[Mono-dev] dcraw.net performance

[Sam777]

That was 2.4.3.  2.6 is faster in some places but slower overall (ignore
reading input):

Reading input : 30ms
Loader: dcraw.Loaders.LosslessJpegLoader : 3288ms
Filter: dcraw.ColourScaler : 414ms
AHD Green : 381ms
AHD Red Blue : 6212ms
AHD Homogeneity map : 2378ms
AHD image production : 1046ms
Filter: dcraw.Demosaicing.AHD : 2610ms
Filter: dcraw.Filters.Median : 0ms
Filter: dcraw.ColourSpace : 2206ms
Writer: dcraw.Tiff.write_ppm_tiff : 739ms

Mono is close and I suspect SIMD could more than make up for the difference.

Cheers,
Sam.


Stifu wrote:
> 
> For the record, is this with Mono 2.4 or 2.6?
> Anyway, Mono isn't far behind. :)
> 
> 
> Sam777 wrote:
>> 
>> Hi,
>> 
>> We've ported a large part of  http://www.cybercom.net/~dcoffin/dcraw/
>> dcraw  to C# (enough to convert several Canon cameras) - 
>> https://sourceforge.net/projects/dcrawnet/
>> https://sourceforge.net/projects/dcrawnet/ .  As expected for image
>> processing code, this is slower than the original C implementation but
>> we've managed to claw some performance back by making it easier for the
>> JIT to optimise bounds checking.  Interestingly, Mono didn't get quite as
>> much of a boost from this as the MS implementation:
>> 
>> MS:
>> Reading input : 227ms
>> Loader: dcraw.Loaders.LosslessJpegLoader : 2404ms
>> Filter: dcraw.ColourScaler : 304ms
>> AHD Green : 378ms
>> AHD Red Blue : 3943ms
>> AHD Homogeneity map : 2879ms
>> AHD image production : 777ms
>> Filter: dcraw.Demosaicing.AHD : 2069ms
>> Filter: dcraw.Filters.Median : 1ms
>> Filter: dcraw.ColourSpace : 1069ms
>> Writer: dcraw.Tiff.write_ppm_tiff : 582ms
>> 
>> Mono:
>> Reading input : 237ms
>> Loader: dcraw.Loaders.LosslessJpegLoader : 3191ms
>> Filter: dcraw.ColourScaler : 451ms
>> AHD Green : 385ms
>> AHD Red Blue : 5965ms
>> AHD Homogeneity map : 2367ms
>> AHD image production : 990ms
>> Filter: dcraw.Demosaicing.AHD : 2545ms
>> Filter: dcraw.Filters.Median : 0ms
>> Filter: dcraw.ColourSpace : 2108ms
>> Writer: dcraw.Tiff.write_ppm_tiff : 736ms
>> 
>> I'm afraid I don't have the time to look into this any further but
>> thought it might be an interesting real world data point/benchmark for
>> you.  I'd like to try your SIMD stuff at some point.
>> 
>> Thanks,
>> Sam.
>> 
> 
> 

-- 
View this message in context: http://old.nabble.com/dcraw.net-performance-tp26769069p26769732.html
Sent from the Mono - Dev mailing list archive at Nabble.com.