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(TGSI_PROCESSOR_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);
}
/**
* Make a fragment shader that copies the input color to N output colors.
*/
void *
util_make_fragment_clonecolor_shader(struct pipe_context *pipe, int num_cbufs)
{
struct ureg_program *ureg;
struct ureg_src src;
struct ureg_dst dst[PIPE_MAX_COLOR_BUFS];
int i;
assert(num_cbufs <= PIPE_MAX_COLOR_BUFS);
ureg = ureg_create( TGSI_PROCESSOR_FRAGMENT );
if (ureg == NULL)
return NULL;
src = ureg_DECL_fs_input( ureg, TGSI_SEMANTIC_COLOR, 0,
TGSI_INTERPOLATE_PERSPECTIVE );
for (i = 0; i < num_cbufs; i++)
dst[i] = ureg_DECL_output( ureg, TGSI_SEMANTIC_COLOR, i );
for (i = 0; i < num_cbufs; i++)
ureg_MOV( ureg, dst[i], src );
ureg_END( ureg );
return ureg_create_shader_and_destroy( ureg, pipe );
}
/**
* Make a fragment shader that copies the input color to N output colors.
*/
void *
util_make_fragment_cloneinput_shader(struct pipe_context *pipe, int num_cbufs,
int input_semantic,
int input_interpolate)
{
struct ureg_program *ureg;
struct ureg_src src;
struct ureg_dst dst[PIPE_MAX_COLOR_BUFS];
int i;
assert(num_cbufs <= PIPE_MAX_COLOR_BUFS);
ureg = ureg_create( PIPE_SHADER_FRAGMENT );
if (!ureg)
return NULL;
src = ureg_DECL_fs_input( ureg, input_semantic, 0,
input_interpolate );
for (i = 0; i < num_cbufs; i++)
dst[i] = ureg_DECL_output( ureg, TGSI_SEMANTIC_COLOR, i );
for (i = 0; i < num_cbufs; i++)
ureg_MOV( ureg, dst[i], src );
ureg_END( ureg );
return ureg_create_shader_and_destroy( ureg, pipe );
}
static struct ureg_dst
calc_position(struct vl_mc *r, struct ureg_program *shader, struct ureg_src block_scale)
{
struct ureg_src vrect, vpos;
struct ureg_dst t_vpos;
struct ureg_dst o_vpos;
vrect = ureg_DECL_vs_input(shader, VS_I_RECT);
vpos = ureg_DECL_vs_input(shader, VS_I_VPOS);
t_vpos = ureg_DECL_temporary(shader);
o_vpos = ureg_DECL_output(shader, TGSI_SEMANTIC_POSITION, VS_O_VPOS);
/*
* block_scale = (VL_MACROBLOCK_WIDTH, VL_MACROBLOCK_HEIGHT) / (dst.width, dst.height)
*
* t_vpos = (vpos + vrect) * block_scale
* o_vpos.xy = t_vpos
* o_vpos.zw = vpos
*/
ureg_ADD(shader, ureg_writemask(t_vpos, TGSI_WRITEMASK_XY), vpos, vrect);
ureg_MUL(shader, ureg_writemask(t_vpos, TGSI_WRITEMASK_XY), ureg_src(t_vpos), block_scale);
ureg_MOV(shader, ureg_writemask(o_vpos, TGSI_WRITEMASK_XY), ureg_src(t_vpos));
ureg_MOV(shader, ureg_writemask(o_vpos, TGSI_WRITEMASK_ZW), ureg_imm1f(shader, 1.0f));
return t_vpos;
}
void *
util_make_vertex_passthrough_shader_with_so(struct pipe_context *pipe,
uint num_attribs,
const uint *semantic_names,
const uint *semantic_indexes,
bool window_space,
const struct pipe_stream_output_info *so)
{
struct ureg_program *ureg;
uint i;
ureg = ureg_create( PIPE_SHADER_VERTEX );
if (!ureg)
return NULL;
if (window_space)
ureg_property(ureg, TGSI_PROPERTY_VS_WINDOW_SPACE_POSITION, TRUE);
for (i = 0; i < num_attribs; i++) {
struct ureg_src src;
struct ureg_dst dst;
src = ureg_DECL_vs_input( ureg, i );
dst = ureg_DECL_output( ureg,
semantic_names[i],
semantic_indexes[i]);
ureg_MOV( ureg, dst, src );
}
ureg_END( ureg );
return ureg_create_shader_with_so_and_destroy( ureg, pipe, so );
}
void *
util_make_vertex_passthrough_shader_with_so(struct pipe_context *pipe,
uint num_attribs,
const uint *semantic_names,
const uint *semantic_indexes,
const struct pipe_stream_output_info *so)
{
struct ureg_program *ureg;
uint i;
ureg = ureg_create( TGSI_PROCESSOR_VERTEX );
if (ureg == NULL)
return NULL;
for (i = 0; i < num_attribs; i++) {
struct ureg_src src;
struct ureg_dst dst;
src = ureg_DECL_vs_input( ureg, i );
dst = ureg_DECL_output( ureg,
semantic_names[i],
semantic_indexes[i]);
ureg_MOV( ureg, dst, src );
}
ureg_END( ureg );
return ureg_create_shader_with_so_and_destroy( ureg, pipe, so );
}
static void *
create_frag_shader_rgba(struct vl_compositor *c)
{
struct ureg_program *shader;
struct ureg_src tc, color, sampler;
struct ureg_dst texel, fragment;
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);
color = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_COLOR, VS_O_COLOR, TGSI_INTERPOLATE_LINEAR);
sampler = ureg_DECL_sampler(shader, 0);
texel = ureg_DECL_temporary(shader);
fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
/*
* fragment = tex(tc, sampler)
*/
ureg_TEX(shader, texel, TGSI_TEXTURE_2D, tc, sampler);
ureg_MUL(shader, fragment, ureg_src(texel), color);
ureg_END(shader);
return ureg_create_shader_and_destroy(shader, c->pipe);
}
/**
* If we fail to compile a fragment shader (because it uses too many
* registers, for example) we'll use a dummy/fallback shader that
* simply emits a constant color (red for debug, black for release).
