#include "pipe/p_defines.h"
#include "pipe/p_shader_tokens.h"
#include "util/u_math.h"
-#include "util/u_format.h"
+#include "util/format/u_format.h"
#include "util/u_memory.h"
#include "util/u_inlines.h"
#include "sp_quad.h" /* only for #define QUAD_* tokens */
{
/* s limited to [0,1) */
/* i limited to [0,size-1] */
- int i = util_ifloor(s * size);
+ const int i = util_ifloor(s * size);
*icoord = repeat(i + offset, size);
}
wrap_linear_repeat(float s, unsigned size, int offset,
int *icoord0, int *icoord1, float *w)
{
- float u = s * size - 0.5F;
+ const float u = s * size - 0.5F;
*icoord0 = repeat(util_ifloor(u) + offset, size);
*icoord1 = repeat(*icoord0 + 1, size);
*w = frac(u);
wrap_linear_clamp(float s, unsigned size, int offset,
int *icoord0, int *icoord1, float *w)
{
- float u = CLAMP(s * size + offset, 0.0F, (float)size);
+ const float u = CLAMP(s * size + offset, 0.0F, (float)size) - 0.5f;
- u = u - 0.5f;
*icoord0 = util_ifloor(u);
*icoord1 = *icoord0 + 1;
*w = frac(u);
wrap_linear_clamp_to_edge(float s, unsigned size, int offset,
int *icoord0, int *icoord1, float *w)
{
- float u = CLAMP(s * size + offset, 0.0F, (float)size);
- u = u - 0.5f;
+ const float u = CLAMP(s * size + offset, 0.0F, (float)size) - 0.5f;
*icoord0 = util_ifloor(u);
*icoord1 = *icoord0 + 1;
if (*icoord0 < 0)
wrap_linear_clamp_to_border(float s, unsigned size, int offset,
int *icoord0, int *icoord1, float *w)
{
- const float min = -0.5F;
+ const float min = -1.0F;
const float max = (float)size + 0.5F;
- float u = CLAMP(s * size + offset, min, max);
- u = u - 0.5f;
+ const float u = CLAMP(s * size + offset, min, max) - 0.5f;
*icoord0 = util_ifloor(u);
*icoord1 = *icoord0 + 1;
*w = frac(u);
{
int flr;
float u;
+ bool no_mirror;
s += (float)offset / size;
flr = util_ifloor(s);
+ no_mirror = !(flr & 1);
+
u = frac(s);
- if (flr & 1)
+ if (no_mirror) {
+ u = u * size - 0.5F;
+ } else {
u = 1.0F - u;
- u = u * size - 0.5F;
+ u = u * size + 0.5F;
+ }
+
*icoord0 = util_ifloor(u);
- *icoord1 = *icoord0 + 1;
+ *icoord1 = (no_mirror) ? *icoord0 + 1 : *icoord0 - 1;
+
if (*icoord0 < 0)
- *icoord0 = 0;
+ *icoord0 = 1 + *icoord0;
+ if (*icoord0 >= (int) size)
+ *icoord0 = size - 1;
+
if (*icoord1 >= (int) size)
*icoord1 = size - 1;
- *w = frac(u);
+ if (*icoord1 < 0)
+ *icoord1 = 1 + *icoord1;
+
+ *w = (no_mirror) ? frac(u) : frac(1.0f - u);
}
{
const float min = -0.5F;
const float max = size + 0.5F;
- float u = fabsf(s * size + offset);
- if (u <= min)
- u = min;
- else if (u >= max)
- u = max;
- u -= 0.5F;
+ const float t = fabsf(s * size + offset);
+ const float u = CLAMP(t, min, max) - 0.5F;
*icoord0 = util_ifloor(u);
*icoord1 = *icoord0 + 1;
*w = frac(u);
static void
wrap_nearest_unorm_clamp(float s, unsigned size, int offset, int *icoord)
{
- int i = util_ifloor(s);
+ const int i = util_ifloor(s);
*icoord = CLAMP(i + offset, 0, (int) size-1);
}
int *icoord0, int *icoord1, float *w)
{
/* Not exactly what the spec says, but it matches NVIDIA output */
- float u = CLAMP(s + offset - 0.5F, 0.0f, (float) size - 1.0f);
+ const float u = CLAMP(s + offset - 0.5F, 0.0f, (float) size - 1.0f);
*icoord0 = util_ifloor(u);
*icoord1 = *icoord0 + 1;
*w = frac(u);
wrap_linear_unorm_clamp_to_border(float s, unsigned size, int offset,
int *icoord0, int *icoord1, float *w)
{
- float u = CLAMP(s + offset, -0.5F, (float) size + 0.5F);
- u -= 0.5F;
+ const float u = CLAMP(s + offset, -0.5F, (float) size + 0.5F) - 0.5F;
*icoord0 = util_ifloor(u);
*icoord1 = *icoord0 + 1;
if (*icoord1 > (int) size - 1)
wrap_linear_unorm_clamp_to_edge(float s, unsigned size, int offset,
int *icoord0, int *icoord1, float *w)
{
- float u = CLAMP(s + offset, +0.5F, (float) size - 0.5F);
- u -= 0.5F;
+ const float u = CLAMP(s + offset, +0.5F, (float) size - 0.5F) - 0.5F;
*icoord0 = util_ifloor(u);
*icoord1 = *icoord0 + 1;
if (*icoord1 > (int) size - 1)
static inline int
coord_to_layer(float coord, unsigned first_layer, unsigned last_layer)
{
- int c = util_ifloor(coord + 0.5F);
+ const int c = util_ifloor(coord + 0.5F);
return CLAMP(c, (int)first_layer, (int)last_layer);
}
+static void
+compute_gradient_1d(const float s[TGSI_QUAD_SIZE],
+ const float t[TGSI_QUAD_SIZE],
+ const float p[TGSI_QUAD_SIZE],
+ float derivs[3][2][TGSI_QUAD_SIZE])
+{
+ memset(derivs, 0, 6 * TGSI_QUAD_SIZE * sizeof(float));
+ derivs[0][0][0] = s[QUAD_BOTTOM_RIGHT] - s[QUAD_BOTTOM_LEFT];
+ derivs[0][1][0] = s[QUAD_TOP_LEFT] - s[QUAD_BOTTOM_LEFT];
+}
+
+static float
+compute_lambda_1d_explicit_gradients(const struct sp_sampler_view *sview,
+ const float derivs[3][2][TGSI_QUAD_SIZE],
+ uint quad)
+{
+ const struct pipe_resource *texture = sview->base.texture;
+ const float dsdx = fabsf(derivs[0][0][quad]);
+ const float dsdy = fabsf(derivs[0][1][quad]);
+ const float rho = MAX2(dsdx, dsdy) * u_minify(texture->width0, sview->base.u.tex.first_level);
+ return util_fast_log2(rho);
+}
+
/**
* Examine the quad's texture coordinates to compute the partial
const float t[TGSI_QUAD_SIZE],
const float p[TGSI_QUAD_SIZE])
{
- const struct pipe_resource *texture = sview->base.texture;
- float dsdx = fabsf(s[QUAD_BOTTOM_RIGHT] - s[QUAD_BOTTOM_LEFT]);
- float dsdy = fabsf(s[QUAD_TOP_LEFT] - s[QUAD_BOTTOM_LEFT]);
- float rho = MAX2(dsdx, dsdy) * u_minify(texture->width0, sview->base.u.tex.first_level);
+ float derivs[3][2][TGSI_QUAD_SIZE];
+ compute_gradient_1d(s, t, p, derivs);
+ return compute_lambda_1d_explicit_gradients(sview, derivs, 0);
+}
+
+
+static void
+compute_gradient_2d(const float s[TGSI_QUAD_SIZE],
+ const float t[TGSI_QUAD_SIZE],
+ const float p[TGSI_QUAD_SIZE],
+ float derivs[3][2][TGSI_QUAD_SIZE])
+{
+ memset(derivs, 0, 6 * TGSI_QUAD_SIZE * sizeof(float));
+ derivs[0][0][0] = s[QUAD_BOTTOM_RIGHT] - s[QUAD_BOTTOM_LEFT];
+ derivs[0][1][0] = s[QUAD_TOP_LEFT] - s[QUAD_BOTTOM_LEFT];
+ derivs[1][0][0] = t[QUAD_BOTTOM_RIGHT] - t[QUAD_BOTTOM_LEFT];
+ derivs[1][1][0] = t[QUAD_TOP_LEFT] - t[QUAD_BOTTOM_LEFT];
+}
+static float
+compute_lambda_2d_explicit_gradients(const struct sp_sampler_view *sview,
+ const float derivs[3][2][TGSI_QUAD_SIZE],
+ uint quad)
+{
+ const struct pipe_resource *texture = sview->base.texture;
+ const float dsdx = fabsf(derivs[0][0][quad]);
+ const float dsdy = fabsf(derivs[0][1][quad]);
+ const float dtdx = fabsf(derivs[1][0][quad]);
+ const float dtdy = fabsf(derivs[1][1][quad]);
+ const float maxx = MAX2(dsdx, dsdy) * u_minify(texture->width0, sview->base.u.tex.first_level);
+ const float maxy = MAX2(dtdx, dtdy) * u_minify(texture->height0, sview->base.u.tex.first_level);
+ const float rho = MAX2(maxx, maxy);
return util_fast_log2(rho);
}
const float s[TGSI_QUAD_SIZE],
const float t[TGSI_QUAD_SIZE],
const float p[TGSI_QUAD_SIZE])
+{
+ float derivs[3][2][TGSI_QUAD_SIZE];
+ compute_gradient_2d(s, t, p, derivs);
+ return compute_lambda_2d_explicit_gradients(sview, derivs, 0);
+}
+
+
+static void
+compute_gradient_3d(const float s[TGSI_QUAD_SIZE],
+ const float t[TGSI_QUAD_SIZE],
+ const float p[TGSI_QUAD_SIZE],
+ float derivs[3][2][TGSI_QUAD_SIZE])
+{
+ memset(derivs, 0, 6 * TGSI_QUAD_SIZE * sizeof(float));
+ derivs[0][0][0] = fabsf(s[QUAD_BOTTOM_RIGHT] - s[QUAD_BOTTOM_LEFT]);
+ derivs[0][1][0] = fabsf(s[QUAD_TOP_LEFT] - s[QUAD_BOTTOM_LEFT]);
+ derivs[1][0][0] = fabsf(t[QUAD_BOTTOM_RIGHT] - t[QUAD_BOTTOM_LEFT]);
+ derivs[1][1][0] = fabsf(t[QUAD_TOP_LEFT] - t[QUAD_BOTTOM_LEFT]);
+ derivs[2][0][0] = fabsf(p[QUAD_BOTTOM_RIGHT] - p[QUAD_BOTTOM_LEFT]);
+ derivs[2][1][0] = fabsf(p[QUAD_TOP_LEFT] - p[QUAD_BOTTOM_LEFT]);
+}
+
+static float
+compute_lambda_3d_explicit_gradients(const struct sp_sampler_view *sview,
+ const float derivs[3][2][TGSI_QUAD_SIZE],
+ uint quad)
{
const struct pipe_resource *texture = sview->base.texture;
- float dsdx = fabsf(s[QUAD_BOTTOM_RIGHT] - s[QUAD_BOTTOM_LEFT]);
- float dsdy = fabsf(s[QUAD_TOP_LEFT] - s[QUAD_BOTTOM_LEFT]);
- float dtdx = fabsf(t[QUAD_BOTTOM_RIGHT] - t[QUAD_BOTTOM_LEFT]);
- float dtdy = fabsf(t[QUAD_TOP_LEFT] - t[QUAD_BOTTOM_LEFT]);
- float maxx = MAX2(dsdx, dsdy) * u_minify(texture->width0, sview->base.u.tex.first_level);
- float maxy = MAX2(dtdx, dtdy) * u_minify(texture->height0, sview->base.u.tex.first_level);
- float rho = MAX2(maxx, maxy);
+ const float dsdx = fabsf(derivs[0][0][quad]);
+ const float dsdy = fabsf(derivs[0][1][quad]);
+ const float dtdx = fabsf(derivs[1][0][quad]);
+ const float dtdy = fabsf(derivs[1][1][quad]);
+ const float dpdx = fabsf(derivs[2][0][quad]);
+ const float dpdy = fabsf(derivs[2][1][quad]);
+ const float maxx = MAX2(dsdx, dsdy) * u_minify(texture->width0, sview->base.u.tex.first_level);
+ const float maxy = MAX2(dtdx, dtdy) * u_minify(texture->height0, sview->base.u.tex.first_level);
+ const float maxz = MAX2(dpdx, dpdy) * u_minify(texture->depth0, sview->base.u.tex.first_level);
+ const float rho = MAX3(maxx, maxy, maxz);
return util_fast_log2(rho);
}
const float s[TGSI_QUAD_SIZE],
const float t[TGSI_QUAD_SIZE],
const float p[TGSI_QUAD_SIZE])
+{
+ float derivs[3][2][TGSI_QUAD_SIZE];
+ compute_gradient_3d(s, t, p, derivs);
+ return compute_lambda_3d_explicit_gradients(sview, derivs, 0);
+}
+
+
+static float
+compute_lambda_cube_explicit_gradients(const struct sp_sampler_view *sview,
+ const float derivs[3][2][TGSI_QUAD_SIZE],
+ uint quad)
{
const struct pipe_resource *texture = sview->base.texture;
- float dsdx = fabsf(s[QUAD_BOTTOM_RIGHT] - s[QUAD_BOTTOM_LEFT]);
- float dsdy = fabsf(s[QUAD_TOP_LEFT] - s[QUAD_BOTTOM_LEFT]);
- float dtdx = fabsf(t[QUAD_BOTTOM_RIGHT] - t[QUAD_BOTTOM_LEFT]);
- float dtdy = fabsf(t[QUAD_TOP_LEFT] - t[QUAD_BOTTOM_LEFT]);
- float dpdx = fabsf(p[QUAD_BOTTOM_RIGHT] - p[QUAD_BOTTOM_LEFT]);
- float dpdy = fabsf(p[QUAD_TOP_LEFT] - p[QUAD_BOTTOM_LEFT]);
- float maxx = MAX2(dsdx, dsdy) * u_minify(texture->width0, sview->base.u.tex.first_level);
- float maxy = MAX2(dtdx, dtdy) * u_minify(texture->height0, sview->base.u.tex.first_level);
