| Issue |
63842
|
| Summary |
APFloatTest: convert test accidentally ended up with `x87DoubleExtended` -> `x87DoubleExtended` noops.
|
| Labels |
new issue
|
| Assignees |
|
| Reporter |
eddyb
|
It looks like back in 2013 (https://github.com/llvm/llvm-project/commit/b361adbb64535677d617e2680b06cc0283b07d1a), tests were added for these conversions:
|`IEEEsingle` -> `x87DoubleExtended`|`x87DoubleExtended` -> `x87DoubleExtended`|
|-|-|
|`APFloat::getSNaN(APFloat::IEEEsingle)`|`APFloat::getSNaN(APFloat::x87DoubleExtended)`|
|`APFloat::getQNaN(APFloat::IEEEsingle)`|`APFloat::getQNaN(APFloat::x87DoubleExtended)`|
The first column (`IEEEsingle` NaNs) makes sense, and it fits with the commit description:
> APFloat: Make sure that we get a well-formed x87 NaN when converting from a smaller type.
But the second column looks like a mistake - the simplest explanation I can think of is they were meant to be `IEEEdouble` (which is still smaller than `x87DoubleExtended`, and so it would also fit the commit description) *not* noops.
cc @d0k
![]()
_______________________________________________
llvm-bugs mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
