Can a 4x4-Filter produce better quality than common TF?

[tjl]

Grid displaced 1/4 pixel in x- and y-axis:

1  MAD R0, H0.xyxy, {0, 0, 0.5, 0.5}, f[TEX0].xyxy;

2  MAD R0, H1.xyxy, { 0.5, 0.5, 0.5, 0.5}, R0;

Very nice! Thanks, that should be a LOT faster on my GF FX 5600...

Like I said, I was too tired.

Shaders are fun - it's like assembly language optimization all over again.

 

[tjl]

Grid displaced 1/4 pixel in x- and y-axis:

1  MAD R0, H0.xyxy, {0, 0, 0.5, 0.5}, f[TEX0].xyxy;

2  MAD R0, H1.xyxy, { 0.5, 0.5, 0.5, 0.5}, R0;

Grid displaced 1/4 pixel in y-axis:

1  MAD R0, H0.xyxy, {-0.25, -0.25, 0.25, 0.25}, f[TEX0].xyxy;

2  MAD R0, H1.xyxy, { 0.5, 0.5, 0.5, 0.5}, R0;

Grid not displaced:

1  MAD R0, H0.xyxy, {-0.25, -0.25, 0.25, 0.25},  f[TEX0].xyxy;
   MAD R0, H1.xyxy, {-0.25, -0.25, -0.25, -0.25}, R0;

2  MAD R0, H1.xyxy, { 0.5, 0.5, 0.5, 0.5}, R0;

Now I feel I'm entering the twilight zone. When I put your optimizations to test on a GF FX 5600, they were all approx. 20-30% slower than my original code. I have no explanation for this, except that I used more half floats, and that doesn't make sense for the last program... except it does a full-precision MAD. Changing R0 to H3 did make it faster. (Linux drivers 44.96).

I *REALLY* wish NV would publish real optimization guidelines.