void
vl_idct_stage2_frag_shader(struct vl_idct *idct, struct ureg_program *shader,
unsigned first_input, struct ureg_dst fragment)
{
struct ureg_src l_addr[2], r_addr[2];
struct ureg_dst l[2], r[2];
--first_input;
l_addr[0] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, first_input + VS_O_L_ADDR0, TGSI_INTERPOLATE_LINEAR);
l_addr[1] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, first_input + VS_O_L_ADDR1, TGSI_INTERPOLATE_LINEAR);
r_addr[0] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, first_input + VS_O_R_ADDR0, TGSI_INTERPOLATE_LINEAR);
r_addr[1] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, first_input + VS_O_R_ADDR1, TGSI_INTERPOLATE_LINEAR);
l[0] = ureg_DECL_temporary(shader);
l[1] = ureg_DECL_temporary(shader);
r[0] = ureg_DECL_temporary(shader);
r[1] = ureg_DECL_temporary(shader);
fetch_four(shader, l, l_addr, ureg_DECL_sampler(shader, 1), false);
fetch_four(shader, r, r_addr, ureg_DECL_sampler(shader, 0), true);
matrix_mul(shader, fragment, l, r);
ureg_release_temporary(shader, l[0]);
ureg_release_temporary(shader, l[1]);
ureg_release_temporary(shader, r[0]);
ureg_release_temporary(shader, r[1]);
}
static INLINE void
xrender_tex(struct ureg_program *ureg,
struct ureg_dst dst,
struct ureg_src coords,
struct ureg_src sampler,
struct ureg_src imm0,
boolean repeat_none, boolean swizzle, boolean set_alpha)
{
if (repeat_none) {
struct ureg_dst tmp0 = ureg_DECL_temporary(ureg);
struct ureg_dst tmp1 = ureg_DECL_temporary(ureg);
ureg_SGT(ureg, tmp1, ureg_swizzle(coords,
TGSI_SWIZZLE_X,
TGSI_SWIZZLE_Y,
TGSI_SWIZZLE_X,
TGSI_SWIZZLE_Y), ureg_scalar(imm0,
TGSI_SWIZZLE_X));
ureg_SLT(ureg, tmp0,
ureg_swizzle(coords, TGSI_SWIZZLE_X, TGSI_SWIZZLE_Y,
TGSI_SWIZZLE_X, TGSI_SWIZZLE_Y), ureg_scalar(imm0,
TGSI_SWIZZLE_W));
ureg_MIN(ureg, tmp0, ureg_src(tmp0), ureg_src(tmp1));
ureg_MIN(ureg, tmp0, ureg_scalar(ureg_src(tmp0), TGSI_SWIZZLE_X),
ureg_scalar(ureg_src(tmp0), TGSI_SWIZZLE_Y));
ureg_TEX(ureg, tmp1, TGSI_TEXTURE_2D, coords, sampler);
if (swizzle)
ureg_MOV(ureg, tmp1, ureg_swizzle(ureg_src(tmp1),
TGSI_SWIZZLE_Z,
TGSI_SWIZZLE_Y, TGSI_SWIZZLE_X,
TGSI_SWIZZLE_W));
if (set_alpha)
ureg_MOV(ureg,
ureg_writemask(tmp1, TGSI_WRITEMASK_W),
ureg_scalar(imm0, TGSI_SWIZZLE_W));
ureg_MUL(ureg, dst, ureg_src(tmp1), ureg_src(tmp0));
ureg_release_temporary(ureg, tmp0);
ureg_release_temporary(ureg, tmp1);
} else {
if (swizzle) {
struct ureg_dst tmp = ureg_DECL_temporary(ureg);
ureg_TEX(ureg, tmp, TGSI_TEXTURE_2D, coords, sampler);
ureg_MOV(ureg, dst, ureg_swizzle(ureg_src(tmp),
TGSI_SWIZZLE_Z,
TGSI_SWIZZLE_Y, TGSI_SWIZZLE_X,
TGSI_SWIZZLE_W));
ureg_release_temporary(ureg, tmp);
} else {
ureg_TEX(ureg, dst, TGSI_TEXTURE_2D, coords, sampler);
}
if (set_alpha)
ureg_MOV(ureg,
ureg_writemask(dst, TGSI_WRITEMASK_W),
ureg_scalar(imm0, TGSI_SWIZZLE_W));
}
}
static void *
create_yuv_shader(struct pipe_context *pipe, struct ureg_program *ureg)
{
struct ureg_src y_sampler, u_sampler, v_sampler;
struct ureg_src pos;
struct ureg_src matrow0, matrow1, matrow2;
struct ureg_dst y, u, v, rgb;
struct ureg_dst out = ureg_DECL_output(ureg,
TGSI_SEMANTIC_COLOR,
0);
pos = ureg_DECL_fs_input(ureg,
TGSI_SEMANTIC_GENERIC, 0,
TGSI_INTERPOLATE_PERSPECTIVE);
rgb = ureg_DECL_temporary(ureg);
y = ureg_DECL_temporary(ureg);
u = ureg_DECL_temporary(ureg);
v = ureg_DECL_temporary(ureg);
y_sampler = ureg_DECL_sampler(ureg, 0);
u_sampler = ureg_DECL_sampler(ureg, 1);
v_sampler = ureg_DECL_sampler(ureg, 2);
matrow0 = ureg_DECL_constant(ureg, 0);
matrow1 = ureg_DECL_constant(ureg, 1);
matrow2 = ureg_DECL_constant(ureg, 2);
ureg_TEX(ureg, y, TGSI_TEXTURE_2D, pos, y_sampler);
ureg_TEX(ureg, u, TGSI_TEXTURE_2D, pos, u_sampler);
ureg_TEX(ureg, v, TGSI_TEXTURE_2D, pos, v_sampler);
ureg_SUB(ureg, u, ureg_src(u), ureg_scalar(matrow0, TGSI_SWIZZLE_W));
ureg_SUB(ureg, v, ureg_src(v), ureg_scalar(matrow0, TGSI_SWIZZLE_W));
ureg_MUL(ureg, rgb, ureg_scalar(ureg_src(y), TGSI_SWIZZLE_X), matrow0);
ureg_MAD(ureg, rgb,
ureg_scalar(ureg_src(u), TGSI_SWIZZLE_X), matrow1, ureg_src(rgb));
ureg_MAD(ureg, rgb,
ureg_scalar(ureg_src(v), TGSI_SWIZZLE_X), matrow2, ureg_src(rgb));
/* rgb.a = 1; */
ureg_MOV(ureg, ureg_writemask(rgb, TGSI_WRITEMASK_W),
ureg_scalar(matrow0, TGSI_SWIZZLE_X));
ureg_MOV(ureg, out, ureg_src(rgb));
ureg_release_temporary(ureg, rgb);
ureg_release_temporary(ureg, y);
ureg_release_temporary(ureg, u);
ureg_release_temporary(ureg, v);
ureg_END(ureg);
return ureg_create_shader_and_destroy(ureg, pipe);
}
static void *
create_copy_frag_shader(struct vl_deint_filter *filter, unsigned field)
{
struct ureg_program *shader;
struct ureg_src i_vtex;
struct ureg_src sampler;
struct ureg_dst o_fragment;
struct ureg_dst t_tex;
shader = ureg_create(PIPE_SHADER_FRAGMENT);
if (!shader) {
return NULL;
}
t_tex = ureg_DECL_temporary(shader);
i_vtex = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_VTEX, TGSI_INTERPOLATE_LINEAR);
sampler = ureg_DECL_sampler(shader, 2);
o_fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
ureg_MOV(shader, t_tex, i_vtex);
if (field) {
ureg_MOV(shader, ureg_writemask(t_tex, TGSI_WRITEMASK_ZW),
ureg_imm4f(shader, 0, 0, 1.0f, 0));
} else {
ureg_MOV(shader, ureg_writemask(t_tex, TGSI_WRITEMASK_ZW),
ureg_imm1f(shader, 0));
}
ureg_TEX(shader, o_fragment, TGSI_TEXTURE_2D_ARRAY, ureg_src(t_tex), sampler);
ureg_release_temporary(shader, t_tex);
ureg_END(shader);
return ureg_create_shader_and_destroy(shader, filter->pipe);
}
static void *
create_ref_vert_shader(struct vl_mc *r)
{
struct ureg_program *shader;
struct ureg_src mv_scale;
struct ureg_src vmv[2];
struct ureg_dst t_vpos;
struct ureg_dst o_vmv[2];
unsigned i;
shader = ureg_create(TGSI_PROCESSOR_VERTEX);
if (!shader)
return NULL;
vmv[0] = ureg_DECL_vs_input(shader, VS_I_MV_TOP);
vmv[1] = ureg_DECL_vs_input(shader, VS_I_MV_BOTTOM);
t_vpos = calc_position(r, shader, ureg_imm2f(shader,
(float)VL_MACROBLOCK_WIDTH / r->buffer_width,
(float)VL_MACROBLOCK_HEIGHT / r->buffer_height)
);
o_vmv[0] = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_VTOP);
o_vmv[1] = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_VBOTTOM);
/*
* mv_scale.xy = 0.5 / (dst.width, dst.height);
* mv_scale.z = 1.0f / 4.0f
* mv_scale.w = 1.0f / 255.0f
*
* // Apply motion vectors
* o_vmv[0..1].xy = vmv[0..1] * mv_scale + t_vpos
* o_vmv[0..1].zw = vmv[0..1] * mv_scale
*
*/
mv_scale = ureg_imm4f(shader,
0.5f / r->buffer_width,
0.5f / r->buffer_height,
1.0f / 4.0f,
1.0f / PIPE_VIDEO_MV_WEIGHT_MAX);
for (i = 0; i < 2; ++i) {
