Contents | < Browse | Browse >
%% European Outlook by Jesper Juul %%
%% (firstname.lastname@example.org) %%
Our future with HP-PA Risc
OK. So Commodore won't _tell_ us which RISC processor they're going to
put inside next years Amigas. But they do have a policy of _hinting_
that it is going to be Hewlett-Packards HP-PA chip, and indicating the
truth of this it seems HP and Commodore have agreed to some kind of
technology-share deal. (See AR 1.38.)
I guess HP-PA is coming, and at this stage a lot of people (including
myself) would be wise to skip assembler programming and take up C++
instead. In a year or two, I may in fact be writing my fractal routines
for HP-PA Risc in C++, what a strange thought!
What seems to concern most everybody about HPs chip is: 1) Speed and 2)
emulations. Plenty of purely speculative speed-comparisons have been
put forward. Advocates of the PowerPC have been eager to point to the
number of emulators allegedly being developed for that chip. Here's a
few facts to sort things out:
1) In the 1994/2 issue of Danish computer magazine "Datatid", a
Hewlett-Packard workstation with a HP PA-7100LC 80Mhz CPU is benched
again an IBM workstation with 66Mhz PowerPC 601, a Siemens-Nixdorf with
60Mhz Pentium, and a 50Mhz Sun SparcStation. The test was a program
written in AutoCad. (Autocad is avaliable on each of these machines.)
Results are in seconds taken to perform some calculations and some
HP-PA PowerPC Pentium Sparc
Calc. 2 2.9 3.9 6
Draw 4.25 6 6.25 10.25
This looks rather promising! The HP-PA calculates nearly twice as fast
as the Pentium, nearly 50% faster than the PowerPC. And drawing speed
doesn't look too bad either. The Sparc is trailing far behind for some
reason. Sure, there's a million things to look out for, like how
efficient the compiler is, memory speed and so on. But real-life tests
do give the best indication of real-life speed!
The HP-PA includes on-chip instructions for packing/depacking JPEG and
MPEG graphics. It also replaces some of the other chips of the average
computer; the motherboard is about half the size of a '486.
2) For the Hewlett-Packard workstation, a program called "SoftWindows"
emulates Dos and Windows on the processor level. And a program called
"WABI" - later to come - emulates Windows itself, meaning that OS-bound
operations will perform at full RISC speed. (There are some doubts
concerning the compatibility of this one.)
Although these two emulators are written on top of UNIX and thus would
require adjustment to work with AmigaDos, it does mean that all the
low-level emulation programming has been completed, and that fast,
well-tested Dos/Windows emulation could quickly appear on a HP-PA based
So HP-PA may after all be a much better choice than PowerPC.
Well done, Commodore. When is it going to be official?
Full-Motion video, a short review
During the last days of January, a still limited supply of CD32 FMV
modules finally reached Northern Europe. I went to look at it, the
store carried an "Assorted Arias" or something opera CD, "Black Rain"
by Ridley Scott, and "Star Trek IV", all in the Philips CD-I video
format. Price was 199 DKR ($29), a bit steep for old movies.
I watched Star Trek IV for a while: The sound is great, the colours
are great, the still pictures are great. About BetaCam quality, none of
the usual VHS disturbances. But still, I wasn't all that impressed. I
may not be the average user (who is?), having spent a good part of my
teens writing demos on the C64, learning to spot single 1/50th second
delays in screen updates, but there was one thing I really thought
didn't measure up: Movement. To understand my problem, you'll need no
know a bit about how FMV works. -It _is_ completely incredible that
they've been able to pack motion video and digital stereo sound into
the mere 150K/second a CD offers. A combination of three different
strategies does it:
1) Standard compression. Like IFF, GIF, TIFF etc., a picture can be
compressed to take up less disk space, then later loaded from
disk and decompressed to its original state.
2) Key and delta frames. Some of the frames ("key frames") on the CD
are stored as complete pictures. But most of the frames ("delta
frames") are stored only as their difference to the previous frame
- this saves a huge amount of space. For example, a close-up of
Mr. Spocks facial reactions to some unanticipated event might be a
completely still picture, with the exception of his left eyebrow
being raised. The delta frame needs then only contain enough
information to describe the eyebrows' new whereabouts. The rest of
the picture can remain unchanged.
3) Lossy compression. Where standard compression attempts to keep all
the information of the original picture intact, lossy compression
draws heavily on the fact that our sensory apparatus is a bit slow
at times. So a lossy compression scheme is always based on certain
ideas as to which two shades of red are indistinguishable to our
eyes, which two bits of the sound of a violin are
indistinguishable to our ears, which tiny facial movement is
unnoticed by most people. The unnecessary information is
consequently discarded. (Hence the term "lossy".)
