static struct ureg calculate_light_attenuation( struct tnl_program *p,
GLuint i,
struct ureg VPpli,
struct ureg dist )
{
struct ureg attenuation = register_param3(p, STATE_LIGHT, i,
STATE_ATTENUATION);
struct ureg att = undef;
/* Calculate spot attenuation:
*/
if (!p->state->unit[i].light_spotcutoff_is_180) {
struct ureg spot_dir_norm = register_param3(p, STATE_INTERNAL,
STATE_LIGHT_SPOT_DIR_NORMALIZED, i);
struct ureg spot = get_temp(p);
struct ureg slt = get_temp(p);
att = get_temp(p);
emit_op2(p, OPCODE_DP3, spot, 0, negate(VPpli), spot_dir_norm);
emit_op2(p, OPCODE_SLT, slt, 0, swizzle1(spot_dir_norm,W), spot);
emit_op1(p, OPCODE_ABS, spot, 0, spot);
emit_op2(p, OPCODE_POW, spot, 0, spot, swizzle1(attenuation, W));
emit_op2(p, OPCODE_MUL, att, 0, slt, spot);
release_temp(p, spot);
release_temp(p, slt);
}
/* Calculate distance attenuation(See formula (2.4) at glspec 2.1 page 62):
*
* Skip the calucation when _dist_ is undefined(light_eyepos3_is_zero)
*/
if (p->state->unit[i].light_attenuated && !is_undef(dist)) {
if (is_undef(att))
att = get_temp(p);
/* 1/d,d,d,1/d */
emit_op1(p, OPCODE_RCP, dist, WRITEMASK_YZ, dist);
/* 1,d,d*d,1/d */
emit_op2(p, OPCODE_MUL, dist, WRITEMASK_XZ, dist, swizzle1(dist,Y));
/* 1/dist-atten */
emit_op2(p, OPCODE_DP3, dist, 0, attenuation, dist);
if (!p->state->unit[i].light_spotcutoff_is_180) {
/* dist-atten */
emit_op1(p, OPCODE_RCP, dist, 0, dist);
/* spot-atten * dist-atten */
emit_op2(p, OPCODE_MUL, att, 0, dist, att);
}
else {
/* dist-atten */
emit_op1(p, OPCODE_RCP, att, 0, dist);
}
}
return att;
}
static struct ureg calculate_light_attenuation( struct tnl_program *p,
GLuint i,
struct ureg VPpli,
struct ureg dist )
{
struct ureg attenuation = register_param3(p, STATE_LIGHT, i,
STATE_ATTENUATION);
struct ureg att = get_temp(p);
/* Calculate spot attenuation:
*/
if (!p->state->unit[i].light_spotcutoff_is_180) {
struct ureg spot_dir_norm = register_param3(p, STATE_INTERNAL,
STATE_LIGHT_SPOT_DIR_NORMALIZED, i);
struct ureg spot = get_temp(p);
struct ureg slt = get_temp(p);
emit_op2(p, OPCODE_DP3, spot, 0, negate(VPpli), spot_dir_norm);
emit_op2(p, OPCODE_SLT, slt, 0, swizzle1(spot_dir_norm,W), spot);
emit_op2(p, OPCODE_POW, spot, 0, spot, swizzle1(attenuation, W));
emit_op2(p, OPCODE_MUL, att, 0, slt, spot);
release_temp(p, spot);
release_temp(p, slt);
}
/* Calculate distance attenuation:
*/
if (p->state->unit[i].light_attenuated) {
/* 1/d,d,d,1/d */
emit_op1(p, OPCODE_RCP, dist, WRITEMASK_YZ, dist);
/* 1,d,d*d,1/d */
emit_op2(p, OPCODE_MUL, dist, WRITEMASK_XZ, dist, swizzle1(dist,Y));