- float maxz = MAX2(dpdx, dpdy) * u_minify(texture->depth0, sview->base.u.tex.first_level);
- float rho;
-
- rho = MAX2(maxx, maxy);
- rho = MAX2(rho, maxz);
+ const float dsdx = fabsf(derivs[0][0][quad]);
+ const float dsdy = fabsf(derivs[0][1][quad]);
+ const float dtdx = fabsf(derivs[1][0][quad]);
+ const float dtdy = fabsf(derivs[1][1][quad]);
+ const float dpdx = fabsf(derivs[2][0][quad]);
+ const float dpdy = fabsf(derivs[2][1][quad]);
+ const float maxx = MAX2(dsdx, dsdy);
+ const float maxy = MAX2(dtdx, dtdy);
+ const float maxz = MAX2(dpdx, dpdy);
+ const float rho = MAX3(maxx, maxy, maxz) * u_minify(texture->width0, sview->base.u.tex.first_level) / 2.0f;
return util_fast_log2(rho);
}
+static float
+compute_lambda_cube(const struct sp_sampler_view *sview,
+ const float s[TGSI_QUAD_SIZE],
+ const float t[TGSI_QUAD_SIZE],
+ const float p[TGSI_QUAD_SIZE])
+{
+ float derivs[3][2][TGSI_QUAD_SIZE];
+ compute_gradient_3d(s, t, p, derivs);
+ return compute_lambda_cube_explicit_gradients(sview, derivs, 0);
+}
/**
* Compute lambda for a vertex texture sampler.
}
+compute_lambda_from_grad_func
+softpipe_get_lambda_from_grad_func(const struct pipe_sampler_view *view,
+ enum pipe_shader_type shader)
+{
+ switch (view->target) {
+ case PIPE_BUFFER:
+ case PIPE_TEXTURE_1D:
+ case PIPE_TEXTURE_1D_ARRAY:
+ return compute_lambda_1d_explicit_gradients;
+ case PIPE_TEXTURE_2D:
+ case PIPE_TEXTURE_2D_ARRAY:
+ case PIPE_TEXTURE_RECT:
+ return compute_lambda_2d_explicit_gradients;
+ case PIPE_TEXTURE_CUBE:
+ case PIPE_TEXTURE_CUBE_ARRAY:
+ return compute_lambda_cube_explicit_gradients;
+ case PIPE_TEXTURE_3D:
+ return compute_lambda_3d_explicit_gradients;
+ default:
+ assert(0);
+ return compute_lambda_1d_explicit_gradients;
+ }
+}
+
/**
* Get a texel from a texture, using the texture tile cache.
+static inline const float *
+get_texel_buffer_no_border(const struct sp_sampler_view *sp_sview,
+ union tex_tile_address addr, int x, unsigned elmsize)
+{
+ const struct softpipe_tex_cached_tile *tile;
+ addr.bits.x = x * elmsize / TEX_TILE_SIZE;
+ assert(x * elmsize / TEX_TILE_SIZE == addr.bits.x);
+
+ x %= TEX_TILE_SIZE / elmsize;
+
+ tile = sp_get_cached_tile_tex(sp_sview->cache, addr);
+
+ return &tile->data.color[0][x][0];
+}
+
static inline const float *
get_texel_2d_no_border(const struct sp_sampler_view *sp_sview,
union tex_tile_address addr, int x, int y)
{
const struct pipe_resource *texture = sp_sview->base.texture;
- unsigned level = addr.bits.level;
+ const unsigned level = addr.bits.level;
if (x < 0 || x >= (int) u_minify(texture->width0, level) ||
y < 0 || y >= (int) u_minify(texture->height0, level)) {
- return sp_samp->base.border_color.f;
+ return sp_sview->border_color.f;
}
else {
return get_texel_2d_no_border( sp_sview, addr, x, y );
out[3] = get_texel_2d_no_border( sp_sview, addr, x1, y1 );
}
-/* Can involve a lot of unnecessary checks for border color:
- */
-static inline void
-get_texel_quad_2d(const struct sp_sampler_view *sp_sview,
- const struct sp_sampler *sp_samp,
- union tex_tile_address addr,
- int x0, int y0,
- int x1, int y1,
- const float *out[4])
-{
- out[0] = get_texel_2d( sp_sview, sp_samp, addr, x0, y0 );
- out[1] = get_texel_2d( sp_sview, sp_samp, addr, x1, y0 );
- out[3] = get_texel_2d( sp_sview, sp_samp, addr, x1, y1 );
- out[2] = get_texel_2d( sp_sview, sp_samp, addr, x0, y1 );
-}
-
-
/* 3d variants:
*/
union tex_tile_address addr, int x, int y, int z)
{
const struct pipe_resource *texture = sp_sview->base.texture;
- unsigned level = addr.bits.level;
+ const unsigned level = addr.bits.level;
if (x < 0 || x >= (int) u_minify(texture->width0, level) ||
y < 0 || y >= (int) u_minify(texture->height0, level) ||
z < 0 || z >= (int) u_minify(texture->depth0, level)) {
- return sp_samp->base.border_color.f;
+ return sp_sview->border_color.f;
}
else {
return get_texel_3d_no_border( sp_sview, addr, x, y, z );
union tex_tile_address addr, int x, int y)
{
const struct pipe_resource *texture = sp_sview->base.texture;
- unsigned level = addr.bits.level;
+ const unsigned level = addr.bits.level;
if (x < 0 || x >= (int) u_minify(texture->width0, level)) {
- return sp_samp->base.border_color.f;
+ return sp_sview->border_color.f;
}
else {
return get_texel_2d_no_border(sp_sview, addr, x, y);
union tex_tile_address addr, int x, int y, int layer)
{
const struct pipe_resource *texture = sp_sview->base.texture;
- unsigned level = addr.bits.level;
+ const unsigned level = addr.bits.level;
assert(layer < (int) texture->array_size);
assert(layer >= 0);
if (x < 0 || x >= (int) u_minify(texture->width0, level) ||
y < 0 || y >= (int) u_minify(texture->height0, level)) {
- return sp_samp->base.border_color.f;
+ return sp_sview->border_color.f;
}
else {
return get_texel_3d_no_border(sp_sview, addr, x, y, layer);
float *corner, int layer, unsigned face)
{
const struct pipe_resource *texture = sp_sview->base.texture;
- unsigned level = addr.bits.level;
+ const unsigned level = addr.bits.level;
int new_x, new_y, max_x;
max_x = (int) u_minify(texture->width0, level);
union tex_tile_address addr, int x, int y, int layer)
{
const struct pipe_resource *texture = sp_sview->base.texture;
- unsigned level = addr.bits.level;
+ const unsigned level = addr.bits.level;
assert(layer < (int) texture->array_size);
assert(layer >= 0);
if (x < 0 || x >= (int) u_minify(texture->width0, level) ||
y < 0 || y >= (int) u_minify(texture->height0, level)) {
- return sp_samp->base.border_color.f;
+ return sp_sview->border_color.f;
}
else {
return get_texel_3d_no_border(sp_sview, addr, x, y, layer);
/* Some image-filter fastpaths:
*/
static inline void
-img_filter_2d_linear_repeat_POT(struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
+img_filter_2d_linear_repeat_POT(const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
const struct img_filter_args *args,
float *rgba)
{
- unsigned xpot = pot_level_size(sp_sview->xpot, args->level);
- unsigned ypot = pot_level_size(sp_sview->ypot, args->level);
- int xmax = (xpot - 1) & (TEX_TILE_SIZE - 1); /* MIN2(TEX_TILE_SIZE, xpot) - 1; */
- int ymax = (ypot - 1) & (TEX_TILE_SIZE - 1); /* MIN2(TEX_TILE_SIZE, ypot) - 1; */
+ const unsigned xpot = pot_level_size(sp_sview->xpot, args->level);
+ const unsigned ypot = pot_level_size(sp_sview->ypot, args->level);
+ const int xmax = (xpot - 1) & (TEX_TILE_SIZE - 1); /* MIN2(TEX_TILE_SIZE, xpot) - 1; */
+ const int ymax = (ypot - 1) & (TEX_TILE_SIZE - 1); /* MIN2(TEX_TILE_SIZE, ypot) - 1; */
union tex_tile_address addr;
int c;
- float u = (args->s * xpot - 0.5F) + args->offset[0];
- float v = (args->t * ypot - 0.5F) + args->offset[1];
+ const float u = (args->s * xpot - 0.5F) + args->offset[0];
+ const float v = (args->t * ypot - 0.5F) + args->offset[1];
- int uflr = util_ifloor(u);
- int vflr = util_ifloor(v);
+ const int uflr = util_ifloor(u);
+ const int vflr = util_ifloor(v);
- float xw = u - (float)uflr;
- float yw = v - (float)vflr;
+ const float xw = u - (float)uflr;
+ const float yw = v - (float)vflr;
- int x0 = uflr & (xpot - 1);
- int y0 = vflr & (ypot - 1);
+ const int x0 = uflr & (xpot - 1);
+ const int y0 = vflr & (ypot - 1);
const float *tx[4];
addr.value = 0;
addr.bits.level = args->level;
+ addr.bits.z = sp_sview->base.u.tex.first_layer;
/* Can we fetch all four at once:
*/
get_texel_quad_2d_no_border_single_tile(sp_sview, addr, x0, y0, tx);
}
else {
- unsigned x1 = (x0 + 1) & (xpot - 1);
- unsigned y1 = (y0 + 1) & (ypot - 1);
+ const unsigned x1 = (x0 + 1) & (xpot - 1);
+ const unsigned y1 = (y0 + 1) & (ypot - 1);
get_texel_quad_2d_no_border(sp_sview, addr, x0, y0, x1, y1, tx);
}
/* interpolate R, G, B, A */
- for (c = 0; c < TGSI_QUAD_SIZE; c++) {
+ for (c = 0; c < TGSI_NUM_CHANNELS; c++) {
rgba[TGSI_NUM_CHANNELS*c] = lerp_2d(xw, yw,
tx[0][c], tx[1][c],
tx[2][c], tx[3][c]);
static inline void
-img_filter_2d_nearest_repeat_POT(struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
+img_filter_2d_nearest_repeat_POT(const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
const struct img_filter_args *args,
- float rgba[TGSI_QUAD_SIZE])
+ float *rgba)
{
- unsigned xpot = pot_level_size(sp_sview->xpot, args->level);
- unsigned ypot = pot_level_size(sp_sview->ypot, args->level);
+ const unsigned xpot = pot_level_size(sp_sview->xpot, args->level);
+ const unsigned ypot = pot_level_size(sp_sview->ypot, args->level);
const float *out;
union tex_tile_address addr;
int c;
- float u = args->s * xpot + args->offset[0];
- float v = args->t * ypot + args->offset[1];
+ const float u = args->s * xpot + args->offset[0];
+ const float v = args->t * ypot + args->offset[1];
- int uflr = util_ifloor(u);
- int vflr = util_ifloor(v);
+ const int uflr = util_ifloor(u);
+ const int vflr = util_ifloor(v);
- int x0 = uflr & (xpot - 1);
- int y0 = vflr & (ypot - 1);
+ const int x0 = uflr & (xpot - 1);
+ const int y0 = vflr & (ypot - 1);
addr.value = 0;
addr.bits.level = args->level;
+ addr.bits.z = sp_sview->base.u.tex.first_layer;
out = get_texel_2d_no_border(sp_sview, addr, x0, y0);
- for (c = 0; c < TGSI_QUAD_SIZE; c++)
+ for (c = 0; c < TGSI_NUM_CHANNELS; c++)
rgba[TGSI_NUM_CHANNELS*c] = out[c];
if (DEBUG_TEX) {
static inline void
-img_filter_2d_nearest_clamp_POT(struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
+img_filter_2d_nearest_clamp_POT(const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
const struct img_filter_args *args,
- float rgba[TGSI_QUAD_SIZE])
+ float *rgba)
{
- unsigned xpot = pot_level_size(sp_sview->xpot, args->level);
- unsigned ypot = pot_level_size(sp_sview->ypot, args->level);
+ const unsigned xpot = pot_level_size(sp_sview->xpot, args->level);
+ const unsigned ypot = pot_level_size(sp_sview->ypot, args->level);
union tex_tile_address addr;
int c;
- float u = args->s * xpot + args->offset[0];