ureg_MAD(shader, ureg_writemask(o_vmv[i], TGSI_WRITEMASK_XY), mv_scale, vmv[i], ureg_src(t_vpos));
ureg_MUL(shader, ureg_writemask(o_vmv[i], TGSI_WRITEMASK_ZW), mv_scale, vmv[i]);
}
ureg_release_temporary(shader, t_vpos);
ureg_END(shader);
return ureg_create_shader_and_destroy(shader, r->pipe);
}
static struct ureg_src
vs_normalize_coords(struct ureg_program *ureg,
struct ureg_src coords,
struct ureg_src const0, struct ureg_src const1)
{
struct ureg_dst tmp = ureg_DECL_temporary(ureg);
struct ureg_src ret;
ureg_MAD(ureg, tmp, coords, const0, const1);
ret = ureg_src(tmp);
ureg_release_temporary(ureg, tmp);
return ret;
}
static void *
create_frag_shader_palette(struct vl_compositor *c, bool include_cc)
{
struct ureg_program *shader;
struct ureg_src csc[3];
struct ureg_src tc;
struct ureg_src sampler;
struct ureg_src palette;
struct ureg_dst texel;
struct ureg_dst fragment;
unsigned i;
shader = ureg_create(TGSI_PROCESSOR_FRAGMENT);
if (!shader)
return false;
for (i = 0; include_cc && i < 3; ++i)
csc[i] = ureg_DECL_constant(shader, i);
tc = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_VTEX, TGSI_INTERPOLATE_LINEAR);
sampler = ureg_DECL_sampler(shader, 0);
palette = ureg_DECL_sampler(shader, 1);
texel = ureg_DECL_temporary(shader);
fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
/*
* texel = tex(tc, sampler)
* fragment.xyz = tex(texel, palette) * csc
* fragment.a = texel.a
*/
ureg_TEX(shader, texel, TGSI_TEXTURE_2D, tc, sampler);
ureg_MOV(shader, ureg_writemask(fragment, TGSI_WRITEMASK_W), ureg_src(texel));
if (include_cc) {
ureg_TEX(shader, texel, TGSI_TEXTURE_1D, ureg_src(texel), palette);
for (i = 0; i < 3; ++i)
ureg_DP4(shader, ureg_writemask(fragment, TGSI_WRITEMASK_X << i), csc[i], ureg_src(texel));
} else {
ureg_TEX(shader, ureg_writemask(fragment, TGSI_WRITEMASK_XYZ),
TGSI_TEXTURE_1D, ureg_src(texel), palette);
}
ureg_release_temporary(shader, texel);
ureg_END(shader);
return ureg_create_shader_and_destroy(shader, c->pipe);
}
static void *
create_mismatch_vert_shader(struct vl_idct *idct)
{
struct ureg_program *shader;
struct ureg_src vpos;
struct ureg_src scale;
struct ureg_dst t_tex;
struct ureg_dst o_vpos, o_addr[2];
shader = ureg_create(TGSI_PROCESSOR_VERTEX);
if (!shader)
return NULL;
vpos = ureg_DECL_vs_input(shader, VS_I_VPOS);
t_tex = ureg_DECL_temporary(shader);
o_vpos = ureg_DECL_output(shader, TGSI_SEMANTIC_POSITION, VS_O_VPOS);
o_addr[0] = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_L_ADDR0);
o_addr[1] = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_L_ADDR1);
/*
* scale = (VL_BLOCK_WIDTH, VL_BLOCK_HEIGHT) / (dst.width, dst.height)
*
* t_vpos = vpos + 7 / VL_BLOCK_WIDTH
* o_vpos.xy = t_vpos * scale
*
* o_addr = calc_addr(...)
*
*/
scale = ureg_imm2f(shader,
(float)VL_BLOCK_WIDTH / idct->buffer_width,
(float)VL_BLOCK_HEIGHT / idct->buffer_height);
ureg_MAD(shader, ureg_writemask(o_vpos, TGSI_WRITEMASK_XY), vpos, scale, scale);
ureg_MOV(shader, ureg_writemask(o_vpos, TGSI_WRITEMASK_ZW), ureg_imm1f(shader, 1.0f));
ureg_MUL(shader, ureg_writemask(t_tex, TGSI_WRITEMASK_XY), vpos, scale);
calc_addr(shader, o_addr, ureg_src(t_tex), ureg_src(t_tex), false, false, idct->buffer_width / 4);
ureg_release_temporary(shader, t_tex);
ureg_END(shader);
return ureg_create_shader_and_destroy(shader, idct->pipe);
}
static void
matrix_mul(struct ureg_program *shader, struct ureg_dst dst, struct ureg_dst l[2], struct ureg_dst r[2])
{
struct ureg_dst tmp;
tmp = ureg_DECL_temporary(shader);
/*
* tmp.xy = dot4(m[0][0..1], m[1][0..1])
* dst = tmp.x + tmp.y
*/
ureg_DP4(shader, ureg_writemask(tmp, TGSI_WRITEMASK_X), ureg_src(l[0]), ureg_src(r[0]));
ureg_DP4(shader, ureg_writemask(tmp, TGSI_WRITEMASK_Y), ureg_src(l[1]), ureg_src(r[1]));
ureg_ADD(shader, dst,
ureg_scalar(ureg_src(tmp), TGSI_SWIZZLE_X),
ureg_scalar(ureg_src(tmp), TGSI_SWIZZLE_Y));
ureg_release_temporary(shader, tmp);
}
static void
linear_gradient(struct ureg_program *ureg,
struct ureg_dst out,
struct ureg_src pos,
struct ureg_src sampler,
struct ureg_src coords,
struct ureg_src const0124,
struct ureg_src matrow0,
struct ureg_src matrow1, struct ureg_src matrow2)
{
struct ureg_dst temp0 = ureg_DECL_temporary(ureg);
struct ureg_dst temp1 = ureg_DECL_temporary(ureg);
struct ureg_dst temp2 = ureg_DECL_temporary(ureg);
struct ureg_dst temp3 = ureg_DECL_temporary(ureg);
struct ureg_dst temp4 = ureg_DECL_temporary(ureg);
struct ureg_dst temp5 = ureg_DECL_temporary(ureg);
ureg_MOV(ureg, ureg_writemask(temp0, TGSI_WRITEMASK_XY), pos);
ureg_MOV(ureg,
ureg_writemask(temp0, TGSI_WRITEMASK_Z),
ureg_scalar(const0124, TGSI_SWIZZLE_Y));
ureg_DP3(ureg, temp1, matrow0, ureg_src(temp0));
ureg_DP3(ureg, temp2, matrow1, ureg_src(temp0));
ureg_DP3(ureg, temp3, matrow2, ureg_src(temp0));
ureg_RCP(ureg, temp3, ureg_src(temp3));
ureg_MUL(ureg, temp1, ureg_src(temp1), ureg_src(temp3));
ureg_MUL(ureg, temp2, ureg_src(temp2), ureg_src(temp3));
ureg_MOV(ureg, ureg_writemask(temp4, TGSI_WRITEMASK_X), ureg_src(temp1));
ureg_MOV(ureg, ureg_writemask(temp4, TGSI_WRITEMASK_Y), ureg_src(temp2));
ureg_MUL(ureg, temp0,
ureg_scalar(coords, TGSI_SWIZZLE_Y),
ureg_scalar(ureg_src(temp4), TGSI_SWIZZLE_Y));
ureg_MAD(ureg, temp1,
ureg_scalar(coords, TGSI_SWIZZLE_X),
ureg_scalar(ureg_src(temp4), TGSI_SWIZZLE_X), ureg_src(temp0));
ureg_MUL(ureg, temp2, ureg_src(temp1), ureg_scalar(coords, TGSI_SWIZZLE_Z));
ureg_TEX(ureg, out, TGSI_TEXTURE_1D, ureg_src(temp2), sampler);
ureg_release_temporary(ureg, temp0);
ureg_release_temporary(ureg, temp1);
ureg_release_temporary(ureg, temp2);
ureg_release_temporary(ureg, temp3);
ureg_release_temporary(ureg, temp4);
ureg_release_temporary(ureg, temp5);
}
static void *
create_frag_shader_video_buffer(struct vl_compositor *c)
{
struct ureg_program *shader;
struct ureg_src tc;
struct ureg_src csc[3];
struct ureg_src sampler[3];
struct ureg_dst texel;
struct ureg_dst fragment;
unsigned i;
shader = ureg_create(TGSI_PROCESSOR_FRAGMENT);
if (!shader)
return false;
tc = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_VTEX, TGSI_INTERPOLATE_LINEAR);
for (i = 0; i < 3; ++i) {
csc[i] = ureg_DECL_constant(shader, i);
sampler[i] = ureg_DECL_sampler(shader, i);
}
texel = ureg_DECL_temporary(shader);
fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
/*
* texel.xyz = tex(tc, sampler[i])
* fragment = csc * texel
*/
for (i = 0; i < 3; ++i)
ureg_TEX(shader, ureg_writemask(texel, TGSI_WRITEMASK_X << i), TGSI_TEXTURE_3D, tc, sampler[i]);