/* 1/dist-atten */
emit_op2(p, OPCODE_DP3, dist, 0, attenuation, dist);
if (!p->state->unit[i].light_spotcutoff_is_180) {
/* dist-atten */
emit_op1(p, OPCODE_RCP, dist, 0, dist);
/* spot-atten * dist-atten */
emit_op2(p, OPCODE_MUL, att, 0, dist, att);
}
else {
/* dist-atten */
emit_op1(p, OPCODE_RCP, att, 0, dist);
}
}
return att;
}
/* Seems like it could be tighter:
*/
static void build_pointsize( struct tnl_program *p )
{
struct ureg eye = get_eye_position(p);
struct ureg state_size = register_param1(p, STATE_POINT_SIZE);
struct ureg state_attenuation = register_param1(p, STATE_POINT_ATTENUATION);
struct ureg out = register_output(p, VERT_RESULT_PSIZ);
struct ureg ut = get_temp(p);
/* 1, Z, Z * Z, 1 */
emit_op1(p, OPCODE_MOV, ut, WRITEMASK_XW, swizzle1(get_identity_param(p), W));
emit_op1(p, OPCODE_ABS, ut, WRITEMASK_YZ, swizzle1(eye, Z));
emit_op2(p, OPCODE_MUL, ut, WRITEMASK_Z, ut, ut);
/* p1 + p2 * dist + p3 * dist * dist, 0 */
emit_op2(p, OPCODE_DP3, ut, WRITEMASK_X, ut, state_attenuation);
/* 1 / sqrt(factor) */
emit_op1(p, OPCODE_RSQ, ut, WRITEMASK_X, ut );
/* ut = pointSize / factor */
emit_op2(p, OPCODE_MUL, ut, WRITEMASK_X, ut, state_size);
/* Clamp to min/max - state_size.[yz]
*/
emit_op2(p, OPCODE_MAX, ut, WRITEMASK_X, ut, swizzle1(state_size, Y));
emit_op2(p, OPCODE_MIN, out, 0, swizzle1(ut, X), swizzle1(state_size, Z));
release_temp(p, ut);
}
static void build_pointsize( struct tnl_program *p )
{
struct ureg eye = get_eye_position(p);
struct ureg state_size = register_param1(p, STATE_POINT_SIZE);
struct ureg state_attenuation = register_param1(p, STATE_POINT_ATTENUATION);
struct ureg out = register_output(p, VERT_RESULT_PSIZ);
struct ureg ut = get_temp(p);
/* dist = |eyez| */
emit_op1(p, OPCODE_ABS, ut, WRITEMASK_Y, swizzle1(eye, Z));
/* p1 + dist * (p2 + dist * p3); */
emit_op3(p, OPCODE_MAD, ut, WRITEMASK_X, swizzle1(ut, Y),
swizzle1(state_attenuation, Z), swizzle1(state_attenuation, Y));
emit_op3(p, OPCODE_MAD, ut, WRITEMASK_X, swizzle1(ut, Y),
ut, swizzle1(state_attenuation, X));
/* 1 / sqrt(factor) */
emit_op1(p, OPCODE_RSQ, ut, WRITEMASK_X, ut );
#if 1
/* out = pointSize / sqrt(factor) */
emit_op2(p, OPCODE_MUL, out, WRITEMASK_X, ut, state_size);
#else
/* not sure, might make sense to do clamping here,
but it's not done in t_vb_points neither */
emit_op2(p, OPCODE_MUL, ut, WRITEMASK_X, ut, state_size);
emit_op2(p, OPCODE_MAX, ut, WRITEMASK_X, ut, swizzle1(state_size, Y));
emit_op2(p, OPCODE_MIN, out, WRITEMASK_X, ut, swizzle1(state_size, Z));
#endif
release_temp(p, ut);
}
/**
* Point size attenuation computation.
