There seems to be a jz4750 model on the market, next to all the ones with secretive ARM chips. Finally a successor to the old jz4730-based Skytones, so probably without Ethernet.
(Friday, 30 April)
The Ramos W7 (rk2808) IMP is now listed at House of Dap for $236. A very special price indeed.
(Saturday, 20 march)
Still no sub-$100 Android tablets on sale, so let's kill some time.
ZSP G1 opcodes seem to be mostly 4 bit wide. The op16/2 table uses 8 bit for simplicity/speed. The unused bits will be passed to the opcode's handler, but then something is needed to avoid having to enter very similar entries:
?- between(0,15,I),op16(Op,f16t4rk(I)),write(op16(Op,f16t4rk(I))),nl,fail. op16(160, f16t4rk(0)) op16(161, f16t4rk(1)) op16(162, f16t4rk(2)) op16(163, f16t4rk(3)) op16(164, f16t4rk(4)) op16(165, f16t4rk(5)) op16(166, f16t4rk(6)) op16(167, f16t4rk(7)) op16(168, f16t4rk(8)) op16(169, f16t4rk(9)) op16(170, f16t4rk(10)) op16(171, f16t4rk(11)) op16(172, f16t4rk(12)) op16(173, f16t4rk(13)) op16(174, f16t4rk(14)) op16(175, f16t4rk(15)) false.
It's called define in m4, term_expansion/2 in Prolog:
term_expansion(m4k4_s_e,L) :- findall(op16(Op,f16t4rk(I)),(between(0,15,I),Op is 0xa0+I),L).
The full power of the language can be used in the macro body, much like Forth or Scheme and unlike C. Then call the macro:
m4k4_s_e.
This fact is replaced by the 16 shown at compile-time.
(Friday, 12 march)
The ZSP400 (ZSP G1 instruction set, 16-bit instructions) is source-level compatible with the ZSP500 (G2, 16-bit and 32-bit), while the ZSP500 is binary compatible with the ZSP800 (G3). AFAIK the rk2608 has a ZSP400, the rk270[56] a ZSP500 and the rk280[68] a ZSP800, but I haven't verified this yet.
It seems reasonable to assume that many, perhaps most, 16-bit instructions still have the same instruction format, although they might have reassigned opcodes (ld is 0x5x in G2 and 0x7x in G1?).
We should first improve the disassembler before adding new instructions. If we add a table indexed by opcode we no longer have to search for its handler. Of course each opcode could keep its own private handler but if we also identify instruction formats it becomes much easier to add similar instructions.
The dumb but readable approach:
instr16(I) --> ld(I) ; st(I).
ld(ld(r(R1),r(R2))) --> word(0x55,B),{nibbles(B,R1,R2)}.
st(st(r(R1),r(R2))) --> word(0x5d,B),{nibbles(B,R1,R2)}.
The normal approach:
instr16(I) --> word(Op,B),{op16(Op,F),call(F,B,I)}.
op16(0x55,f16rr(ld)).
op16(0x5d,f16rr(st)).
f16rr(Id,B,I) :-
nibbles(B,R1,R2),
I =.. [Id,r(R1),r(R2)].
(Monday, 1 march)
Finding a small loop, as in the ZSP500 test, is fairly easy. Input N is an upper limit on loop size (not counting the branch) in words. Output L is a list of instructions in the loop, including the branch:
br_back(N,L) :- N > 0, imem(E,br(M)), M < 0, -M =< N, B is E+2*M, imem(B,_), %make sure branch target exists findall(PC:I,(imem(PC,I),PC >= B,PC =< E),L).
To allow an unknown 16-bit instruction (skip16) as branch target, the handle/2 filter must be fixed first.
(Saturday, 27 february)
Because my posts were getting deleted after only a cursory glance and without any explanation that makes sense, I'll just try to answer your question here where I don't have to fear losing my work.
Yes, the cc1600 unfortunately doesn't have a MIPS core. I hoped it would have, because the SoC supported USB OTG. I once posted some disassembled ARM code over on mp4brickers to show it was different from jz47xx, to no avail as we have seen.
About ZSP (more interesting than an old ARM core anyway): here is a rudimentary (75 lines) disassembler I wrote. The idea is to use a modern Prolog (finite domain constraints might be useful) to search for small loops (FIR filters, etc) and other patterns to figure out the instruction set. Individual instructions are converted to Prolog facts and constraints would be defined over ranges of these facts.
