thomas Harte
36bc558798
Converted all 'Components' to postfix underscores.
2016-12-03 10:51:09 -05:00
Thomas Harte
81ee834530
As well as a bunch of logging, reinstated rotation position preservation across tracks.
2016-12-02 18:36:47 -05:00
Thomas Harte
93c573bfa9
Implemented missing status bits (other than the index hole), and a head loading delay for the Microdisc.
2016-12-01 21:13:16 -05:00
Thomas Harte
0a0775c3bd
Removed earlier hacky solution.
2016-12-01 20:16:11 -05:00
Thomas Harte
442986ee2c
Introduced a head loading path for 1793 machines.
2016-12-01 20:12:22 -05:00
Thomas Harte
82899f2f47
Ensured flag setting is atomic, removed duplication of interrupt request versus busy, found better names for the personality testers, unified delegate protocol.
2016-12-01 07:41:52 -05:00
Thomas Harte
b31fd11470
Fixed reporting of data request line, initial status values.
2016-11-30 22:39:55 -05:00
Thomas Harte
2222cb65d6
Split the status up into flags, assembled into a register upon demand. Attempted to implement some of the differences between the 1770/1772 and 1773/1793. Albeit with a motor fix still in place.
2016-11-30 22:26:02 -05:00
Thomas Harte
84cb07613d
Checked some documentation more thoroughly; the 1793 has quite different spin-up (/head load) semantics. So it's another distinct personality. Grrr.
2016-11-27 20:39:08 -08:00
Thomas Harte
02ba1f220f
The '72 seems to be a '70 with altered timing. So worth differentiating.
2016-11-27 21:06:17 +08:00
Thomas Harte
2c01f9dbed
Added meaningful TODOs.
2016-11-27 08:42:39 +08:00
Thomas Harte
2f459690d4
It would appear the 1770 and 1773 actually differ in relation to the (non-sensical) ability not to spin-up for a Type 2, and whether a side compare can occur. So the WD1770 class now requires a personality to be specified. Which it singly fails to honour.
2016-11-26 23:29:30 +08:00
Thomas Harte
d8ecc52de8
Temporarily disabled spin-down as harmful to the status register if following anything other than a Type 1 command.
2016-11-26 22:27:20 +08:00
Thomas Harte
b9677c9927
Consolidated interrupt request setting.
2016-11-26 09:41:53 +08:00
Thomas Harte
d5f9e0aa3b
Ensured there's no such thing as a zero-cycle operation, even if i don't yet know exactly what I should be doing.
2016-11-25 21:24:25 +08:00
Thomas Harte
4af678d2ed
Gave the Microdisc a clock signal, added just enough of force interrupt to avoid a spurious belief that a type 3 command has started.
2016-11-25 20:51:39 +08:00
Thomas Harte
d4a1961378
Added getters for the IRQ and DRQ lines plus a delegate to receive changes; adjusted code so that the two lines signal.
2016-11-21 13:21:49 +08:00
Thomas Harte
7eeaac23e7
Reversed myself. I once again do not think the clock is divided by 256 for envelopes.
2016-11-11 20:31:48 -05:00
Thomas Harte
77987bf31e
Decided to go with divide by 256 for the envelope counter after all.
2016-11-09 21:51:56 -05:00
Thomas Harte
77ce200fbb
Simplified/corrected AY tone/noise mixer logic, and made a new guess at the effect of reading registers that are smaller than 8 bits.
2016-11-09 21:21:17 -05:00
Thomas Harte
fa65cc2058
Resolved type conversion error.
2016-11-05 12:57:01 -04:00
Thomas Harte
30c670f8de
Ensured programmatic setting of the timers occurs during phase 2 _instead_ of counting.
2016-11-04 21:30:18 -04:00
Thomas Harte
21604376e6
Reintroduced clocking of the AY and boxed in the range of the master divider a little further.
2016-10-30 22:51:08 -04:00
Thomas Harte
ad00304e8a
Fixed 6522 countdown.
2016-10-28 21:05:42 -04:00
Thomas Harte
4fab794747
Added a direct-to-two-cycles emulation path for 6522 owners.
2016-10-27 21:13:25 -04:00
Thomas Harte
2eda0b3c86
Attempted to simplify the logic behind the most common 6522 usage.
2016-10-27 21:06:31 -04:00
Thomas Harte
fd823dc222
Settled on terminology.
2016-10-23 20:42:49 -04:00
Thomas Harte
b12f2f2796
Switched to more straightforward version of two-step loop, dealing with my mistaken dealing of when _master_divider&15 == 0 upon entry without adding an extra sanity check. Am also temporarily on non-modulo logic for tone generation, for a profiling test.
2016-10-23 20:32:48 -04:00
Thomas Harte
583db88299
Added a dispatch queue-powered Apple implementation of the async task queue, removed any mention of skip_samples in the AY since it isn't implemented.
2016-10-22 21:58:45 -04:00
Thomas Harte
33e628a096
Made an attempt to eliminate what amounts to manual division.
2016-10-21 22:16:44 -04:00
Thomas Harte
46a3c0922f
Slightly simplified code, fixed divider.
2016-10-21 22:12:44 -04:00
Thomas Harte
d7c0c49715
Might as well be consistent with divider loads.
2016-10-21 20:07:14 -04:00
Thomas Harte
782ef960e1
Sought both to [start to] optimise the AY and correct divider reloads. It turns out that conditionals aren't that troubling. But I can probably eliminate the counters.
2016-10-21 20:05:38 -04:00
Thomas Harte
cd59eb5f43
Implemented non-linear volume.
2016-10-19 23:07:51 -04:00
Thomas Harte
b59da7d4bc
Added some documentation.
2016-10-19 22:47:44 -04:00
Thomas Harte
08275c6241
Fixed mixer IO bit usage.
2016-10-19 22:22:15 -04:00
Thomas Harte
59162228ef
Reduced mask for clarity.
2016-10-19 22:14:05 -04:00
Thomas Harte
319d7c2b12
Fixed premature wrapping of the envelope, played about with whether that should be subject to a predivision by 256. It feels unlikely?
2016-10-19 22:12:51 -04:00
Thomas Harte
6073906c39
Commented and fixed mistake that would equate any noise divider > 32768 with 0.
2016-10-19 21:51:35 -04:00
Thomas Harte
101f168ea4
Made an attempt to tidy up.
2016-10-19 21:43:18 -04:00
Thomas Harte
ada37abe23
Made an attempt to implement noise and envelopes. Not quite right yet.
2016-10-19 21:13:22 -04:00