- float v = args->t * ypot + args->offset[1];
+ const float u = args->s * xpot + args->offset[0];
+ const float v = args->t * ypot + args->offset[1];
int x0, y0;
const float *out;
addr.value = 0;
addr.bits.level = args->level;
+ addr.bits.z = sp_sview->base.u.tex.first_layer;
x0 = util_ifloor(u);
if (x0 < 0)
y0 = ypot - 1;
out = get_texel_2d_no_border(sp_sview, addr, x0, y0);
- for (c = 0; c < TGSI_QUAD_SIZE; c++)
+ for (c = 0; c < TGSI_NUM_CHANNELS; c++)
rgba[TGSI_NUM_CHANNELS*c] = out[c];
if (DEBUG_TEX) {
static void
-img_filter_1d_nearest(struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
+img_filter_1d_nearest(const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
const struct img_filter_args *args,
- float rgba[TGSI_QUAD_SIZE])
+ float *rgba)
{
const struct pipe_resource *texture = sp_sview->base.texture;
- int width;
+ const int width = u_minify(texture->width0, args->level);
int x;
union tex_tile_address addr;
const float *out;
int c;
- width = u_minify(texture->width0, args->level);
-
assert(width > 0);
addr.value = 0;
sp_samp->nearest_texcoord_s(args->s, width, args->offset[0], &x);
- out = get_texel_2d(sp_sview, sp_samp, addr, x, 0);
- for (c = 0; c < TGSI_QUAD_SIZE; c++)
+ out = get_texel_1d_array(sp_sview, sp_samp, addr, x,
+ sp_sview->base.u.tex.first_layer);
+ for (c = 0; c < TGSI_NUM_CHANNELS; c++)
rgba[TGSI_NUM_CHANNELS*c] = out[c];
if (DEBUG_TEX) {
static void
-img_filter_1d_array_nearest(struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
+img_filter_1d_array_nearest(const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
const struct img_filter_args *args,
float *rgba)
{
const struct pipe_resource *texture = sp_sview->base.texture;
- int width;
- int x, layer;
+ const int width = u_minify(texture->width0, args->level);
+ const int layer = coord_to_layer(args->t, sp_sview->base.u.tex.first_layer,
+ sp_sview->base.u.tex.last_layer);
+ int x;
union tex_tile_address addr;
const float *out;
int c;
- width = u_minify(texture->width0, args->level);
-
assert(width > 0);
addr.value = 0;
addr.bits.level = args->level;
sp_samp->nearest_texcoord_s(args->s, width, args->offset[0], &x);
- layer = coord_to_layer(args->t, sp_sview->base.u.tex.first_layer,
- sp_sview->base.u.tex.last_layer);
out = get_texel_1d_array(sp_sview, sp_samp, addr, x, layer);
- for (c = 0; c < TGSI_QUAD_SIZE; c++)
+ for (c = 0; c < TGSI_NUM_CHANNELS; c++)
rgba[TGSI_NUM_CHANNELS*c] = out[c];
if (DEBUG_TEX) {
static void
-img_filter_2d_nearest(struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
+img_filter_2d_nearest(const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
const struct img_filter_args *args,
float *rgba)
{
const struct pipe_resource *texture = sp_sview->base.texture;
- int width, height;
+ const int width = u_minify(texture->width0, args->level);
+ const int height = u_minify(texture->height0, args->level);
int x, y;
union tex_tile_address addr;
const float *out;
int c;
- width = u_minify(texture->width0, args->level);
- height = u_minify(texture->height0, args->level);
-
assert(width > 0);
assert(height > 0);
addr.value = 0;
addr.bits.level = args->level;
+ addr.bits.z = sp_sview->base.u.tex.first_layer;
sp_samp->nearest_texcoord_s(args->s, width, args->offset[0], &x);
sp_samp->nearest_texcoord_t(args->t, height, args->offset[1], &y);
out = get_texel_2d(sp_sview, sp_samp, addr, x, y);
- for (c = 0; c < TGSI_QUAD_SIZE; c++)
+ for (c = 0; c < TGSI_NUM_CHANNELS; c++)
rgba[TGSI_NUM_CHANNELS*c] = out[c];
if (DEBUG_TEX) {
static void
-img_filter_2d_array_nearest(struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
+img_filter_2d_array_nearest(const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
const struct img_filter_args *args,
float *rgba)
{
const struct pipe_resource *texture = sp_sview->base.texture;
- int width, height;
- int x, y, layer;
+ const int width = u_minify(texture->width0, args->level);
+ const int height = u_minify(texture->height0, args->level);
+ const int layer = coord_to_layer(args->p, sp_sview->base.u.tex.first_layer,
+ sp_sview->base.u.tex.last_layer);
+ int x, y;
union tex_tile_address addr;
const float *out;
int c;
- width = u_minify(texture->width0, args->level);
- height = u_minify(texture->height0, args->level);
-
assert(width > 0);
assert(height > 0);
sp_samp->nearest_texcoord_s(args->s, width, args->offset[0], &x);
sp_samp->nearest_texcoord_t(args->t, height, args->offset[1], &y);
- layer = coord_to_layer(args->p, sp_sview->base.u.tex.first_layer,
- sp_sview->base.u.tex.last_layer);
out = get_texel_2d_array(sp_sview, sp_samp, addr, x, y, layer);
- for (c = 0; c < TGSI_QUAD_SIZE; c++)
+ for (c = 0; c < TGSI_NUM_CHANNELS; c++)
rgba[TGSI_NUM_CHANNELS*c] = out[c];
if (DEBUG_TEX) {
static void
-img_filter_cube_nearest(struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
+img_filter_cube_nearest(const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
const struct img_filter_args *args,
float *rgba)
{
const struct pipe_resource *texture = sp_sview->base.texture;
- int width, height;
- int x, y, layerface;
+ const int width = u_minify(texture->width0, args->level);
+ const int height = u_minify(texture->height0, args->level);
+ const int layerface = args->face_id + sp_sview->base.u.tex.first_layer;
+ int x, y;
union tex_tile_address addr;
const float *out;
int c;
- width = u_minify(texture->width0, args->level);
- height = u_minify(texture->height0, args->level);
-
assert(width > 0);
assert(height > 0);
sp_samp->nearest_texcoord_t(args->t, height, args->offset[1], &y);
}
- layerface = args->face_id + sp_sview->base.u.tex.first_layer;
out = get_texel_cube_array(sp_sview, sp_samp, addr, x, y, layerface);
- for (c = 0; c < TGSI_QUAD_SIZE; c++)
+ for (c = 0; c < TGSI_NUM_CHANNELS; c++)
rgba[TGSI_NUM_CHANNELS*c] = out[c];
if (DEBUG_TEX) {
}
static void
-img_filter_cube_array_nearest(struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
+img_filter_cube_array_nearest(const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
const struct img_filter_args *args,
float *rgba)
{
const struct pipe_resource *texture = sp_sview->base.texture;
- int width, height;
- int x, y, layerface;
+ const int width = u_minify(texture->width0, args->level);
+ const int height = u_minify(texture->height0, args->level);
+ const int layerface = CLAMP(6 * util_ifloor(args->p + 0.5f) + sp_sview->base.u.tex.first_layer,
+ sp_sview->base.u.tex.first_layer,
+ sp_sview->base.u.tex.last_layer - 5) + args->face_id;
+ int x, y;
union tex_tile_address addr;
const float *out;
int c;
- width = u_minify(texture->width0, args->level);
- height = u_minify(texture->height0, args->level);
-
assert(width > 0);
assert(height > 0);
sp_samp->nearest_texcoord_s(args->s, width, args->offset[0], &x);
sp_samp->nearest_texcoord_t(args->t, height, args->offset[1], &y);
- layerface = coord_to_layer(6 * args->p + sp_sview->base.u.tex.first_layer,
- sp_sview->base.u.tex.first_layer,
- sp_sview->base.u.tex.last_layer - 5) + args->face_id;
out = get_texel_cube_array(sp_sview, sp_samp, addr, x, y, layerface);
- for (c = 0; c < TGSI_QUAD_SIZE; c++)
+ for (c = 0; c < TGSI_NUM_CHANNELS; c++)
rgba[TGSI_NUM_CHANNELS*c] = out[c];
if (DEBUG_TEX) {
}
static void
-img_filter_3d_nearest(struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
+img_filter_3d_nearest(const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
const struct img_filter_args *args,
float *rgba)
{
const struct pipe_resource *texture = sp_sview->base.texture;
- int width, height, depth;
+ const int width = u_minify(texture->width0, args->level);
+ const int height = u_minify(texture->height0, args->level);
+ const int depth = u_minify(texture->depth0, args->level);
int x, y, z;
union tex_tile_address addr;
const float *out;
int c;
- width = u_minify(texture->width0, args->level);
- height = u_minify(texture->height0, args->level);
- depth = u_minify(texture->depth0, args->level);
-
assert(width > 0);
assert(height > 0);
assert(depth > 0);
addr.bits.level = args->level;
out = get_texel_3d(sp_sview, sp_samp, addr, x, y, z);
- for (c = 0; c < TGSI_QUAD_SIZE; c++)
+ for (c = 0; c < TGSI_NUM_CHANNELS; c++)
rgba[TGSI_NUM_CHANNELS*c] = out[c];
}
static void
-img_filter_1d_linear(struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
+img_filter_1d_linear(const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
const struct img_filter_args *args,
float *rgba)
{
const struct pipe_resource *texture = sp_sview->base.texture;
- int width;
+ const int width = u_minify(texture->width0, args->level);
int x0, x1;
float xw; /* weights */
union tex_tile_address addr;
const float *tx0, *tx1;
int c;
- width = u_minify(texture->width0, args->level);
-
assert(width > 0);
addr.value = 0;
sp_samp->linear_texcoord_s(args->s, width, args->offset[0], &x0, &x1, &xw);
- tx0 = get_texel_2d(sp_sview, sp_samp, addr, x0, 0);
- tx1 = get_texel_2d(sp_sview, sp_samp, addr, x1, 0);
+ tx0 = get_texel_1d_array(sp_sview, sp_samp, addr, x0,
+ sp_sview->base.u.tex.first_layer);
+ tx1 = get_texel_1d_array(sp_sview, sp_samp, addr, x1,
+ sp_sview->base.u.tex.first_layer);
/* interpolate R, G, B, A */
- for (c = 0; c < TGSI_QUAD_SIZE; c++)
+ for (c = 0; c < TGSI_NUM_CHANNELS; c++)
rgba[TGSI_NUM_CHANNELS*c] = lerp(xw, tx0[c], tx1[c]);
}
static void
-img_filter_1d_array_linear(struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
+img_filter_1d_array_linear(const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
const struct img_filter_args *args,
float *rgba)
{
const struct pipe_resource *texture = sp_sview->base.texture;
- int width;
- int x0, x1, layer;
+ const int width = u_minify(texture->width0, args->level);
+ const int layer = coord_to_layer(args->t, sp_sview->base.u.tex.first_layer,
+ sp_sview->base.u.tex.last_layer);
+ int x0, x1;
float xw; /* weights */
union tex_tile_address addr;
