peterhillman

Perhaps the library could support compiling with something like `-DFORCE_HALF_EXPLICIT_CAST` I totally agree it would have been better to have half to float an explicit cast from the beginning for...

Indeed only making float-to-half explicit would be less problematic for templated functions, though I suspect could still get you into gotchas. I wonder whether having implicit casts only in one...

The regular scanline InputFile interface can read tiled files, using an internal cache to store a row of tiles. I believe that is the mechanism Nuke uses to read regular...

That code won't read the pixels from the file, but will do a certain amount of memory allocation in preparation for that, which is slow with very large files. This...

Thanks for the report. I don't see any attached poc file (the link is empty) Could you perhaps provide that file? Meanwhile, it appears you are building against an OpenEXR-2.3.X...

You can run IlmImfTest/IlmImfTest directly (rather than via a make build target) using the test name as an argument, to run a single test e.g. `IlmImfTest testDwaCompressorSimd` You can also...

Do you have a use case? Making a one level tiled deep image would be trivial, but the OpenEXR library doesn't support filtering of deep pixels to produce mipmap/ripmap images....

There's nothing ARM specific in IlmBase or libImf that I'm aware of. Are you aware of any operations that would particularly benefit from such optimization? The only SIMD optimization in...

The OpenEXR library itself doesn't rely heavily on Vector and Matrix operations in libImath, so reading/writing images on Arm processors wouldn't be accelerated much by SIMD optimizations. I believe that...

Apologies for the delay in responding. The OpenEXR TSC discussed this today. We're interested whether this remains a required feature for you? If so, it would be helpful if you...