From: Floyd L. Davidson on
nospam <nospam(a)nospam.invalid> wrote:
>In article <87k4q7d7ib.fld(a)apaflo.com>, Floyd L. Davidson
><floyd(a)apaflo.com> wrote:
>
>> >> I've found dcraw to be quick for my situation - it's free.
>> >
>> >dcraw is one of the slowest raw converters.
>>
>> tanana:floyd /u10/p5/2006/jan31a 0>time dcraw -h dsc_7022.nef
>> real 0m0.131s
>
>> You should actually try it before claiming to know
>> something about how it works.
>
>i've compared both and dcraw is a lot slower than camera raw on the
>same hardware.

I'll wait for credible comparisons.

--
Floyd L. Davidson <http://www.apaflo.com/floyd_davidson>
Ukpeagvik (Barrow, Alaska) floyd(a)apaflo.com
From: nospam on
In article <87y6enb8n3.fld(a)apaflo.com>, Floyd L. Davidson
<floyd(a)apaflo.com> wrote:

> >> >> I've found dcraw to be quick for my situation - it's free.
> >> >
> >> >dcraw is one of the slowest raw converters.
> >>
> >> tanana:floyd /u10/p5/2006/jan31a 0>time dcraw -h dsc_7022.nef
> >> real 0m0.131s
> >
> >> You should actually try it before claiming to know
> >> something about how it works.
> >
> >i've compared both and dcraw is a lot slower than camera raw on the
> >same hardware.
>
> I'll wait for credible comparisons.

says the person who provided a comparison with only *one* timing! a
comparison needs at least one *other* timing with which to compare.
that's why it's called a comparison.

furthermore, the -h option outputs a half-size image, which the man
page says is twice as fast as the high-speed low-quality option. in
other words, it's bogus and not a real world representation of a
typical workflow, as if about 1/8th of a second was even believable.

there's also no data on the size of the raw file, how many pixels it
has or what type of hardware was used.
From: Floyd L. Davidson on
nospam <nospam(a)nospam.invalid> wrote:
>In article <87y6enb8n3.fld(a)apaflo.com>, Floyd L. Davidson
><floyd(a)apaflo.com> wrote:
>
>> >> >> I've found dcraw to be quick for my situation - it's free.
>> >> >
>> >> >dcraw is one of the slowest raw converters.
>> >>
>> >> tanana:floyd /u10/p5/2006/jan31a 0>time dcraw -h dsc_7022.nef
>> >> real 0m0.131s
>> >
>> >> You should actually try it before claiming to know
>> >> something about how it works.
>> >
>> >i've compared both and dcraw is a lot slower than camera raw on the
>> >same hardware.
>>
>> I'll wait for credible comparisons.
>
>says the person who provided a comparison with only *one* timing! a
>comparison needs at least one *other* timing with which to compare.
>that's why it's called a comparison.

I've already provided multiple examples of times for
UFRAW running in various configurations. Pay attention,
eh?

>furthermore, the -h option outputs a half-size image, which the man
>page says is twice as fast as the high-speed low-quality option. in
>other words, it's bogus and not a real world representation of a
>typical workflow, as if about 1/8th of a second was even believable.

Not bogus, just optimized for speed. That actually is
very reasonable for someone (such as the OP) who merely
wants a 'default' image for comparison. For production
purposes it might be used for previewing, for example.

Clearly the speed of dcraw just blows you mind away!

>there's also no data on the size of the raw file, how many pixels it
>has or what type of hardware was used.

The point was that *you* can't get another raw converter
to run that fast.

Here are other options for that same file:

tanana:floyd /u10/p5/2006/jan31a 0>time dcraw -h dsc_7022.nef
real 0m0.131s

tanana:floyd /u10/p5/2006/jan31a 0>time dcraw -q 0 dsc_7022.nef
real 0m0.391s

tanana:floyd /u10/p5/2006/jan31a 0>time dcraw -q 2 dsc_7022.nef
real 0m0.459s

tanana:floyd /u10/p5/2006/jan31a 0>time dcraw -q 3 dsc_7022.nef
real 0m1.042s

It is indeed an interesting file too!

File Name : dsc_7022.nef
File Size : 3.9 MB
File Type : NEF
Camera Model Name : NIKON D1
Bits Per Sample : 12
Compression : Uncompressed

Here are the same options on another NEF file:

tanana:floyd /u5/2010/jun8 0>time dcraw -h d3f_0091.nef
real 0m0.243s

tanana:floyd /u5/2010/jun8 0>time dcraw -q 0 d3f_0091.nef
real 0m0.778s

tanana:floyd /u5/2010/jun8 0>time dcraw -q 2 d3f_0091.nef
real 0m0.857s

tanana:floyd /u5/2010/jun8 0>time dcraw -q 3 d3f_0091.nef
real 0m2.026s

The Exif data:

File Name : d3f_0091.nef
File Size : 8.2 MB
File Type : NEF
Camera Model Name : NIKON D3
Bits Per Sample : 12
Compression : Uncompressed

And from yet another file:

tanana:floyd /u5/2010/jun7 0>time dcraw -h d3s_4936.nef
real 0m0.931s

tanana:floyd /u5/2010/jun7 0>time dcraw -q 0 d3s_4936.nef
real 0m2.139s

tanana:floyd /u5/2010/jun7 0>time dcraw -q 2 d3s_4936.nef
real 0m2.395s

tanana:floyd /u5/2010/jun7 0>time dcraw -q 3 d3s_4936.nef
real 0m4.907s

And Exif data:

File Name : d3s_4936.nef
File Size : 14 MB
File Type : NEF
Camera Model Name : NIKON D3S
Bits Per Sample : 14
Compression : Nikon NEF Compressed


Your complaint about no comparisons is invalid, given
the number of data points that I'd already listed for
UFRAW running on different systems. Obviously at it's
best UFRAW is four times slower than dcraw.

