Half-float support #354
Open
Half-float support #354
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Tries to use the compiler's native float16 type, such as _Float16 or __fp16 first. This is the best case, because these conversions can compile to single asm instructions on supported architectures, falling back to calling a generic routine from the runtime library (libgcc or compiler-rt). Failing that, we'll try to call the rt library routines directly. This is tricky, because there are no headers for those, and we must get the ABI right. We must compile and run some code to detect the ABI, which is not always possible when cross-compiling. For that reason, you can set the `float16_rt_abi` property to either `native`, `integer`, or `none` in the cross file. The `native` ABI assumes that float16s are passed as the compiler's _Float16 type (e.g. in xmm registers on x86_64). This is currently not very useful, since on compilers that have _Float16 we'll simply use the intrinsic conversions. The `integer` ABI assumes that float16s are passed as uint16_t. `none` means that no attempt to use the runtime library routines will be made, even if they exist. If we can use neither intrinsic conversions nor the runtime library, we'll fall back to questionable bit hackery. This method currently does not support infinities and NaNs.
These can be optimized into SIMD operations in some cases
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Currently used to reduce bandwidth usage of sprite attribute uploads. Not sure if worth it yet.