*/
static void build_atten_pointsize( struct tnl_program *p )
{
struct ureg eye = get_eye_position_z(p);
struct ureg state_size = register_param1(p, STATE_POINT_SIZE);
struct ureg state_attenuation = register_param1(p, STATE_POINT_ATTENUATION);
struct ureg out = register_output(p, VERT_RESULT_PSIZ);
struct ureg ut = get_temp(p);
/* dist = |eyez| */
emit_op1(p, OPCODE_ABS, ut, WRITEMASK_Y, swizzle1(eye, Z));
/* p1 + dist * (p2 + dist * p3); */
emit_op3(p, OPCODE_MAD, ut, WRITEMASK_X, swizzle1(ut, Y),
swizzle1(state_attenuation, Z), swizzle1(state_attenuation, Y));
emit_op3(p, OPCODE_MAD, ut, WRITEMASK_X, swizzle1(ut, Y),
ut, swizzle1(state_attenuation, X));
/* 1 / sqrt(factor) */
emit_op1(p, OPCODE_RSQ, ut, WRITEMASK_X, ut );
#if 0
/* out = pointSize / sqrt(factor) */
emit_op2(p, OPCODE_MUL, out, WRITEMASK_X, ut, state_size);
#else
/* this is a good place to clamp the point size since there's likely
* no hardware registers to clamp point size at rasterization time.
*/
emit_op2(p, OPCODE_MUL, ut, WRITEMASK_X, ut, state_size);
emit_op2(p, OPCODE_MAX, ut, WRITEMASK_X, ut, swizzle1(state_size, Y));
emit_op2(p, OPCODE_MIN, out, WRITEMASK_X, ut, swizzle1(state_size, Z));
#endif
release_temp(p, ut);
}
void jinit_cpy_arg_res( struct JFunc* jf, struct machine_ops* mop, struct emitter* e, struct machine* m ){
operand rargs[ RA_SIZE ];
prefer_nontemp_acquire_reg( mop, e, m, RA_SIZE, rargs );
syn_min( mop, e, m, rargs[ RA_EXIST ], rargs[ RA_EXIST ], rargs[ RA_EXPECT ] );
// init iterator
operand iter = OP_TARGETREG( acquire_temp( mop, e, m ) );
mop->move( e, m, iter, OP_TARGETIMMED( 0 ) );
do_copying( mop, e, m, iter, rargs[ RA_EXIST ], rargs[ RA_DST ], rargs[ RA_SRC ] );
/*
* TODO: Below logic is incorrect the first is most likely to get spilled. So change enum order. BUT
* then you have to think about prefer saved reg and the order there.
* TODO: add error prefer_nontemps to error when not all live simultenously
*
* RA_EXPECT is last register therefore its the most likely to be spilled. So to stop repeat
* spill/unspill move it to exist.
*/
mop->move( e, m, rargs[ RA_SIZE ], rargs[ RA_EXIST ] );
do_nilling( mop, e, m, iter, rargs[ RA_EXIST ], rargs[ RA_DST ] );
release_temp( mop, e, m );
prefer_nontemp_release_reg( mop, e, m, RA_SIZE );
mop->ret( e, m );
}
static void build_fog( struct tnl_program *p )
{
struct ureg fog = register_output(p, VERT_RESULT_FOGC);
struct ureg input;
GLuint useabs = p->state->fog_source_is_depth && p->state->fog_option &&
(p->state->fog_option != FOG_EXP2);
if (p->state->fog_source_is_depth) {
input = swizzle1(get_eye_position(p), Z);
}
else {
input = swizzle1(register_input(p, VERT_ATTRIB_FOG), X);
}
if (p->state->fog_option &&
p->state->tnl_do_vertex_fog) {
struct ureg params = register_param2(p, STATE_INTERNAL,
STATE_FOG_PARAMS_OPTIMIZED);
struct ureg tmp = get_temp(p);
struct ureg id = get_identity_param(p);
emit_op1(p, OPCODE_MOV, fog, 0, id);
if (useabs) {
emit_op1(p, OPCODE_ABS, tmp, 0, input);
}
switch (p->state->fog_option) {
case FOG_LINEAR: {
emit_op3(p, OPCODE_MAD, tmp, 0, useabs ? tmp : input,
swizzle1(params,X), swizzle1(params,Y));
emit_op2(p, OPCODE_MAX, tmp, 0, tmp, swizzle1(id,X)); /* saturate */
emit_op2(p, OPCODE_MIN, fog, WRITEMASK_X, tmp, swizzle1(id,W));
break;
}
case FOG_EXP:
emit_op2(p, OPCODE_MUL, tmp, 0, useabs ? tmp : input,
swizzle1(params,Z));
emit_op1(p, OPCODE_EX2, fog, WRITEMASK_X, ureg_negate(tmp));
break;
case FOG_EXP2:
emit_op2(p, OPCODE_MUL, tmp, 0, input, swizzle1(params,W));
emit_op2(p, OPCODE_MUL, tmp, 0, tmp, tmp);
emit_op1(p, OPCODE_EX2, fog, WRITEMASK_X, ureg_negate(tmp));
break;
}
release_temp(p, tmp);
}
else {
/* results = incoming fog coords (compute fog per-fragment later)
*
* KW: Is it really necessary to do anything in this case?
