[VM][FMTOWNS][MEMORY] Temporally revert value around NMI.

[csp-qt/common_source_project-fm7.git] / doc / ABOUT_eFM7_1.en.txt

2016 New Year's Anniversary & eFM-7 Transplant One Year Anniversary
"eFM-7 and everything else."
Jan 1,2016 K.Ohta <whatisthis.sowhat _at_ gmail.com>

* Sorry, this document is written in Japanese only,
I can't use english well, if you interest of this
and you can translate from Japanese to any languages,
please translate this.
This document is released with GFDL, you can copy
and redistribute this document under GFDL.
This document is written with UTF-8.
This document is in line with the GNU Free Document License (GFDL).
You can copy and redistribute.
As I am not good at English, I will write only in Japanese. Sorry (´・ω・｀)
If there is anyone who can translate from Japanese to other languages, please do so.
Yes m(_ _)m


1) What is this (literal translation)
The Common Source Code Project (CSP) is a pre-stage to eFM-7.
To Qt + OpenGL on December 30th last year.
So, maybe in about 30 years, it might be worth something historically.
I thought that I don't know. (^ _ ^ ; I will summarize various things until now.
I thought so.

2. Prehistory : XM7/SDL
There has been a very good FM-7 emulator [1] called XM7.
However, this only worked on Windows.
Whether it was 2008 or 09, this was the South Korean GP2X [2]
I want to play Xanadu on a portable game console. The feeling of [3]
I asked PI. who made the base with XM7 creator Takegami for permission.
Although Mr. PI readily accepted it, I wonder if Mr. Takegami can really do it.
At first, it was not so good, but SDL and Gtk
When I show you the code and screenshots of the temporary port
It was relatively easy to get permission, so we started full-scale work.

[1] http://retropc.net/ryu/xm7/ Development completed at present
[2] https://ja.wikipedia.org/wiki/GP2X
[3] In the end, I gave up porting to the GP2X series due to the problem of the machine spec.

In the beginning, Mr. Gimons ported it to Gtk around 05.
So we started by porting the then latest version of XM7 to Gtk and SDL.
On top of that, just because Gtk doesn't work with GP2X, you can't run it on SDL.
I looked for a GUI toolkit that worked with, and I arrived at something called Agar [4] and adopted it.
That's why.
[4] http://www.libagar.org/

At that time, there were a lot of illnesses, and I was on the verge of falling asleep, so I didn't have a keyboard.
There are not a few days to hit from a futon, and if you think about it now, even if you die of disease
I was in a perfectly normal state, and when I was sick and unable to move my body
At the beginning of 2010, there was something that was reasonably practical.
I'm here.
<B> except that Agar is full of bugs </B>.
Then, after eliminating the bugs one by one and seriously considering load balancing along the way
Offloading from multithreading [5] to OpenCL around 2011.
On top of that, development became sluggish.

[5] This is also leveraged by eFM-7 and CSP/Qt. OpenCL is
I don't think I will do it in the future because the performance doesn't come out.

The original family (Taegami version XM7) has almost completed it.
However, after the Great East Japan Earthquake, I realized the importance of radio.
I have decided to focus on the project called Radio [6] [7].

[6] https://osdn.jp/projects/openi2cradio/
[7] http://sky.geocities.jp/artanejp/

Until the end of 2014, these two projects
Open I2C Radio will be nearly stable by mid- 2013 (I think it was)
The next project is a USB port for 1-segment reception using RTL2832U.
I have done a project (listed in [7]) to convert K to receive short waves and Nakanami.
I was doing it.

3. About XM7
With EM7, the prototype of which was originally created when C++ didn't have much performance
The GUI part of C++, the FM sound source emulator, and the main body of C language.
It's very different, but it does have performance and accuracy.
It's better than my eFM7. On top of that, I used a C++ centric UI and SDL
Add a renderer, and it becomes like a stack of roof houses.
In the middle of 2014, there was a desire to remake it (Sange).
On top of that, there is an emulator collector called Anna Wu from Germany (he said).
Since about the middle of 2014, Mr. Takegami and his wife, Mr. Takegami
But in the end, Mr. Takegami and his surroundings were able to communicate in English.
But he didn't like it and had a trouble.

So I watched the process on Twitter, and I got tired of it, I said
He's a terrible English speaker, but originally a bug in Debian GNU/Linux [8], etc.
We throw reports, and when we need them, Discussions about development in the English community
It is also a person who has become a normal person to do so, so the process is amazing.
I felt something was wrong, and thought, "This is not where I should be."
He also complained to his old friend, and he was working for a foreign manufacturer.
(I can't name him and maybe he's not famous, but the technology is domestic.
I think it's top-of-the-line. complaints … "Well, it's good to do it in English.
He said "ATA-RI-MAE-YAN (^^)" and pressed his back.

