-
The 16-bit floating point type, what IEEE 754 calls BINARY16, will become _de rigeur_ sooner rather than later. Note that I am not talking about BFLOAT16 (a.k.a. BF16) which is not an IEEE 754 format.…
-
Further reference:
- https://en.wikipedia.org/wiki/IEEE_754-1985
- https://docs.oracle.com/cd/E19957-01/806-3568/ncg_math.html
Quoting [RTS gitter](https://gitter.im/RedisLabs/RedisTimeSeries):
…
-
`purs` compiles the following:
```
(1.0 < nan) == false
```
to
```
1.0 < nan === false
```
but if we instead go through the `Ord` API we can observe different behavior that is not compat…
-
Here's my partial review of the spec as of October 2024:
### Introduction
* I like the removal of the quantum. This greatly simplified the proposal!
* [x] "Only a subset of the values specified in …
-
On Z, fmin/fmax/dmin/dmax intrinsics were changed to support the Java spec (+0.0 compares as strictly greater than -0.0, and NaNs are returned unchanged) on the OpenJ9 side, and to support the IEEE-75…
-
In accuracy.rst, correct rounding is required for add, sub, mul, and divide. I propose adding sqrt to this list.
Note that sqrt is not listed later among the mathematical functions whose accuracy i…
-
-
Current version of the spec includes NaNs, infinities and signed zeroes. So, we would like to ask a question: why they was introduced originally? Here is the answer.
First, they was designed for ca…
-
## Checklist
> Please ensure the following tasks are completed before filing a bug report.
- [x] Read and understood the [Code of Conduct][code-of-conduct].
- [x] Searched for existing …
-
## 目录
1. 双精度二进制浮点数
2. [指数编码](https://github.com/anjia/blog/issues/87#issuecomment-983560820)
3. [二进制浮点数的精度损失,你怎么看?](https://github.com/anjia/blog/issues/87#issuecomment-984785462)
4. [五种浮点异常](ht…