const float *tx0, *tx1;
int c;
- width = u_minify(texture->width0, args->level);
-
assert(width > 0);
addr.value = 0;
addr.bits.level = args->level;
sp_samp->linear_texcoord_s(args->s, width, args->offset[0], &x0, &x1, &xw);
- layer = coord_to_layer(args->t, sp_sview->base.u.tex.first_layer,
- sp_sview->base.u.tex.last_layer);
tx0 = get_texel_1d_array(sp_sview, sp_samp, addr, x0, layer);
tx1 = get_texel_1d_array(sp_sview, sp_samp, addr, x1, layer);
/* interpolate R, G, B, A */
- for (c = 0; c < TGSI_QUAD_SIZE; c++)
+ for (c = 0; c < TGSI_NUM_CHANNELS; c++)
rgba[TGSI_NUM_CHANNELS*c] = lerp(xw, tx0[c], tx1[c]);
}
/* get correct result using the channel and swizzle */
switch (swizzle) {
- case PIPE_SWIZZLE_ZERO:
+ case PIPE_SWIZZLE_0:
return 0.0;
- case PIPE_SWIZZLE_ONE:
+ case PIPE_SWIZZLE_1:
return 1.0;
default:
return tx[chan][swizzle];
static void
-img_filter_2d_linear(struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
+img_filter_2d_linear(const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
const struct img_filter_args *args,
float *rgba)
{
const struct pipe_resource *texture = sp_sview->base.texture;
- int width, height;
+ const int width = u_minify(texture->width0, args->level);
+ const int height = u_minify(texture->height0, args->level);
int x0, y0, x1, y1;
float xw, yw; /* weights */
union tex_tile_address addr;
const float *tx[4];
int c;
- width = u_minify(texture->width0, args->level);
- height = u_minify(texture->height0, args->level);
-
assert(width > 0);
assert(height > 0);
addr.value = 0;
addr.bits.level = args->level;
+ addr.bits.z = sp_sview->base.u.tex.first_layer;
sp_samp->linear_texcoord_s(args->s, width, args->offset[0], &x0, &x1, &xw);
sp_samp->linear_texcoord_t(args->t, height, args->offset[1], &y0, &y1, &yw);
tx[3] = get_texel_2d(sp_sview, sp_samp, addr, x1, y1);
if (args->gather_only) {
- for (c = 0; c < TGSI_QUAD_SIZE; c++)
+ for (c = 0; c < TGSI_NUM_CHANNELS; c++)
rgba[TGSI_NUM_CHANNELS*c] = get_gather_value(sp_sview, c,
args->gather_comp,
tx);
} else {
/* interpolate R, G, B, A */
- for (c = 0; c < TGSI_QUAD_SIZE; c++)
+ for (c = 0; c < TGSI_NUM_CHANNELS; c++)
rgba[TGSI_NUM_CHANNELS*c] = lerp_2d(xw, yw,
tx[0][c], tx[1][c],
tx[2][c], tx[3][c]);
static void
-img_filter_2d_array_linear(struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
+img_filter_2d_array_linear(const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
const struct img_filter_args *args,
float *rgba)
{
const struct pipe_resource *texture = sp_sview->base.texture;
- int width, height;
- int x0, y0, x1, y1, layer;
+ const int width = u_minify(texture->width0, args->level);
+ const int height = u_minify(texture->height0, args->level);
+ const int layer = coord_to_layer(args->p, sp_sview->base.u.tex.first_layer,
+ sp_sview->base.u.tex.last_layer);
+ int x0, y0, x1, y1;
float xw, yw; /* weights */
union tex_tile_address addr;
const float *tx[4];
int c;
- width = u_minify(texture->width0, args->level);
- height = u_minify(texture->height0, args->level);
-
assert(width > 0);
assert(height > 0);
sp_samp->linear_texcoord_s(args->s, width, args->offset[0], &x0, &x1, &xw);
sp_samp->linear_texcoord_t(args->t, height, args->offset[1], &y0, &y1, &yw);
- layer = coord_to_layer(args->p, sp_sview->base.u.tex.first_layer,
- sp_sview->base.u.tex.last_layer);
tx[0] = get_texel_2d_array(sp_sview, sp_samp, addr, x0, y0, layer);
tx[1] = get_texel_2d_array(sp_sview, sp_samp, addr, x1, y0, layer);
tx[3] = get_texel_2d_array(sp_sview, sp_samp, addr, x1, y1, layer);
if (args->gather_only) {
- for (c = 0; c < TGSI_QUAD_SIZE; c++)
+ for (c = 0; c < TGSI_NUM_CHANNELS; c++)
rgba[TGSI_NUM_CHANNELS*c] = get_gather_value(sp_sview, c,
args->gather_comp,
tx);
} else {
/* interpolate R, G, B, A */
- for (c = 0; c < TGSI_QUAD_SIZE; c++)
+ for (c = 0; c < TGSI_NUM_CHANNELS; c++)
rgba[TGSI_NUM_CHANNELS*c] = lerp_2d(xw, yw,
tx[0][c], tx[1][c],
tx[2][c], tx[3][c]);
static void
-img_filter_cube_linear(struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
+img_filter_cube_linear(const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
const struct img_filter_args *args,
float *rgba)
{
const struct pipe_resource *texture = sp_sview->base.texture;
- int width, height;
- int x0, y0, x1, y1, layer;
+ const int width = u_minify(texture->width0, args->level);
+ const int height = u_minify(texture->height0, args->level);
+ const int layer = sp_sview->base.u.tex.first_layer;
+ int x0, y0, x1, y1;
float xw, yw; /* weights */
union tex_tile_address addr;
const float *tx[4];
corner2[TGSI_QUAD_SIZE], corner3[TGSI_QUAD_SIZE];
int c;
- width = u_minify(texture->width0, args->level);
- height = u_minify(texture->height0, args->level);
-
assert(width > 0);
assert(height > 0);
sp_samp->linear_texcoord_t(args->t, height, args->offset[1], &y0, &y1, &yw);
}
- layer = sp_sview->base.u.tex.first_layer;
-
if (sp_samp->base.seamless_cube_map) {
tx[0] = get_texel_cube_seamless(sp_sview, addr, x0, y0, corner0, layer, args->face_id);
tx[1] = get_texel_cube_seamless(sp_sview, addr, x1, y0, corner1, layer, args->face_id);
}
if (args->gather_only) {
- for (c = 0; c < TGSI_QUAD_SIZE; c++)
+ for (c = 0; c < TGSI_NUM_CHANNELS; c++)
rgba[TGSI_NUM_CHANNELS*c] = get_gather_value(sp_sview, c,
args->gather_comp,
tx);
} else {
/* interpolate R, G, B, A */
- for (c = 0; c < TGSI_QUAD_SIZE; c++)
+ for (c = 0; c < TGSI_NUM_CHANNELS; c++)
rgba[TGSI_NUM_CHANNELS*c] = lerp_2d(xw, yw,
tx[0][c], tx[1][c],
tx[2][c], tx[3][c]);
static void
-img_filter_cube_array_linear(struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
+img_filter_cube_array_linear(const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
const struct img_filter_args *args,
float *rgba)
{
const struct pipe_resource *texture = sp_sview->base.texture;
- int width, height;
- int x0, y0, x1, y1, layer;
+ const int width = u_minify(texture->width0, args->level);
+ const int height = u_minify(texture->height0, args->level);
+
+ const int layer = CLAMP(6 * util_ifloor(args->p + 0.5f) + sp_sview->base.u.tex.first_layer,
+ sp_sview->base.u.tex.first_layer,
+ sp_sview->base.u.tex.last_layer - 5);
+
+ int x0, y0, x1, y1;
float xw, yw; /* weights */
union tex_tile_address addr;
const float *tx[4];
corner2[TGSI_QUAD_SIZE], corner3[TGSI_QUAD_SIZE];
int c;
- width = u_minify(texture->width0, args->level);
- height = u_minify(texture->height0, args->level);
-
assert(width > 0);
assert(height > 0);
sp_samp->linear_texcoord_t(args->t, height, args->offset[1], &y0, &y1, &yw);
}
- layer = coord_to_layer(6 * args->p + sp_sview->base.u.tex.first_layer,
- sp_sview->base.u.tex.first_layer,
- sp_sview->base.u.tex.last_layer - 5);
-
if (sp_samp->base.seamless_cube_map) {
tx[0] = get_texel_cube_seamless(sp_sview, addr, x0, y0, corner0, layer, args->face_id);
tx[1] = get_texel_cube_seamless(sp_sview, addr, x1, y0, corner1, layer, args->face_id);
}
if (args->gather_only) {
- for (c = 0; c < TGSI_QUAD_SIZE; c++)
+ for (c = 0; c < TGSI_NUM_CHANNELS; c++)
rgba[TGSI_NUM_CHANNELS*c] = get_gather_value(sp_sview, c,
args->gather_comp,
tx);
} else {
/* interpolate R, G, B, A */
- for (c = 0; c < TGSI_QUAD_SIZE; c++)
+ for (c = 0; c < TGSI_NUM_CHANNELS; c++)
rgba[TGSI_NUM_CHANNELS*c] = lerp_2d(xw, yw,
tx[0][c], tx[1][c],
tx[2][c], tx[3][c]);
}
static void
-img_filter_3d_linear(struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
+img_filter_3d_linear(const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
const struct img_filter_args *args,
float *rgba)
{
const struct pipe_resource *texture = sp_sview->base.texture;
- int width, height, depth;
+ const int width = u_minify(texture->width0, args->level);
+ const int height = u_minify(texture->height0, args->level);
+ const int depth = u_minify(texture->depth0, args->level);
int x0, x1, y0, y1, z0, z1;
float xw, yw, zw; /* interpolation weights */
union tex_tile_address addr;
const float *tx00, *tx01, *tx02, *tx03, *tx10, *tx11, *tx12, *tx13;
int c;
- width = u_minify(texture->width0, args->level);
- height = u_minify(texture->height0, args->level);
- depth = u_minify(texture->depth0, args->level);
-
addr.value = 0;
addr.bits.level = args->level;
tx13 = get_texel_3d(sp_sview, sp_samp, addr, x1, y1, z1);
/* interpolate R, G, B, A */
- for (c = 0; c < TGSI_QUAD_SIZE; c++)
+ for (c = 0; c < TGSI_NUM_CHANNELS; c++)
rgba[TGSI_NUM_CHANNELS*c] = lerp_3d(xw, yw, zw,
tx00[c], tx01[c],
tx02[c], tx03[c],
const float lod_in[TGSI_QUAD_SIZE],
float lod[TGSI_QUAD_SIZE])
{
- float min_lod = sampler->min_lod;
- float max_lod = sampler->max_lod;
+ const float min_lod = sampler->min_lod;
+ const float max_lod = sampler->max_lod;
uint i;
switch (control) {
case TGSI_SAMPLER_LOD_NONE:
case TGSI_SAMPLER_LOD_ZERO:
- /* XXX FIXME */
- case TGSI_SAMPLER_DERIVS_EXPLICIT:
lod[0] = lod[1] = lod[2] = lod[3] = CLAMP(biased_lambda, min_lod, max_lod);
break;
+ case TGSI_SAMPLER_DERIVS_EXPLICIT:
+ for (i = 0; i < TGSI_QUAD_SIZE; i++)
+ lod[i] = lod_in[i];
+ break;
case TGSI_SAMPLER_LOD_BIAS:
for (i = 0; i < TGSI_QUAD_SIZE; i++) {
lod[i] = biased_lambda + lod_in[i];
const float s[TGSI_QUAD_SIZE],
const float t[TGSI_QUAD_SIZE],
const float p[TGSI_QUAD_SIZE],
+ const float derivs[3][2][TGSI_QUAD_SIZE],
const float lod_in[TGSI_QUAD_SIZE],
enum tgsi_sampler_control control,
float lod[TGSI_QUAD_SIZE])
switch (control) {
case TGSI_SAMPLER_LOD_NONE:
- /* XXX FIXME */
- case TGSI_SAMPLER_DERIVS_EXPLICIT:
lambda = sp_sview->compute_lambda(sp_sview, s, t, p) + lod_bias;
lod[0] = lod[1] = lod[2] = lod[3] = lambda;
break;
+ case TGSI_SAMPLER_DERIVS_EXPLICIT:
+ for (i = 0; i < TGSI_QUAD_SIZE; i++)
+ lod[i] = sp_sview->compute_lambda_from_grad(sp_sview, derivs, i);
+ break;
case TGSI_SAMPLER_LOD_BIAS:
lambda = sp_sview->compute_lambda(sp_sview, s, t, p) + lod_bias;
for (i = 0; i < TGSI_QUAD_SIZE; i++) {
* \param lod results per-fragment lod.
