<veltas>
Fixed point is much more forthy though, faster, and uses *way* less code in embedded
<cmtptr>
veltas, but if you need that precision then can't you just choose your units and bit width appropriately? the same problem is there with floats, it just mysteriously works at some scales and doesn't at others
<veltas>
With floats the precision is less, but the scale is always optimal.
<veltas>
I mean units are always optimal
<veltas>
Usually within one calculation a number of different units are optimal for different numbers, floating point will handle that accurately, fixed does not
<veltas>
If you don't believe me try it out yourself, it's obvious very quickly that fixed is a lot of effort and less accurate. But still less effort than writing floating in software and 'faster'
<veltas>
But if you've got highly parallel work and a normal computer with streaming floating operations then it's not even faster...
<cmtptr>
i have tried it out. i'm using fixed point in my own software, it seems way simpler to me because i know what to expect
<veltas>
And compare the results with floating point
<cmtptr>
i have. with floats you get weird artifacts because you lose sigfigs at higher scales and you gain sigfigs at lower scales
<cmtptr>
how is that ever what you'd want??
<veltas>
You don't gain or lose significant figures with floats, fixed point has more significant figures available at the optimal unit but less in most units
<veltas>
The weird artifacts are often much less severe than the loss of accuracy of fixed point in typical calculations
<veltas>
If it works for you then it's because of the kind of calculation you're doing, floating point is much more flexible, obviously I wouldn't claim it's better in all situations.
Zarutian_HTC has quit [Quit: Bye]
X-Scale` has joined #forth
X-Scale has quit [Ping timeout: 246 seconds]
X-Scale` is now known as X-Scale
jedb has quit [*.net *.split]
WilhelmVonWeiner has quit [*.net *.split]
catern has quit [*.net *.split]
djinni has quit [*.net *.split]
tolja has quit [*.net *.split]
WilhelmVonWeiner has joined #forth
djinni has joined #forth
jedb has joined #forth
tolja has joined #forth
catern has joined #forth
jedb_ has joined #forth
proteus-guy has joined #forth
WilhelmV1nWeiner has joined #forth
<cmtptr>
well i will accept that it's possible i just haven't been faced with the right situation yet to see the value
djinni_ has joined #forth
<cmtptr>
i know that sometimes i can be dense to things until i experience it firsthand
lonjil2 has joined #forth
fiddlerwoaroof has quit [*.net *.split]
pareidolia has quit [*.net *.split]
bluekelp has quit [*.net *.split]
lonjil has quit [*.net *.split]
jedb has quit [*.net *.split]
WilhelmVonWeiner has quit [*.net *.split]
djinni has quit [*.net *.split]
tolja has quit [*.net *.split]
fiddlerwoaroof has joined #forth
bluekelp has joined #forth
pareidolia has joined #forth
boru` has joined #forth
boru has quit [Disconnected by services]
boru` is now known as boru
tolja has joined #forth
_whitelogger has joined #forth
proteus-guy has quit [Remote host closed the connection]
sts-q has quit [Ping timeout: 260 seconds]
sts-q has joined #forth
cantstanya has quit [Ping timeout: 240 seconds]
cantstanya has joined #forth
zolk3ri has joined #forth
gravicappa has joined #forth
birdwing has joined #forth
_whitelogger has joined #forth
dave0 has joined #forth
_whitelogger has joined #forth
dave0 has quit [Quit: dave's not here]
birdwing has quit [Remote host closed the connection]
birdwing has joined #forth
<veltas>
cmtptr: Not understanding it isn't a crime, but what really irritates me is when people *do* see it and then refuse to accept it
<veltas>
Not that that's happened here...
<veltas>
Too much of people understanding something and then closing their mind off and pretending it's not true at the moment
_whitelogger has joined #forth
jedb_ has quit [Ping timeout: 260 seconds]
gravicappa has quit [Ping timeout: 256 seconds]
jedb has joined #forth
jedb has quit [Remote host closed the connection]
jedb_ has joined #forth
jedb_ has quit [Remote host closed the connection]
jedb has joined #forth
gravicappa has joined #forth
zolk3ri has quit [Write error: Connection reset by peer]
cantstanya has quit [Write error: Connection reset by peer]
cantstanya has joined #forth
zolk3ri has joined #forth
zolk3ri has quit [Ping timeout: 240 seconds]
zolk3ri has joined #forth
cantstanya has quit [Ping timeout: 240 seconds]
cantstanya has joined #forth
<TangentDelta>
Oh wow, the 6809 is a beast of a processor.
<TangentDelta>
Its like a 16-bit processor that is pretending to be 8-bit.
WickedShell has joined #forth
WilhelmV1nWeiner has quit [Quit: leaving]
WilhelmVonWeiner has joined #forth
zolk3ri has quit [Ping timeout: 240 seconds]
zolk3ri has joined #forth
spoofer has quit [Read error: Connection reset by peer]
spoofer has joined #forth
gravicappa has quit [Ping timeout: 260 seconds]
gravicappa has joined #forth
birdwing has quit [Ping timeout: 268 seconds]
<MrMobius>
TangentDelta, check out the 6309
<TangentDelta>
Yeah, it's even more crazy.
<TangentDelta>
The 4 8-bit accumulators can be combined into 1 32-bit accumulator :P
<MrMobius>
kind of a dead end though since you dont want to be doing too many 32-bit calculations with an 8 bit memory bus
<TangentDelta>
I don't know how accessing that 32-bit register works. I wonder if you can just use it to accumulate multiple arithmatic operations.
<MrMobius>
it works by combining the existing 8 bit registers
<TangentDelta>
There a 16-bit multiplication and summing accumulator chip with a 35-bit accumulator. I've thought about interfacing that with an 8-bit microprocessor.
<MrMobius>
which is kind of neat. you have the 4 bytes each split up into its own register which may be useful
<MrMobius>
neat
<MrMobius>
one of the 8051s I was playing with had a 40-bit multiply/accumulate thing in it
<TangentDelta>
Wow
WickedShell has quit [Remote host closed the connection]