Thanks for addressing this, some minor issues.
Am 27.05.2015 um 03:14 schrieb Dave Airlie:
> Roland pointed out my previous attempt was lacking, so I enhanced the
> texwrap piglit test, and tested them. This fixes the offset calculations
> in a number of areas by adding the offset first, it also fixes the fastpaths,
> which I forgot to address in the previous commit.
>
> Signed-off-by: Dave Airlie <airl...@redhat.com>
> ---
> src/gallium/drivers/softpipe/sp_tex_sample.c | 76
> +++++++++++++---------------
> 1 file changed, 34 insertions(+), 42 deletions(-)
>
> diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c
> b/src/gallium/drivers/softpipe/sp_tex_sample.c
> index 4ac3498..59bbd86 100644
> --- a/src/gallium/drivers/softpipe/sp_tex_sample.c
> +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c
> @@ -145,14 +145,13 @@ wrap_nearest_clamp(float s, unsigned size, int offset,
> int *icoord)
> {
> /* s limited to [0,1] */
> /* i limited to [0,size-1] */
> + s += (float)offset / size;
Ok softpipe isn't about performance, but an (unconditional) float div
here is imho stretching that "we don't care about performance" line...
Don't forget float divs are usually non-pipelined, 20+ cycles
operations, I think this could possibly make a very real performance
difference.
I think either doing it conditionally based on offset != 0 or something like
s = s * (float)size;
if (offset)
s += (float)offset;
if (s <= 0.0F)
*icoord = 0;
else if (s >= size)
*icoord = size - 1;
else
*icoord = util_itrunc(s);
Would probably be much better...
(And the same for the other wrap modes)
(You could also fix the compute_lambda_lod issue where gather needs to
use lod_zero and not lod_none case while you're here though I guess it's
sort of an orthogonal issue.)
Roland
> if (s <= 0.0F)
> *icoord = 0;
> else if (s >= 1.0F)
> *icoord = size - 1;
> else
> *icoord = util_ifloor(s * size);
> - if (offset)
> - *icoord = CLAMP(*icoord + offset, 0, size - 1);
> }
>
>
> @@ -164,14 +163,13 @@ wrap_nearest_clamp_to_edge(float s, unsigned size, int
> offset, int *icoord)
> const float min = 1.0F / (2.0F * size);
> const float max = 1.0F - min;
>
> + s += (float)offset / size;
> if (s < min)
> *icoord = 0;
> else if (s > max)
> *icoord = size - 1;
> else
> *icoord = util_ifloor(s * size);
> - if (offset)
> - *icoord = CLAMP(*icoord + offset, 0, size - 1);
> }
>
>
> @@ -182,14 +180,14 @@ wrap_nearest_clamp_to_border(float s, unsigned size,
> int offset, int *icoord)
> /* i limited to [-1, size] */
> const float min = -1.0F / (2.0F * size);
> const float max = 1.0F - min;
> +
> + s += (float)offset / size;
> if (s <= min)
> *icoord = -1;
> else if (s >= max)
> *icoord = size;
> else
> *icoord = util_ifloor(s * size);
> - if (offset)
> - *icoord = CLAMP(*icoord + offset, 0, size - 1);
> }
>
>
> @@ -198,8 +196,12 @@ wrap_nearest_mirror_repeat(float s, unsigned size, int
> offset, int *icoord)
> {
> const float min = 1.0F / (2.0F * size);
> const float max = 1.0F - min;
> - const int flr = util_ifloor(s);
> - float u = frac(s);
> + int flr;
> + float u;
> +
> + s += (float)offset / size;
> + flr = util_ifloor(s);
> + u = frac(s);
> if (flr & 1)
> u = 1.0F - u;
> if (u < min)
> @@ -208,8 +210,6 @@ wrap_nearest_mirror_repeat(float s, unsigned size, int
> offset, int *icoord)
> *icoord = size - 1;
> else
> *icoord = util_ifloor(u * size);
> - if (offset)
> - *icoord = CLAMP(*icoord + offset, 0, size - 1);
> }
>
>
> @@ -218,15 +218,13 @@ wrap_nearest_mirror_clamp(float s, unsigned size, int
> offset, int *icoord)
> {
> /* s limited to [0,1] */
> /* i limited to [0,size-1] */
> - const float u = fabsf(s);
> + const float u = fabsf(s + (float)offset / size);
> if (u <= 0.0F)
> *icoord = 0;
> else if (u >= 1.0F)
> *icoord = size - 1;
> else
> *icoord = util_ifloor(u * size);
> - if (offset)
> - *icoord = CLAMP(*icoord + offset, 0, size - 1);
> }
>
>
> @@ -237,15 +235,14 @@ wrap_nearest_mirror_clamp_to_edge(float s, unsigned
> size, int offset, int *icoor
> /* i limited to [0, size-1] */
> const float min = 1.0F / (2.0F * size);
> const float max = 1.0F - min;
> - const float u = fabsf(s);
> + const float u = fabsf(s + (float)offset / size);
> +
> if (u < min)
> *icoord = 0;
> else if (u > max)
> *icoord = size - 1;
> else
> *icoord = util_ifloor(u * size);
> - if (offset)
> - *icoord = CLAMP(*icoord + offset, 0, size - 1);
> }
>
>