* We hit this with the Unigine/Heaven demo when Shaders = High.
* With black, the demo still looks good.
*/
static const struct tgsi_token *
get_dummy_fragment_shader(void)
{
#ifdef DEBUG
static const float color[4] = { 1.0, 0.0, 0.0, 0.0 }; /* red */
#else
static const float color[4] = { 0.0, 0.0, 0.0, 0.0 }; /* black */
#endif
struct ureg_program *ureg;
const struct tgsi_token *tokens;
struct ureg_src src;
struct ureg_dst dst;
unsigned num_tokens;
ureg = ureg_create(TGSI_PROCESSOR_FRAGMENT);
if (!ureg)
return NULL;
dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_COLOR, 0);
src = ureg_DECL_immediate(ureg, color, 4);
ureg_MOV(ureg, dst, src);
ureg_END(ureg);
tokens = ureg_get_tokens(ureg, &num_tokens);
ureg_destroy(ureg);
return tokens;
}
/**
* If we fail to compile a vertex shader we'll use a dummy/fallback shader
* that simply emits a (0,0,0,1) vertex position.
*/
static const struct tgsi_token *
get_dummy_vertex_shader(void)
{
static const float zero[4] = { 0.0, 0.0, 0.0, 1.0 };
struct ureg_program *ureg;
const struct tgsi_token *tokens;
struct ureg_src src;
struct ureg_dst dst;
unsigned num_tokens;
ureg = ureg_create(TGSI_PROCESSOR_VERTEX);
if (!ureg)
return NULL;
dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_POSITION, 0);
src = ureg_DECL_immediate(ureg, zero, 4);
ureg_MOV(ureg, dst, src);
ureg_END(ureg);
tokens = ureg_get_tokens(ureg, &num_tokens);
ureg_destroy(ureg);
return tokens;
}
/**
* Make a simple fragment texture shader which reads a texture and writes it
* as stencil.
*/
void *
util_make_fragment_tex_shader_writestencil(struct pipe_context *pipe,
unsigned tex_target,
unsigned interp_mode,
bool load_level_zero,
bool use_txf)
{
struct ureg_program *ureg;
struct ureg_src stencil_sampler;
struct ureg_src tex;
struct ureg_dst out, stencil;
struct ureg_src imm;
ureg = ureg_create( PIPE_SHADER_FRAGMENT );
if (!ureg)
return NULL;
stencil_sampler = ureg_DECL_sampler( ureg, 0 );
ureg_DECL_sampler_view(ureg, 0, tex_target,
TGSI_RETURN_TYPE_UINT,
TGSI_RETURN_TYPE_UINT,
TGSI_RETURN_TYPE_UINT,
TGSI_RETURN_TYPE_UINT);
tex = ureg_DECL_fs_input( ureg,
TGSI_SEMANTIC_GENERIC, 0,
interp_mode );
out = ureg_DECL_output( ureg,
TGSI_SEMANTIC_COLOR,
0 );
stencil = ureg_DECL_output( ureg,
TGSI_SEMANTIC_STENCIL,
0 );
imm = ureg_imm4f( ureg, 0, 0, 0, 1 );
ureg_MOV( ureg, out, imm );
ureg_load_tex(ureg, ureg_writemask(stencil, TGSI_WRITEMASK_Y), tex,
stencil_sampler, tex_target, load_level_zero, use_txf);
ureg_END( ureg );
return ureg_create_shader_and_destroy( ureg, pipe );
}
void *
util_make_fs_msaa_resolve(struct pipe_context *pipe,
enum tgsi_texture_type tgsi_tex, unsigned nr_samples,
enum tgsi_return_type stype)
{
struct ureg_program *ureg;
struct ureg_src sampler, coord;
struct ureg_dst out, tmp_sum, tmp_coord, tmp;
unsigned i;
ureg = ureg_create(PIPE_SHADER_FRAGMENT);
if (!ureg)
return NULL;
/* Declarations. */
sampler = ureg_DECL_sampler(ureg, 0);
ureg_DECL_sampler_view(ureg, 0, tgsi_tex, stype, stype, stype, stype);
coord = ureg_DECL_fs_input(ureg, TGSI_SEMANTIC_GENERIC, 0,
TGSI_INTERPOLATE_LINEAR);