ureg_MOV(shader, ureg_writemask(texel, TGSI_WRITEMASK_W), ureg_imm1f(shader, 1.0f));
for (i = 0; i < 3; ++i)
ureg_DP4(shader, ureg_writemask(fragment, TGSI_WRITEMASK_X << i), csc[i], ureg_src(texel));
ureg_MOV(shader, ureg_writemask(fragment, TGSI_WRITEMASK_W), ureg_imm1f(shader, 1.0f));
ureg_release_temporary(shader, texel);
ureg_END(shader);
return ureg_create_shader_and_destroy(shader, c->pipe);
}
static void *
create_frag_shader_weave(struct vl_compositor *c)
{
struct ureg_program *shader;
struct ureg_src i_tc[2];
struct ureg_src csc[3];
struct ureg_src sampler[3];
struct ureg_dst t_tc[2];
struct ureg_dst t_texel[2];
struct ureg_dst o_fragment;
unsigned i, j;
shader = ureg_create(TGSI_PROCESSOR_FRAGMENT);
if (!shader)
return false;
i_tc[0] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_VTOP, TGSI_INTERPOLATE_LINEAR);
i_tc[1] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_VBOTTOM, TGSI_INTERPOLATE_LINEAR);
for (i = 0; i < 3; ++i) {
csc[i] = ureg_DECL_constant(shader, i);
sampler[i] = ureg_DECL_sampler(shader, i);
}
for (i = 0; i < 2; ++i) {
t_tc[i] = ureg_DECL_temporary(shader);
t_texel[i] = ureg_DECL_temporary(shader);
}
o_fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
/* calculate the texture offsets
* t_tc.x = i_tc.x
* t_tc.y = (round(i_tc.y) + 0.5) / height * 2
*/
for (i = 0; i < 2; ++i) {
ureg_MOV(shader, ureg_writemask(t_tc[i], TGSI_WRITEMASK_X), i_tc[i]);
ureg_ROUND(shader, ureg_writemask(t_tc[i], TGSI_WRITEMASK_YZ), i_tc[i]);
ureg_MOV(shader, ureg_writemask(t_tc[i], TGSI_WRITEMASK_W),
ureg_imm1f(shader, i ? 0.75f : 0.25f));
ureg_ADD(shader, ureg_writemask(t_tc[i], TGSI_WRITEMASK_YZ),
ureg_src(t_tc[i]), ureg_imm1f(shader, 0.5f));
ureg_MUL(shader, ureg_writemask(t_tc[i], TGSI_WRITEMASK_Y),
ureg_src(t_tc[i]), ureg_scalar(i_tc[0], TGSI_SWIZZLE_W));
ureg_MUL(shader, ureg_writemask(t_tc[i], TGSI_WRITEMASK_Z),
ureg_src(t_tc[i]), ureg_scalar(i_tc[1], TGSI_SWIZZLE_W));
}
/* fetch the texels
* texel[0..1].x = tex(t_tc[0..1][0])
* texel[0..1].y = tex(t_tc[0..1][1])
* texel[0..1].z = tex(t_tc[0..1][2])
*/
for (i = 0; i < 2; ++i)
for (j = 0; j < 3; ++j) {
struct ureg_src src = ureg_swizzle(ureg_src(t_tc[i]),
TGSI_SWIZZLE_X, j ? TGSI_SWIZZLE_Z : TGSI_SWIZZLE_Y, TGSI_SWIZZLE_W, TGSI_SWIZZLE_W);
ureg_TEX(shader, ureg_writemask(t_texel[i], TGSI_WRITEMASK_X << j),
TGSI_TEXTURE_3D, src, sampler[j]);
}
/* calculate linear interpolation factor
* factor = |round(i_tc.y) - i_tc.y| * 2
*/
ureg_ROUND(shader, ureg_writemask(t_tc[0], TGSI_WRITEMASK_YZ), i_tc[0]);
ureg_ADD(shader, ureg_writemask(t_tc[0], TGSI_WRITEMASK_YZ),
ureg_src(t_tc[0]), ureg_negate(i_tc[0]));
ureg_MUL(shader, ureg_writemask(t_tc[0], TGSI_WRITEMASK_XY),
ureg_abs(ureg_src(t_tc[0])), ureg_imm1f(shader, 2.0f));
ureg_LRP(shader, t_texel[0], ureg_swizzle(ureg_src(t_tc[0]),
TGSI_SWIZZLE_Y, TGSI_SWIZZLE_Z, TGSI_SWIZZLE_Z, TGSI_SWIZZLE_Z),
ureg_src(t_texel[1]), ureg_src(t_texel[0]));
/* and finally do colour space transformation
* fragment = csc * texel
*/
ureg_MOV(shader, ureg_writemask(t_texel[0], TGSI_WRITEMASK_W), ureg_imm1f(shader, 1.0f));
for (i = 0; i < 3; ++i)
ureg_DP4(shader, ureg_writemask(o_fragment, TGSI_WRITEMASK_X << i), csc[i], ureg_src(t_texel[0]));
ureg_MOV(shader, ureg_writemask(o_fragment, TGSI_WRITEMASK_W), ureg_imm1f(shader, 1.0f));
for (i = 0; i < 2; ++i) {
ureg_release_temporary(shader, t_texel[i]);
ureg_release_temporary(shader, t_tc[i]);
}
ureg_END(shader);
return ureg_create_shader_and_destroy(shader, c->pipe);
}
static void *
create_deint_frag_shader(struct vl_deint_filter *filter, unsigned field,
struct vertex2f *sizes, bool spatial_filter)
{
struct ureg_program *shader;
struct ureg_src i_vtex;
struct ureg_src sampler_cur;
struct ureg_src sampler_prevprev;
struct ureg_src sampler_prev;
struct ureg_src sampler_next;
struct ureg_dst o_fragment;
struct ureg_dst t_tex;
struct ureg_dst t_comp_top, t_comp_bot;
struct ureg_dst t_diff;
struct ureg_dst t_a, t_b;
struct ureg_dst t_weave, t_linear;
shader = ureg_create(PIPE_SHADER_FRAGMENT);
if (!shader) {
return NULL;
}
t_tex = ureg_DECL_temporary(shader);
t_comp_top = ureg_DECL_temporary(shader);
t_comp_bot = ureg_DECL_temporary(shader);
t_diff = ureg_DECL_temporary(shader);
t_a = ureg_DECL_temporary(shader);
t_b = ureg_DECL_temporary(shader);
t_weave = ureg_DECL_temporary(shader);
t_linear = ureg_DECL_temporary(shader);
i_vtex = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_VTEX, TGSI_INTERPOLATE_LINEAR);
sampler_prevprev = ureg_DECL_sampler(shader, 0);
sampler_prev = ureg_DECL_sampler(shader, 1);
sampler_cur = ureg_DECL_sampler(shader, 2);
sampler_next = ureg_DECL_sampler(shader, 3);
o_fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
// we don't care about ZW interpolation (allows better optimization)
ureg_MOV(shader, t_tex, i_vtex);
ureg_MOV(shader, ureg_writemask(t_tex, TGSI_WRITEMASK_ZW),
ureg_imm1f(shader, 0));
// sample between texels for cheap lowpass
ureg_ADD(shader, t_comp_top, ureg_src(t_tex),
ureg_imm4f(shader, sizes->x * 0.5f, sizes->y * -0.5f, 0, 0));
ureg_ADD(shader, t_comp_bot, ureg_src(t_tex),
ureg_imm4f(shader, sizes->x * -0.5f, sizes->y * 0.5f, 1.0f, 0));
if (field == 0) {
/* interpolating top field -> current field is a bottom field */
// cur vs prev2
ureg_TEX(shader, t_a, TGSI_TEXTURE_2D_ARRAY, ureg_src(t_comp_bot), sampler_cur);
ureg_TEX(shader, t_b, TGSI_TEXTURE_2D_ARRAY, ureg_src(t_comp_bot), sampler_prevprev);
ureg_ADD(shader, ureg_writemask(t_diff, TGSI_WRITEMASK_X), ureg_src(t_a), ureg_negate(ureg_src(t_b)));
// prev vs next
ureg_TEX(shader, t_a, TGSI_TEXTURE_2D_ARRAY, ureg_src(t_comp_top), sampler_prev);
ureg_TEX(shader, t_b, TGSI_TEXTURE_2D_ARRAY, ureg_src(t_comp_top), sampler_next);
ureg_ADD(shader, ureg_writemask(t_diff, TGSI_WRITEMASK_Y), ureg_src(t_a), ureg_negate(ureg_src(t_b)));
} else {
/* interpolating bottom field -> current field is a top field */
// cur vs prev2
ureg_TEX(shader, t_a, TGSI_TEXTURE_2D_ARRAY, ureg_src(t_comp_top), sampler_cur);
ureg_TEX(shader, t_b, TGSI_TEXTURE_2D_ARRAY, ureg_src(t_comp_top), sampler_prevprev);
ureg_ADD(shader, ureg_writemask(t_diff, TGSI_WRITEMASK_X), ureg_src(t_a), ureg_negate(ureg_src(t_b)));
// prev vs next
ureg_TEX(shader, t_a, TGSI_TEXTURE_2D_ARRAY, ureg_src(t_comp_bot), sampler_prev);
ureg_TEX(shader, t_b, TGSI_TEXTURE_2D_ARRAY, ureg_src(t_comp_bot), sampler_next);
ureg_ADD(shader, ureg_writemask(t_diff, TGSI_WRITEMASK_Y), ureg_src(t_a), ureg_negate(ureg_src(t_b)));
}
// absolute maximum of differences