Anyhow, I got distracted before implementing a FIR recognizer or porting it from swi-prolog to gnu-prolog. The current version can only disassemble the few instructions I found:
$ swi-prolog Welcome to SWI-Prolog (Multi-threaded, 32 bits, Version 5.9.7) Copyright (c) 1990-2009 University of Amsterdam. SWI-Prolog comes with ABSOLUTELY NO WARRANTY. This is free software, and you are welcome to redistribute it under certain conditions. Please visit http://www.swi-prolog.org for details. For help, use ?- help(Topic). or ?- apropos(Word). ?- ['gram.txt']. % gram.txt compiled 0.03 sec, 9,712 bytes true. ?- test. 0 mov(r(1), 256) 4 st(r(0), r(1)) 6 ld(r(2), r(1)) 8 add(r(0), 1) a br(-3) e nop Disassembled 16 out of 16 bytes true. ?- findall(I,(imem(PC,I),PC>=4,PC<8),L). L = [st(r(0), r(1)), ld(r(2), r(1))].
PS this is a new version to fix a problem with tab-stops and to restore the support for "all" instructions in the test case.
(Thursday, 25 february)
Despite using one since Christmas, Prithvi Shenoy has been unable to
It's not clear if the "4hrs of continuous use" is with Wifi on. I assume it is, but that's still not good enough with such a gargantuan battery (5500 mAh). The SoC might be one of those Realtek or Sigma chips used in hard disk media players. It's 500 MHz so that rules out an Ingenic chip (for the time being). Quite a disappointment if you also take into account the meagre amount of RAM.
(Wednesday, 24 february)
If you're in the EU you can get ripped off for 99 euro, while other countries pay $99 for what is now a rather dated (jz4720) electronic dictionary with a tiny keyboard, a replaceable battery and Linux.
(Wednesday, 17 february)
Since my last post atomik_hippo unfortunately got trapped inside the daomingjin-maderin1 reality distortion field. Luckily for him his player was sufficiently idiot-proof and he was able to restore the old firmware.
Let's see what fanciful flight of imagination is needed to explain this failure away:
About that root filesystem. Our ARM9 Linux "creator" daomingjin must have spent some time looking at its files before copying them. Which obvious clues did he miss?
bin/busybox: ELF 32-bit LSB executable, MIPS, MIPS-I version 1 (SYSV), for GNU/Linux 2.4.0, statically linked, stripped
Machine: MIPS R3000
(Saturday, 13 february)
Over at mp4brickers we learn some interesting "facts". Did you know that a jz4740 is the same as a cc1600? Or that the processor in a Cube B53WIFI uses the same instruction set as an ARM9? With his head firmly up his posterior a poster is now trying to entice people into installing Linux for MIPS on their ARM devices.
Of course this is completely insane, not that another confused poster (maderin1) allowed anyone to warn people about it. He directly contradicted the one poster who got it right. Another poster then tried to repeat these dire warnings but was banned for being rude and inhelpful (sic).
So there you have it. Some clueless greenhorn receives the blessing ("it seems it is definitely cc1600 as daomingjin rightly said") from an equally clueless "cant be much help im afraid, all a bit above my head nowadays" moderator.
Nowadays? I'm afraid the penny never dropped.
This disassembled code from B53WIFI firmware (usb_boot.bin) shows perfectly normal MIPS code (nops in branch delay slots, loading a 32 bit constant in register 1)
0: 09000010 b 0x28
4: 00000000 nop
...
28: 3f00013c lui at,0x3f
2c: f0ff2134 ori at,at,0xfff0
30: 20e88102 add sp,s4,at
34: 40009922 addi t9,s4,64
38: 08002003 jr t9
3c: 00000000 nop
Any programmer worth his salt checks his assumptions and doesn't release completely untested crap.
(Thursday, 11 february)
House of DAP now sells the Teclast K3 for $229. As usual their technical information is incorrect. It doesn't have an "Ingenic 360MHz" but an rk2706 and is probably the same as one sold under the Merlin brand, if the presence of a merlin.pdf file is anything to go by.
Now you can buy a paleolithic rk2706 player from tinydeal for about $30, so they make a lot of profit on the E-ink screen. Sure it's more expensive than a simple LCD screen, but not that expensive.
(Tuesday, 9 february)
Some old, but still interesting news. This explains the improvement in audio of more recent Ingenic SoCs, but shatters the illusion that their chips are 100% Chinese-designed by including Dolphin IP. Dolphin is (or was) also used in Rockchip SoCs.
I assume this is going to be the jz4760 audio codec.
(Sunday, 7 february)
According to Linuxfordevices there's an upcoming Android LIMP from Hivision with a Rockchip rk2808, described as an ARM9 processor.
