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_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 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);
}
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
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_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);
}
/**
* 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 a SWZ instruction into a MOV, MUL or MAD instruction. EG:
*
* SWZ dst, src.x-y10
*
* becomes:
*
* MAD dst {1,-1,0,0}, src.xyxx, {0,0,1,0}
*/
static void emit_swz( struct st_translate *t,
struct ureg_dst dst,
const struct prog_src_register *SrcReg )
{
struct ureg_program *ureg = t->ureg;
struct ureg_src src = src_register( t, SrcReg->File, SrcReg->Index );
unsigned negate_mask = SrcReg->Negate;
unsigned one_mask = ((GET_SWZ(SrcReg->Swizzle, 0) == SWIZZLE_ONE) << 0 |
(GET_SWZ(SrcReg->Swizzle, 1) == SWIZZLE_ONE) << 1 |
(GET_SWZ(SrcReg->Swizzle, 2) == SWIZZLE_ONE) << 2 |
(GET_SWZ(SrcReg->Swizzle, 3) == SWIZZLE_ONE) << 3);
unsigned zero_mask = ((GET_SWZ(SrcReg->Swizzle, 0) == SWIZZLE_ZERO) << 0 |
(GET_SWZ(SrcReg->Swizzle, 1) == SWIZZLE_ZERO) << 1 |
(GET_SWZ(SrcReg->Swizzle, 2) == SWIZZLE_ZERO) << 2 |
(GET_SWZ(SrcReg->Swizzle, 3) == SWIZZLE_ZERO) << 3);
unsigned negative_one_mask = one_mask & negate_mask;
unsigned positive_one_mask = one_mask & ~negate_mask;
struct ureg_src imm;
unsigned i;
unsigned mul_swizzle[4] = {0,0,0,0};
unsigned add_swizzle[4] = {0,0,0,0};
unsigned src_swizzle[4] = {0,0,0,0};
boolean need_add = FALSE;
boolean need_mul = FALSE;
if (dst.WriteMask == 0)
return;
/* Is this just a MOV?
*/
if (zero_mask == 0 &&
one_mask == 0 &&
(negate_mask == 0 || negate_mask == TGSI_WRITEMASK_XYZW))
{
ureg_MOV( ureg, dst, translate_src( t, SrcReg ));
return;
}
#define IMM_ZERO 0
#define IMM_ONE 1
#define IMM_NEG_ONE 2
imm = ureg_imm3f( ureg, 0, 1, -1 );
for (i = 0; i < 4; i++) {
unsigned bit = 1 << i;
if (dst.WriteMask & bit) {
if (positive_one_mask & bit) {
mul_swizzle[i] = IMM_ZERO;
add_swizzle[i] = IMM_ONE;
need_add = TRUE;
}
else if (negative_one_mask & bit) {
mul_swizzle[i] = IMM_ZERO;
add_swizzle[i] = IMM_NEG_ONE;
need_add = TRUE;
}
else if (zero_mask & bit) {
mul_swizzle[i] = IMM_ZERO;
add_swizzle[i] = IMM_ZERO;
need_add = TRUE;
}
else {
add_swizzle[i] = IMM_ZERO;
src_swizzle[i] = GET_SWZ(SrcReg->Swizzle, i);
need_mul = TRUE;
if (negate_mask & bit) {
mul_swizzle[i] = IMM_NEG_ONE;
}
else {
mul_swizzle[i] = IMM_ONE;
}
}
}
}
if (need_mul && need_add) {
ureg_MAD( ureg,
dst,
swizzle_4v( src, src_swizzle ),
swizzle_4v( imm, mul_swizzle ),
swizzle_4v( imm, add_swizzle ) );
}
else if (need_mul) {
ureg_MUL( ureg,
dst,
swizzle_4v( src, src_swizzle ),
swizzle_4v( imm, mul_swizzle ) );
}
else if (need_add) {
ureg_MOV( ureg,
dst,
swizzle_4v( imm, add_swizzle ) );