out = ureg_DECL_output(ureg, TGSI_SEMANTIC_COLOR, 0);
tmp_sum = ureg_DECL_temporary(ureg);
tmp_coord = ureg_DECL_temporary(ureg);
tmp = ureg_DECL_temporary(ureg);
/* Instructions. */
ureg_MOV(ureg, tmp_sum, ureg_imm1f(ureg, 0));
ureg_F2U(ureg, tmp_coord, coord);
for (i = 0; i < nr_samples; i++) {
/* Read one sample. */
ureg_MOV(ureg, ureg_writemask(tmp_coord, TGSI_WRITEMASK_W),
ureg_imm1u(ureg, i));
ureg_TXF(ureg, tmp, tgsi_tex, ureg_src(tmp_coord), sampler);
if (stype == TGSI_RETURN_TYPE_UINT)
ureg_U2F(ureg, tmp, ureg_src(tmp));
else if (stype == TGSI_RETURN_TYPE_SINT)
ureg_I2F(ureg, tmp, ureg_src(tmp));
/* Add it to the sum.*/
ureg_ADD(ureg, tmp_sum, ureg_src(tmp_sum), ureg_src(tmp));
}
/* Calculate the average and return. */
ureg_MUL(ureg, tmp_sum, ureg_src(tmp_sum),
ureg_imm1f(ureg, 1.0 / nr_samples));
if (stype == TGSI_RETURN_TYPE_UINT)
ureg_F2U(ureg, out, ureg_src(tmp_sum));
else if (stype == TGSI_RETURN_TYPE_SINT)
ureg_F2I(ureg, out, ureg_src(tmp_sum));
else
ureg_MOV(ureg, out, ureg_src(tmp_sum));
ureg_END(ureg);
return ureg_create_shader_and_destroy(ureg, pipe);
}
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);
}
void *
st_pbo_create_gs(struct st_context *st)
{
static const int zero = 0;
struct ureg_program *ureg;
struct ureg_dst out_pos;
struct ureg_dst out_layer;
struct ureg_src in_pos;
struct ureg_src imm;
unsigned i;
ureg = ureg_create(PIPE_SHADER_GEOMETRY);
if (!ureg)
return NULL;
ureg_property(ureg, TGSI_PROPERTY_GS_INPUT_PRIM, PIPE_PRIM_TRIANGLES);
ureg_property(ureg, TGSI_PROPERTY_GS_OUTPUT_PRIM, PIPE_PRIM_TRIANGLE_STRIP);
ureg_property(ureg, TGSI_PROPERTY_GS_MAX_OUTPUT_VERTICES, 3);
out_pos = ureg_DECL_output(ureg, TGSI_SEMANTIC_POSITION, 0);
out_layer = ureg_DECL_output(ureg, TGSI_SEMANTIC_LAYER, 0);
in_pos = ureg_DECL_input(ureg, TGSI_SEMANTIC_POSITION, 0, 0, 1);
imm = ureg_DECL_immediate_int(ureg, &zero, 1);
for (i = 0; i < 3; ++i) {
struct ureg_src in_pos_vertex = ureg_src_dimension(in_pos, i);
/* out_pos = in_pos[i] */
ureg_MOV(ureg, out_pos, in_pos_vertex);
/* out_layer.x = f2i(in_pos[i].z) */
ureg_F2I(ureg, ureg_writemask(out_layer, TGSI_WRITEMASK_X),
ureg_scalar(in_pos_vertex, TGSI_SWIZZLE_Z));
ureg_EMIT(ureg, ureg_scalar(imm, TGSI_SWIZZLE_X));
}
ureg_END(ureg);
return ureg_create_shader_and_destroy(ureg, st->pipe);
}
void *
st_pbo_create_vs(struct st_context *st)
{
struct ureg_program *ureg;
struct ureg_src in_pos;
struct ureg_src in_instanceid;
struct ureg_dst out_pos;
struct ureg_dst out_layer;
ureg = ureg_create(PIPE_SHADER_VERTEX);
if (!ureg)
return NULL;
in_pos = ureg_DECL_vs_input(ureg, TGSI_SEMANTIC_POSITION);
out_pos = ureg_DECL_output(ureg, TGSI_SEMANTIC_POSITION, 0);
if (st->pbo.layers) {
in_instanceid = ureg_DECL_system_value(ureg, TGSI_SEMANTIC_INSTANCEID, 0);
if (!st->pbo.use_gs)
out_layer = ureg_DECL_output(ureg, TGSI_SEMANTIC_LAYER, 0);
}
/* out_pos = in_pos */
ureg_MOV(ureg, out_pos, in_pos);
if (st->pbo.layers) {
if (st->pbo.use_gs) {
/* out_pos.z = i2f(gl_InstanceID) */
ureg_I2F(ureg, ureg_writemask(out_pos, TGSI_WRITEMASK_Z),
ureg_scalar(in_instanceid, TGSI_SWIZZLE_X));