ureg_MAX(shader, ureg_writemask(t_diff, TGSI_WRITEMASK_X), ureg_abs(ureg_src(t_diff)),
ureg_scalar(ureg_abs(ureg_src(t_diff)), TGSI_SWIZZLE_Y));
if (field == 0) {
/* weave with prev top field */
ureg_TEX(shader, t_weave, TGSI_TEXTURE_2D_ARRAY, ureg_src(t_tex), sampler_prev);
/* get linear interpolation from current bottom field */
ureg_ADD(shader, t_comp_top, ureg_src(t_tex), ureg_imm4f(shader, 0, sizes->y * -1.0f, 1.0f, 0));
ureg_TEX(shader, t_linear, TGSI_TEXTURE_2D_ARRAY, ureg_src(t_comp_top), sampler_cur);
} else {
/* weave with prev bottom field */
ureg_ADD(shader, t_comp_bot, ureg_src(t_tex), ureg_imm4f(shader, 0, 0, 1.0f, 0));
ureg_TEX(shader, t_weave, TGSI_TEXTURE_2D_ARRAY, ureg_src(t_comp_bot), sampler_prev);
/* get linear interpolation from current top field */
ureg_ADD(shader, t_comp_bot, ureg_src(t_tex), ureg_imm4f(shader, 0, sizes->y * 1.0f, 0, 0));
ureg_TEX(shader, t_linear, TGSI_TEXTURE_2D_ARRAY, ureg_src(t_comp_bot), sampler_cur);
}
// mix between weave and linear
// fully weave if diff < 6 (0.02353), fully interpolate if diff > 14 (0.05490)
ureg_ADD(shader, ureg_writemask(t_diff, TGSI_WRITEMASK_X), ureg_src(t_diff),
ureg_imm4f(shader, -0.02353f, 0, 0, 0));
ureg_MUL(shader, ureg_saturate(ureg_writemask(t_diff, TGSI_WRITEMASK_X)),
ureg_src(t_diff), ureg_imm4f(shader, 31.8750f, 0, 0, 0));
ureg_LRP(shader, ureg_writemask(t_tex, TGSI_WRITEMASK_X), ureg_src(t_diff),
ureg_src(t_linear), ureg_src(t_weave));
ureg_MOV(shader, o_fragment, ureg_scalar(ureg_src(t_tex), TGSI_SWIZZLE_X));
ureg_release_temporary(shader, t_tex);
ureg_release_temporary(shader, t_comp_top);
ureg_release_temporary(shader, t_comp_bot);
ureg_release_temporary(shader, t_diff);
ureg_release_temporary(shader, t_a);
ureg_release_temporary(shader, t_b);
ureg_release_temporary(shader, t_weave);
ureg_release_temporary(shader, t_linear);
ureg_END(shader);
return ureg_create_shader_and_destroy(shader, filter->pipe);
}
static void *
create_fs(struct st_context *st, bool download, enum pipe_texture_target target)
{
struct pipe_context *pipe = st->pipe;
struct pipe_screen *screen = pipe->screen;
struct ureg_program *ureg;
bool have_layer;
struct ureg_dst out;
struct ureg_src sampler;
struct ureg_src pos;
struct ureg_src layer;
struct ureg_src const0;
struct ureg_src const1;
struct ureg_dst temp0;
have_layer =
st->pbo.layers &&
(!download || target == PIPE_TEXTURE_1D_ARRAY
|| target == PIPE_TEXTURE_2D_ARRAY
|| target == PIPE_TEXTURE_3D
|| target == PIPE_TEXTURE_CUBE
|| target == PIPE_TEXTURE_CUBE_ARRAY);
ureg = ureg_create(PIPE_SHADER_FRAGMENT);
if (!ureg)
return NULL;
if (!download) {
out = ureg_DECL_output(ureg, TGSI_SEMANTIC_COLOR, 0);
} else {
struct ureg_src image;
/* writeonly images do not require an explicitly given format. */
image = ureg_DECL_image(ureg, 0, TGSI_TEXTURE_BUFFER, PIPE_FORMAT_NONE,
true, false);
out = ureg_dst(image);
}
sampler = ureg_DECL_sampler(ureg, 0);
if (screen->get_param(screen, PIPE_CAP_TGSI_FS_POSITION_IS_SYSVAL)) {
pos = ureg_DECL_system_value(ureg, TGSI_SEMANTIC_POSITION, 0);
} else {
pos = ureg_DECL_fs_input(ureg, TGSI_SEMANTIC_POSITION, 0,
TGSI_INTERPOLATE_LINEAR);
}
if (have_layer) {
layer = ureg_DECL_fs_input(ureg, TGSI_SEMANTIC_LAYER, 0,
TGSI_INTERPOLATE_CONSTANT);
}
const0 = ureg_DECL_constant(ureg, 0);
const1 = ureg_DECL_constant(ureg, 1);
temp0 = ureg_DECL_temporary(ureg);
/* Note: const0 = [ -xoffset + skip_pixels, -yoffset, stride, image_height ] */
/* temp0.xy = f2i(temp0.xy) */
ureg_F2I(ureg, ureg_writemask(temp0, TGSI_WRITEMASK_XY),
ureg_swizzle(pos,
TGSI_SWIZZLE_X, TGSI_SWIZZLE_Y,
TGSI_SWIZZLE_Y, TGSI_SWIZZLE_Y));
/* temp0.xy = temp0.xy + const0.xy */
ureg_UADD(ureg, ureg_writemask(temp0, TGSI_WRITEMASK_XY),
ureg_swizzle(ureg_src(temp0),
TGSI_SWIZZLE_X, TGSI_SWIZZLE_Y,
TGSI_SWIZZLE_Y, TGSI_SWIZZLE_Y),
ureg_swizzle(const0,
TGSI_SWIZZLE_X, TGSI_SWIZZLE_Y,
TGSI_SWIZZLE_Y, TGSI_SWIZZLE_Y));
/* temp0.x = const0.z * temp0.y + temp0.x */
ureg_UMAD(ureg, ureg_writemask(temp0, TGSI_WRITEMASK_X),
ureg_scalar(const0, TGSI_SWIZZLE_Z),
ureg_scalar(ureg_src(temp0), TGSI_SWIZZLE_Y),
ureg_scalar(ureg_src(temp0), TGSI_SWIZZLE_X));
if (have_layer) {
/* temp0.x = const0.w * layer + temp0.x */
ureg_UMAD(ureg, ureg_writemask(temp0, TGSI_WRITEMASK_X),
ureg_scalar(const0, TGSI_SWIZZLE_W),
ureg_scalar(layer, TGSI_SWIZZLE_X),
ureg_scalar(ureg_src(temp0), TGSI_SWIZZLE_X));
}
/* temp0.w = 0 */
ureg_MOV(ureg, ureg_writemask(temp0, TGSI_WRITEMASK_W), ureg_imm1u(ureg, 0));
if (download) {
struct ureg_dst temp1;
struct ureg_src op[2];
temp1 = ureg_DECL_temporary(ureg);
/* temp1.xy = pos.xy */
ureg_F2I(ureg, ureg_writemask(temp1, TGSI_WRITEMASK_XY), pos);
/* temp1.zw = 0 */
ureg_MOV(ureg, ureg_writemask(temp1, TGSI_WRITEMASK_ZW), ureg_imm1u(ureg, 0));
if (have_layer) {
struct ureg_dst temp1_layer =
ureg_writemask(temp1, target == PIPE_TEXTURE_1D_ARRAY ? TGSI_WRITEMASK_Y
: TGSI_WRITEMASK_Z);
/* temp1.y/z = layer */
ureg_MOV(ureg, temp1_layer, ureg_scalar(layer, TGSI_SWIZZLE_X));
if (target == PIPE_TEXTURE_3D) {
/* temp1.z += layer_offset */
ureg_UADD(ureg, temp1_layer,
ureg_scalar(ureg_src(temp1), TGSI_SWIZZLE_Z),
ureg_scalar(const1, TGSI_SWIZZLE_X));
}
}
/* temp1 = txf(sampler, temp1) */
ureg_TXF(ureg, temp1, util_pipe_tex_to_tgsi_tex(target, 1),
ureg_src(temp1), sampler);
/* store(out, temp0, temp1) */
op[0] = ureg_src(temp0);
op[1] = ureg_src(temp1);
ureg_memory_insn(ureg, TGSI_OPCODE_STORE, &out, 1, op, 2, 0,
TGSI_TEXTURE_BUFFER, PIPE_FORMAT_NONE);
ureg_release_temporary(ureg, temp1);
} else {
/* out = txf(sampler, temp0.x) */
ureg_TXF(ureg, out, TGSI_TEXTURE_BUFFER, ureg_src(temp0), sampler);
}
ureg_release_temporary(ureg, temp0);
ureg_END(ureg);
return ureg_create_shader_and_destroy(ureg, pipe);
}
static void *
create_ycbcr_vert_shader(struct vl_mc *r, vl_mc_ycbcr_vert_shader vs_callback, void *callback_priv)
{
struct ureg_program *shader;
struct ureg_src vrect, vpos;
struct ureg_dst t_vpos, t_vtex;
struct ureg_dst o_vpos, o_flags;
struct vertex2f scale = {
(float)VL_BLOCK_WIDTH / r->buffer_width * VL_MACROBLOCK_WIDTH / r->macroblock_size,
(float)VL_BLOCK_HEIGHT / r->buffer_height * VL_MACROBLOCK_HEIGHT / r->macroblock_size
};
unsigned label;
shader = ureg_create(TGSI_PROCESSOR_VERTEX);
if (!shader)
return NULL;
vrect = ureg_DECL_vs_input(shader, VS_I_RECT);
vpos = ureg_DECL_vs_input(shader, VS_I_VPOS);
t_vpos = calc_position(r, shader, ureg_imm2f(shader, scale.x, scale.y));
t_vtex = ureg_DECL_temporary(shader);
o_vpos = ureg_DECL_output(shader, TGSI_SEMANTIC_POSITION, VS_O_VPOS);
