[PATCH] LNT ARM sqlite3
Renato Golin
renato.golin at linaro.org
Thu Mar 28 06:25:43 PDT 2013
On 28 March 2013 09:58, Jiong Wang <jiwang at tilera.com> wrote:
> these minor differences accumulated, and thus cause the the final digit
> difference on TILE-Gx and X86/ARM.
>
Yup, each architecture accumulates in a slightly different way. That's
what's wrong with the code. ;)
And I found in your sqlite3.patch, you are trying to skip those tricky
> float point transformation to keep realvalue untouched. I guess it better
> to assign "temp" after the negative number processed.
>
Good point.
Anyway, I've applied both patches and I'm getting round errors like:
< 9|428249.777777
---
> 9|428249.777776999
Even when I specifically requested "round(avg(b), 6)". (that also happens
when I round for 5, 6, 7, 8, etc). I guess the rounding code is expecting
broken FP values to work, which I'm not willing to fix.
Attached is the same patch again, rounding to 5 digits, works on x86 and
ARM. Hopefully it'll also work on TILE-Gx.
cheers,
--renato
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20130328/40943b32/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: sqlite3-results.patch
Type: application/octet-stream
Size: 104952 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20130328/40943b32/attachment.obj>
More information about the llvm-commits
mailing list