Displaying 2 results from an estimated 2 matches for "d46276".
2018 May 16
0
Rotates, once again
...ested in PR31274 [1]?
>
> [1] https://bugs.llvm.org/show_bug.cgi?id=31274
>
>
Yes, I think everyone agrees that we need to improve the cost model
interface. Note that there's a proposal for review for the 1 IR -> many
machine ops variant of this question:
https://reviews.llvm.org/D46276
> In all these cases (rotate, min, max, abs, fma, add-with-overflow, and
> probably many more) there's a tradeoff between elaborating them as simpler
> IR instructions and modelling them as its own instruction / intrinsic. A
> big disadvantage of introducing new instructions / i...
2018 May 16
2
Rotates, once again
On 2018-05-16 00:34, Sanjay Patel via llvm-dev wrote:
> Vectorization goes overboard because the throughput cost model used by
> the
> vectorizers doesn't match the 6 IR instructions that correspond to 1
> x86
> rotate instruction. Instead, we have:
>
> [...]
>
> The broken cost model also affects unrolling and inlining. Size costs
> are
> overestimated