} else {
/* out_layer = gl_InstanceID */
ureg_MOV(ureg, out_layer, in_instanceid);
}
}
ureg_END(ureg);
return ureg_create_shader_and_destroy(ureg, st->pipe);
}
static void *
create_frag_shader(struct vl_matrix_filter *filter, unsigned num_offsets,
struct vertex2f *offsets, const float *matrix_values)
{
struct ureg_program *shader;
struct ureg_src i_vtex;
struct ureg_src sampler;
struct ureg_dst tmp;
struct ureg_dst t_sum;
struct ureg_dst o_fragment;
unsigned i;
shader = ureg_create(PIPE_SHADER_FRAGMENT);
if (!shader) {
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);
tmp = ureg_DECL_temporary(shader);
t_sum = ureg_DECL_temporary(shader);
o_fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
ureg_MOV(shader, t_sum, ureg_imm1f(shader, 0.0f));
for (i = 0; i < num_offsets; ++i) {
if (matrix_values[i] == 0.0f)
continue;
if (!is_vec_zero(offsets[i])) {
ureg_ADD(shader, ureg_writemask(tmp, TGSI_WRITEMASK_XY),
i_vtex, ureg_imm2f(shader, offsets[i].x, offsets[i].y));
ureg_MOV(shader, ureg_writemask(tmp, TGSI_WRITEMASK_ZW),
ureg_imm1f(shader, 0.0f));
ureg_TEX(shader, tmp, TGSI_TEXTURE_2D, ureg_src(tmp), sampler);
} else {
ureg_TEX(shader, tmp, TGSI_TEXTURE_2D, i_vtex, sampler);
}
ureg_MAD(shader, t_sum, ureg_src(tmp), ureg_imm1f(shader, matrix_values[i]),
ureg_src(t_sum));
}
ureg_MOV(shader, o_fragment, ureg_src(t_sum));
ureg_END(shader);
return ureg_create_shader_and_destroy(shader, filter->pipe);
}
/**
* Make a simple fragment texture shader which reads a texture and writes it
* as stencil.
*/
void *
util_make_fragment_tex_shader_writestencil(struct pipe_context *pipe,
unsigned tex_target,
unsigned interp_mode)
{
struct ureg_program *ureg;
struct ureg_src stencil_sampler;
struct ureg_src tex;
struct ureg_dst out, stencil;
struct ureg_src imm;
ureg = ureg_create( TGSI_PROCESSOR_FRAGMENT );
if (ureg == NULL)
return NULL;
stencil_sampler = ureg_DECL_sampler( ureg, 0 );
tex = ureg_DECL_fs_input( ureg,
TGSI_SEMANTIC_GENERIC, 0,
interp_mode );
out = ureg_DECL_output( ureg,
TGSI_SEMANTIC_COLOR,
0 );
stencil = ureg_DECL_output( ureg,
TGSI_SEMANTIC_STENCIL,
0 );
imm = ureg_imm4f( ureg, 0, 0, 0, 1 );
ureg_MOV( ureg, out, imm );
ureg_TEX( ureg,
ureg_writemask(stencil, TGSI_WRITEMASK_Y),
tex_target, tex, stencil_sampler );
ureg_END( ureg );
return ureg_create_shader_and_destroy( ureg, pipe );
}
static void *
create_vert_shader(struct vl_deint_filter *filter)
{
struct ureg_program *shader;
struct ureg_src i_vpos;
struct ureg_dst o_vpos, o_vtex;
shader = ureg_create(PIPE_SHADER_VERTEX);
if (!shader)
return NULL;
i_vpos = ureg_DECL_vs_input(shader, 0);
o_vpos = ureg_DECL_output(shader, TGSI_SEMANTIC_POSITION, VS_O_VPOS);
o_vtex = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_VTEX);
ureg_MOV(shader, o_vpos, i_vpos);
ureg_MOV(shader, o_vtex, i_vpos);
ureg_END(shader);
return ureg_create_shader_and_destroy(shader, filter->pipe);
}
/**
* This is used when TCS is NULL in the VS->TCS->TES chain. In this case,
* VS passes its outputs to TES directly, so the fixed-function shader only
* has to write TESSOUTER and TESSINNER.
