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);
}
/**
* Make a simple fragment texture shader which reads the texture unit 0 and 1
* and writes it as depth and stencil, respectively.
*/
void *
util_make_fragment_tex_shader_writedepthstencil(struct pipe_context *pipe,
unsigned tex_target,
unsigned interp_mode,
bool load_level_zero,
bool use_txf)
{
struct ureg_program *ureg;
struct ureg_src depth_sampler, stencil_sampler;
struct ureg_src tex;
struct ureg_dst out, depth, stencil;
struct ureg_src imm;
ureg = ureg_create( PIPE_SHADER_FRAGMENT );
if (!ureg)
return NULL;
depth_sampler = ureg_DECL_sampler( ureg, 0 );
ureg_DECL_sampler_view(ureg, 0, tex_target,
TGSI_RETURN_TYPE_FLOAT,
TGSI_RETURN_TYPE_FLOAT,
TGSI_RETURN_TYPE_FLOAT,
TGSI_RETURN_TYPE_FLOAT);
stencil_sampler = ureg_DECL_sampler( ureg, 1 );
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 );
depth = ureg_DECL_output( ureg,
TGSI_SEMANTIC_POSITION,
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(depth, TGSI_WRITEMASK_Z), tex,
depth_sampler, tex_target, load_level_zero, use_txf);
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 );
}
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
apply_swizzle(struct st_translate *t,
struct ureg_src src, GLuint swizzle)
{
if (swizzle == GL_SWIZZLE_STR_ATI) {
return src;
} else if (swizzle == GL_SWIZZLE_STQ_ATI) {
return ureg_swizzle(src,
TGSI_SWIZZLE_X,
TGSI_SWIZZLE_Y,
TGSI_SWIZZLE_W,
TGSI_SWIZZLE_Z);
} else {
struct ureg_dst tmp[2];
struct ureg_src imm[3];
tmp[0] = get_temp(t, MAX_NUM_FRAGMENT_REGISTERS_ATI);
tmp[1] = get_temp(t, MAX_NUM_FRAGMENT_REGISTERS_ATI + 1);
imm[0] = src;
imm[1] = ureg_imm4f(t->ureg, 1.0f, 1.0f, 0.0f, 0.0f);
imm[2] = ureg_imm4f(t->ureg, 0.0f, 0.0f, 1.0f, 1.0f);
ureg_insn(t->ureg, TGSI_OPCODE_MAD, &tmp[0], 1, imm, 3);
if (swizzle == GL_SWIZZLE_STR_DR_ATI) {
imm[0] = ureg_scalar(src, TGSI_SWIZZLE_Z);
} else {
imm[0] = ureg_scalar(src, TGSI_SWIZZLE_W);
}
ureg_insn(t->ureg, TGSI_OPCODE_RCP, &tmp[1], 1, &imm[0], 1);
imm[0] = ureg_src(tmp[0]);
imm[1] = ureg_src(tmp[1]);
ureg_insn(t->ureg, TGSI_OPCODE_MUL, &tmp[0], 1, imm, 2);
return ureg_src(tmp[0]);
}
}
/**
* 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);
}
/**
* Make a simple fragment texture shader which reads the texture unit 0 and 1
* and writes it as depth and stencil, respectively.
*/
void *
util_make_fragment_tex_shader_writedepthstencil(struct pipe_context *pipe,
unsigned tex_target,
unsigned interp_mode)
{
struct ureg_program *ureg;
struct ureg_src depth_sampler, stencil_sampler;
struct ureg_src tex;
struct ureg_dst out, depth, stencil;
struct ureg_src imm;
ureg = ureg_create( TGSI_PROCESSOR_FRAGMENT );
if (ureg == NULL)
return NULL;
depth_sampler = ureg_DECL_sampler( ureg, 0 );
stencil_sampler = ureg_DECL_sampler( ureg, 1 );
tex = ureg_DECL_fs_input( ureg,
TGSI_SEMANTIC_GENERIC, 0,
interp_mode );
out = ureg_DECL_output( ureg,
TGSI_SEMANTIC_COLOR,
0 );
depth = ureg_DECL_output( ureg,
TGSI_SEMANTIC_POSITION,
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(depth, TGSI_WRITEMASK_Z),
tex_target, tex, depth_sampler );
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 );
}
/**
* 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_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 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 *
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);
}
/**
* Emit the TGSI instructions for inverting and adjusting WPOS.
* This code is unavoidable because it also depends on whether
* a FBO is bound (STATE_FB_WPOS_Y_TRANSFORM).
*/
static void
emit_wpos_adjustment( struct st_translate *t,
const struct gl_program *program,
boolean invert,
GLfloat adjX, GLfloat adjY[2])
{
struct ureg_program *ureg = t->ureg;
/* Fragment program uses fragment position input.
* Need to replace instances of INPUT[WPOS] with temp T
* where T = INPUT[WPOS] by y is inverted.