Incidentally, the system for all of the times has been
one running a pair of dual core Opteron AMD 275
processors in 64 bit mode with a 2.2Ghz clock. (The
process of course runs only on one processor as it does
not run in a threaded mode.) Clearly this is no where
near the fastest hardware available.

--
Floyd L. Davidson <http://www.apaflo.com/floyd_davidson>
Ukpeagvik (Barrow, Alaska) floyd(a)apaflo.com
From: nospam on
In article <87pqzzb25j.fld(a)apaflo.com>, Floyd L. Davidson
<floyd(a)apaflo.com> wrote:

> The point was that *you* can't get another raw converter
> to run that fast.

first of all, your timings are not elapsed time, and second, i can
easily process a raw *much* faster than dcraw can.

here's a comparison, using a nikon d700 image (12 mp) on a 2.6 ghz core
2 duo laptop (not the fastest these days either):

% time dcraw DSC_0052.NEF
16.456u 0.374s 0:17.49 96.1% 0+0k 0+21io 1pf+0w

% time dcraw -h DSC_0052.NEF
3.877u 0.116s 0:04.09 97.3% 0+0k 0+0io 0pf+0w

in high speed mode, dcraw is just over 4 seconds, and in default mode,
it's 17.5 seconds!

photoshop & camera raw is about 2 seconds to open the raw. there's no
way to run time on it.

thus, camera raw is twice as fast as the half-size low-quality mode of
dcraw and almost *nine* times as fast as normal mode.

dcraw is *slow* (and worse than i recall too).

> Your complaint about no comparisons is invalid,

it's invalid if you don't have timings from other raw converters.

> given
> the number of data points that I'd already listed for
> UFRAW running on different systems. Obviously at it's
> best UFRAW is four times slower than dcraw.

ufraw is actually a more appropriate comparison with camera raw, so my
test ends up being very biased *for* dcraw, yet it still lost.

> Incidentally, the system for all of the times has been
> one running a pair of dual core Opteron AMD 275
> processors in 64 bit mode with a 2.2Ghz clock. (The
> process of course runs only on one processor as it does
> not run in a threaded mode.)

that's one reason why it's so slow. why isn't it threaded?

> Clearly this is no where
> near the fastest hardware available.

nor was mine.
From: Floyd L. Davidson on
nospam <nospam(a)nospam.invalid> wrote:
>In article <87pqzzb25j.fld(a)apaflo.com>, Floyd L. Davidson
><floyd(a)apaflo.com> wrote:
>
>> The point was that *you* can't get another raw converter
>> to run that fast.
>
>first of all, your timings are not elapsed time, and second, i can
>easily process a raw *much* faster than dcraw can.

Eh? If they are not "elapsed time", what are they???

>here's a comparison, using a nikon d700 image (12 mp) on a 2.6 ghz core
>2 duo laptop (not the fastest these days either):
>
>% time dcraw DSC_0052.NEF
>16.456u 0.374s 0:17.49 96.1% 0+0k 0+21io 1pf+0w
>
>% time dcraw -h DSC_0052.NEF
>3.877u 0.116s 0:04.09 97.3% 0+0k 0+0io 0pf+0w
>
>in high speed mode, dcraw is just over 4 seconds, and in default mode,
>it's 17.5 seconds!

Okay, you've got a very slow system.

Incidentally, the 17.5 seconds time that you are using
is *exactly* the same measurement that I provided and
you said above was not elapsed time. (Try reading the
man page for the time command.)

>photoshop & camera raw is about 2 seconds to open the raw. there's no
>way to run time on it.
>
>thus, camera raw is twice as fast as the half-size low-quality mode of
>dcraw and almost *nine* times as fast as normal mode.
>
>dcraw is *slow* (and worse than i recall too).

So you can't show times to support your claims?

Sounds like a personal problem.

>> Your complaint about no comparisons is invalid,
>
>it's invalid if you don't have timings from other raw converters.

Which I had already provided, so what's your problem?

>> given
>> the number of data points that I'd already listed for
>> UFRAW running on different systems. Obviously at it's
>> best UFRAW is four times slower than dcraw.
>
>ufraw is actually a more appropriate comparison with camera raw, so my
>test ends up being very biased *for* dcraw, yet it still lost.

It seems to have won.

>> Incidentally, the system for all of the times has been
>> one running a pair of dual core Opteron AMD 275
>> processors in 64 bit mode with a 2.2Ghz clock. (The
>> process of course runs only on one processor as it does
>> not run in a threaded mode.)
>
>that's one reason why it's so slow. why isn't it threaded?

Do you have a clue?

>> Clearly this is no where
>> near the fastest hardware available.
>
>nor was mine.

Well, we agree on that.

--
Floyd L. Davidson <http://www.apaflo.com/floyd_davidson>
Ukpeagvik (Barrow, Alaska) floyd(a)apaflo.com