*/
static inline void
-compute_lambda_lod(struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
+compute_lambda_lod(const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
const float s[TGSI_QUAD_SIZE],
const float t[TGSI_QUAD_SIZE],
const float p[TGSI_QUAD_SIZE],
+ float derivs[3][2][TGSI_QUAD_SIZE],
const float lod_in[TGSI_QUAD_SIZE],
enum tgsi_sampler_control control,
float lod[TGSI_QUAD_SIZE])
int i;
compute_lambda_lod_unclamped(sp_sview, sp_samp,
- s, t, p, lod_in, control, lod);
+ s, t, p, derivs, lod_in, control, lod);
for (i = 0; i < TGSI_QUAD_SIZE; i++) {
lod[i] = CLAMP(lod[i], min_lod, max_lod);
}
}
static void
-mip_filter_linear(struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
+mip_filter_linear(const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
img_filter_func min_filter,
img_filter_func mag_filter,
const float s[TGSI_QUAD_SIZE],
const float t[TGSI_QUAD_SIZE],
const float p[TGSI_QUAD_SIZE],
- const float c0[TGSI_QUAD_SIZE],
- const float lod_in[TGSI_QUAD_SIZE],
+ int gather_comp,
+ const float lod[TGSI_QUAD_SIZE],
const struct filter_args *filt_args,
float rgba[TGSI_NUM_CHANNELS][TGSI_QUAD_SIZE])
{
const struct pipe_sampler_view *psview = &sp_sview->base;
int j;
- float lod[TGSI_QUAD_SIZE];
struct img_filter_args args;
- compute_lambda_lod(sp_sview, sp_samp, s, t, p, lod_in, filt_args->control, lod);
-
args.offset = filt_args->offset;
args.gather_only = filt_args->control == TGSI_SAMPLER_GATHER;
- args.gather_comp = get_gather_component(lod_in);
+ args.gather_comp = gather_comp;
for (j = 0; j < TGSI_QUAD_SIZE; j++) {
- int level0 = psview->u.tex.first_level + (int)lod[j];
+ const int level0 = psview->u.tex.first_level + (int)lod[j];
args.s = s[j];
args.t = t[j];
args.p = p[j];
args.face_id = filt_args->faces[j];
- if (lod[j] < 0.0) {
+ if (lod[j] <= 0.0 && !args.gather_only) {
args.level = psview->u.tex.first_level;
mag_filter(sp_sview, sp_samp, &args, &rgba[0][j]);
}
* \param c0 the LOD bias factors, or absolute LODs (depending on control)
*/
static void
-mip_filter_nearest(struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
+mip_filter_nearest(const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
img_filter_func min_filter,
img_filter_func mag_filter,
const float s[TGSI_QUAD_SIZE],
const float t[TGSI_QUAD_SIZE],
const float p[TGSI_QUAD_SIZE],
- const float c0[TGSI_QUAD_SIZE],
- const float lod_in[TGSI_QUAD_SIZE],
+ int gather_component,
+ const float lod[TGSI_QUAD_SIZE],
const struct filter_args *filt_args,
float rgba[TGSI_NUM_CHANNELS][TGSI_QUAD_SIZE])
{
const struct pipe_sampler_view *psview = &sp_sview->base;
- float lod[TGSI_QUAD_SIZE];
int j;
struct img_filter_args args;
args.offset = filt_args->offset;
args.gather_only = filt_args->control == TGSI_SAMPLER_GATHER;
- args.gather_comp = get_gather_component(lod_in);
-
- compute_lambda_lod(sp_sview, sp_samp, s, t, p, lod_in, filt_args->control, lod);
+ args.gather_comp = gather_component;
for (j = 0; j < TGSI_QUAD_SIZE; j++) {
args.s = s[j];
args.p = p[j];
args.face_id = filt_args->faces[j];
- if (lod[j] < 0.0) {
+ if (lod[j] <= 0.0f && !args.gather_only) {
args.level = psview->u.tex.first_level;
mag_filter(sp_sview, sp_samp, &args, &rgba[0][j]);
} else {
- int level = psview->u.tex.first_level + (int)(lod[j] + 0.5F);
+ const int level = psview->u.tex.first_level + (int)(lod[j] + 0.5F);
args.level = MIN2(level, (int)psview->u.tex.last_level);
min_filter(sp_sview, sp_samp, &args, &rgba[0][j]);
}
}
static void
-mip_filter_none(struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
+mip_filter_none(const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
img_filter_func min_filter,
img_filter_func mag_filter,
const float s[TGSI_QUAD_SIZE],
const float t[TGSI_QUAD_SIZE],
const float p[TGSI_QUAD_SIZE],
- const float c0[TGSI_QUAD_SIZE],
- const float lod_in[TGSI_QUAD_SIZE],
+ int gather_component,
+ const float lod[TGSI_QUAD_SIZE],
const struct filter_args *filt_args,
float rgba[TGSI_NUM_CHANNELS][TGSI_QUAD_SIZE])
{
- float lod[TGSI_QUAD_SIZE];
int j;
struct img_filter_args args;
args.level = sp_sview->base.u.tex.first_level;
args.offset = filt_args->offset;
args.gather_only = filt_args->control == TGSI_SAMPLER_GATHER;
-
- compute_lambda_lod(sp_sview, sp_samp, s, t, p, lod_in, filt_args->control, lod);
+ args.gather_comp = gather_component;
for (j = 0; j < TGSI_QUAD_SIZE; j++) {
args.s = s[j];
args.t = t[j];
args.p = p[j];
args.face_id = filt_args->faces[j];
- if (lod[j] < 0.0) {
+ if (lod[j] <= 0.0f && !args.gather_only) {
mag_filter(sp_sview, sp_samp, &args, &rgba[0][j]);
}
else {
}
static void
-mip_filter_none_no_filter_select(struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
+mip_filter_none_no_filter_select(const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
img_filter_func min_filter,
img_filter_func mag_filter,
const float s[TGSI_QUAD_SIZE],
const float t[TGSI_QUAD_SIZE],
const float p[TGSI_QUAD_SIZE],
- const float c0[TGSI_QUAD_SIZE],
+ int gather_comp,
const float lod_in[TGSI_QUAD_SIZE],
const struct filter_args *filt_args,
float rgba[TGSI_NUM_CHANNELS][TGSI_QUAD_SIZE])
args.level = sp_sview->base.u.tex.first_level;
args.offset = filt_args->offset;
args.gather_only = filt_args->control == TGSI_SAMPLER_GATHER;
+ args.gather_comp = gather_comp;
for (j = 0; j < TGSI_QUAD_SIZE; j++) {
args.s = s[j];
args.t = t[j];
/* For anisotropic filtering */
#define WEIGHT_LUT_SIZE 1024
-static float *weightLut = NULL;
+static const float *weightLut = NULL;
/**
* Creates the look-up table used to speed-up EWA sampling
{
unsigned i;
if (!weightLut) {
- weightLut = (float *) MALLOC(WEIGHT_LUT_SIZE * sizeof(float));
+ float *lut = (float *) MALLOC(WEIGHT_LUT_SIZE * sizeof(float));
for (i = 0; i < WEIGHT_LUT_SIZE; ++i) {
- float alpha = 2;
- float r2 = (float) i / (float) (WEIGHT_LUT_SIZE - 1);
- float weight = (float) exp(-alpha * r2);
- weightLut[i] = weight;
+ const float alpha = 2;
+ const float r2 = (float) i / (float) (WEIGHT_LUT_SIZE - 1);
+ const float weight = (float) expf(-alpha * r2);
+ lut[i] = weight;
}
+ weightLut = lut;
}
}
* "Fundamentals of Texture Mapping and Image Warping" (1989)
*/
static void
-img_filter_2d_ewa(struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
+img_filter_2d_ewa(const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
img_filter_func min_filter,
img_filter_func mag_filter,
const float s[TGSI_QUAD_SIZE],
const float t[TGSI_QUAD_SIZE],
const float p[TGSI_QUAD_SIZE],
- const float faces[TGSI_QUAD_SIZE],
+ const uint faces[TGSI_QUAD_SIZE],
+ const int8_t *offset,
unsigned level,
const float dudx, const float dvdx,
const float dudy, const float dvdy,
const struct pipe_resource *texture = sp_sview->base.texture;
// ??? Won't the image filters blow up if level is negative?
- unsigned level0 = level > 0 ? level : 0;
- float scaling = 1.0f / (1 << level0);
- int width = u_minify(texture->width0, level0);
- int height = u_minify(texture->height0, level0);
+ const unsigned level0 = level > 0 ? level : 0;
+ const float scaling = 1.0f / (1 << level0);
+ const int width = u_minify(texture->width0, level0);
+ const int height = u_minify(texture->height0, level0);
struct img_filter_args args;
- float ux = dudx * scaling;
- float vx = dvdx * scaling;
- float uy = dudy * scaling;
- float vy = dvdy * scaling;
+ const float ux = dudx * scaling;
+ const float vx = dvdx * scaling;
+ const float uy = dudy * scaling;
+ const float vy = dvdy * scaling;
/* compute ellipse coefficients to bound the region:
* A*x*x + B*x*y + C*y*y = F.
