Experienced Forum User, Published Author, Former player
Joined: 2/19/2007
Posts: 424
Location: UK
I have also looked a bit at the rom, finding the definition of the first-generation characters, the classes, bloodlines and love growth values. Most of this is already documented at gamefaqs, though, but I'll put it here in the hope that it will be useful to somebody. It is in the form of a few short bash scripts:
http://folk.uio.no/sigurdkn/fe4_bloodhttp://folk.uio.no/sigurdkn/fe4_classhttp://folk.uio.no/sigurdkn/fe4_charshttp://folk.uio.no/sigurdkn/fe4_affec
These are mostly wrappers around hexdump, which extracts the relevant subset of the rom and formats it as a table. Running each of these on the rom produces:
fe4_blood
Most of the above information can be found in a more convenient format (but without any memory or rom addresses, despite the author clearly having disassembled the rom quite thoroughly) in this gamefaqs article.
Experienced Forum User, Published Author, Former player
Joined: 2/19/2007
Posts: 424
Location: UK
The other script is more directly useful during normal play. It displays the relationship status for the first-generation characters when the Y-button is pressed:
Language: lua
emu = snes9x
function sword(x) if x >= 0x8000 then return x-0x10000 else return x end end
nw, nm = 9, 15
wname = { "brg", "til", "fur", "syl", "lch", "dia", "edn", "ayr", "eth" }
mname = { "sig", "noi", "alc", "adn", "fin", "cua", "mid", "lev", "hol", "azl", "jam", "cld", "beo", "lex", "deu" }
function get_love()
local love_skip = 3
local love_tab = 0x7e3be5
local love = {}
for i = 1, nw do love[i] = {} end
for w = 0, nw-1 do
for m = 0, nm-1 do
love[w+1][m+1] = sword(memory.readword(love_tab+2*(w*(nm+love_skip)+m)))
end
end
return love
end
function love_color(score)
if score >= 500 then return 0x00ff00ff
elseif score > 400 then return 0xff8080
else return 0xffffff00 end
end
-- Display love table, hiding people who are not present
-- (i.e. those with 0 for all scores)
function show_love(love)
local width, height = 16, 7
local bsub = 4
-- Check which are present of each sex, and which are already
-- in love
local mok, wok, mmax, wmax = {}, {}, {}, {}
for w = 1, nw do wok[w] = 0 wmax[w] = 0 end
for m = 1, nm do mok[m] = 0 mmax[m] = 0 end
for w = 1, nw do
for m = 1, nm do
if love[w][m] ~= 0 then
mok[m] = mok[m] + 1
wok[w] = wok[w] + 1
if(love[w][m] > mmax[m]) then mmax[m] = love[w][m] end
if(love[w][m] > wmax[w]) then wmax[w] = love[w][m] end
end
end
end
wi = 0
for w = nw, 1, -1 do
if wok[w] > 0 then
wi = wi + 1
gui.fillbox(width*wi-2,0,width*(wi+1)-bsub,height,love_color(wmax[w]))
gui.text(width*wi,0,string.format("%3s",wname[w]))
mi = 0
for m = 1, nm do
if mok[m] > 0 then
mi = mi + 1
gui.fillbox(width*wi-2,height*mi,width*(wi+1)-bsub,height*(mi+1),love_color(love[w][m]))
gui.text(width*wi,height*mi,string.format("%3d",love[w][m]))
if wi==1 then
gui.fillbox(-2,height*mi,width-bsub,height*(mi+1),love_color(mmax[m]))
gui.text(0,height*mi,string.format("%3s",mname[m]))
end
end
end
end
end
end
while true do
local keys = joypad.get(1)
if keys.Y then
show_love(get_love())
end
emu.frameadvance()
end
Most of this code is simply drawing to the screen. The actual work is done in get_love().