*/
emit_op1(p, useabs ? OPCODE_ABS : OPCODE_MOV, fog, 0, input);
}
}
static void build_fog( struct tnl_program *p )
{
struct ureg fog = register_output(p, VERT_RESULT_FOGC);
struct ureg input;
if (p->state->fog_source_is_depth) {
input = swizzle1(get_eye_position(p), Z);
}
else {
input = swizzle1(register_input(p, VERT_ATTRIB_FOG), X);
}
if (p->state->fog_mode && p->state->tnl_do_vertex_fog) {
struct ureg params = register_param2(p, STATE_INTERNAL,
STATE_FOG_PARAMS_OPTIMIZED);
struct ureg tmp = get_temp(p);
GLboolean useabs = (p->state->fog_mode != FOG_EXP2);
if (useabs) {
emit_op1(p, OPCODE_ABS, tmp, 0, input);
}
switch (p->state->fog_mode) {
case FOG_LINEAR: {
struct ureg id = get_identity_param(p);
emit_op3(p, OPCODE_MAD, tmp, 0, useabs ? tmp : input,
swizzle1(params,X), swizzle1(params,Y));
emit_op2(p, OPCODE_MAX, tmp, 0, tmp, swizzle1(id,X)); /* saturate */
emit_op2(p, OPCODE_MIN, fog, WRITEMASK_X, tmp, swizzle1(id,W));
break;
}
case FOG_EXP:
emit_op2(p, OPCODE_MUL, tmp, 0, useabs ? tmp : input,
swizzle1(params,Z));
emit_op1(p, OPCODE_EX2, fog, WRITEMASK_X, negate(tmp));
break;
case FOG_EXP2:
emit_op2(p, OPCODE_MUL, tmp, 0, input, swizzle1(params,W));
emit_op2(p, OPCODE_MUL, tmp, 0, tmp, tmp);
emit_op1(p, OPCODE_EX2, fog, WRITEMASK_X, negate(tmp));
break;
}
release_temp(p, tmp);
}
else {
/* results = incoming fog coords (compute fog per-fragment later)
*
* KW: Is it really necessary to do anything in this case?
* BP: Yes, we always need to compute the absolute value, unless
* we want to push that down into the fragment program...