o_flags = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_FLAGS);
/*
* o_vtex.xy = t_vpos
* o_flags.z = intra * 0.5
*
* if(interlaced) {
* t_vtex.xy = vrect.y ? { 0, scale.y } : { -scale.y : 0 }
* t_vtex.z = vpos.y % 2
* t_vtex.y = t_vtex.z ? t_vtex.x : t_vtex.y
* o_vpos.y = t_vtex.y + t_vpos.y
*
* o_flags.w = t_vtex.z ? 0 : 1
* }
*
*/
vs_callback(callback_priv, r, shader, VS_O_VTEX, t_vpos);
ureg_MUL(shader, ureg_writemask(o_flags, TGSI_WRITEMASK_Z),
ureg_scalar(vpos, TGSI_SWIZZLE_Z), ureg_imm1f(shader, 0.5f));
ureg_MOV(shader, ureg_writemask(o_flags, TGSI_WRITEMASK_W), ureg_imm1f(shader, -1.0f));
if (r->macroblock_size == VL_MACROBLOCK_HEIGHT) { //TODO
ureg_IF(shader, ureg_scalar(vpos, TGSI_SWIZZLE_W), &label);
ureg_CMP(shader, ureg_writemask(t_vtex, TGSI_WRITEMASK_XY),
ureg_negate(ureg_scalar(vrect, TGSI_SWIZZLE_Y)),
ureg_imm2f(shader, 0.0f, scale.y),
ureg_imm2f(shader, -scale.y, 0.0f));
ureg_MUL(shader, ureg_writemask(t_vtex, TGSI_WRITEMASK_Z),
ureg_scalar(vpos, TGSI_SWIZZLE_Y), ureg_imm1f(shader, 0.5f));
ureg_FRC(shader, ureg_writemask(t_vtex, TGSI_WRITEMASK_Z), ureg_src(t_vtex));
ureg_CMP(shader, ureg_writemask(t_vtex, TGSI_WRITEMASK_Y),
ureg_negate(ureg_scalar(ureg_src(t_vtex), TGSI_SWIZZLE_Z)),
ureg_scalar(ureg_src(t_vtex), TGSI_SWIZZLE_X),
ureg_scalar(ureg_src(t_vtex), TGSI_SWIZZLE_Y));
ureg_ADD(shader, ureg_writemask(o_vpos, TGSI_WRITEMASK_Y),
ureg_src(t_vpos), ureg_src(t_vtex));
ureg_CMP(shader, ureg_writemask(o_flags, TGSI_WRITEMASK_W),
ureg_negate(ureg_scalar(ureg_src(t_vtex), TGSI_SWIZZLE_Z)),
ureg_imm1f(shader, 0.0f), ureg_imm1f(shader, 1.0f));
ureg_fixup_label(shader, label, ureg_get_instruction_number(shader));
ureg_ENDIF(shader);
}
ureg_release_temporary(shader, t_vtex);
ureg_release_temporary(shader, t_vpos);
ureg_END(shader);
return ureg_create_shader_and_destroy(shader, r->pipe);
}
static void *
create_ref_frag_shader(struct vl_mc *r)
{
const float y_scale =
r->buffer_height / 2 *
r->macroblock_size / VL_MACROBLOCK_HEIGHT;
struct ureg_program *shader;
struct ureg_src tc[2], sampler;
struct ureg_dst ref, field;
struct ureg_dst fragment;
unsigned label;
shader = ureg_create(TGSI_PROCESSOR_FRAGMENT);
if (!shader)
return NULL;
tc[0] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_VTOP, TGSI_INTERPOLATE_LINEAR);
tc[1] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_VBOTTOM, TGSI_INTERPOLATE_LINEAR);
sampler = ureg_DECL_sampler(shader, 0);
ref = ureg_DECL_temporary(shader);
fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
field = calc_line(shader);
/*
* ref = field.z ? tc[1] : tc[0]
*
* // Adjust tc acording to top/bottom field selection
* if (|ref.z|) {
* ref.y *= y_scale
* ref.y = floor(ref.y)
* ref.y += ref.z
* ref.y /= y_scale
* }
* fragment.xyz = tex(ref, sampler[0])
*/
ureg_CMP(shader, ureg_writemask(ref, TGSI_WRITEMASK_XYZ),
ureg_negate(ureg_scalar(ureg_src(field), TGSI_SWIZZLE_Y)),
tc[1], tc[0]);
ureg_CMP(shader, ureg_writemask(fragment, TGSI_WRITEMASK_W),
ureg_negate(ureg_scalar(ureg_src(field), TGSI_SWIZZLE_Y)),
tc[1], tc[0]);
ureg_IF(shader, ureg_scalar(ureg_src(ref), TGSI_SWIZZLE_Z), &label);
ureg_MUL(shader, ureg_writemask(ref, TGSI_WRITEMASK_Y),
ureg_src(ref), ureg_imm1f(shader, y_scale));
ureg_FLR(shader, ureg_writemask(ref, TGSI_WRITEMASK_Y), ureg_src(ref));
ureg_ADD(shader, ureg_writemask(ref, TGSI_WRITEMASK_Y),
ureg_src(ref), ureg_scalar(ureg_src(ref), TGSI_SWIZZLE_Z));
ureg_MUL(shader, ureg_writemask(ref, TGSI_WRITEMASK_Y),
ureg_src(ref), ureg_imm1f(shader, 1.0f / y_scale));
ureg_fixup_label(shader, label, ureg_get_instruction_number(shader));
ureg_ENDIF(shader);
ureg_TEX(shader, ureg_writemask(fragment, TGSI_WRITEMASK_XYZ), TGSI_TEXTURE_2D, ureg_src(ref), sampler);
ureg_release_temporary(shader, ref);
ureg_release_temporary(shader, field);
ureg_END(shader);
return ureg_create_shader_and_destroy(shader, r->pipe);
}
static void *
combine_shaders(const struct shader_asm_info *shaders[SHADER_STAGES], int num_shaders,
struct pipe_context *pipe,
struct pipe_shader_state *shader)
{
VGboolean declare_input = VG_FALSE;
VGint start_const = -1, end_const = 0;
VGint start_temp = -1, end_temp = 0;
VGint start_sampler = -1, end_sampler = 0;
VGint i, current_shader = 0;
VGint num_consts, num_temps, num_samplers;
struct ureg_program *ureg;
struct ureg_src in[2];
struct ureg_src *sampler = NULL;
struct ureg_src *constant = NULL;
struct ureg_dst out, *temp = NULL;
void *p = NULL;
for (i = 0; i < num_shaders; ++i) {
if (shaders[i]->num_consts)
start_const = range_min(start_const, shaders[i]->start_const);
if (shaders[i]->num_temps)
start_temp = range_min(start_temp, shaders[i]->start_temp);
if (shaders[i]->num_samplers)
start_sampler = range_min(start_sampler, shaders[i]->start_sampler);
end_const = range_max(end_const, shaders[i]->start_const +
shaders[i]->num_consts);
end_temp = range_max(end_temp, shaders[i]->start_temp +
shaders[i]->num_temps);
end_sampler = range_max(end_sampler, shaders[i]->start_sampler +
shaders[i]->num_samplers);
if (shaders[i]->needs_position)
declare_input = VG_TRUE;
}
/* if they're still unitialized, initialize them */
if (start_const < 0)
start_const = 0;
if (start_temp < 0)
start_temp = 0;
if (start_sampler < 0)
start_sampler = 0;
num_consts = end_const - start_const;
num_temps = end_temp - start_temp;
num_samplers = end_sampler - start_sampler;
ureg = ureg_create(TGSI_PROCESSOR_FRAGMENT);
if (!ureg)
return NULL;
if (declare_input) {
in[0] = ureg_DECL_fs_input(ureg,
TGSI_SEMANTIC_POSITION,
0,
TGSI_INTERPOLATE_LINEAR);
in[1] = ureg_DECL_fs_input(ureg,
TGSI_SEMANTIC_GENERIC,
0,
TGSI_INTERPOLATE_PERSPECTIVE);
}
/* we always have a color output */
out = ureg_DECL_output(ureg, TGSI_SEMANTIC_COLOR, 0);
if (num_consts >= 1) {
constant = (struct ureg_src *) malloc(sizeof(struct ureg_src) * end_const);
for (i = start_const; i < end_const; i++) {
constant[i] = ureg_DECL_constant(ureg, i);
}
}
if (num_temps >= 1) {
temp = (struct ureg_dst *) malloc(sizeof(struct ureg_dst) * end_temp);
for (i = start_temp; i < end_temp; i++) {
temp[i] = ureg_DECL_temporary(ureg);
}
}
if (num_samplers >= 1) {
sampler = (struct ureg_src *) malloc(sizeof(struct ureg_src) * end_sampler);
for (i = start_sampler; i < end_sampler; i++) {
sampler[i] = ureg_DECL_sampler(ureg, i);
}
}
while (current_shader < num_shaders) {
if ((current_shader + 1) == num_shaders) {
shaders[current_shader]->func(ureg,
&out,
in,
sampler,
temp,
constant);
} else {
shaders[current_shader]->func(ureg,