Experienced Forum User, Published Author, Former player
Joined: 2/19/2007
Posts: 424
Location: UK
I've made a couple of scripts for Fire Emblem 4 that might be useful to people. The first script contains an implementation of the game's RNG, and uses it to predict the next 16 values:
Language: lua
emu = snes9x
function get_rng()
return {
step = memory.readbyte(0x7e04a2),
table = memory.readbyterange(0x7e046b,0x37)
}
end
function rand(rng)
rng.step = rng.step + 1
if rng.step >= 0x37 then
for i = 0, 0x36 do
local j if i < 0x18 then j = i+0x1f else j = i-0x18 end
rng.table[i+1] = rng.table[i+1] - rng.table[j+1]
if rng.table[i+1] < 0 then rng.table[i+1] = rng.table[i+1] + 100 end
end
rng.step = 0
end
local res = rng.table[rng.step+1]-1
if res < 0 then res = 99 end
return res
end
while true do
local rng = get_rng()
for i = 1, 0x10 do
gui.text(i*12, 0,string.format("%02x",rng.step))
gui.text(i*12,10,string.format("%2u",rand(rng)))
end
emu.frameadvance()
end
The forum won't let me paste another code block in the same post due to a bug, so the other script will come in the next post.
Edit: The reason for the code block problem was just me forgetting to disable HTML in the post.
I should note that though the above RNG is simple, the actual code in the game is quite a bit scarier, with code paths for much more shuffling of the random table. However, this code never kicks in - the function that should randomly enable this extra shuffling depends solely on ROM values, and therefore always returns the same result, which leads to the extra shuffling always being skipped. If you want to see it for yourself, the RNG is at 80A73A, and a commented disassembly of it can be seen here.
Experienced Forum User, Published Author, Former player
Joined: 2/19/2007
Posts: 424
Location: UK
Have you looked at the current TAS (http://www.youtube.com/watch?v=6sklnNBZvyc)? After a by-eye comparison, it seems like you lose about 10 seconds just between exiting your house and entering the second floor of Hyrule castle, and a similar amount from there to the Triforce room. This seems to be mostly due to an inefficient way of walking, but also due to general sloppiness.
I was going to point you towards our link to the past tricks page to help you improve, but it seems we don't have a resource page for this game.
PS. You should upload your videos to microstorage rather than mediafire.
Experienced Forum User, Published Author, Former player
Joined: 2/19/2007
Posts: 424
Location: UK
YaLTeR wrote:
How do I make my gui.drawRectangle function paint a rectangle that will stay while emulation is paused (like in FCEUX or like gui.text)?
I am not familiar with Bizhawk, but if it works like other lua-enabled emualtors, then a function passed to gui.register() will be called every time the screen needs to be redrawn, whether paused or not. If so, you can get a rectangle to stay by doing:
Experienced Forum User, Published Author, Former player
Joined: 2/19/2007
Posts: 424
Location: UK
feos wrote:
P.JBoy wrote:
Here's a much simpler script
Language: lua
while true do
if memory.readbyte(0x03000BDE) ~= 0001 or memory.readbyte(0x03000BE0) ~= 0002 then
avi.pause()
else
avi.resume()
end
vba.frameadvance()
end
Oh god this was really obvious! We don't need a frame perfect video after all! The Guide shall have this script for all cutsceneless GBA encodes. I also guess some other emuators allow that.
I use this for super metroid in snes9x:
Language: lua
prev = -1
while true do
igframe = memory.readword(0x7e09da)
if igframe == prev then emu.speedmode("maximum") else emu.speedmode("normal") end
prev = igframe
emu.frameadvance()
end
This is similar to the above, but has the advantage that the same script can be used for skipping cutscenes while playing, not just while dumping. However, I think this method may not work with AVI dumping on windows with snes9x due to a bug which causes the audio dumping to ignore the speedmode.
I have also implemented support for emu.speedmode in my copy of VBA (for the linux version), which is what I used to create the cutsceneless encode. For that, I used this lua script:
Language: lua
pigf = -1
while true do
local igf = memory.readbyte(0x0300013B)
if igf ~= pigf then emu.speedmode("normal") else emu.speedmode("maximum") end
pigf = igf
emu.frameadvance()
end
Experienced Forum User, Published Author, Former player
Joined: 2/19/2007
Posts: 424
Location: UK
Mister wrote:
amaurea wrote:
Mister wrote:
Looking through two movies (and ISM's), I came to the conclusion that Panga did actually save a frame in entering the pipe over ISM, despite his loss from lag. So all I did is to imitate Panga's pipe entry.