}
else {
debug_assert(0);
}
#undef IMM_ZERO
#undef IMM_ONE
#undef IMM_NEG_ONE
}
static void *
create_vert_shader(struct vl_compositor *c)
{
struct ureg_program *shader;
struct ureg_src vpos, vtex, color;
struct ureg_dst tmp;
struct ureg_dst o_vpos, o_vtex, o_color;
struct ureg_dst o_vtop, o_vbottom;
shader = ureg_create(TGSI_PROCESSOR_VERTEX);
if (!shader)
return false;
vpos = ureg_DECL_vs_input(shader, 0);
vtex = ureg_DECL_vs_input(shader, 1);
color = ureg_DECL_vs_input(shader, 2);
tmp = ureg_DECL_temporary(shader);
o_vpos = ureg_DECL_output(shader, TGSI_SEMANTIC_POSITION, VS_O_VPOS);
o_color = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, VS_O_COLOR);
o_vtex = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_VTEX);
o_vtop = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_VTOP);
o_vbottom = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_VBOTTOM);
/*
* o_vpos = vpos
* o_vtex = vtex
* o_color = color
*/
ureg_MOV(shader, o_vpos, vpos);
ureg_MOV(shader, o_vtex, vtex);
ureg_MOV(shader, o_color, color);
/*
* tmp.x = vtex.w / 2
* tmp.y = vtex.w / 4
*
* o_vtop.x = vtex.x
* o_vtop.y = vtex.y * tmp.x + 0.25f
* o_vtop.z = vtex.y * tmp.y + 0.25f
* o_vtop.w = 1 / tmp.x
*
* o_vbottom.x = vtex.x
* o_vbottom.y = vtex.y * tmp.x - 0.25f
* o_vbottom.z = vtex.y * tmp.y - 0.25f
* o_vbottom.w = 1 / tmp.y
*/
ureg_MUL(shader, ureg_writemask(tmp, TGSI_WRITEMASK_X),
ureg_scalar(vtex, TGSI_SWIZZLE_W), ureg_imm1f(shader, 0.5f));
ureg_MUL(shader, ureg_writemask(tmp, TGSI_WRITEMASK_Y),
ureg_scalar(vtex, TGSI_SWIZZLE_W), ureg_imm1f(shader, 0.25f));
ureg_MOV(shader, ureg_writemask(o_vtop, TGSI_WRITEMASK_X), vtex);
ureg_MAD(shader, ureg_writemask(o_vtop, TGSI_WRITEMASK_Y), ureg_scalar(vtex, TGSI_SWIZZLE_Y),
ureg_scalar(ureg_src(tmp), TGSI_SWIZZLE_X), ureg_imm1f(shader, 0.25f));
ureg_MAD(shader, ureg_writemask(o_vtop, TGSI_WRITEMASK_Z), ureg_scalar(vtex, TGSI_SWIZZLE_Y),
ureg_scalar(ureg_src(tmp), TGSI_SWIZZLE_Y), ureg_imm1f(shader, 0.25f));
ureg_RCP(shader, ureg_writemask(o_vtop, TGSI_WRITEMASK_W),
ureg_scalar(ureg_src(tmp), TGSI_SWIZZLE_X));
ureg_MOV(shader, ureg_writemask(o_vbottom, TGSI_WRITEMASK_X), vtex);
ureg_MAD(shader, ureg_writemask(o_vbottom, TGSI_WRITEMASK_Y), ureg_scalar(vtex, TGSI_SWIZZLE_Y),
ureg_scalar(ureg_src(tmp), TGSI_SWIZZLE_X), ureg_imm1f(shader, -0.25f));
ureg_MAD(shader, ureg_writemask(o_vbottom, TGSI_WRITEMASK_Z), ureg_scalar(vtex, TGSI_SWIZZLE_Y),
ureg_scalar(ureg_src(tmp), TGSI_SWIZZLE_Y), ureg_imm1f(shader, -0.25f));
ureg_RCP(shader, ureg_writemask(o_vbottom, TGSI_WRITEMASK_W),
ureg_scalar(ureg_src(tmp), TGSI_SWIZZLE_Y));
ureg_END(shader);
return ureg_create_shader_and_destroy(shader, c->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);
}
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);
}