&temp[0],
in,
sampler,
temp,
constant);
}
current_shader++;
}
ureg_END(ureg);
shader->tokens = ureg_finalize(ureg);
if(!shader->tokens)
return NULL;
p = pipe->create_fs_state(pipe, shader);
ureg_destroy(ureg);
if (num_temps >= 1) {
for (i = start_temp; i < end_temp; i++) {
ureg_release_temporary(ureg, temp[i]);
}
}
if (temp)
free(temp);
if (constant)
free(constant);
if (sampler)
free(sampler);
return p;
}
static void *
create_stage1_frag_shader(struct vl_idct *idct)
{
struct ureg_program *shader;
struct ureg_src l_addr[2], r_addr[2];
struct ureg_dst l[4][2], r[2];
struct ureg_dst *fragment;
int i, j;
shader = ureg_create(TGSI_PROCESSOR_FRAGMENT);
if (!shader)
return NULL;
fragment = MALLOC(idct->nr_of_render_targets * sizeof(struct ureg_dst));
l_addr[0] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_L_ADDR0, TGSI_INTERPOLATE_LINEAR);
l_addr[1] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_L_ADDR1, TGSI_INTERPOLATE_LINEAR);
r_addr[0] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_R_ADDR0, TGSI_INTERPOLATE_LINEAR);
r_addr[1] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_R_ADDR1, TGSI_INTERPOLATE_LINEAR);
for (i = 0; i < idct->nr_of_render_targets; ++i)
fragment[i] = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, i);
for (i = 0; i < 4; ++i) {
l[i][0] = ureg_DECL_temporary(shader);
l[i][1] = ureg_DECL_temporary(shader);
}
r[0] = ureg_DECL_temporary(shader);
r[1] = ureg_DECL_temporary(shader);
for (i = 0; i < 4; ++i) {
increment_addr(shader, l[i], l_addr, false, false, i - 2, idct->buffer_height);
}
for (i = 0; i < 4; ++i) {
struct ureg_src s_addr[2];
s_addr[0] = ureg_src(l[i][0]);
s_addr[1] = ureg_src(l[i][1]);
fetch_four(shader, l[i], s_addr, ureg_DECL_sampler(shader, 0), false);
}
for (i = 0; i < idct->nr_of_render_targets; ++i) {
struct ureg_src s_addr[2];
increment_addr(shader, r, r_addr, true, true, i - (signed)idct->nr_of_render_targets / 2, VL_BLOCK_HEIGHT);
s_addr[0] = ureg_src(r[0]);
s_addr[1] = ureg_src(r[1]);
fetch_four(shader, r, s_addr, ureg_DECL_sampler(shader, 1), false);
for (j = 0; j < 4; ++j) {
matrix_mul(shader, ureg_writemask(fragment[i], TGSI_WRITEMASK_X << j), l[j], r);
}
}
for (i = 0; i < 4; ++i) {
ureg_release_temporary(shader, l[i][0]);
ureg_release_temporary(shader, l[i][1]);
}
ureg_release_temporary(shader, r[0]);
ureg_release_temporary(shader, r[1]);
ureg_END(shader);
FREE(fragment);
return ureg_create_shader_and_destroy(shader, idct->pipe);
}
static void *
create_mismatch_frag_shader(struct vl_idct *idct)
{
struct ureg_program *shader;
struct ureg_src addr[2];
struct ureg_dst m[8][2];
struct ureg_dst fragment;
unsigned i;
shader = ureg_create(TGSI_PROCESSOR_FRAGMENT);
if (!shader)
return NULL;
addr[0] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_L_ADDR0, TGSI_INTERPOLATE_LINEAR);
addr[1] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_L_ADDR1, TGSI_INTERPOLATE_LINEAR);
fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
for (i = 0; i < 8; ++i) {
m[i][0] = ureg_DECL_temporary(shader);
m[i][1] = ureg_DECL_temporary(shader);
}
for (i = 0; i < 8; ++i) {
increment_addr(shader, m[i], addr, false, false, i, idct->buffer_height);
}
for (i = 0; i < 8; ++i) {
struct ureg_src s_addr[2];
s_addr[0] = ureg_src(m[i][0]);
s_addr[1] = ureg_src(m[i][1]);
fetch_four(shader, m[i], s_addr, ureg_DECL_sampler(shader, 0), false);
}
for (i = 1; i < 8; ++i) {
ureg_ADD(shader, m[0][0], ureg_src(m[0][0]), ureg_src(m[i][0]));
ureg_ADD(shader, m[0][1], ureg_src(m[0][1]), ureg_src(m[i][1]));
}
ureg_ADD(shader, m[0][0], ureg_src(m[0][0]), ureg_src(m[0][1]));
ureg_DP4(shader, m[0][0], ureg_abs(ureg_src(m[0][0])), ureg_imm1f(shader, 1 << 14));
ureg_MUL(shader, ureg_writemask(m[0][0], TGSI_WRITEMASK_W), ureg_abs(ureg_src(m[7][1])), ureg_imm1f(shader, 1 << 14));
ureg_FRC(shader, m[0][0], ureg_src(m[0][0]));
ureg_SGT(shader, m[0][0], ureg_imm1f(shader, 0.5f), ureg_abs(ureg_src(m[0][0])));
ureg_CMP(shader, ureg_writemask(m[0][0], TGSI_WRITEMASK_W), ureg_negate(ureg_src(m[0][0])),
ureg_imm1f(shader, 1.0f / (1 << 15)), ureg_imm1f(shader, -1.0f / (1 << 15)));
ureg_MUL(shader, ureg_writemask(m[0][0], TGSI_WRITEMASK_W), ureg_src(m[0][0]),
ureg_scalar(ureg_src(m[0][0]), TGSI_SWIZZLE_X));
ureg_MOV(shader, ureg_writemask(fragment, TGSI_WRITEMASK_XYZ), ureg_src(m[7][1]));
ureg_ADD(shader, ureg_writemask(fragment, TGSI_WRITEMASK_W), ureg_src(m[0][0]), ureg_src(m[7][1]));
for (i = 0; i < 8; ++i) {
ureg_release_temporary(shader, m[i][0]);
ureg_release_temporary(shader, m[i][1]);
}
ureg_END(shader);
return ureg_create_shader_and_destroy(shader, idct->pipe);
}
static void *
create_frag_shader(struct vl_median_filter *filter,
struct vertex2f *offsets,
unsigned num_offsets)
{
struct pipe_screen *screen = filter->pipe->screen;
struct ureg_program *shader;
struct ureg_src i_vtex;
struct ureg_src sampler;
struct ureg_dst *t_array = MALLOC(sizeof(struct ureg_dst) * num_offsets);
struct ureg_dst o_fragment;
const unsigned median = num_offsets >> 1;
unsigned i, j;
assert(num_offsets & 1); /* we need an odd number of offsets */
if (!(num_offsets & 1)) { /* yeah, we REALLY need an odd number of offsets!!! */
FREE(t_array);
return NULL;
}
if (num_offsets > screen->get_shader_param(
screen, PIPE_SHADER_FRAGMENT, PIPE_SHADER_CAP_MAX_TEMPS)) {
FREE(t_array);
return NULL;
}
shader = ureg_create(PIPE_SHADER_FRAGMENT);
if (!shader) {
FREE(t_array);
return NULL;
}
i_vtex = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_VTEX, TGSI_INTERPOLATE_LINEAR);
sampler = ureg_DECL_sampler(shader, 0);
ureg_DECL_sampler_view(shader, 0, TGSI_TEXTURE_2D,
TGSI_RETURN_TYPE_FLOAT,
TGSI_RETURN_TYPE_FLOAT,
TGSI_RETURN_TYPE_FLOAT,
TGSI_RETURN_TYPE_FLOAT);
for (i = 0; i < num_offsets; ++i)
t_array[i] = ureg_DECL_temporary(shader);
o_fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
/*
* t_array[0..*] = vtex + offset[0..*]
* t_array[0..*] = tex(t_array[0..*], sampler)
* result = partial_bubblesort(t_array)[mid]
*/
for (i = 0; i < num_offsets; ++i) {
if (!is_vec_zero(offsets[i])) {
ureg_ADD(shader, ureg_writemask(t_array[i], TGSI_WRITEMASK_XY),
i_vtex, ureg_imm2f(shader, offsets[i].x, offsets[i].y));
ureg_MOV(shader, ureg_writemask(t_array[i], TGSI_WRITEMASK_ZW),
ureg_imm1f(shader, 0.0f));
}
}
for (i = 0; i < num_offsets; ++i) {
struct ureg_src src = is_vec_zero(offsets[i]) ? i_vtex : ureg_src(t_array[i]);
ureg_TEX(shader, t_array[i], TGSI_TEXTURE_2D, src, sampler);
}
// TODO: Couldn't this be improved even more?