*/
void *si_create_fixed_func_tcs(struct si_context *sctx)
{
struct ureg_src outer, inner;
struct ureg_dst tessouter, tessinner;
struct ureg_program *ureg = ureg_create(PIPE_SHADER_TESS_CTRL);
if (!ureg)
return NULL;
outer = ureg_DECL_system_value(ureg,
TGSI_SEMANTIC_DEFAULT_TESSOUTER_SI, 0);
inner = ureg_DECL_system_value(ureg,
TGSI_SEMANTIC_DEFAULT_TESSINNER_SI, 0);
tessouter = ureg_DECL_output(ureg, TGSI_SEMANTIC_TESSOUTER, 0);
tessinner = ureg_DECL_output(ureg, TGSI_SEMANTIC_TESSINNER, 0);
ureg_MOV(ureg, tessouter, outer);
ureg_MOV(ureg, tessinner, inner);
ureg_END(ureg);
return ureg_create_shader_and_destroy(ureg, &sctx->b);
}
void
vl_idct_stage2_vert_shader(struct vl_idct *idct, struct ureg_program *shader,
unsigned first_output, struct ureg_dst tex)
{
struct ureg_src vrect, vpos;
struct ureg_src scale;
struct ureg_dst t_start;
struct ureg_dst o_l_addr[2], o_r_addr[2];
vrect = ureg_DECL_vs_input(shader, VS_I_RECT);
vpos = ureg_DECL_vs_input(shader, VS_I_VPOS);
t_start = ureg_DECL_temporary(shader);
--first_output;
o_l_addr[0] = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, first_output + VS_O_L_ADDR0);
o_l_addr[1] = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, first_output + VS_O_L_ADDR1);
o_r_addr[0] = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, first_output + VS_O_R_ADDR0);
o_r_addr[1] = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, first_output + VS_O_R_ADDR1);
scale = ureg_imm2f(shader,
(float)VL_BLOCK_WIDTH / idct->buffer_width,
(float)VL_BLOCK_HEIGHT / idct->buffer_height);
ureg_MUL(shader, ureg_writemask(tex, TGSI_WRITEMASK_Z),
ureg_scalar(vrect, TGSI_SWIZZLE_X),
ureg_imm1f(shader, VL_BLOCK_WIDTH / idct->nr_of_render_targets));
ureg_MUL(shader, ureg_writemask(t_start, TGSI_WRITEMASK_XY), vpos, scale);
calc_addr(shader, o_l_addr, vrect, ureg_imm1f(shader, 0.0f), false, false, VL_BLOCK_WIDTH / 4);
calc_addr(shader, o_r_addr, ureg_src(tex), ureg_src(t_start), true, false, idct->buffer_height / 4);
ureg_MOV(shader, ureg_writemask(o_r_addr[0], TGSI_WRITEMASK_Z), ureg_src(tex));
ureg_MOV(shader, ureg_writemask(o_r_addr[1], TGSI_WRITEMASK_Z), ureg_src(tex));
}
void *
util_make_geometry_passthrough_shader(struct pipe_context *pipe,
uint num_attribs,
const ubyte *semantic_names,
const ubyte *semantic_indexes)
{
static const unsigned zero[4] = {0, 0, 0, 0};
struct ureg_program *ureg;
struct ureg_dst dst[PIPE_MAX_SHADER_OUTPUTS];
struct ureg_src src[PIPE_MAX_SHADER_INPUTS];
struct ureg_src imm;
unsigned i;
ureg = ureg_create(PIPE_SHADER_GEOMETRY);
if (!ureg)
return NULL;
ureg_property(ureg, TGSI_PROPERTY_GS_INPUT_PRIM, PIPE_PRIM_POINTS);
ureg_property(ureg, TGSI_PROPERTY_GS_OUTPUT_PRIM, PIPE_PRIM_POINTS);
ureg_property(ureg, TGSI_PROPERTY_GS_MAX_OUTPUT_VERTICES, 1);
ureg_property(ureg, TGSI_PROPERTY_GS_INVOCATIONS, 1);
imm = ureg_DECL_immediate_uint(ureg, zero, 4);
/**
* Loop over all the attribs and declare the corresponding
* declarations in the geometry shader
*/
for (i = 0; i < num_attribs; i++) {
src[i] = ureg_DECL_input(ureg, semantic_names[i],
semantic_indexes[i], 0, 1);
src[i] = ureg_src_dimension(src[i], 0);
dst[i] = ureg_DECL_output(ureg, semantic_names[i], semantic_indexes[i]);
}
/* MOV dst[i] src[i] */
for (i = 0; i < num_attribs; i++) {
ureg_MOV(ureg, dst[i], src[i]);
}
/* EMIT IMM[0] */
ureg_insn(ureg, TGSI_OPCODE_EMIT, NULL, 0, &imm, 1, 0);
/* END */
ureg_END(ureg);
return ureg_create_shader_and_destroy(ureg, pipe);
}
/**
* Create a simple fragment shader that sets the color to white.