*/
GLboolean useabs = GL_TRUE;
emit_op1(p, useabs ? OPCODE_ABS : OPCODE_MOV, fog, WRITEMASK_X, input);
}
}
static void prefer_nontemp_release_reg( struct machine_ops* mop, struct emitter* e, struct machine* m, int n ){
const int nr_temps = m->nr_temp_regs;
const int nr_nontemps = m->nr_reg - nr_temps;
for( int i = nr_nontemps; i < n; i++ ){
release_temp( mop, e, m );
}
enable_spill( m );
}
static void emit_normalize_vec3( struct tnl_program *p,
struct ureg dest,
struct ureg src )
{
struct ureg tmp = get_temp(p);
emit_op2(p, OPCODE_DP3, tmp, WRITEMASK_X, src, src);
emit_op1(p, OPCODE_RSQ, tmp, WRITEMASK_X, tmp);
emit_op2(p, OPCODE_MUL, dest, 0, src, swizzle1(tmp, X));
release_temp(p, tmp);
}
static void build_sphere_texgen( struct tnl_program *p,
struct ureg dest,
GLuint writemask )
{
struct ureg normal = get_transformed_normal(p);
struct ureg eye_hat = get_eye_position_normalized(p);
struct ureg tmp = get_temp(p);
struct ureg half = register_scalar_const(p, .5);
struct ureg r = get_temp(p);
struct ureg inv_m = get_temp(p);
struct ureg id = get_identity_param(p);
/* Could share the above calculations, but it would be
* a fairly odd state for someone to set (both sphere and
* reflection active for different texture coordinate
* components. Of course - if two texture units enable
* reflect and/or sphere, things start to tilt in favour
* of seperating this out:
*/
/* n.u */
emit_op2(p, OPCODE_DP3, tmp, 0, normal, eye_hat);
/* 2n.u */
emit_op2(p, OPCODE_ADD, tmp, 0, tmp, tmp);
/* (-2n.u)n + u */
emit_op3(p, OPCODE_MAD, r, 0, negate(tmp), normal, eye_hat);
/* r + 0,0,1 */
emit_op2(p, OPCODE_ADD, tmp, 0, r, swizzle(id,X,Y,W,Z));
/* rx^2 + ry^2 + (rz+1)^2 */
emit_op2(p, OPCODE_DP3, tmp, 0, tmp, tmp);
/* 2/m */
emit_op1(p, OPCODE_RSQ, tmp, 0, tmp);
/* 1/m */
emit_op2(p, OPCODE_MUL, inv_m, 0, tmp, half);
/* r/m + 1/2 */
emit_op3(p, OPCODE_MAD, dest, writemask, r, inv_m, half);
release_temp(p, tmp);
release_temp(p, r);
release_temp(p, inv_m);
}
static void
configure_sensor_trip_points(void)
{
int trigger_temperature = limit_temp(throttled_bin);
int release_temperature = release_temp(throttled_bin);
pr_info("msm_thermal: setting trip range %d..%d on sensor %d.\n", release_temperature, trigger_temperature, msm_thermal_info.sensor_id);
if (trigger_temperature != NO_TRIGGER_TEMPERATURE)
tsens_set_tz_warm_temp_degC(msm_thermal_info.sensor_id, trigger_temperature, &trip_work);
if (release_temperature != NO_RELEASE_TEMPERATURE)
tsens_set_tz_cool_temp_degC(msm_thermal_info.sensor_id, release_temperature, &trip_work);
}
static int
select_throttled_bin(unsigned temp)
{
int i;
int new_bin = -1;
for (i = 0; i < N_TEMP_LIMITS; i++) {
if (temp >= limit_temp_degC[i]) new_bin = i;
}
if (new_bin > throttled_bin) return new_bin;
if (temp <= release_temp(throttled_bin)) return new_bin;
return throttled_bin;
}
static void emit_normalize_vec3( struct tnl_program *p,
struct ureg dest,
struct ureg src )
{
#if 0
/* XXX use this when drivers are ready for NRM3 */
emit_op1(p, OPCODE_NRM3, dest, WRITEMASK_XYZ, src);
#else
struct ureg tmp = get_temp(p);
emit_op2(p, OPCODE_DP3, tmp, WRITEMASK_X, src, src);
emit_op1(p, OPCODE_RSQ, tmp, WRITEMASK_X, tmp);
emit_op2(p, OPCODE_MUL, dest, 0, src, swizzle1(tmp, X));
release_temp(p, tmp);
#endif
}
static void build_reflect_texgen( struct tnl_program *p,
struct ureg dest,
GLuint writemask )
{
struct ureg normal = get_transformed_normal(p);
struct ureg eye_hat = get_eye_position_normalized(p);
struct ureg tmp = get_temp(p);
/* n.u */
emit_op2(p, OPCODE_DP3, tmp, 0, normal, eye_hat);
/* 2n.u */
emit_op2(p, OPCODE_ADD, tmp, 0, tmp, tmp);
/* (-2n.u)n + u */
emit_op3(p, OPCODE_MAD, dest, writemask, negate(tmp), normal, eye_hat);
release_temp(p, tmp);
}
/**
* Emit code for TXP.
