Eric Biggers

I am thinking about submitting a workaround that would update the CMake-based build system to disable some optimized code in cases where the binutils is too old. Can you please...

https://github.com/ebiggers/libdeflate/pull/389 has been submitted.

I mainly focus on x86 and arm; the PowerPC support isn't something I've really worked on yet. From searching around it sounds like all PowerPC processors support unaligned memory accesses....

> By the way, how could I test if adding unaligned access support for PowerPC actually works here or at least does not break anything or introduce any unwanted effects...

> FWIW, here are tests on PowerPC. There is no meaningful difference, and time between runs with the same code can differ more that between some runs with different code...

No. It's pretty straightforward, though; just build the benchmark program (`-DLIBDEFLATE_BUILD_TESTS` needs to be passed to cmake) and pass it some large-ish files. For what you're trying to test, the...

I think you are requesting that support for this be added to [the `benchmark` program](https://github.com/ebiggers/libdeflate/blob/master/programs/benchmark.c)? Note that it already links to zlib, and this can be any version or fork...

As per the previous comment, I don't think anything needs to be done here. If there is a specific metric where libdeflate is not doing well vs. another implementation, then...

Thanks for looking into this! This is a part of the code where it feels like there should be something better, but it's been hard to find something that can...

For quick testing I just use an amalgamation of a bunch of different files. For more detailed testing I do look at different types of files specifically, including types that...