*/
static void *create_white_fs(struct pipe_context *pipe)
{
struct ureg_program *ureg;
struct ureg_dst out;
struct ureg_src imm;
ureg = ureg_create(TGSI_PROCESSOR_FRAGMENT);
out = ureg_DECL_output(ureg, TGSI_SEMANTIC_COLOR, 0);
imm = ureg_imm4f(ureg, 1.0f, 1.0f, 1.0f, 1.0f);
ureg_MOV(ureg, out, imm);
ureg_END(ureg);
return ureg_create_shader_and_destroy(ureg, pipe);
}
/**
* Create a simple fragment shader that sets the depth to 0.0f.
*/
static void *create_scissor_fs(struct pipe_context *pipe)
{
struct ureg_program *ureg;
struct ureg_dst out;
struct ureg_src imm;
ureg = ureg_create(TGSI_PROCESSOR_FRAGMENT);
out = ureg_DECL_output(ureg, TGSI_SEMANTIC_POSITION, 0);
imm = ureg_imm4f(ureg, 0.0f, 0.0f, 0.0f, 0.0f);
ureg_MOV(ureg, ureg_writemask(out, TGSI_WRITEMASK_Z), imm);
ureg_END(ureg);
return ureg_create_shader_and_destroy(ureg, pipe);
}
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_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
mc_vert_shader_callback(void *priv, struct vl_mc *mc,
struct ureg_program *shader,
unsigned first_output,
struct ureg_dst tex)
{
struct vl_mpeg12_decoder *dec = priv;
struct ureg_dst o_vtex;
assert(priv && mc);
assert(shader);
if (dec->base.entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT) {
struct vl_idct *idct = mc == &dec->mc_y ? &dec->idct_y : &dec->idct_c;
vl_idct_stage2_vert_shader(idct, shader, first_output, tex);
} else {
o_vtex = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, first_output);
ureg_MOV(shader, ureg_writemask(o_vtex, TGSI_WRITEMASK_XY), ureg_src(tex));
}
}
/**
* Make simple fragment texture shader:
* IMM {0,0,0,1} // (if writemask != 0xf)
* MOV OUT[0], IMM[0] // (if writemask != 0xf)
* TEX OUT[0].writemask, IN[0], SAMP[0], 2D;
* END;
*
* \param tex_target one of PIPE_TEXTURE_x
* \parma interp_mode either TGSI_INTERPOLATE_LINEAR or PERSPECTIVE
* \param writemask mask of TGSI_WRITEMASK_x
*/
void *
util_make_fragment_tex_shader_writemask(struct pipe_context *pipe,
unsigned tex_target,
unsigned interp_mode,
unsigned writemask )
{
struct ureg_program *ureg;
struct ureg_src sampler;
struct ureg_src tex;
struct ureg_dst out;
assert(interp_mode == TGSI_INTERPOLATE_LINEAR ||
interp_mode == TGSI_INTERPOLATE_PERSPECTIVE);
ureg = ureg_create( TGSI_PROCESSOR_FRAGMENT );
if (ureg == NULL)
return NULL;
sampler = ureg_DECL_sampler( ureg, 0 );
tex = ureg_DECL_fs_input( ureg,
TGSI_SEMANTIC_GENERIC, 0,
interp_mode );
out = ureg_DECL_output( ureg,
TGSI_SEMANTIC_COLOR,
0 );
if (writemask != TGSI_WRITEMASK_XYZW) {
struct ureg_src imm = ureg_imm4f( ureg, 0, 0, 0, 1 );
ureg_MOV( ureg, out, imm );
}
ureg_TEX( ureg,
ureg_writemask(out, writemask),
tex_target, tex, sampler );
ureg_END( ureg );
return ureg_create_shader_and_destroy( ureg, pipe );
}
static void r300_dummy_fragment_shader(
struct r300_context* r300,
struct r300_fragment_shader_code* shader)
{
struct pipe_shader_state state;
struct ureg_program *ureg;
struct ureg_dst out;
struct ureg_src imm;
/* Make a simple fragment shader which outputs (0, 0, 0, 1) */
ureg = ureg_create(TGSI_PROCESSOR_FRAGMENT);
out = ureg_DECL_output(ureg, TGSI_SEMANTIC_COLOR, 0);
imm = ureg_imm4f(ureg, 0, 0, 0, 1);
ureg_MOV(ureg, out, imm);
ureg_END(ureg);
state.tokens = ureg_finalize(ureg);
shader->dummy = TRUE;
r300_translate_fragment_shader(r300, shader, state.tokens);
ureg_destroy(ureg);
}
static void r300_dummy_vertex_shader(
struct r300_context* r300,
struct r300_vertex_shader* shader)
{
struct ureg_program *ureg;
struct ureg_dst dst;
struct ureg_src imm;
/* Make a simple vertex shader which outputs (0, 0, 0, 1),
* effectively rendering nothing. */
ureg = ureg_create(TGSI_PROCESSOR_VERTEX);
dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_POSITION, 0);
imm = ureg_imm4f(ureg, 0, 0, 0, 1);
ureg_MOV(ureg, dst, imm);
ureg_END(ureg);
shader->state.tokens = tgsi_dup_tokens(ureg_finalize(ureg));
ureg_destroy(ureg);
shader->dummy = TRUE;
r300_init_vs_outputs(r300, shader);
r300_translate_vertex_shader(r300, shader);
}
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);
}
/**
* Translate Mesa program to TGSI format.