> @@ -256,15 +253,14 @@ wrap_nearest_mirror_clamp_to_border(float s, unsigned
> size, int offset, int *ico
> /* i limited to [0, size-1] */
> const float min = -1.0F / (2.0F * size);
> const float max = 1.0F - min;
> - const float u = fabsf(s);
> + const float u = fabsf(s + (float)offset / size);
> +
> if (u < min)
> *icoord = -1;
> else if (u > max)
> *icoord = size;
> else
> *icoord = util_ifloor(u * size);
> - if (offset)
> - *icoord = CLAMP(*icoord + offset, 0, size - 1);
> }
>
>
> @@ -293,14 +289,10 @@ static void
> wrap_linear_clamp(float s, unsigned size, int offset,
> int *icoord0, int *icoord1, float *w)
> {
> - float u = CLAMP(s, 0.0F, 1.0F);
> + float u = CLAMP(s + (float)offset / size, 0.0F, 1.0F);
> u = u * size - 0.5f;
> *icoord0 = util_ifloor(u);
> *icoord1 = *icoord0 + 1;
> - if (offset) {
> - *icoord0 = CLAMP(*icoord0 + offset, 0, size - 1);
> - *icoord1 = CLAMP(*icoord1 + offset, 0, size - 1);
> - }
> *w = frac(u);
> }
>
> @@ -309,7 +301,7 @@ static void
> wrap_linear_clamp_to_edge(float s, unsigned size, int offset,
> int *icoord0, int *icoord1, float *w)
> {
> - float u = CLAMP(s, 0.0F, 1.0F);
> + float u = CLAMP(s + (float)offset / size, 0.0F, 1.0F);
> u = u * size - 0.5f;
> *icoord0 = util_ifloor(u);
> *icoord1 = *icoord0 + 1;
> @@ -317,10 +309,6 @@ wrap_linear_clamp_to_edge(float s, unsigned size, int
> offset,
> *icoord0 = 0;
> if (*icoord1 >= (int) size)
> *icoord1 = size - 1;
> - if (offset) {
> - *icoord0 = CLAMP(*icoord0 + offset, 0, size - 1);
> - *icoord1 = CLAMP(*icoord1 + offset, 0, size - 1);
> - }
> *w = frac(u);
> }
>
> @@ -331,7 +319,7 @@ wrap_linear_clamp_to_border(float s, unsigned size, int
> offset,
> {
> const float min = -1.0F / (2.0F * size);
> const float max = 1.0F - min;
> - float u = CLAMP(s, min, max);
> + float u = CLAMP(s + (float)offset / size, min, max);
> u = u * size - 0.5f;
> *icoord0 = util_ifloor(u);
> *icoord1 = *icoord0 + 1;
> @@ -343,8 +331,12 @@ static void
> wrap_linear_mirror_repeat(float s, unsigned size, int offset,
> int *icoord0, int *icoord1, float *w)
> {
> - const int flr = util_ifloor(s);
> - float u = frac(s);
> + int flr;
> + float u;
> +
> + s += (float)offset / size;
> + flr = util_ifloor(s);
> + u = frac(s);
> if (flr & 1)
> u = 1.0F - u;
> u = u * size - 0.5F;
> @@ -362,13 +354,13 @@ static void
> wrap_linear_mirror_clamp(float s, unsigned size, int offset,
> int *icoord0, int *icoord1, float *w)
> {
> - float u = fabsf(s);
> + float u = fabsf(s + (float)size / offset);
> if (u >= 1.0F)
> u = (float) size;
> else
> u *= size;
> u -= 0.5F;
> - *icoord0 = util_ifloor(u);
> + *icoord0 = util_ifloor(u) + offset;
> *icoord1 = *icoord0 + 1;
> *w = frac(u);
> }
> @@ -378,7 +370,7 @@ static void
> wrap_linear_mirror_clamp_to_edge(float s, unsigned size, int offset,
> int *icoord0, int *icoord1, float *w)
> {
> - float u = fabsf(s);
> + float u = fabsf(s + (float)offset / size);
> if (u >= 1.0F)
> u = (float) size;
> else
> @@ -400,7 +392,7 @@ wrap_linear_mirror_clamp_to_border(float s, unsigned
> size, int offset,
> {
> const float min = -1.0F / (2.0F * size);
> const float max = 1.0F - min;
> - float u = fabsf(s);
> + float u = fabsf(s + (float)offset / size);
> if (u <= min)
> u = min * size;
> else if (u >= max)
> @@ -1040,8 +1032,8 @@ img_filter_2d_linear_repeat_POT(struct sp_sampler_view
> *sp_sview,
> union tex_tile_address addr;
> int c;
>
> - float u = args->s * xpot - 0.5F;
> - float v = args->t * ypot - 0.5F;
> + float u = (args->s * xpot - 0.5F) + args->offset[0];
> + float v = (args->t * ypot - 0.5F) + args->offset[1];
>
> int uflr = util_ifloor(u);
> int vflr = util_ifloor(v);
> @@ -1093,8 +1085,8 @@ img_filter_2d_nearest_repeat_POT(struct sp_sampler_view
> *sp_sview,
> union tex_tile_address addr;
> int c;
>
> - float u = args->s * xpot;
> - float v = args->t * ypot;
> + float u = args->s * xpot + args->offset[0];
> + float v = args->t * ypot + args->offset[1];
>
> int uflr = util_ifloor(u);
> int vflr = util_ifloor(v);
> @@ -1126,8 +1118,8 @@ img_filter_2d_nearest_clamp_POT(struct sp_sampler_view
> *sp_sview,
> union tex_tile_address addr;
> int c;
>
> - float u = args->s * xpot;
> - float v = args->t * ypot;
> + float u = args->s * xpot + args->offset[0];
> + float v = args->t * ypot + args->offset[1];
>
> int x0, y0;
> const float *out;
>
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev
