Re: [ft-devel] Fast bisecting BBox_Cubic_Check

[Alexei Podtelezhnikov]

On Sat, Feb 16, 2013 at 2:58 AM, Werner LEMBERG <address@hidden> wrote:
>
>> Thanks to Werner, benchmarking FT_Outline_Get_BBox is now easy using
>> ftbench. I've added a rotation by 30 degrees to really challenge it.
>
> Please commit this to ftbench.

Pushed out.

>> These are results.
>
> Looks good, thanks!  How do you check that the new code produces
> identical bboxes?

I used src/tools/test_bbox.c. I had to add some hexadecimal output,
because the original output was just identical. As you'll see below
the results are not bit-identical. I wish I knew the true answer. One
possibility is to add a third outline to test_bbox with known
theoretical bbox.

OLD CODE:
address@hidden ft-test]$ ./bbox
outline #1
time =  0.13 cbox = [-37.88 535.32 455.89 786.22]
cbox_hex =  [FFDA1F9E 021750FF 01C7E381 0312387F]
time =  3.14 bbox = [115.62 535.32 412.08 672.25]
bbox_hex =  [00739FDF 021750FF 019C14EE 02A04028]
outline #2
time =  0.13 cbox = [100.00 100.00 200.00 200.00]
cbox_hex =  [00640000 00640000 00C80000 00C80000]
time =  1.45 bbox = [100.00 100.00 200.00 179.74]
bbox_hex =  [00640000 00640000 00C80000 00B3BD88]

NEW CODE:
address@hidden ft-test]$ ./bbox
outline #1
time =  0.13 cbox = [-37.88 535.32 455.89 786.22]
cbox_hex =  [FFDA1F9E 021750FF 01C7E381 0312387F]
time =  7.34 bbox = [115.62 535.32 412.08 672.25]
bbox_hex =  [00739F5D 021750FF 019C143F 02A03FD5]
outline #2
time =  0.13 cbox = [100.00 100.00 200.00 200.00]
cbox_hex =  [00640000 00640000 00C80000 00C80000]
time =  3.30 bbox = [100.00 100.00 200.00 179.74]
bbox_hex =  [00640000 00640000 00C80000 00B3BDAC]