Now the ARM core replaced a Z80
8051 in the rk2608 and that should tell you something.
IF this SoC is satisfactory for Android you can bet it has everything to do with either a certain superscalar DSP (550 MHz in the rk2808, 450 MHz in the rk2806) or a boatload of hardware accelerators to mask the inadequacies of a cheap and slow ARM9 core (or both).
More interesting is the question if the PWS700B has a jz4750 or jz4755.
(Thursday, 4 february)
Most likely another fata morgana: 6", capacitive touch, wifi. It is supposed to be cheaper ($150) than the Teclast K3 ($225). Well, it could easily be since making e-readers is like printing money.
(Friday, 29 january)
Since yesterday the speed of Tradestead's new slugbook has dropped from 600 to 400 MHz. Is it true that VIA does ARM these days? This one has rather similar specs and does have an ARM.
Tomorrow Wince 5.0? Friday 64 MB ROM + 64 MB RAM? Next week the old 266 MHz Anyka ARM SoC? You never know with these guys.
PS I have seen the name of Wondermedia before, but didn't know they were part of VIA. There is a VT8500, WM8505 and a new WM8510 SoC. Can't say I'm interested in anything VIA does because of their secrecy. Ingenic is much better, but I wish they had released their jz4760 by now.
(Wednesday, 27 january)
The old slugbook link doesn't work anymore, so here is a new one.
It's cheaper than the Cherrypal Africa: $89.39 instead of $99 and you will get an x86 (?) VIA instead of the Anyka ARM in the slugbook. The battery capacity is pathetic and it probably has a fan so it's not much of an improvement and it's not clear if a Wifi dongle (internal or external) is included.
But I expect them at least to deliver, in sharp contrast with what I've read about Cherrypal recently.
(Tuesday, 26 january)
Rip-off galore. The most reasonable of the lot is the
Yifang M5 IMP. It's supposedly $128 (MOQ unknown).
But would you buy your IMP from a company that actually sells ugly toy
MP3 players?
(Monday, 25 january)
Yinhu is now shipping Vimicro (3 new models), next to a boatload of jz4725b, while Onda is treading water with rk2705.
Unfortunately I haven't been able to find a shop selling them with the exception of an old model in Singapore.
(Monday, 4 january)
Onda will make an (Android) IMP this year? I remember similar stories from two years ago.
[ ] It's about bloody time [X] Don't hold your breath
(Sunday, 3 january)
Happy new year and let's hope we will see more designs like this one. Surely if there are $99 netbooks with 7" screen plus keyboard then a simple IMP with 4" screen but without keyboard should cost less.
(Friday, 1 january)
I thought they had switched to sochip but here is a new one with a jz4755. The imp3.net advertorial mentions how easy it is to disassemble a unit, which should give pause for thought.
(Thursday, 24 december)
A keyboard with a Taiwanese 1 GHz
x86 chip
(including fpu).
Based on the Emtec laptop with Loongson 2F (4 W) and fan experience this steers dangerously close to FI=IO land. How hot will your fingers get?
(Friday, 18 december)
The $99 Cherrypal Africa is powered with an 400 MHz processor and 256 MB DDR, according to this.
So what's it going to be, the usual 336 MHz jz4730 with 128 MB SDR SDRAM, a 400 MHz Samsung ARM, a PPC or something else? Please note the only 400 MHz Ingenic processor at the moment is the jz4755 and none of them support DDR RAM. In addition, the 256 MB Skytone Alpha-400 (Elonex One-T, etc, etc) never materialized probably because higher-capacity SDR chips are prohibitively expensive.
It seems they want to keep budding sub-Saharan open source programmers in the dark, because later on I read they won't tell you which processor you will get but it might be XBurst = Ingenic. In fact I asked and am still waiting for an answer.
I consider them less than trustworthy.
(Thursday, 17 december)
The Odroid at $349 is the perfect DIY solution. It comes with its own debug board. Just apply electricity.
(Thursday, 10 december)
That's what they call the ARM core inside Teclast's t8100 chip.
I agree, the ARM9 is advanced. In age.
In the same way, a very advanced 130 nm manufacturing process is used for this chip.
(Wednesday, 9 december)
It seems Focalprice has started to sell other brands than Onda. I kind of missed that because of their terrible browser support.
The Benss BX-57 is a 4.3" player with 8 GB flash and sells for $73.35, so it's probably a better deal than the jz4755 players Ownta has on sale. The cheapest ones there are 4.0" Mahdi and Bmorn players with 4 GB for $79.80.
