<< 63 double.low = x.low >> 1 // If the most-significant bit was set before shifting then it, // conceptually, becomes a term of x^128. This is greater than the // irreducible polynomial so the result has to be reduced. The // irreducible polynomial is 1+x+x^2+x^7+x^128. We can subtract that to // eliminate the term at x^128 which also means subtracting the other // four terms. In characteristic 2 fields, subtraction == addition == // XOR. if msbSet { double.low ^= 0xe100000000000000 } return } var...

<< 63 double.low = x.low >> 1 // If the most-significant bit was set before shifting then it, // conceptually, becomes a term of x^128. This is greater than the // irreducible polynomial so the result has to be reduced. The // irreducible polynomial is 1+x+x^2+x^7+x^128. We can subtract that to // eliminate the term at x^128 which also means subtracting the other // four terms. In characteristic 2 fields, subtraction == addition == // XOR. if msbSet { double.low ^= 0xe100000000000000 } return } var...

* `tf.lite`
    * LiteRT announced a [new release](https://developers.googleblog.com/en/litert-maximum-performance-simplified/) at Google IO ‘25 that improves upon TFLite, particularly in terms of NPU and GPU hardware acceleration and performance for on-device ML and AI applications. The APIs are available in Kotlin and C++.

    * It's now possible to use [Starlette's `StreamingResponse`](https://www.starlette.dev/responses/#streamingresponse) with iterators, like [file-like](https://docs.python.org/3/glossary.html#term-file-like-object) objects (as those returned by `open()`).
    * It's now possible to use the low level utility `iterate_in_threadpool` from `starlette.concurrency` (for advanced scenarios).