*/
static void precalc_txp( struct brw_wm_compile *c,
const struct prog_instruction *inst )
{
struct prog_src_register src0 = inst->SrcReg[0];
if (projtex(c, inst)) {
struct prog_dst_register tmp = get_temp(c);
struct prog_instruction tmp_inst;
/* tmp0.w = RCP inst.arg[0][3]
*/
emit_op(c,
OPCODE_RCP,
dst_mask(tmp, WRITEMASK_W),
0,
src_swizzle1(src0, GET_SWZ(src0.Swizzle, W)),
src_undef(),
src_undef());
/* tmp0.xyz = MUL inst.arg[0], tmp0.wwww
*/
emit_op(c,
OPCODE_MUL,
dst_mask(tmp, WRITEMASK_XYZ),
0,
src0,
src_swizzle1(src_reg_from_dst(tmp), W),
src_undef());
/* dst = precalc(TEX tmp0)
*/
tmp_inst = *inst;
tmp_inst.SrcReg[0] = src_reg_from_dst(tmp);
precalc_tex(c, &tmp_inst);
release_temp(c, tmp);
}
else
{
/* dst = precalc(TEX src0)
*/
precalc_tex(c, inst);
}
}
/**
* Emit code for TXP.
*/
static void precalc_txp( struct brw_wm_compile *c,
struct brw_fp_dst dst,
unsigned target,
unsigned unit,
struct brw_fp_src src0,
struct brw_fp_src sampler )
{
if (projtex(c, target, src0)) {
struct brw_fp_dst tmp = get_temp(c);
/* tmp0.w = RCP inst.arg[0][3]
*/
emit_op1(c,
TGSI_OPCODE_RCP,
dst_mask(tmp, BRW_WRITEMASK_W),
src_scalar(src0, W));
/* tmp0.xyz = MUL inst.arg[0], tmp0.wwww
*/
emit_op2(c,
TGSI_OPCODE_MUL,
dst_mask(tmp, BRW_WRITEMASK_XYZ),
src0,
src_scalar(src_reg_from_dst(tmp), W));
/* dst = TEX tmp0
*/
precalc_tex(c,
dst,
target,
unit,
src_reg_from_dst(tmp),
sampler );
release_temp(c, tmp);
}
else
{
/* dst = TEX src0
*/
precalc_tex(c, dst, target, unit, src0, sampler);
}
}
/* This version is much easier to implement if writemasks are not
* supported natively on the target or (like SSE), the target doesn't
* have a clean/obvious dotproduct implementation.
*/
static void emit_transpose_matrix_transform_vec4( struct tnl_program *p,
struct ureg dest,
const struct ureg *mat,
struct ureg src)
{
struct ureg tmp;
if (dest.file != PROGRAM_TEMPORARY)
tmp = get_temp(p);
else
tmp = dest;
emit_op2(p, OPCODE_MUL, tmp, 0, swizzle1(src,X), mat[0]);
emit_op3(p, OPCODE_MAD, tmp, 0, swizzle1(src,Y), mat[1], tmp);
emit_op3(p, OPCODE_MAD, tmp, 0, swizzle1(src,Z), mat[2], tmp);
emit_op3(p, OPCODE_MAD, dest, 0, swizzle1(src,W), mat[3], tmp);
if (dest.file != PROGRAM_TEMPORARY)
release_temp(p, tmp);
}
/* Need to add some addtional parameters to allow lighting in object
* space - STATE_SPOT_DIRECTION and STATE_HALF_VECTOR implicitly assume eye
* space lighting.