* \param program the program to translate
* \param numInputs number of input registers used
* \param inputMapping maps Mesa fragment program inputs to TGSI generic
* input indexes
* \param inputSemanticName the TGSI_SEMANTIC flag for each input
* \param inputSemanticIndex the semantic index (ex: which texcoord) for
* each input
* \param interpMode the TGSI_INTERPOLATE_LINEAR/PERSP mode for each input
* \param numOutputs number of output registers used
* \param outputMapping maps Mesa fragment program outputs to TGSI
* generic outputs
* \param outputSemanticName the TGSI_SEMANTIC flag for each output
* \param outputSemanticIndex the semantic index (ex: which texcoord) for
* each output
*
* \return PIPE_OK or PIPE_ERROR_OUT_OF_MEMORY
*/
enum pipe_error
st_translate_mesa_program(
struct gl_context *ctx,
uint procType,
struct ureg_program *ureg,
const struct gl_program *program,
GLuint numInputs,
const GLuint inputMapping[],
const ubyte inputSemanticName[],
const ubyte inputSemanticIndex[],
const GLuint interpMode[],
GLuint numOutputs,
const GLuint outputMapping[],
const ubyte outputSemanticName[],
const ubyte outputSemanticIndex[],
boolean passthrough_edgeflags,
boolean clamp_color)
{
struct st_translate translate, *t;
unsigned i;
enum pipe_error ret = PIPE_OK;
assert(numInputs <= ARRAY_SIZE(t->inputs));
assert(numOutputs <= ARRAY_SIZE(t->outputs));
t = &translate;
memset(t, 0, sizeof *t);
t->procType = procType;
t->inputMapping = inputMapping;
t->outputMapping = outputMapping;
t->ureg = ureg;
/*_mesa_print_program(program);*/
/*
* Declare input attributes.
*/
if (procType == TGSI_PROCESSOR_FRAGMENT) {
for (i = 0; i < numInputs; i++) {
t->inputs[i] = ureg_DECL_fs_input(ureg,
inputSemanticName[i],
inputSemanticIndex[i],
interpMode[i]);
}
if (program->InputsRead & VARYING_BIT_POS) {
/* Must do this after setting up t->inputs, and before
* emitting constant references, below:
*/
emit_wpos(st_context(ctx), t, program, ureg);
}
if (program->InputsRead & VARYING_BIT_FACE) {
emit_face_var( t, program );
}
/*
* Declare output attributes.
*/
for (i = 0; i < numOutputs; i++) {
switch (outputSemanticName[i]) {
case TGSI_SEMANTIC_POSITION:
t->outputs[i] = ureg_DECL_output( ureg,
TGSI_SEMANTIC_POSITION, /* Z / Depth */
outputSemanticIndex[i] );
t->outputs[i] = ureg_writemask( t->outputs[i],
TGSI_WRITEMASK_Z );
break;
case TGSI_SEMANTIC_STENCIL:
t->outputs[i] = ureg_DECL_output( ureg,
TGSI_SEMANTIC_STENCIL, /* Stencil */
outputSemanticIndex[i] );
t->outputs[i] = ureg_writemask( t->outputs[i],
TGSI_WRITEMASK_Y );
break;
case TGSI_SEMANTIC_COLOR:
t->outputs[i] = ureg_DECL_output( ureg,
TGSI_SEMANTIC_COLOR,
outputSemanticIndex[i] );
break;
default:
debug_assert(0);
return 0;
}
}
}
else if (procType == TGSI_PROCESSOR_GEOMETRY) {
for (i = 0; i < numInputs; i++) {
t->inputs[i] = ureg_DECL_input(ureg,
inputSemanticName[i],
inputSemanticIndex[i], 0, 1);
}
for (i = 0; i < numOutputs; i++) {
t->outputs[i] = ureg_DECL_output( ureg,
outputSemanticName[i],
outputSemanticIndex[i] );
}
}
else {
assert(procType == TGSI_PROCESSOR_VERTEX);
for (i = 0; i < numInputs; i++) {
t->inputs[i] = ureg_DECL_vs_input(ureg, i);
}
for (i = 0; i < numOutputs; i++) {
t->outputs[i] = ureg_DECL_output( ureg,
outputSemanticName[i],
outputSemanticIndex[i] );
if (outputSemanticName[i] == TGSI_SEMANTIC_FOG) {
/* force register to contain a fog coordinate in the form (F, 0, 0, 1). */
ureg_MOV(ureg,
ureg_writemask(t->outputs[i], TGSI_WRITEMASK_YZW),
ureg_imm4f(ureg, 0.0f, 0.0f, 0.0f, 1.0f));
t->outputs[i] = ureg_writemask(t->outputs[i], TGSI_WRITEMASK_X);
}
}
if (passthrough_edgeflags)
emit_edgeflags( t, program );
}
/* Declare address register.