/* assert(F > 0.0); */
/* Compute the ellipse's (u,v) bounding box in texture space */
- float d = -B*B+4.0f*C*A;
- float box_u = 2.0f / d * sqrtf(d*C*F); /* box_u -> half of bbox with */
- float box_v = 2.0f / d * sqrtf(A*d*F); /* box_v -> half of bbox height */
+ const float d = -B*B+4.0f*C*A;
+ const float box_u = 2.0f / d * sqrtf(d*C*F); /* box_u -> half of bbox with */
+ const float box_v = 2.0f / d * sqrtf(A*d*F); /* box_v -> half of bbox height */
float rgba_temp[TGSI_NUM_CHANNELS][TGSI_QUAD_SIZE];
float s_buffer[TGSI_QUAD_SIZE];
float t_buffer[TGSI_QUAD_SIZE];
float weight_buffer[TGSI_QUAD_SIZE];
- unsigned buffer_next;
int j;
- float den; /* = 0.0F; */
- float ddq;
- float U; /* = u0 - tex_u; */
- int v;
-
- /* Scale ellipse formula to directly index the Filter Lookup Table.
- * i.e. scale so that F = WEIGHT_LUT_SIZE-1
- */
- double formScale = (double) (WEIGHT_LUT_SIZE - 1) / F;
- A *= formScale;
- B *= formScale;
- C *= formScale;
- /* F *= formScale; */ /* no need to scale F as we don't use it below here */
/* For each quad, the du and dx values are the same and so the ellipse is
* also the same. Note that texel/image access can only be performed using
* using the s_buffer/t_buffer and weight_buffer. Only when the buffer is
* full, then the pixel values are read from the image.
*/
- ddq = 2 * A;
+ const float ddq = 2 * A;
+
+ /* Scale ellipse formula to directly index the Filter Lookup Table.
+ * i.e. scale so that F = WEIGHT_LUT_SIZE-1
+ */
+ const double formScale = (double) (WEIGHT_LUT_SIZE - 1) / F;
+ A *= formScale;
+ B *= formScale;
+ C *= formScale;
+ /* F *= formScale; */ /* no need to scale F as we don't use it below here */
args.level = level;
+ args.offset = offset;
+
for (j = 0; j < TGSI_QUAD_SIZE; j++) {
/* Heckbert MS thesis, p. 59; scan over the bounding box of the ellipse
* and incrementally update the value of Ax^2+Bxy*Cy^2; when this
* value, q, is less than F, we're inside the ellipse
*/
- float tex_u = -0.5F + s[j] * texture->width0 * scaling;
- float tex_v = -0.5F + t[j] * texture->height0 * scaling;
+ const float tex_u = -0.5F + s[j] * texture->width0 * scaling;
+ const float tex_v = -0.5F + t[j] * texture->height0 * scaling;
- int u0 = (int) floorf(tex_u - box_u);
- int u1 = (int) ceilf(tex_u + box_u);
- int v0 = (int) floorf(tex_v - box_v);
- int v1 = (int) ceilf(tex_v + box_v);
+ const int u0 = (int) floorf(tex_u - box_u);
+ const int u1 = (int) ceilf(tex_u + box_u);
+ const int v0 = (int) floorf(tex_v - box_v);
+ const int v1 = (int) ceilf(tex_v + box_v);
+ const float U = u0 - tex_u;
float num[4] = {0.0F, 0.0F, 0.0F, 0.0F};
- buffer_next = 0;
- den = 0;
+ unsigned buffer_next = 0;
+ float den = 0;
+ int v;
args.face_id = faces[j];
- U = u0 - tex_u;
for (v = v0; v <= v1; ++v) {
- float V = v - tex_v;
+ const float V = v - tex_v;
float dq = A * (2 * U + 1) + B * V;
float q = (C * V + B * U) * V + A * U * U;
* should not happen, though
*/
const int qClamped = q >= 0.0F ? q : 0;
- float weight = weightLut[qClamped];
+ const float weight = weightLut[qClamped];
weight_buffer[buffer_next] = weight;
s_buffer[buffer_next] = u / ((float) width);
* Sample 2D texture using an anisotropic filter.
*/
static void
-mip_filter_linear_aniso(struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
+mip_filter_linear_aniso(const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
img_filter_func min_filter,
img_filter_func mag_filter,
const float s[TGSI_QUAD_SIZE],
const float t[TGSI_QUAD_SIZE],
const float p[TGSI_QUAD_SIZE],
- const float c0[TGSI_QUAD_SIZE],
+ UNUSED int gather_comp,
const float lod_in[TGSI_QUAD_SIZE],
const struct filter_args *filt_args,
float rgba[TGSI_NUM_CHANNELS][TGSI_QUAD_SIZE])
float lambda;
float lod[TGSI_QUAD_SIZE];
- float s_to_u = u_minify(texture->width0, psview->u.tex.first_level);
- float t_to_v = u_minify(texture->height0, psview->u.tex.first_level);
- float dudx = (s[QUAD_BOTTOM_RIGHT] - s[QUAD_BOTTOM_LEFT]) * s_to_u;
- float dudy = (s[QUAD_TOP_LEFT] - s[QUAD_BOTTOM_LEFT]) * s_to_u;
- float dvdx = (t[QUAD_BOTTOM_RIGHT] - t[QUAD_BOTTOM_LEFT]) * t_to_v;
- float dvdy = (t[QUAD_TOP_LEFT] - t[QUAD_BOTTOM_LEFT]) * t_to_v;
+ const float s_to_u = u_minify(texture->width0, psview->u.tex.first_level);
+ const float t_to_v = u_minify(texture->height0, psview->u.tex.first_level);
+ const float dudx = (s[QUAD_BOTTOM_RIGHT] - s[QUAD_BOTTOM_LEFT]) * s_to_u;
+ const float dudy = (s[QUAD_TOP_LEFT] - s[QUAD_BOTTOM_LEFT]) * s_to_u;
+ const float dvdx = (t[QUAD_BOTTOM_RIGHT] - t[QUAD_BOTTOM_LEFT]) * t_to_v;
+ const float dvdy = (t[QUAD_TOP_LEFT] - t[QUAD_BOTTOM_LEFT]) * t_to_v;
struct img_filter_args args;
+ args.offset = filt_args->offset;
+
if (filt_args->control == TGSI_SAMPLER_LOD_BIAS ||
filt_args->control == TGSI_SAMPLER_LOD_NONE ||
/* XXX FIXME */
/* note: instead of working with Px and Py, we will use the
* squared length instead, to avoid sqrt.
*/
- float Px2 = dudx * dudx + dvdx * dvdx;
- float Py2 = dudy * dudy + dvdy * dvdy;
+ const float Px2 = dudx * dudx + dvdx * dvdx;
+ const float Py2 = dudy * dudy + dvdy * dvdy;
float Pmax2;
float Pmin2;
args.p = p[j];
args.level = psview->u.tex.last_level;
args.face_id = filt_args->faces[j];
+ /*
+ * XXX: we overwrote any linear filter with nearest, so this
+ * isn't right (albeit if last level is 1x1 and no border it
+ * will work just the same).
+ */
min_filter(sp_sview, sp_samp, &args, &rgba[0][j]);
}
}
* seem to be worth the extra running time.
*/
img_filter_2d_ewa(sp_sview, sp_samp, min_filter, mag_filter,
- s, t, p, filt_args->faces, level0,
- dudx, dvdx, dudy, dvdy, rgba);
+ s, t, p, filt_args->faces, filt_args->offset,
+ level0, dudx, dvdx, dudy, dvdy, rgba);
}
if (DEBUG_TEX) {
*/
static void
mip_filter_linear_2d_linear_repeat_POT(
- struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
+ const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
img_filter_func min_filter,
img_filter_func mag_filter,
const float s[TGSI_QUAD_SIZE],
const float t[TGSI_QUAD_SIZE],
const float p[TGSI_QUAD_SIZE],
- const float c0[TGSI_QUAD_SIZE],
- const float lod_in[TGSI_QUAD_SIZE],
+ int gather_comp,
+ const float lod[TGSI_QUAD_SIZE],
const struct filter_args *filt_args,
float rgba[TGSI_NUM_CHANNELS][TGSI_QUAD_SIZE])
{
const struct pipe_sampler_view *psview = &sp_sview->base;
int j;
- float lod[TGSI_QUAD_SIZE];
-
- compute_lambda_lod(sp_sview, sp_samp, s, t, p, lod_in, filt_args->control, lod);
for (j = 0; j < TGSI_QUAD_SIZE; j++) {
- int level0 = psview->u.tex.first_level + (int)lod[j];
+ const int level0 = psview->u.tex.first_level + (int)lod[j];
struct img_filter_args args;
/* Catches both negative and large values of level0:
*/
args.face_id = filt_args->faces[j];
args.offset = filt_args->offset;
args.gather_only = filt_args->control == TGSI_SAMPLER_GATHER;
+ args.gather_comp = gather_comp;
if ((unsigned)level0 >= psview->u.tex.last_level) {
if (level0 < 0)
args.level = psview->u.tex.first_level;
}
else {
- float levelBlend = frac(lod[j]);
+ const float levelBlend = frac(lod[j]);
float rgbax[TGSI_NUM_CHANNELS][TGSI_QUAD_SIZE];
int c;
* Do shadow/depth comparisons.
*/
static void
-sample_compare(struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
- const float s[TGSI_QUAD_SIZE],
- const float t[TGSI_QUAD_SIZE],
- const float p[TGSI_QUAD_SIZE],
+sample_compare(const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
const float c0[TGSI_QUAD_SIZE],
- const float c1[TGSI_QUAD_SIZE],
enum tgsi_sampler_control control,
float rgba[TGSI_NUM_CHANNELS][TGSI_QUAD_SIZE])
{
int j, v;
int k[TGSI_NUM_CHANNELS][TGSI_QUAD_SIZE];
float pc[4];
- const struct util_format_description *format_desc;
- unsigned chan_type;
- bool is_gather = (control == TGSI_SAMPLER_GATHER);
+ const struct util_format_description *format_desc =
+ util_format_description(sp_sview->base.format);
+ /* not entirely sure we couldn't end up with non-valid swizzle here */
+ const unsigned chan_type =
+ format_desc->swizzle[0] <= PIPE_SWIZZLE_W ?
+ format_desc->channel[format_desc->swizzle[0]].type :
+ UTIL_FORMAT_TYPE_FLOAT;
+ const bool is_gather = (control == TGSI_SAMPLER_GATHER);
/**
* Compare texcoord 'p' (aka R) against texture value 'rgba[0]'
* RGBA channels. We look at the red channel here.
*/
- if (sp_sview->base.target == PIPE_TEXTURE_2D_ARRAY ||
- sp_sview->base.target == PIPE_TEXTURE_CUBE) {
- pc[0] = c0[0];
- pc[1] = c0[1];
- pc[2] = c0[2];
- pc[3] = c0[3];
- } else if (sp_sview->base.target == PIPE_TEXTURE_CUBE_ARRAY) {
- pc[0] = c1[0];
- pc[1] = c1[1];
- pc[2] = c1[2];
- pc[3] = c1[3];
- } else {
- pc[0] = p[0];
- pc[1] = p[1];
- pc[2] = p[2];
- pc[3] = p[3];
- }
- format_desc = util_format_description(sp_sview->base.format);
- /* not entirely sure we couldn't end up with non-valid swizzle here */
- chan_type = format_desc->swizzle[0] <= UTIL_FORMAT_SWIZZLE_W ?
- format_desc->channel[format_desc->swizzle[0]].type :
- UTIL_FORMAT_TYPE_FLOAT;
+
if (chan_type != UTIL_FORMAT_TYPE_FLOAT) {
/*
* clamping is a result of conversion to texture format, hence
* doesn't happen with floats. Technically also should do comparison
* in texture format (quantization!).