*/
static void build_lighting( struct tnl_program *p )
{
const GLboolean twoside = p->state->light_twoside;
const GLboolean separate = p->state->separate_specular;
GLuint nr_lights = 0, count = 0;
struct ureg normal = get_transformed_normal(p);
struct ureg lit = get_temp(p);
struct ureg dots = get_temp(p);
struct ureg _col0 = undef, _col1 = undef;
struct ureg _bfc0 = undef, _bfc1 = undef;
GLuint i;
/*
* NOTE:
* dots.x = dot(normal, VPpli)
* dots.y = dot(normal, halfAngle)
* dots.z = back.shininess
* dots.w = front.shininess
*/
for (i = 0; i < MAX_LIGHTS; i++)
if (p->state->unit[i].light_enabled)
nr_lights++;
set_material_flags(p);
{
if (!p->state->material_shininess_is_zero) {
struct ureg shininess = get_material(p, 0, STATE_SHININESS);
emit_op1(p, OPCODE_MOV, dots, WRITEMASK_W, swizzle1(shininess,X));
release_temp(p, shininess);
}
_col0 = make_temp(p, get_scenecolor(p, 0));
if (separate)
_col1 = make_temp(p, get_identity_param(p));
else
_col1 = _col0;
}
if (twoside) {
if (!p->state->material_shininess_is_zero) {
/* Note that we negate the back-face specular exponent here.
* The negation will be un-done later in the back-face code below.
*/
struct ureg shininess = get_material(p, 1, STATE_SHININESS);
emit_op1(p, OPCODE_MOV, dots, WRITEMASK_Z,
negate(swizzle1(shininess,X)));
release_temp(p, shininess);
}
_bfc0 = make_temp(p, get_scenecolor(p, 1));
if (separate)
_bfc1 = make_temp(p, get_identity_param(p));
else
_bfc1 = _bfc0;
}
/* If no lights, still need to emit the scenecolor.
*/
{
struct ureg res0 = register_output( p, VERT_RESULT_COL0 );
emit_op1(p, OPCODE_MOV, res0, 0, _col0);
}
if (separate) {
struct ureg res1 = register_output( p, VERT_RESULT_COL1 );
emit_op1(p, OPCODE_MOV, res1, 0, _col1);
}
if (twoside) {
struct ureg res0 = register_output( p, VERT_RESULT_BFC0 );
emit_op1(p, OPCODE_MOV, res0, 0, _bfc0);
}
if (twoside && separate) {
struct ureg res1 = register_output( p, VERT_RESULT_BFC1 );
emit_op1(p, OPCODE_MOV, res1, 0, _bfc1);
}
if (nr_lights == 0) {
release_temps(p);
return;
}
for (i = 0; i < MAX_LIGHTS; i++) {
if (p->state->unit[i].light_enabled) {
struct ureg half = undef;
struct ureg att = undef, VPpli = undef;
count++;
if (p->state->unit[i].light_eyepos3_is_zero) {
/* Can used precomputed constants in this case.
* Attenuation never applies to infinite lights.
*/
VPpli = register_param3(p, STATE_INTERNAL,
STATE_LIGHT_POSITION_NORMALIZED, i);
if (!p->state->material_shininess_is_zero) {
if (p->state->light_local_viewer) {
struct ureg eye_hat = get_eye_position_normalized(p);
half = get_temp(p);
emit_op2(p, OPCODE_SUB, half, 0, VPpli, eye_hat);
emit_normalize_vec3(p, half, half);
}
else {
half = register_param3(p, STATE_INTERNAL,
STATE_LIGHT_HALF_VECTOR, i);
}
}
}
else {
struct ureg Ppli = register_param3(p, STATE_INTERNAL,
STATE_LIGHT_POSITION, i);
struct ureg V = get_eye_position(p);
struct ureg dist = get_temp(p);
VPpli = get_temp(p);
/* Calculate VPpli vector
*/
emit_op2(p, OPCODE_SUB, VPpli, 0, Ppli, V);
/* Normalize VPpli. The dist value also used in
* attenuation below.