*/
if (program->NumAddressRegs > 0) {
debug_assert( program->NumAddressRegs == 1 );
t->address[0] = ureg_DECL_address( ureg );
}
/* Declare misc input registers
*/
{
GLbitfield sysInputs = program->SystemValuesRead;
unsigned numSys = 0;
for (i = 0; sysInputs; i++) {
if (sysInputs & (1 << i)) {
unsigned semName = _mesa_sysval_to_semantic[i];
t->systemValues[i] = ureg_DECL_system_value(ureg, numSys, semName, 0);
if (semName == TGSI_SEMANTIC_INSTANCEID ||
semName == TGSI_SEMANTIC_VERTEXID) {
/* From Gallium perspective, these system values are always
* integer, and require native integer support. However, if
* native integer is supported on the vertex stage but not the
* pixel stage (e.g, i915g + draw), Mesa will generate IR that
* assumes these system values are floats. To resolve the
* inconsistency, we insert a U2F.
*/
struct st_context *st = st_context(ctx);
struct pipe_screen *pscreen = st->pipe->screen;
assert(procType == TGSI_PROCESSOR_VERTEX);
assert(pscreen->get_shader_param(pscreen, PIPE_SHADER_VERTEX, PIPE_SHADER_CAP_INTEGERS));
(void) pscreen; /* silence non-debug build warnings */
if (!ctx->Const.NativeIntegers) {
struct ureg_dst temp = ureg_DECL_local_temporary(t->ureg);
ureg_U2F( t->ureg, ureg_writemask(temp, TGSI_WRITEMASK_X), t->systemValues[i]);
t->systemValues[i] = ureg_scalar(ureg_src(temp), 0);
}
}
numSys++;
sysInputs &= ~(1 << i);
}
}
}
if (program->IndirectRegisterFiles & (1 << PROGRAM_TEMPORARY)) {
/* If temps are accessed with indirect addressing, declare temporaries
* in sequential order. Else, we declare them on demand elsewhere.
*/
for (i = 0; i < program->NumTemporaries; i++) {
/* XXX use TGSI_FILE_TEMPORARY_ARRAY when it's supported by ureg */
t->temps[i] = ureg_DECL_temporary( t->ureg );
}
}
/* Emit constants and immediates. Mesa uses a single index space
* for these, so we put all the translated regs in t->constants.
*/
if (program->Parameters) {
t->constants = calloc( program->Parameters->NumParameters,
sizeof t->constants[0] );
if (t->constants == NULL) {
ret = PIPE_ERROR_OUT_OF_MEMORY;
goto out;
}
for (i = 0; i < program->Parameters->NumParameters; i++) {
switch (program->Parameters->Parameters[i].Type) {
case PROGRAM_STATE_VAR:
case PROGRAM_UNIFORM:
t->constants[i] = ureg_DECL_constant( ureg, i );
break;
/* Emit immediates only when there's no indirect addressing of
* the const buffer.
* FIXME: Be smarter and recognize param arrays:
* indirect addressing is only valid within the referenced
* array.
*/
case PROGRAM_CONSTANT:
if (program->IndirectRegisterFiles & PROGRAM_ANY_CONST)
t->constants[i] = ureg_DECL_constant( ureg, i );
else
t->constants[i] =
ureg_DECL_immediate( ureg,
(const float*) program->Parameters->ParameterValues[i],
4 );
break;
default:
break;
}
}
}
/* texture samplers */
for (i = 0; i < ctx->Const.Program[MESA_SHADER_FRAGMENT].MaxTextureImageUnits; i++) {
if (program->SamplersUsed & (1 << i)) {
t->samplers[i] = ureg_DECL_sampler( ureg, i );
}
}
/* Emit each instruction in turn:
*/
for (i = 0; i < program->NumInstructions; i++) {
set_insn_start( t, ureg_get_instruction_number( ureg ));
compile_instruction( ctx, t, &program->Instructions[i], clamp_color );
}
/* Fix up all emitted labels:
*/
for (i = 0; i < t->labels_count; i++) {
ureg_fixup_label( ureg,
t->labels[i].token,
t->insn[t->labels[i].branch_target] );
}
out:
free(t->insn);
free(t->labels);
free(t->constants);
if (t->error) {
debug_printf("%s: translate error flag set\n", __func__);
}
return ret;
}