*/
static const gl_state_index wposTransformState[STATE_LENGTH]
= { STATE_INTERNAL, STATE_FB_WPOS_Y_TRANSFORM, 0, 0, 0 };
/* XXX: note we are modifying the incoming shader here! Need to
* do this before emitting the constant decls below, or this
* will be missed:
*/
unsigned wposTransConst = _mesa_add_state_reference(program->Parameters,
wposTransformState);
struct ureg_src wpostrans = ureg_DECL_constant( ureg, wposTransConst );
struct ureg_dst wpos_temp = ureg_DECL_temporary( ureg );
struct ureg_src wpos_input = t->inputs[t->inputMapping[VARYING_SLOT_POS]];
/* First, apply the coordinate shift: */
if (adjX || adjY[0] || adjY[1]) {
if (adjY[0] != adjY[1]) {
/* Adjust the y coordinate by adjY[1] or adjY[0] respectively
* depending on whether inversion is actually going to be applied
* or not, which is determined by testing against the inversion
* state variable used below, which will be either +1 or -1.
*/
struct ureg_dst adj_temp = ureg_DECL_temporary(ureg);
ureg_CMP(ureg, adj_temp,
ureg_scalar(wpostrans, invert ? 2 : 0),
ureg_imm4f(ureg, adjX, adjY[0], 0.0f, 0.0f),
ureg_imm4f(ureg, adjX, adjY[1], 0.0f, 0.0f));
ureg_ADD(ureg, wpos_temp, wpos_input, ureg_src(adj_temp));
} else {
ureg_ADD(ureg, wpos_temp, wpos_input,
ureg_imm4f(ureg, adjX, adjY[0], 0.0f, 0.0f));
}
wpos_input = ureg_src(wpos_temp);
} else {
/* MOV wpos_temp, input[wpos]
*/
ureg_MOV( ureg, wpos_temp, wpos_input );
}
/* Now the conditional y flip: STATE_FB_WPOS_Y_TRANSFORM.xy/zw will be
* inversion/identity, or the other way around if we're drawing to an FBO.
*/
if (invert) {
/* MAD wpos_temp.y, wpos_input, wpostrans.xxxx, wpostrans.yyyy
*/
ureg_MAD( ureg,
ureg_writemask(wpos_temp, TGSI_WRITEMASK_Y ),
wpos_input,
ureg_scalar(wpostrans, 0),
ureg_scalar(wpostrans, 1));
} else {
/* MAD wpos_temp.y, wpos_input, wpostrans.zzzz, wpostrans.wwww
*/
ureg_MAD( ureg,
ureg_writemask(wpos_temp, TGSI_WRITEMASK_Y ),
wpos_input,
ureg_scalar(wpostrans, 2),
ureg_scalar(wpostrans, 3));
}
/* Use wpos_temp as position input from here on:
*/
t->inputs[t->inputMapping[VARYING_SLOT_POS]] = ureg_src(wpos_temp);
}
/**
* 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;
}
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);
}
/**
* Make simple fragment texture shader:
* IMM {0,0,0,1} // (if writemask != 0xf)
* MOV TEMP[0], IMM[0] // (if writemask != 0xf)
* TEX TEMP[0].writemask, IN[0], SAMP[0], 2D;
* .. optional SINT <-> UINT clamping ..
* MOV OUT[0], TEMP[0]
* 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,
enum tgsi_return_type stype,
enum tgsi_return_type dtype,
bool load_level_zero,
bool use_txf)
{
struct ureg_program *ureg;
struct ureg_src sampler;
struct ureg_src tex;
struct ureg_dst temp;
struct ureg_dst out;
assert((stype == TGSI_RETURN_TYPE_FLOAT) == (dtype == TGSI_RETURN_TYPE_FLOAT));
assert(interp_mode == TGSI_INTERPOLATE_LINEAR ||
interp_mode == TGSI_INTERPOLATE_PERSPECTIVE);
ureg = ureg_create( PIPE_SHADER_FRAGMENT );
if (!ureg)
return NULL;
sampler = ureg_DECL_sampler( ureg, 0 );
ureg_DECL_sampler_view(ureg, 0, tex_target, stype, stype, stype, stype);
tex = ureg_DECL_fs_input( ureg,
TGSI_SEMANTIC_GENERIC, 0,
interp_mode );
out = ureg_DECL_output( ureg,
TGSI_SEMANTIC_COLOR,
0 );
temp = ureg_DECL_temporary(ureg);
if (writemask != TGSI_WRITEMASK_XYZW) {
struct ureg_src imm = ureg_imm4f( ureg, 0, 0, 0, 1 );
ureg_MOV( ureg, out, imm );
}
if (tex_target == TGSI_TEXTURE_BUFFER)
ureg_TXF(ureg,
ureg_writemask(temp, writemask),
tex_target, tex, sampler);
else
ureg_load_tex(ureg, ureg_writemask(temp, writemask), tex, sampler,
tex_target, load_level_zero, use_txf);
if (stype != dtype) {
if (stype == TGSI_RETURN_TYPE_SINT) {
assert(dtype == TGSI_RETURN_TYPE_UINT);
ureg_IMAX(ureg, temp, ureg_src(temp), ureg_imm1i(ureg, 0));
} else {
assert(stype == TGSI_RETURN_TYPE_UINT);
assert(dtype == TGSI_RETURN_TYPE_SINT);
ureg_UMIN(ureg, temp, ureg_src(temp), ureg_imm1u(ureg, (1u << 31) - 1));
}
}
ureg_MOV(ureg, out, ureg_src(temp));
ureg_END( ureg );
return ureg_create_shader_and_destroy( ureg, pipe );
}