Furthermore, the Bmorn BM-720 looks remarkably similar to the 4.0", 4 GB Benss BX-56 ($56.85), while the same can be said about the UI of the two Benssen.
But there's one big problem: pigeonware. At least for Mahdi I know where to look.
PS Benss BX-57 = Bmorn BM586 or I'll eat a hat.
(Monday, 7 december)
Jiongtang has the Cube
H800 listed for $229 with 0 units in stock. A competitive price it
is not. According to the latest news, you'll get two pieces of hardware
(if you are able to buy it in the first place), one of which is this
awe-inspiring dock:
There was a lot of hype about Dual Linux, but it simply means two different desktop environments: one for mobile use and one for workstation use. You would think, then, that perhaps there's RAM in this 34 pin bus extender monstrosity. After all, 128 MB for a Linux workstation doesn't cut it anymore. Well, maybe there is, but so far I've only read this box contains Ethernet and a serial port and, get this, the other box (now the less mobile part, ironically) has no built-in WiFI. Supposedly you can use its USB Host (I would be surprised if they added circuitry for OTG) to attach a WiFI dongle.
What on Earth have they been smoking?
(Sunday, 6 december)
Just follow this easy recipe:
(Saturday, 5 december)
I have never wanted to use a game console, yet the Korean Odroid does match a large part of my shoppinglist.
Good points:
Bad points:
The Chinese could have produced something like this by combining parts of the Skytone Alpha-400, Dingoo A320 and Cube B53WIFI designs. In fact they could have done all of that TWO YEARS AGO.
Wait for JXD, then, to make a cheap knock-off?
(Saturday, 5 december)
There seems to be an IMP with an ARM-A8, the W1060. It looks better than a SmartQ V7, however you will probably end up with the ARM-11 mentioned here because it's cheaper ($220) than their W1050 ($258). This shop is also unsure/inconsistent about Android. If I asked if it supported NetBSD 6.1 they would no doubt say yes without blinking.
(Friday, 4 december)
In the beginning there was Bluetooth. Now there is HDMI, a second umbilical cord.
I would prefer players to devolve into books with ZigBee rather than VCRs with HDMI.
(Friday, 4 december)
Onda announced a 3" player with HDMI, the vx520, limiting battery size compared to their larger offerings and not solving the problem of having to carry a clumsy HDMI cable.
(Thursday, 3 december)
Trying to find out which chip was used in an $89 Onda vx585hd (seems to be cc1600, unfortunately) I found a dime a dozen seller posting the following specs:
OTC must be some new feature thrown in for free. What happened to OTG?
I don't think the audio chip is the best solution for decoding video.
Why stop at only 4 cores and not count all hardware accelerators? Don't forget the DMA controller and the timers. To us die-hard traditionalists OTOH a functional unit hardly qualifies as an independent SMP core. The 1.6 GHz and comparison with an Intel chip is also absurd. The ARM core in Chinese chips is still single-issue. The real clockspeed must be somewhere between 300 and 600 MHz (probably closer to 300).
Perhaps this player can be had for $75.48 but it's unclear because (in that case) sloppy Focalprice manages to fubar by showing the wrong picture.
Oh yes, House of DAP thinks this player has a Sigmatel 3710 chip (wrong), a 5.3" screen (wrong) and a resolution of 800x480 (wrong).
PS it might be a sc8600, not cc1600, player because it has HDMI.
(Saturday, 28 november)
Almost a whole week went by without a new Cube announcement. Can't have that. Here we go, the C30, because last week's gimmick didn't sell well enough.
It seems EVERY possible combination of features is produced and sold EXCEPT the one I would have even the SLIGHTEST interest in.
Why can't I buy a customized Cube? Here's my shoppinglist:
It worked for Dell.
(Friday, 27 november)
The Ramos T17 for $85 sets, for now, the lowest upper limit on price for an rk2728 player we might be able to buy in a month or two from Tinydeal.com for much less.
(Wednesday, 25 november)
Everyone seems to be jumping on the Telechips tcc8901 bandwagon because it can play 1080p video. It now seems one of the first players with this chip, the Chuwi P7, can only play 2 hours of HD video and 5 hours of music on a 1700 mAh battery.
(Monday, 23 november)
With Christmas approaching, Ownta has announced Worldwide Shipping. It would have been a nice gesture if only they hadn't increased prices by up to 13 dollar at the same time. For instance, the Mahdi T951 was $86.57 and the T552 $67.17. Now they are $99.20 and $79.80. Of course these low-quality players are not worth more than let's say $47.90 (the underutilized jz4755 chip is not at fault here, rather the miserable Skinner-box around it plus accompanying pigeonware).