-The first problem concerns point two: With every cut or change of
camera angle in a movie, a new key frame has to be introduced. Since a
key frame takes up a lot of space on the CD, there's not much room
delta frames, and the whole picture remains frozen for a few frames,
waiting for the new key frame to be loaded. _With every cut all
movement freezes._ Quite disturbing to watch.
-The second problem lies in point three. To fit a movie in 150K/second,
_a lot_ of subtle changes of facial expression are judged as
unnoticable. The result is that when Mr. Spock raises his eyebrow, the
rest of his face is completely - and I do mean completely - static.
Characters silently listening to somebody else are very silent indeed,
they look like still pictures (great quality) and then suddenly move
(not to great).
-What I'm saying is that the MPEG standard is a bit optimistic about
our eyes' _inability_ to spot unsteady movement. Yes, I am nitpicking
and yes, I am looking for problems, but I wasn't nearly as impressed as
I thought I'd be. And here's hoping that my eyesight is vastly superior
to the average users'. But to like watching digital video, it would
probably be a good idea to completely forget what I've said here. I for
one couldn't stop paying attention to stillness once I'd noticed.
On the other hand, this is hopefully just the beginning. Perhaps
VideoCD (the standard Commodore has subscribed to) is much, much
better, perhaps even CD-I video will improve once they learn to turn
all the knobs just right at recording time. And perhaps I'll just
forget all about it after three minutes of "Blade Runner".
I think I have to get used to movies on CD, but then it may be easier
than getting used to VHS movies - or CD sound!
The Steven Spielberg/Amiga produced SeaQuest DSV series can now be seen
on my side of the Atlantic: Starting January 26th, German television
station RTL airs SeaQuest Wednesdays at 21:15. Though certainly
bringing Amiga effects to a large audience - I think most Europeans
with cable TV can watch the channel - RTL may not be the station we
want our computer to be associated with; it is mostly known for
demented gameshows sporting lightly clad females. Furthermore, they
usually employ abysmal actors to dub all non-German speech. Of course
we needn't care; we're only in it for the graphics...
RTL will later on stage an _Amiga_ animation competition in connection
A-Max IV - "the other" Macintosh Emulator - is selling for DM 998 (US$
570). It appararently multitasks, reads 800K Mac disks (this is
surprising), uses Amiga ports and SCSI, supports sound, and let's you
access the Mac Clipboard from the Amiga side. An A-Max vs. Emplant test
would be really interesting now.
Final Copy II/Final Writer: A quick and unjust review
In a slightly personal way: When the wordprocessor Final Copy II was
released, I tried it out because it looked good in the ads. And I _was_
impressed; nice fonts, OK speed, nice facilities. But when pressing
right amiga + i to get italics, nothing happened, why? - Because you
have to select an italic _font_ to get italics. This is very
cumbersome, and it also means that there's no way to write a text and
then decide that you want to use another font or pitch; this would
require manual replacement of font specifications at each and every
italicised word or quote! As a special bonus the programmers didn't
include footnote facilites either, making sure that Final Copy II is
utterly useless for any serious purpose.
Like this, with italics unusuable and no footnotes, I'd rate Final Copy
II as exactly 0 out of 10 quills. The company itself - Softwood -
scores 10 out of 10 empty looks for selective incompetence.
Some time has passed, an expanded version called Final Writer has been
released. Updates are a wonderful thing of the software world but
- (sound of trumpets) - guess what wasn't updated this time! Still
unusable italics! Still no footnotes! And the prices has gone up!
I'd rate Final Writer as -10 out of 10 quills. Softwoods incompetence
rating explodes the scale.
I don't mean to sound harsh (oh well, I guess I do), but how is it
possible to spend months programming adding lots of nifty features,
without for five minutes considering the question: What does the user
need? -Now watch out for Final Paint, the revolutionary program that
loads, edits, and displays 24-bit graphics on your Amiga 500! (It
doesn't actually _save_ graphics, though.)
As you may know, AmiWrite and several other Amiga wordprocessors do
take the same approach to italics, but that's no excuse at all: It is
plainly stupid. And as for not including footnotes, why not forget
capital letters or printing as well?
Please understand that I only find it worth being less than nice to
Softwood because Final Copy II and Final Writer _could be_ great and
professional programs. Easily. Now do it.