3 ms·
This seems very usefull but only when you have gigabytes of compliant ascii text of floating point numbers. I do wonder how well this performs on non-compliant
by Out_of_Characte 6y ago
This seems very usefull but only when you have gigabytes of compliant ascii text of floating point numbers. I do wonder how well this performs on non-compliant text and wether there are systems out there that are limited by the 130MB transfer speeds of standard libraries.
Now that I think about it, Excel spreadsheets, Json, XML,textfiles are all mayor contributors to sometimes very flawed ascii-based workloads that should have a complementary binary backing.
- the8472 6y agoNot all optimization work consists of attacking the dominating function. Sometimes most low-hanging fruit already are plucked and you'll have to speed up dozens of smaller things and float parsing can be one of those.
- maccard 6y agoJust because something isn't the bottleneck doesn't mean it's not worth optimising.if you spend 10% decoding, 80% working and 10% saving the results, saving that first or last 10% is definitely worthwhile. My $400 consumer motherboard has a 10Gb network card, and my SSD reads at over 50Gb/s. Anything that brings IO closer to those speeds is welcome.