for (i = 0; i <= median; ++i) {
for (j = 1; j < (num_offsets - i - 1); ++j) {
struct ureg_dst tmp = ureg_DECL_temporary(shader);
ureg_MOV(shader, tmp, ureg_src(t_array[j]));
ureg_MAX(shader, t_array[j], ureg_src(t_array[j]), ureg_src(t_array[j - 1]));
ureg_MIN(shader, t_array[j - 1], ureg_src(tmp), ureg_src(t_array[j - 1]));
ureg_release_temporary(shader, tmp);
}
if (i == median)
ureg_MAX(shader, t_array[j], ureg_src(t_array[j]), ureg_src(t_array[j - 1]));
else
ureg_MIN(shader, t_array[j - 1], ureg_src(t_array[j]), ureg_src(t_array[j - 1]));
}
ureg_MOV(shader, o_fragment, ureg_src(t_array[median]));
ureg_END(shader);
FREE(t_array);
return ureg_create_shader_and_destroy(shader, filter->pipe);
}
static void *
create_fs(struct pipe_context *pipe, unsigned fs_traits)
{
struct ureg_program *ureg;
struct ureg_src /*dst_sampler, */ src_sampler, mask_sampler;
struct ureg_src /*dst_pos, */ src_input, mask_pos;
struct ureg_dst src, mask;
struct ureg_dst out;
struct ureg_src imm0 = { 0 };
unsigned has_mask = (fs_traits & FS_MASK) != 0;
unsigned is_fill = (fs_traits & FS_FILL) != 0;
unsigned is_composite = (fs_traits & FS_COMPOSITE) != 0;
unsigned is_solid = (fs_traits & FS_SOLID_FILL) != 0;
unsigned is_lingrad = (fs_traits & FS_LINGRAD_FILL) != 0;
unsigned is_radgrad = (fs_traits & FS_RADGRAD_FILL) != 0;
unsigned comp_alpha_mask = fs_traits & FS_COMPONENT_ALPHA;
unsigned is_yuv = (fs_traits & FS_YUV) != 0;
unsigned src_repeat_none = (fs_traits & FS_SRC_REPEAT_NONE) != 0;
unsigned mask_repeat_none = (fs_traits & FS_MASK_REPEAT_NONE) != 0;
unsigned src_swizzle = (fs_traits & FS_SRC_SWIZZLE_RGB) != 0;
unsigned mask_swizzle = (fs_traits & FS_MASK_SWIZZLE_RGB) != 0;
unsigned src_set_alpha = (fs_traits & FS_SRC_SET_ALPHA) != 0;
unsigned mask_set_alpha = (fs_traits & FS_MASK_SET_ALPHA) != 0;
unsigned src_luminance = (fs_traits & FS_SRC_LUMINANCE) != 0;
unsigned mask_luminance = (fs_traits & FS_MASK_LUMINANCE) != 0;
unsigned dst_luminance = (fs_traits & FS_DST_LUMINANCE) != 0;
#if 0
print_fs_traits(fs_traits);
#else
(void)print_fs_traits;
#endif
ureg = ureg_create(TGSI_PROCESSOR_FRAGMENT);
if (ureg == NULL)
return 0;
/* it has to be either a fill, a composite op or a yuv conversion */
debug_assert((is_fill ^ is_composite) ^ is_yuv);
(void)is_yuv;
out = ureg_DECL_output(ureg, TGSI_SEMANTIC_COLOR, 0);
if (src_repeat_none || mask_repeat_none ||
src_set_alpha || mask_set_alpha || src_luminance) {
imm0 = ureg_imm4f(ureg, 0, 0, 0, 1);
}
if (is_composite) {
src_sampler = ureg_DECL_sampler(ureg, 0);
src_input = ureg_DECL_fs_input(ureg,
TGSI_SEMANTIC_GENERIC, 0,
TGSI_INTERPOLATE_PERSPECTIVE);
} else if (is_fill) {
if (is_solid)
src_input = ureg_DECL_fs_input(ureg,
TGSI_SEMANTIC_COLOR, 0,
TGSI_INTERPOLATE_PERSPECTIVE);
else
src_input = ureg_DECL_fs_input(ureg,
TGSI_SEMANTIC_POSITION, 0,
TGSI_INTERPOLATE_PERSPECTIVE);
} else {
debug_assert(is_yuv);
return create_yuv_shader(pipe, ureg);
}
if (has_mask) {
mask_sampler = ureg_DECL_sampler(ureg, 1);
mask_pos = ureg_DECL_fs_input(ureg,
TGSI_SEMANTIC_GENERIC, 1,
TGSI_INTERPOLATE_PERSPECTIVE);
}
#if 0 /* unused right now */
dst_sampler = ureg_DECL_sampler(ureg, 2);
dst_pos = ureg_DECL_fs_input(ureg,
TGSI_SEMANTIC_POSITION, 2,
TGSI_INTERPOLATE_PERSPECTIVE);
#endif
if (is_composite) {
if (has_mask || src_luminance || dst_luminance)
src = ureg_DECL_temporary(ureg);
else
src = out;
xrender_tex(ureg, src, src_input, src_sampler, imm0,
src_repeat_none, src_swizzle, src_set_alpha);
} else if (is_fill) {
if (is_solid) {
if (has_mask || src_luminance || dst_luminance)
src = ureg_dst(src_input);
else
ureg_MOV(ureg, out, src_input);
} else if (is_lingrad || is_radgrad) {
struct ureg_src coords, const0124, matrow0, matrow1, matrow2;
if (has_mask || src_luminance || dst_luminance)
src = ureg_DECL_temporary(ureg);
else
src = out;
coords = ureg_DECL_constant(ureg, 0);
const0124 = ureg_DECL_constant(ureg, 1);
matrow0 = ureg_DECL_constant(ureg, 2);
matrow1 = ureg_DECL_constant(ureg, 3);
matrow2 = ureg_DECL_constant(ureg, 4);
if (is_lingrad) {
linear_gradient(ureg, src,
src_input, src_sampler,
coords, const0124, matrow0, matrow1, matrow2);
} else if (is_radgrad) {
radial_gradient(ureg, src,
src_input, src_sampler,
coords, const0124, matrow0, matrow1, matrow2);
}
} else
debug_assert(!"Unknown fill type!");
}
if (src_luminance) {
ureg_MOV(ureg, src, ureg_scalar(ureg_src(src), TGSI_SWIZZLE_X));
ureg_MOV(ureg, ureg_writemask(src, TGSI_WRITEMASK_XYZ),
ureg_scalar(imm0, TGSI_SWIZZLE_X));
if (!has_mask && !dst_luminance)
ureg_MOV(ureg, out, ureg_src(src));
}
if (has_mask) {
mask = ureg_DECL_temporary(ureg);
xrender_tex(ureg, mask, mask_pos, mask_sampler, imm0,
mask_repeat_none, mask_swizzle, mask_set_alpha);
/* src IN mask */
src_in_mask(ureg, (dst_luminance) ? src : out, ureg_src(src),
ureg_src(mask),
comp_alpha_mask, mask_luminance);
ureg_release_temporary(ureg, mask);
}
if (dst_luminance) {
/*
* Make sure the alpha channel goes into the output L8 surface.