It was December 28, 2014.
In the end, we turned the rudder to put the FM-7 series on the CSP and git on December 30th, 2014.
We built a repository [9] and put in it the latest CSP source code of the day.
SWISH

[8] Strangely, on 28 December 2015, Mr. Ian Murdok was beaten to death by the police.
He died at the age of 42.
[9] https://github.com/Artanejp/common_source_project-fm7

4. Start of CSP transplantation

For porting, I initially used Agar as a toolkit.
<B> It will work for now, because it's easy to do verification work </B>.
So we started with the X1 Turbo and the PC-8801MA. [10] Then
Agar has a lot of bugs, portability, and other problems, so it's more general.
We decided to use the toolkit. [11]

[10] There was a lot of software to verify the operation (^_^;
[11] I've given up running it on GP2X (^_^;

So, what should I use? I thought, but Gtk has become 3 and I can get more and more.
In addition to the need for it, Gnome 3 has a lot of bugs and it didn't make a good impression.
So <B> Well, I've never used it, but let's make it Qt. </B>
I looked into it, and I didn't have voice and joystick as standard.
So let's leave it to SDL. OpenGL-required around the drawing. There is no mercy.
That's why we made a rough policy and transplanted it.
This work itself is very dirty and inconsistent code that uses Qt groping.
By March 2015, most VMs will be available.
I got it.
If you look at the git log
・ On January 8th, X1 on Agar was almost passed by the compiler and roughly operated.
(commit fc894319809e05692fd447fda5f232a46ebdf17c7 )
・ On January 10th, we decided to switch to Qt by raising the sound of the failure of Agar.
(commit 98ac2645ec6a3798072445ac003f08d529d6065c )
On January 11th, the compiler went through in Qt.
(commit 1033cc86e36b8e1d38771ef404e7d43ad07ebb39 )
・ On January 12th, the drawing system and the basics of the GUI were completed and the VM worked? (but there are bugs)
(commit e5887aa7464c7276eef9d2578ea880988f3da60d )
・ On January 13th, the keyboard system was established and the drawing system bug was removed once.
(commit 6deec5abac027eca1a71fa89e485650d3c9735da )
Start on December 30 December 30th, and then Agar for the next 10 days.

The design was done, and within a week the basic parts of what we have were built.
That's why.
Until January 28th, we will be running a number of VMs.
Go [12] and from there you can make your existing VMs work and implement FM-7 in earnest.
[13]

[12] commit ce574213769b2685ead0df84c81f193e519d3906 or later
[13] We tried to implement FM-7 on December 31st, 2014.
(after commit e409be1add5f6b9ddf76bada7c24469cb35ae2bc)
I've had a lot to do, so apply Agar and Qt to the home first.
But.

5.eFM-7
Now, if we come to this point, it is the transplantation of FM-7. In the 80 s, he was protected from his parents' house.
Oh! FM magazine and FM-7 I/O map scattered on the web and VM code of the original CSP.
I took the content of the Tips-like hack in the published XM7 source code and made it my own.
I rebuilt the code (I don't want to infringe copyright …)
We implemented FM-7 from February to March 2015.
And even if it doesn't move on March 11th (this is the fourth anniversary of the Great East Japan Earthquake!)
The link goes through [14], 6809 MPU description error [15], disk. cpp/mb8877. cpp
At the end of March, a number of FM-7 sofs are available, fixing the corresponding part of [16].
[17]

[14] commit 65fe797cd5a681ba25c14a88566d7c80d0a8cd9a
[15] This is due to the old version of the underlying MAME code.
Of course, miscellaneous undefined instructions were not implemented, and flag behavior is also a problem.
Here it is.
The correction work around here was dragged until November 2015. Kanagawa prefectural government
I went to Kawasaki Library and obtained a lot of materials (which were destroyed at my parents' house).
We repeated the correction including the part other than MC6809.
I think I can say that the current mc0. 6809. cpp is a completed version. It is dirty (^ _ ^ ;
[16] MB8877 converts F5, F6, and F7 in the data string at the time of formatting into commands to the FDC.
Since it has been allocated, if you don't do something special, it is difficult to write sector numbers F5 to F7.
I can't do it. In the' 80 s, computers used this software for copy protection
At that time, the character code of F5 to F7 was "hour, minute and second".
This kind of special format was called "hour, minute and second format".
Check the hour, minute and second code written in the sector and gap of this "hour, minute and second" ID
Or important programs to prevent copying = protect.
However, many copy tools (such as CopyBoy7) require only a user's disk drive.
It is now possible to reproduce the min-sec format.
We put special data in a sector, and read the sector to see if the data changes
A set of protection that checks "right" and time, minute and second protection.
It has come to be used.
[17] commit 3c875fb911e484c15bfdad2f67dadb9bdedd5e75

The next target was the FM77AV.
It contains proprietary devices such as MMU and ALU.
The basic command strings for ALU were inspired by XM7.
The biggest reason for the dispute between Anna and Takegami is that Mess is inside the devices around here.
Despite a protest through Anna over the full ripping of the description from XM7, MESS
That there was no response from the community at all.
So I thought I wrote this part more carefully than the other parts. [18]
At the moment, it works as designed, and it may not be able to hack fine ALU.
(like drawing Siobredo).
There are some problems with the timing calculation, so I am looking for an opportunity to fix it.

[18] Internal structure of ALU, written using C division operation with linear interpolation function
Actually, that is the reason.

Including around here, 2 channel's "old PC" board, PC emulator comprehensive wear.
I often throw binaries and ask all the Sleazy residents for bug reports and improvements.
I think it started around here. At that time it didn't work on Windows.
(you'll be able to move it later).
This work has been difficult in various ways, so that most of the software will work without any discrepancy.
It was around June 16th. It took three months after all. [19]

[19] commit 1b4320ef6229e75013478fe1ec66ed1eb3c47bd5

Next, I made the FM-77. It went without so much pain. However
It is by giving up the 400 line part of FM-77L4 which does not have the actual product.
However, the boot sequence and memory map are slightly different between FM-7 and 77 av.
There were things that I noticed later when I almost missed it. especially around boot RAM.

Now, it is inevitable to implement the DMAC of FM77AV40EX and so on.
For a reason. if the number of running software increases, various bugs will come out and need to adjust routines a lot.
It will be gone, so it took several months even here for some reasons.
Many of the results of this period were fed back into the work of Mr. Takeda, the original author of the CSP.
Thank you very much, Takeda-san m(_ _)m

… So on September 14th, the AV40EX was almost OK [20], and from there Windows application
And port all VMs.
As for FM-8, I have done a fraud, but I am thinking of doing it.

[20] commit bbf5ed7132304d6e4c19da3cb2d634923eb55a5b

And in less than a month, we built a cross-development environment for Windows using MinGW.
The Win32 version is also included in the binary package as much as possible to absorb the differences in the API of the library and the API of the library.
While having a discussion with Takeda-san through 2channel
Translation.

5) Finally
Both CSP and Qt adaptation are still evolving. Some suggestions from my side, Takeda
There are also the ones that incorporate the suggestions of Ms. and the ones that incorporate the suggestions of the Sure-tenant of 2channel.
I think it's really nice to be able to do something kind of thing which is the best part of FOSS.
As for the FM-7 series, it cannot be denied that the verification of the AV series is especially lenient, so I have been doing various things.
There will be a lot of trouble, but, well, I will continue to do it at my own pace.
Thank you everyone including Takeda-san and users. m(_ _)m

2016 1.1 K.Ohta <whatisthis.sowhat _at_ gmail.com>

* Postscript :

6. Bonus 1 : About the "1/60 second problem".
The emulator has a "1/60 second problem".
The difference in vertical retrace speed between the actual display and the emulated display.
At the very least, the actual display will be out of sync.
How do you deal with a gap of 1/60 second per frame?
If you don't, there will be a lot of discussion.
There are emulators that sell the deviation within 1/60 seconds.
My position is, "It can't be helped even if you miss 2 frames."
What this means is that the one frame delay comes from the difference between the machines mentioned above.
The delay of one frame is due to the fact that Direct X or OpenGL is in between.
The structure is commonly provided on the OS side of the host machine.
These areas can be difficult to touch directly from the application.
In other words, at worst, the shift of about two frames (a little more when the vertical blanking frequency is significantly different)
Vertical blanking of the display Let's consider that it cannot be avoided as long as the frequency is different.
It's not something a normal application can do to go that far, and it's relaxed by setting on the OS side.
It is possible in some cases.
In the end, the CSP should synchronize itself with the VBLANK interrupt provided by the OS.
I personally think that I don't feel like thinking about such a thing. There's a lot of CSP
Try it for those of you who want to customize the base (maybe GL, Direct X or Vulkan pretty much).
The code will change …). That's why GPLv2 is there. and so on.
(^_^;


Revision History :
2022/06/22 FOSS License Clear, National Institute of Information and Communications
I retranslated it with "Minna no Machine Translation@TexTra®" ( https://mt-auto-minhon-mlt.ucri.jgn-x.jp ).
This still haven't finish yet.

2016/01/05 Based on what you pointed out in the 2 channel emulator general thread
Obtain permission Revised part of the description of the facts. Added paragraph 6 after that.

In response to the indication about the expression of hour, minute and second protection / format on the board in 2016-01-07 ↑
Postscript corrected (scary)

09/15/2016 Added Google translated Japanese sentence as ABOUT_eFM7_1.en.txt.
W E   H A V E N ' T   F I N I S H E D   Y E T .