(Saturday, 21 november)
Today was Cube's Big Day, hyped for all it was worth this last week.
Finally an interesting follow-up to the B53WIFI?
Nope, Cube has discovered a new gimmick to exploit (earphones) and has
introduced yet another ADHD MP4 player, the H200HD
H200HDS H200HDSC H200HDH.
Talking about earphones reminds me of the censuring that's going on in Little Kim's worker's paradise:
I really don't think it's okay to delete a thread just like that. I merely described a common issue with the Betas to provide an information resource for people with this problem. So why did you delete it?
Very disappointing. If there's a problem with your products maby YOU should SOLVE it and not delete topics about this very problem!
(Friday, 20 november)
On jiongtang.com there's a fascinating product description:
While mp4nation.net's top banner freeship.jpg states "free shipping worldwide", below it we find this gem:
Slightly rephrased this says they ship anywhere in the world but at cost. Wait, what?
(Tuesday, 17 november)
The firmware for each MP4 player is abandoned almost as soon as the new, "improved" one is announced (10M, USB OTG, 720P, HDMI, 1080P, etc), so it remains perpetually bug-ridden. There are very few alternatives (Rockbox/Linux) and they trail far behind or require much more RAM than the current 16 MB standard.
What to select for the lowest level if you want to avoid the well-trodden (C/C++) path? DSSP, Erlang, Flat Concurrent Prolog, Gobol? The latter one looks somewhat interesting for practical reasons (fast compilation, BSD-style license), but the footprint doesn't look good at the moment. Maybe native code isn't a good idea (compared to let's say Smalltalk-style dynamic translation). The "concurrency-oriented" systems I mentioned all use some form of bytecode or threaded code. The same for Mozart. The one exception (compiling down to C) is KLIC, FCP's sibling. It has the same problem.
(Wednesday, 11 november)
Most of the Chinese webshops we are forced to do business with either don't know what they are selling or want to confuse us, in Sun-Tzu tradition. For instance, I have been told, after inquiring about the specific SoC in some MP4 player, that it had a Samsung chip. Points for effort. Sure enough, almost every one of them has a Samsung flash memory chip.
Another example is in yesterday's link: there is no XBurst 266 MHz processor. This is presumably a leftover from last year when so-called "XBurst 400 MHz" processors were being sold to unsuspecting customers thinking they were getting a great deal on an Asus EEEPC clone. Of course there were no 400 MHz Ingenic processors until very recently and the jz4730 was only a 336 MHz chip.
Yet another example is a shop selling the rk2706 as a 400 mHz chip. Even after correcting the obvious error (an ancient 4 MHz Z80 would be faster) it's still nonsense.
Unfortunately, the problem is more prevalent when shopping for the cheaper, generic brands. With a brand such as Onda you can be reasonably sure what you're getting (if you've done your homework), although you still have to be careful when there are many similar models at one time and the merchants are trying to get rid of the old ones (Ramos T11, T11RK, T11TE, etc).
(Tuesday, 10 november)
It will blow these 266 MHz ARM/128 MB/Wince slugbooks for $121.06 out of the water.
Make it so.
(Monday, 9 november)
It has been discovered that Sochip (sc8600 and its variants) has an ARM9 core. Also, a recent game console, Benss BX-96, uses the same core if you look at its firmware. It also uses eCos, just as JXD game consoles do, pointing indirectly to Sunplus 8000, one of the first Chinese chips to use an ARM9 (but at a rather low 300 MHz clockspeed), soon followed by the cc1600.
ARM7EJ-S is still being used in newer Rockchips such as the rk2705, a faster rk2706, while only recently (rk28xx, rk2728) they began to use ARM9.
OTOH, higher-end ARM11 cores are found only in Korean chips (Samsung, Telechips). Even the new Vimicro VC0831 appears to be only ARM9.
(Saturday, 7 november)
If you've seen my posts at mp4nation.net you know I have a soft spot for Ingenic's MIPS32 processors. The Cube H800 (jz4750, Wifi, open Linux) showed some promise, but the latest news is it will only be sold to a select few for an outrageous amount of money.
The first 7" SmartQ was 1200 yuan and they want 1299 (at least $190) for this thing, so they are basically saying "get a SmartQ V5 instead", $180 here. The V5 has twice the amount of memory (256 MB RAM). In addition it's DDR2 RAM while Ingenic is still stuck at slow SDR RAM, together with its lack of USB OTG a severe drawback.
In other words, the H800 is going to get killed without fanfare.
(Thursday, 5 november)