*/
ureg_MOV(ureg, out, ureg_scalar(ureg_src(src), TGSI_SWIZZLE_W));
}
ureg_END(ureg);
return ureg_create_shader_and_destroy(ureg, pipe);
}
static void *
create_frag_shader(struct vl_zscan *zscan)
{
struct ureg_program *shader;
struct ureg_src *vtex;
struct ureg_src samp_src, samp_scan, samp_quant;
struct ureg_dst *tmp;
struct ureg_dst quant, fragment;
unsigned i;
shader = ureg_create(PIPE_SHADER_FRAGMENT);
if (!shader)
return NULL;
vtex = MALLOC(zscan->num_channels * sizeof(struct ureg_src));
tmp = MALLOC(zscan->num_channels * sizeof(struct ureg_dst));
for (i = 0; i < zscan->num_channels; ++i)
vtex[i] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_VTEX + i, TGSI_INTERPOLATE_LINEAR);
samp_src = ureg_DECL_sampler(shader, 0);
samp_scan = ureg_DECL_sampler(shader, 1);
samp_quant = ureg_DECL_sampler(shader, 2);
for (i = 0; i < zscan->num_channels; ++i)
tmp[i] = ureg_DECL_temporary(shader);
quant = ureg_DECL_temporary(shader);
fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
/*
* tmp.x = tex(vtex, 1)
* tmp.y = vtex.z
* fragment = tex(tmp, 0) * quant
*/
for (i = 0; i < zscan->num_channels; ++i)
ureg_TEX(shader, ureg_writemask(tmp[i], TGSI_WRITEMASK_X), TGSI_TEXTURE_2D, vtex[i], samp_scan);
for (i = 0; i < zscan->num_channels; ++i)
ureg_MOV(shader, ureg_writemask(tmp[i], TGSI_WRITEMASK_Y), ureg_scalar(vtex[i], TGSI_SWIZZLE_W));
for (i = 0; i < zscan->num_channels; ++i) {
ureg_TEX(shader, ureg_writemask(tmp[0], TGSI_WRITEMASK_X << i), TGSI_TEXTURE_2D, ureg_src(tmp[i]), samp_src);
ureg_TEX(shader, ureg_writemask(quant, TGSI_WRITEMASK_X << i), TGSI_TEXTURE_3D, vtex[i], samp_quant);
}
ureg_MUL(shader, quant, ureg_src(quant), ureg_imm1f(shader, 16.0f));
ureg_MUL(shader, fragment, ureg_src(tmp[0]), ureg_src(quant));
for (i = 0; i < zscan->num_channels; ++i)
ureg_release_temporary(shader, tmp[i]);
ureg_END(shader);
FREE(vtex);
FREE(tmp);
return ureg_create_shader_and_destroy(shader, zscan->pipe);
}
static void *
create_ycbcr_frag_shader(struct vl_mc *r, float scale, bool invert,
vl_mc_ycbcr_frag_shader fs_callback, void *callback_priv)
{
struct ureg_program *shader;
struct ureg_src flags;
struct ureg_dst tmp;
struct ureg_dst fragment;
unsigned label;
shader = ureg_create(TGSI_PROCESSOR_FRAGMENT);
if (!shader)
return NULL;
flags = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_FLAGS, TGSI_INTERPOLATE_LINEAR);
fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
tmp = calc_line(shader);
/*
* if (field == tc.w)
* kill();
* else {
* fragment.xyz = tex(tc, sampler) * scale + tc.z
* fragment.w = 1.0f
* }
*/
ureg_SEQ(shader, ureg_writemask(tmp, TGSI_WRITEMASK_Y),
ureg_scalar(flags, TGSI_SWIZZLE_W), ureg_src(tmp));
ureg_IF(shader, ureg_scalar(ureg_src(tmp), TGSI_SWIZZLE_Y), &label);
ureg_KILP(shader);
ureg_fixup_label(shader, label, ureg_get_instruction_number(shader));
ureg_ELSE(shader, &label);
fs_callback(callback_priv, r, shader, VS_O_VTEX, tmp);
if (scale != 1.0f)
ureg_MAD(shader, ureg_writemask(tmp, TGSI_WRITEMASK_XYZ),
ureg_src(tmp), ureg_imm1f(shader, scale),
ureg_scalar(flags, TGSI_SWIZZLE_Z));
else
ureg_ADD(shader, ureg_writemask(tmp, TGSI_WRITEMASK_XYZ),
ureg_src(tmp), ureg_scalar(flags, TGSI_SWIZZLE_Z));
ureg_MUL(shader, ureg_writemask(fragment, TGSI_WRITEMASK_XYZ), ureg_src(tmp), ureg_imm1f(shader, invert ? -1.0f : 1.0f));
ureg_MOV(shader, ureg_writemask(fragment, TGSI_WRITEMASK_W), ureg_imm1f(shader, 1.0f));
ureg_fixup_label(shader, label, ureg_get_instruction_number(shader));
ureg_ENDIF(shader);
ureg_release_temporary(shader, tmp);
ureg_END(shader);
return ureg_create_shader_and_destroy(shader, r->pipe);
}
static void *
create_vert_shader(struct vl_zscan *zscan)
{
struct ureg_program *shader;
struct ureg_src scale;
struct ureg_src vrect, vpos, block_num;
struct ureg_dst tmp;
struct ureg_dst o_vpos;
struct ureg_dst *o_vtex;
unsigned i;
shader = ureg_create(PIPE_SHADER_VERTEX);
if (!shader)
return NULL;
o_vtex = MALLOC(zscan->num_channels * sizeof(struct ureg_dst));
scale = ureg_imm2f(shader,
(float)VL_BLOCK_WIDTH / zscan->buffer_width,
(float)VL_BLOCK_HEIGHT / zscan->buffer_height);
vrect = ureg_DECL_vs_input(shader, VS_I_RECT);
vpos = ureg_DECL_vs_input(shader, VS_I_VPOS);
block_num = ureg_DECL_vs_input(shader, VS_I_BLOCK_NUM);
tmp = ureg_DECL_temporary(shader);
o_vpos = ureg_DECL_output(shader, TGSI_SEMANTIC_POSITION, VS_O_VPOS);
for (i = 0; i < zscan->num_channels; ++i)
o_vtex[i] = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_VTEX + i);
/*
* o_vpos.xy = (vpos + vrect) * scale
* o_vpos.zw = 1.0f
*
* tmp.xy = InstanceID / blocks_per_line
* tmp.x = frac(tmp.x)
* tmp.y = floor(tmp.y)
*
* o_vtex.x = vrect.x / blocks_per_line + tmp.x
* o_vtex.y = vrect.y
* o_vtex.z = tmp.z * blocks_per_line / blocks_total
*/
ureg_ADD(shader, ureg_writemask(tmp, TGSI_WRITEMASK_XY), vpos, vrect);
ureg_MUL(shader, ureg_writemask(o_vpos, TGSI_WRITEMASK_XY), ureg_src(tmp), scale);
ureg_MOV(shader, ureg_writemask(o_vpos, TGSI_WRITEMASK_ZW), ureg_imm1f(shader, 1.0f));
ureg_MUL(shader, ureg_writemask(tmp, TGSI_WRITEMASK_XW), ureg_scalar(block_num, TGSI_SWIZZLE_X),
ureg_imm1f(shader, 1.0f / zscan->blocks_per_line));
ureg_FRC(shader, ureg_writemask(tmp, TGSI_WRITEMASK_Y), ureg_scalar(ureg_src(tmp), TGSI_SWIZZLE_X));
ureg_FLR(shader, ureg_writemask(tmp, TGSI_WRITEMASK_W), ureg_src(tmp));
for (i = 0; i < zscan->num_channels; ++i) {
ureg_ADD(shader, ureg_writemask(tmp, TGSI_WRITEMASK_X), ureg_scalar(ureg_src(tmp), TGSI_SWIZZLE_Y),
ureg_imm1f(shader, 1.0f / (zscan->blocks_per_line * VL_BLOCK_WIDTH)
* (i - (signed)zscan->num_channels / 2)));
ureg_MAD(shader, ureg_writemask(o_vtex[i], TGSI_WRITEMASK_X), vrect,
ureg_imm1f(shader, 1.0f / zscan->blocks_per_line), ureg_src(tmp));
ureg_MOV(shader, ureg_writemask(o_vtex[i], TGSI_WRITEMASK_Y), vrect);
ureg_MOV(shader, ureg_writemask(o_vtex[i], TGSI_WRITEMASK_Z), vpos);
ureg_MUL(shader, ureg_writemask(o_vtex[i], TGSI_WRITEMASK_W), ureg_src(tmp),
ureg_imm1f(shader, (float)zscan->blocks_per_line / zscan->blocks_total));
}
ureg_release_temporary(shader, tmp);
ureg_END(shader);
FREE(o_vtex);
return ureg_create_shader_and_destroy(shader, zscan->pipe);
}