*/
- pc[0] = CLAMP(pc[0], 0.0F, 1.0F);
- pc[1] = CLAMP(pc[1], 0.0F, 1.0F);
- pc[2] = CLAMP(pc[2], 0.0F, 1.0F);
- pc[3] = CLAMP(pc[3], 0.0F, 1.0F);
+ pc[0] = CLAMP(c0[0], 0.0F, 1.0F);
+ pc[1] = CLAMP(c0[1], 0.0F, 1.0F);
+ pc[2] = CLAMP(c0[2], 0.0F, 1.0F);
+ pc[3] = CLAMP(c0[3], 0.0F, 1.0F);
+ } else {
+ pc[0] = c0[0];
+ pc[1] = c0[1];
+ pc[2] = c0[2];
+ pc[3] = c0[3];
}
for (v = 0; v < (is_gather ? TGSI_NUM_CHANNELS : 1); v++) {
const unsigned swizzle_g = sview->swizzle_g;
const unsigned swizzle_b = sview->swizzle_b;
const unsigned swizzle_a = sview->swizzle_a;
+ float oneval = util_format_is_pure_integer(sview->format) ? uif(1) : 1.0f;
switch (swizzle_r) {
- case PIPE_SWIZZLE_ZERO:
+ case PIPE_SWIZZLE_0:
for (j = 0; j < 4; j++)
out[0][j] = 0.0f;
break;
- case PIPE_SWIZZLE_ONE:
+ case PIPE_SWIZZLE_1:
for (j = 0; j < 4; j++)
- out[0][j] = 1.0f;
+ out[0][j] = oneval;
break;
default:
assert(swizzle_r < 4);
}
switch (swizzle_g) {
- case PIPE_SWIZZLE_ZERO:
+ case PIPE_SWIZZLE_0:
for (j = 0; j < 4; j++)
out[1][j] = 0.0f;
break;
- case PIPE_SWIZZLE_ONE:
+ case PIPE_SWIZZLE_1:
for (j = 0; j < 4; j++)
- out[1][j] = 1.0f;
+ out[1][j] = oneval;
break;
default:
assert(swizzle_g < 4);
}
switch (swizzle_b) {
- case PIPE_SWIZZLE_ZERO:
+ case PIPE_SWIZZLE_0:
for (j = 0; j < 4; j++)
out[2][j] = 0.0f;
break;
- case PIPE_SWIZZLE_ONE:
+ case PIPE_SWIZZLE_1:
for (j = 0; j < 4; j++)
- out[2][j] = 1.0f;
+ out[2][j] = oneval;
break;
default:
assert(swizzle_b < 4);
}
switch (swizzle_a) {
- case PIPE_SWIZZLE_ZERO:
+ case PIPE_SWIZZLE_0:
for (j = 0; j < 4; j++)
out[3][j] = 0.0f;
break;
- case PIPE_SWIZZLE_ONE:
+ case PIPE_SWIZZLE_1:
for (j = 0; j < 4; j++)
- out[3][j] = 1.0f;
+ out[3][j] = oneval;
break;
default:
assert(swizzle_a < 4);
static inline bool
any_swizzle(const struct pipe_sampler_view *view)
{
- return (view->swizzle_r != PIPE_SWIZZLE_RED ||
- view->swizzle_g != PIPE_SWIZZLE_GREEN ||
- view->swizzle_b != PIPE_SWIZZLE_BLUE ||
- view->swizzle_a != PIPE_SWIZZLE_ALPHA);
+ return (view->swizzle_r != PIPE_SWIZZLE_X ||
+ view->swizzle_g != PIPE_SWIZZLE_Y ||
+ view->swizzle_b != PIPE_SWIZZLE_Z ||
+ view->swizzle_a != PIPE_SWIZZLE_W);
}
}
static void
-sample_mip(struct sp_sampler_view *sp_sview,
- struct sp_sampler *sp_samp,
+sample_mip(const struct sp_sampler_view *sp_sview,
+ const struct sp_sampler *sp_samp,
const float s[TGSI_QUAD_SIZE],
const float t[TGSI_QUAD_SIZE],
const float p[TGSI_QUAD_SIZE],
const float c0[TGSI_QUAD_SIZE],
+ int gather_comp,
const float lod[TGSI_QUAD_SIZE],
const struct filter_args *filt_args,
float rgba[TGSI_NUM_CHANNELS][TGSI_QUAD_SIZE])
&funcs, &min_img_filter, &mag_img_filter);
funcs->filter(sp_sview, sp_samp, min_img_filter, mag_img_filter,
- s, t, p, c0, lod, filt_args, rgba);
+ s, t, p, gather_comp, lod, filt_args, rgba);
if (sp_samp->base.compare_mode != PIPE_TEX_COMPARE_NONE) {
- sample_compare(sp_sview, sp_samp, s, t, p, c0,
- lod, filt_args->control, rgba);
+ sample_compare(sp_sview, sp_samp, c0, filt_args->control, rgba);
}
if (sp_sview->need_swizzle && filt_args->control != TGSI_SAMPLER_GATHER) {
float ssss[TGSI_QUAD_SIZE],
float tttt[TGSI_QUAD_SIZE],
float pppp[TGSI_QUAD_SIZE],
- float faces[TGSI_QUAD_SIZE])
+ uint faces[TGSI_QUAD_SIZE])
{
unsigned j;
* deriviates to compute the LOD. Doing so (near cube edges
* anyway) gives us pretty much random values.
*/
- {
- /* use the average of the four pixel's texcoords to choose the face */
- const float rx = 0.25F * (s[0] + s[1] + s[2] + s[3]);
- const float ry = 0.25F * (t[0] + t[1] + t[2] + t[3]);
- const float rz = 0.25F * (p[0] + p[1] + p[2] + p[3]);
+ for (j = 0; j < TGSI_QUAD_SIZE; j++) {
+ const float rx = s[j], ry = t[j], rz = p[j];
const float arx = fabsf(rx), ary = fabsf(ry), arz = fabsf(rz);
if (arx >= ary && arx >= arz) {
const float sign = (rx >= 0.0F) ? 1.0F : -1.0F;
const uint face = (rx >= 0.0F) ?
PIPE_TEX_FACE_POS_X : PIPE_TEX_FACE_NEG_X;
- for (j = 0; j < TGSI_QUAD_SIZE; j++) {
- const float ima = -0.5F / fabsf(s[j]);
- ssss[j] = sign * p[j] * ima + 0.5F;
- tttt[j] = t[j] * ima + 0.5F;
- faces[j] = face;
- }
+ const float ima = -0.5F / fabsf(s[j]);
+ ssss[j] = sign * p[j] * ima + 0.5F;
+ tttt[j] = t[j] * ima + 0.5F;
+ faces[j] = face;
}
else if (ary >= arx && ary >= arz) {
const float sign = (ry >= 0.0F) ? 1.0F : -1.0F;
const uint face = (ry >= 0.0F) ?
PIPE_TEX_FACE_POS_Y : PIPE_TEX_FACE_NEG_Y;
- for (j = 0; j < TGSI_QUAD_SIZE; j++) {
- const float ima = -0.5F / fabsf(t[j]);
- ssss[j] = -s[j] * ima + 0.5F;
- tttt[j] = sign * -p[j] * ima + 0.5F;
- faces[j] = face;
- }
+ const float ima = -0.5F / fabsf(t[j]);
+ ssss[j] = -s[j] * ima + 0.5F;
+ tttt[j] = sign * -p[j] * ima + 0.5F;
+ faces[j] = face;
}
else {
const float sign = (rz >= 0.0F) ? 1.0F : -1.0F;
const uint face = (rz >= 0.0F) ?
PIPE_TEX_FACE_POS_Z : PIPE_TEX_FACE_NEG_Z;
- for (j = 0; j < TGSI_QUAD_SIZE; j++) {
- const float ima = -0.5F / fabsf(p[j]);
- ssss[j] = sign * -s[j] * ima + 0.5F;
- tttt[j] = t[j] * ima + 0.5F;
- faces[j] = face;
- }
+ const float ima = -0.5F / fabsf(p[j]);
+ ssss[j] = sign * -s[j] * ima + 0.5F;
+ tttt[j] = t[j] * ima + 0.5F;
+ faces[j] = face;
}
}
}
static void
-sp_get_dims(struct sp_sampler_view *sp_sview, int level,
+sp_get_dims(const struct sp_sampler_view *sp_sview,
+ int level,
int dims[4])
{
const struct pipe_sampler_view *view = &sp_sview->base;
const struct pipe_resource *texture = view->texture;
if (view->target == PIPE_BUFFER) {
- dims[0] = (view->u.buf.last_element - view->u.buf.first_element) + 1;
+ dims[0] = view->u.buf.size / util_format_get_blocksize(view->format);
/* the other values are undefined, but let's avoid potential valgrind
* warnings.
*/
* coords to the texture image size.
