/**
* 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);
}
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 *
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_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);
}