Caution from me: Do not overestimate the comparison lua script. It wouldn't always provide informations we need. We sometimes have to run two (or three) emulators and compare two movies by our eyes with detailed data.
Ouch, it sounds like you've been stuck for a long while debugging a problem that turned out to be a bug in my script. Sorry about that. This may be fixable, though, if you can provide a few examples where it happens. I tried to make it frame-accurate, but my level of testing was far less than the level of scrutiny you have been giving it.
Uhh sorry, I didn't mean that. To tell the truth I haven't used the comparison script but I've seen a few times they get stuck because of some problems. I'm not sure what kind of problems make them lose their ways. I would guess it's partially because the script doesn't exactly tell us a subtle difference, coming from frame-by-frame detailed data, such as ghosts' subpositions, speeds, momentum, lag frames, etc, which we still need if we want to find out the reason of such a difference. Of course I know that the screen is too small to exhibit all those data at one time.
Hm, ok. So for the the sub position and extra information part, that would be easy to display on screen, as it is all available internally in the script, as well as the velocity of each mario. So if you are willing to sacrifice some screen real estate, that can easily be added. Information about lag frames for each ghost is also indirectly available, since the ghosts keep track of an in-game frame counter which stops during lag frames.
For the getting stuck part, I'm not sure what could be causing that, so I think I'd need an example to be able to debug it.
I also talked to bahamete on IRC on a related topic which may be of interest:
IRC wrote:
22:18 <bahamete> I think the improvement may have been noticed if a lag frame would change the overall difference
22:18 <bahamete> rather than the real-time one
22:18 <amaurea> Let me see if I get this: On an autoscroll level (where the continuous offset detection does not work), a frame was gained, but the appearance of a lag frame conteracted that, leaving 0 frames total?
22:19 <bahamete> Yup, which is why I was so confused at a change of frame rule
22:19 <amaurea> What happens in the same situation with ingame-timing?
22:20 <amaurea> That should ingnore lag frames, I think
22:20 <amaurea> I haven't tested that myself, though
22:20 <bahamete> what do you mean?
22:21 <amaurea> the script has different timing modes, right? For example realtime and ingame
22:21 <amaurea> Realtime considers all frames, ingame only those where an in-game frame counter ticks
22:22 <bahamete> ugh, I overlooked trying that .__.
22:22 <amaurea> I'm guessing that that the in-game counter does not tick during lag frames. If so, using that mode will only see non-lag savings
22:26 <bahamete> Yes, setting "ingame" mode sees the issue.
22:26 <bahamete> I'll remember that next time.
22:26 <amaurea> Glad to be of help :)
Runs like these are sometimes called 'half-assisted', since they use tools, but only to cover the most obvious mistakes, rather than to aim for perfection. That would make this run a 'half-assisted playthrough'.
Experienced Forum User, Published Author, Former player
Joined: 2/19/2007
Posts: 424
Location: UK
Mister wrote:
Looking through two movies (and ISM's), I came to the conclusion that Panga did actually save a frame in entering the pipe over ISM, despite his loss from lag. So all I did is to imitate Panga's pipe entry.
Caution from me: Do not overestimate the comparison lua script. It wouldn't always provide informations we need. We sometimes have to run two (or three) emulators and compare two movies by our eyes with detailed data.
Ouch, it sounds like you've been stuck for a long while debugging a problem that turned out to be a bug in my script. Sorry about that. This may be fixable, though, if you can provide a few examples where it happens. I tried to make it frame-accurate, but my level of testing was far less than the level of scrutiny you have been giving it.
Experienced Forum User, Published Author, Former player
Joined: 2/19/2007
Posts: 424
Location: UK
I tried to rescale the video framerate slightly to improve the A/V sync, and it got quite a bit better, though it still isn't perfect (and won't be until somebody does a new dump with the correct version of Dolphin). Still, this version should have a desync of ± half a second or so, instead of almost 10 seconds or so as the original had.
Link to video
Ps. I haven't had time to check the sync of the actual youtube upload yet, but the file I had pre-upload had the sort of sync described above.
Experienced Forum User, Published Author, Former player
Joined: 2/19/2007
Posts: 424
Location: UK
I voted yes.
The game is played blisteringly fast
The puzzles are solved in surprising ways totally different from those intended
Many of the solutions are possible only due to the high speed used here
The author spends downtime well, providing entertainment with the cursor
The game is fun to play, and this TAS is fun to watch
Isn't this exactly what we look for in a TAS?
I am a bit dismayed at all the no votes here. Most are from people who have not played the game, and I understand if the action here is too quick for them to see all the glitching, breakage and smart solutions going on here. But the primary audience of a TAS is people who have played the game, and not publishing this TAS will be a disservice to them, as they will be stuck wondering what an optimized completion of this game looks like. And remember, the question is "Should this movie be published?", not "Should this movie be starred?".
Ps. A better encode would be nice, though: By the end of the current one, the A/V desync was 7.5 seconds or so, which is very irritating. Perhaps a simple timecode rescaling using mkvmerge will work?
Experienced Forum User, Published Author, Former player
Joined: 2/19/2007
Posts: 424
Location: UK
Here is a cutsceneless encode:
Link to video
To be precise, this encode shows only in-game frames which are the ones BioSpark aimed to minimize. This leaves out all the dialogue, pause screens and door transitions (though they are relatively fast in this game), but leaves in the item fanfares.
Edit: Apparently the html5 version of this video is broken, showing a length of 0 seconds. The flash version works, though.
Edit2: Replaced the video with a better version. This one should have the correct scaling and non-horrible sound.
Experienced Forum User, Published Author, Former player
Joined: 2/19/2007
Posts: 424
Location: UK
I have ported the comparison script to Yoshi's Island. The script can be found here: http://folk.uio.no/sigurdkn/yicomp.zip
It works much like the previous scripts (Super Metroid, Mega Man X, Super Mario Bros, Super Mario World): First run yirecord.lua while playing back each .smv file you want to compare. Then edit the list of "ghost" files to use at the top of yiplayer.lua (you can also change some simple settings up here, like what information to display). After that, simply play back the .smv file you want to compare the others to while running yiplayer.lua.
A comparison made using this script will eventually be available below:
Link to video
Experienced Forum User, Published Author, Former player
Joined: 2/19/2007
Posts: 424
Location: UK
Due to some encode problems this was a bit delayed, but here is a comparison between this run and its two predecessors:
Link to video
I just watched it now: Yes vote, of course. I especially liked maridia. Now, if only Saturn would submit super metroid redesign too...
Experienced Forum User, Published Author, Former player
Joined: 2/19/2007
Posts: 424
Location: UK
Nice to see that somebody is working on this again. There was some talk earlier about not pickup up the piercing weapon upgrade. I think it would be best to use this upgrade - not only is it generally better, it also looks much better.
Experienced Forum User, Published Author, Former player
Joined: 2/19/2007
Posts: 424
Location: UK
HttpFox shows GET's to baidu.tasvideos.org which receive 0 bytes, with a status of (Error). They are the only https-connections, so the reason is probably that I don't have the site's certificate installed (but I haven't been asked to install it either).
The angular momentum of a star is an important astrophysical quantity related to its internal structure, formation and evolution. On average, helioseismology yields S = 1.92 10^41 kg m^2 s^-1 for the angular momentum of the Sun.
Experienced Forum User, Published Author, Former player
Joined: 2/19/2007
Posts: 424
Location: UK
Nach wrote:
Any chance you're getting some kind of security warnings, or you have some kind of blocker installed?
I use adblock plus too, but turning it off does not change anything. I do not get any obvious warnings, but looking in the error console, there is a bunch of warnings about unbalanced trees being written using document.write. But those are not the kind of security warnings you are thinking about, I think.
Experienced Forum User, Published Author, Former player
Joined: 2/19/2007
Posts: 424
Location: UK
Does not work here in firefox. Same symptoms as I had for nicovideo.
Edit: But if I follow the link baidu itself works (though it stops to buffer a lot, and has choppy playback even when not buffering).