*/
static void
-sp_get_texels(struct sp_sampler_view *sp_sview,
+sp_get_texels(const struct sp_sampler_view *sp_sview,
const int v_i[TGSI_QUAD_SIZE],
const int v_j[TGSI_QUAD_SIZE],
const int v_k[TGSI_QUAD_SIZE],
const struct pipe_resource *texture = sp_sview->base.texture;
int j, c;
const float *tx;
- int width, height, depth;
-
- addr.value = 0;
/* TODO write a better test for LOD */
- addr.bits.level = sp_sview->base.target == PIPE_BUFFER ? 0 :
- CLAMP(lod[0] + sp_sview->base.u.tex.first_level,
- sp_sview->base.u.tex.first_level,
- sp_sview->base.u.tex.last_level);
+ const unsigned level =
+ sp_sview->base.target == PIPE_BUFFER ? 0 :
+ CLAMP(lod[0] + sp_sview->base.u.tex.first_level,
+ sp_sview->base.u.tex.first_level,
+ sp_sview->base.u.tex.last_level);
+ const int width = u_minify(texture->width0, level);
+ const int height = u_minify(texture->height0, level);
+ const int depth = u_minify(texture->depth0, level);
+ unsigned elem_size, first_element, last_element;
- width = u_minify(texture->width0, addr.bits.level);
- height = u_minify(texture->height0, addr.bits.level);
- depth = u_minify(texture->depth0, addr.bits.level);
+ addr.value = 0;
+ addr.bits.level = level;
switch (sp_sview->base.target) {
case PIPE_BUFFER:
+ elem_size = util_format_get_blocksize(sp_sview->base.format);
+ first_element = sp_sview->base.u.buf.offset / elem_size;
+ last_element = (sp_sview->base.u.buf.offset +
+ sp_sview->base.u.buf.size) / elem_size - 1;
+ for (j = 0; j < TGSI_QUAD_SIZE; j++) {
+ const int x = CLAMP(v_i[j] + offset[0] +
+ first_element,
+ first_element,
+ last_element);
+ tx = get_texel_buffer_no_border(sp_sview, addr, x, elem_size);
+ for (c = 0; c < 4; c++) {
+ rgba[c][j] = tx[c];
+ }
+ }
+ break;
case PIPE_TEXTURE_1D:
for (j = 0; j < TGSI_QUAD_SIZE; j++) {
- int x = CLAMP(v_i[j] + offset[0], 0, width - 1);
- tx = get_texel_2d_no_border(sp_sview, addr, x, 0);
+ const int x = CLAMP(v_i[j] + offset[0], 0, width - 1);
+ tx = get_texel_2d_no_border(sp_sview, addr, x,
+ sp_sview->base.u.tex.first_layer);
for (c = 0; c < 4; c++) {
rgba[c][j] = tx[c];
}
break;
case PIPE_TEXTURE_1D_ARRAY:
for (j = 0; j < TGSI_QUAD_SIZE; j++) {
- int x = CLAMP(v_i[j] + offset[0], 0, width - 1);
- int y = CLAMP(v_j[j], sp_sview->base.u.tex.first_layer,
- sp_sview->base.u.tex.last_layer);
+ const int x = CLAMP(v_i[j] + offset[0], 0, width - 1);
+ const int y = CLAMP(v_j[j], sp_sview->base.u.tex.first_layer,
+ sp_sview->base.u.tex.last_layer);
tx = get_texel_2d_no_border(sp_sview, addr, x, y);
for (c = 0; c < 4; c++) {
rgba[c][j] = tx[c];
case PIPE_TEXTURE_2D:
case PIPE_TEXTURE_RECT:
for (j = 0; j < TGSI_QUAD_SIZE; j++) {
- int x = CLAMP(v_i[j] + offset[0], 0, width - 1);
- int y = CLAMP(v_j[j] + offset[1], 0, height - 1);
- tx = get_texel_2d_no_border(sp_sview, addr, x, y);
+ const int x = CLAMP(v_i[j] + offset[0], 0, width - 1);
+ const int y = CLAMP(v_j[j] + offset[1], 0, height - 1);
+ tx = get_texel_3d_no_border(sp_sview, addr, x, y,
+ sp_sview->base.u.tex.first_layer);
for (c = 0; c < 4; c++) {
rgba[c][j] = tx[c];
}
break;
case PIPE_TEXTURE_2D_ARRAY:
for (j = 0; j < TGSI_QUAD_SIZE; j++) {
- int x = CLAMP(v_i[j] + offset[0], 0, width - 1);
- int y = CLAMP(v_j[j] + offset[1], 0, height - 1);
- int layer = CLAMP(v_k[j], sp_sview->base.u.tex.first_layer,
- sp_sview->base.u.tex.last_layer);
+ const int x = CLAMP(v_i[j] + offset[0], 0, width - 1);
+ const int y = CLAMP(v_j[j] + offset[1], 0, height - 1);
+ const int layer = CLAMP(v_k[j], sp_sview->base.u.tex.first_layer,
+ sp_sview->base.u.tex.last_layer);
tx = get_texel_3d_no_border(sp_sview, addr, x, y, layer);
for (c = 0; c < 4; c++) {
rgba[c][j] = tx[c];
}
break;
case PIPE_TEXTURE_CUBE: /* TXF can't work on CUBE according to spec */
+ case PIPE_TEXTURE_CUBE_ARRAY:
default:
assert(!"Unknown or CUBE texture type in TXF processing\n");
break;
compute_lambda_func
-softpipe_get_lambda_func(const struct pipe_sampler_view *view, unsigned shader)
+softpipe_get_lambda_func(const struct pipe_sampler_view *view,
+ enum pipe_shader_type shader)
{
if (shader != PIPE_SHADER_FRAGMENT)
return compute_lambda_vert;
case PIPE_TEXTURE_2D:
case PIPE_TEXTURE_2D_ARRAY:
case PIPE_TEXTURE_RECT:
+ return compute_lambda_2d;
case PIPE_TEXTURE_CUBE:
case PIPE_TEXTURE_CUBE_ARRAY:
- return compute_lambda_2d;
+ return compute_lambda_cube;
case PIPE_TEXTURE_3D:
return compute_lambda_3d;
default:
const struct pipe_sampler_view *templ)
{
struct sp_sampler_view *sview = CALLOC_STRUCT(sp_sampler_view);
- struct softpipe_resource *spr = (struct softpipe_resource *)resource;
+ const struct softpipe_resource *spr = (struct softpipe_resource *)resource;
if (sview) {
struct pipe_sampler_view *view = &sview->base;
#ifdef DEBUG
/*
* This is possibly too lenient, but the primary reason is just
- * to catch state trackers which forget to initialize this, so
+ * to catch gallium frontends which forget to initialize this, so
* it only catches clearly impossible view targets.
*/
if (view->target != resource->target) {
}
+static inline const struct sp_tgsi_sampler *
+sp_tgsi_sampler_cast_c(const struct tgsi_sampler *sampler)
+{
+ return (const struct sp_tgsi_sampler *)sampler;
+}
+
+
static void
sp_tgsi_get_dims(struct tgsi_sampler *tgsi_sampler,
const unsigned sview_index,
int level, int dims[4])
{
- struct sp_tgsi_sampler *sp_samp = (struct sp_tgsi_sampler *)tgsi_sampler;
+ const struct sp_tgsi_sampler *sp_samp =
+ sp_tgsi_sampler_cast_c(tgsi_sampler);
assert(sview_index < PIPE_MAX_SHADER_SAMPLER_VIEWS);
/* always have a view here but texture is NULL if no sampler view was set. */
}
+static void prepare_compare_values(enum pipe_texture_target target,
+ const float p[TGSI_QUAD_SIZE],
+ const float c0[TGSI_QUAD_SIZE],
+ const float c1[TGSI_QUAD_SIZE],
+ float pc[TGSI_QUAD_SIZE])
+{
+ if (target == PIPE_TEXTURE_2D_ARRAY ||
+ target == PIPE_TEXTURE_CUBE) {
+ pc[0] = c0[0];
+ pc[1] = c0[1];
+ pc[2] = c0[2];
+ pc[3] = c0[3];
+ } else if (target == PIPE_TEXTURE_CUBE_ARRAY) {
+ pc[0] = c1[0];
+ pc[1] = c1[1];
+ pc[2] = c1[2];
+ pc[3] = c1[3];
+ } else {
+ pc[0] = p[0];
+ pc[1] = p[1];
+ pc[2] = p[2];
+ pc[3] = p[3];
+ }
+}
+
static void
sp_tgsi_get_samples(struct tgsi_sampler *tgsi_sampler,
const unsigned sview_index,
const float t[TGSI_QUAD_SIZE],
const float p[TGSI_QUAD_SIZE],
const float c0[TGSI_QUAD_SIZE],
- const float lod[TGSI_QUAD_SIZE],
+ const float lod_in[TGSI_QUAD_SIZE],
float derivs[3][2][TGSI_QUAD_SIZE],
const int8_t offset[3],
enum tgsi_sampler_control control,
float rgba[TGSI_NUM_CHANNELS][TGSI_QUAD_SIZE])
{
- struct sp_tgsi_sampler *sp_tgsi_samp = (struct sp_tgsi_sampler *)tgsi_sampler;
- struct sp_sampler_view *sp_sview;
- struct sp_sampler *sp_samp;
+ const struct sp_tgsi_sampler *sp_tgsi_samp =
+ sp_tgsi_sampler_cast_c(tgsi_sampler);
+ struct sp_sampler_view sp_sview;
+ const struct sp_sampler *sp_samp;
struct filter_args filt_args;
+ float compare_values[TGSI_QUAD_SIZE];
+ float lod[TGSI_QUAD_SIZE];
+ int c;
assert(sview_index < PIPE_MAX_SHADER_SAMPLER_VIEWS);
assert(sampler_index < PIPE_MAX_SAMPLERS);
assert(sp_tgsi_samp->sp_sampler[sampler_index]);
- sp_sview = &sp_tgsi_samp->sp_sview[sview_index];
+ memcpy(&sp_sview, &sp_tgsi_samp->sp_sview[sview_index],
+ sizeof(struct sp_sampler_view));
sp_samp = sp_tgsi_samp->sp_sampler[sampler_index];
+
+ if (util_format_is_unorm(sp_sview.base.format)) {
+ for (c = 0; c < TGSI_NUM_CHANNELS; c++)
+ sp_sview.border_color.f[c] = CLAMP(sp_samp->base.border_color.f[c],
+ 0.0f, 1.0f);
+ } else if (util_format_is_snorm(sp_sview.base.format)) {
+ for (c = 0; c < TGSI_NUM_CHANNELS; c++)
+ sp_sview.border_color.f[c] = CLAMP(sp_samp->base.border_color.f[c],
+ -1.0f, 1.0f);
+ } else {
+ memcpy(sp_sview.border_color.f, sp_samp->base.border_color.f,
+ TGSI_NUM_CHANNELS * sizeof(float));
+ }
+
/* always have a view here but texture is NULL if no sampler view was set. */
- if (!sp_sview->base.texture) {
+ if (!sp_sview.base.texture) {
int i, j;
for (j = 0; j < TGSI_NUM_CHANNELS; j++) {
for (i = 0; i < TGSI_QUAD_SIZE; i++) {
return;
}
+ if (sp_samp->base.compare_mode != PIPE_TEX_COMPARE_NONE)
+ prepare_compare_values(sp_sview.base.target, p, c0, lod_in, compare_values);
+
filt_args.control = control;
filt_args.offset = offset;
+ int gather_comp = get_gather_component(lod_in);
+
+ compute_lambda_lod(&sp_sview, sp_samp, s, t, p, derivs, lod_in, control, lod);
- if (sp_sview->need_cube_convert) {
+ if (sp_sview.need_cube_convert) {
float cs[TGSI_QUAD_SIZE];
float ct[TGSI_QUAD_SIZE];
float cp[TGSI_QUAD_SIZE];
- float faces[TGSI_QUAD_SIZE];
+ uint faces[TGSI_QUAD_SIZE];
- convert_cube(sp_sview, sp_samp, s, t, p, c0, cs, ct, cp, faces);
+ convert_cube(&sp_sview, sp_samp, s, t, p, c0, cs, ct, cp, faces);
filt_args.faces = faces;
- sample_mip(sp_sview, sp_samp, cs, ct, cp, c0, lod, &filt_args, rgba);
+ sample_mip(&sp_sview, sp_samp, cs, ct, cp, compare_values, gather_comp, lod, &filt_args, rgba);
} else {
- static const float zero_faces[TGSI_QUAD_SIZE] = {0.0f, 0.0f, 0.0f, 0.0f};
+ static const uint zero_faces[TGSI_QUAD_SIZE] = {0, 0, 0, 0};
filt_args.faces = zero_faces;
- sample_mip(sp_sview, sp_samp, s, t, p, c0, lod, &filt_args, rgba);
+ sample_mip(&sp_sview, sp_samp, s, t, p, compare_values, gather_comp, lod, &filt_args, rgba);
}
}
float lod[TGSI_QUAD_SIZE])
{
static const float lod_in[TGSI_QUAD_SIZE] = { 0.0, 0.0, 0.0, 0.0 };
+ static const float dummy_grad[3][2][TGSI_QUAD_SIZE];
const struct sp_tgsi_sampler *sp_tgsi_samp =
- (const struct sp_tgsi_sampler *)tgsi_sampler;
+ sp_tgsi_sampler_cast_c(tgsi_sampler);
const struct sp_sampler_view *sp_sview;
const struct sp_sampler *sp_samp;
const struct sp_filter_funcs *funcs;
}
return;
}
-
- if (sp_sview->need_cube_convert) {
- float cs[TGSI_QUAD_SIZE];
- float ct[TGSI_QUAD_SIZE];
- float cp[TGSI_QUAD_SIZE];
- float unused_faces[TGSI_QUAD_SIZE];
-
- convert_cube(sp_sview, sp_samp, s, t, p, c0, cs, ct, cp, unused_faces);
- compute_lambda_lod_unclamped(sp_sview, sp_samp,
- cs, ct, cp, lod_in, control, lod);
- } else {
- compute_lambda_lod_unclamped(sp_sview, sp_samp,
- s, t, p, lod_in, control, lod);
- }
+ compute_lambda_lod_unclamped(sp_sview, sp_samp,
+ s, t, p, dummy_grad, lod_in, control, lod);
get_filters(sp_sview, sp_samp, control, &funcs, NULL, NULL);
funcs->relative_level(sp_sview, sp_samp, lod, mipmap);
const int lod[TGSI_QUAD_SIZE], const int8_t offset[3],
float rgba[TGSI_NUM_CHANNELS][TGSI_QUAD_SIZE])
{
- struct sp_tgsi_sampler *sp_samp = (struct sp_tgsi_sampler *)tgsi_sampler;
+ const struct sp_tgsi_sampler *sp_samp =
+ sp_tgsi_sampler_cast_c(tgsi_sampler);
assert(sview_index < PIPE_MAX_SHADER_SAMPLER_VIEWS);
/* always have a view here but texture is NULL if no sampler view was set. */