JPEG loading speedup possible?

[p0nce]

Graillon diffuse takes 34ms to decode, isn't it a bit much for a 1332 x 1276?

• JPGD_SUPPORT_FREQ_DOMAIN_UPSAMPLING adds 7ms by itself

[p0nce]

Using that 34ms time as baseline:

• [ ] Try for Row and Col not to be templated
• [ ] expanded convert seemingly can be optimized
• [ ] Col!4.idct not great at all

[p0nce]

So, there was a small YUV to RGB speedup, we can perhaps remove the tables if generalized, which will win 4kb ram :| The DCT/iDCT not that easy to optimize