*/
emit_op2(p, OPCODE_DP3, dist, 0, VPpli, VPpli);
emit_op1(p, OPCODE_RSQ, dist, 0, dist);
emit_op2(p, OPCODE_MUL, VPpli, 0, VPpli, dist);
/* Calculate attenuation:
*/
if (!p->state->unit[i].light_spotcutoff_is_180 ||
p->state->unit[i].light_attenuated) {
att = calculate_light_attenuation(p, i, VPpli, dist);
}
/* Calculate viewer direction, or use infinite viewer:
*/
if (!p->state->material_shininess_is_zero) {
half = get_temp(p);
if (p->state->light_local_viewer) {
struct ureg eye_hat = get_eye_position_normalized(p);
emit_op2(p, OPCODE_SUB, half, 0, VPpli, eye_hat);
}
else {
struct ureg z_dir = swizzle(get_identity_param(p),X,Y,W,Z);
emit_op2(p, OPCODE_ADD, half, 0, VPpli, z_dir);
}
emit_normalize_vec3(p, half, half);
}
release_temp(p, dist);
}
/* Calculate dot products:
*/
if (p->state->material_shininess_is_zero) {
emit_op2(p, OPCODE_DP3, dots, 0, normal, VPpli);
}
else {
emit_op2(p, OPCODE_DP3, dots, WRITEMASK_X, normal, VPpli);
emit_op2(p, OPCODE_DP3, dots, WRITEMASK_Y, normal, half);
}
/* Front face lighting:
*/
{
struct ureg ambient = get_lightprod(p, i, 0, STATE_AMBIENT);
struct ureg diffuse = get_lightprod(p, i, 0, STATE_DIFFUSE);
struct ureg specular = get_lightprod(p, i, 0, STATE_SPECULAR);
struct ureg res0, res1;
GLuint mask0, mask1;
if (count == nr_lights) {
if (separate) {
mask0 = WRITEMASK_XYZ;
mask1 = WRITEMASK_XYZ;
res0 = register_output( p, VERT_RESULT_COL0 );
res1 = register_output( p, VERT_RESULT_COL1 );
}
else {
mask0 = 0;
mask1 = WRITEMASK_XYZ;
res0 = _col0;
res1 = register_output( p, VERT_RESULT_COL0 );
}
}
else {
mask0 = 0;
mask1 = 0;
res0 = _col0;
res1 = _col1;
}
if (!is_undef(att)) {
/* light is attenuated by distance */
emit_op1(p, OPCODE_LIT, lit, 0, dots);
emit_op2(p, OPCODE_MUL, lit, 0, lit, att);
emit_op3(p, OPCODE_MAD, _col0, 0, swizzle1(lit,X), ambient, _col0);
}
else if (!p->state->material_shininess_is_zero) {
/* there's a non-zero specular term */
emit_op1(p, OPCODE_LIT, lit, 0, dots);
emit_op2(p, OPCODE_ADD, _col0, 0, ambient, _col0);
}
else {
/* no attenutation, no specular */
emit_degenerate_lit(p, lit, dots);
emit_op2(p, OPCODE_ADD, _col0, 0, ambient, _col0);
}
emit_op3(p, OPCODE_MAD, res0, mask0, swizzle1(lit,Y), diffuse, _col0);
emit_op3(p, OPCODE_MAD, res1, mask1, swizzle1(lit,Z), specular, _col1);
release_temp(p, ambient);
release_temp(p, diffuse);
release_temp(p, specular);
}
/* Back face lighting:
*/
if (twoside) {
struct ureg ambient = get_lightprod(p, i, 1, STATE_AMBIENT);
struct ureg diffuse = get_lightprod(p, i, 1, STATE_DIFFUSE);
struct ureg specular = get_lightprod(p, i, 1, STATE_SPECULAR);
struct ureg res0, res1;
GLuint mask0, mask1;
if (count == nr_lights) {
if (separate) {
mask0 = WRITEMASK_XYZ;
mask1 = WRITEMASK_XYZ;
res0 = register_output( p, VERT_RESULT_BFC0 );
res1 = register_output( p, VERT_RESULT_BFC1 );
}
else {
mask0 = 0;
mask1 = WRITEMASK_XYZ;
res0 = _bfc0;
res1 = register_output( p, VERT_RESULT_BFC0 );
}
}
else {
res0 = _bfc0;
res1 = _bfc1;
mask0 = 0;
mask1 = 0;
}
/* For the back face we need to negate the X and Y component
* dot products. dots.Z has the negated back-face specular
* exponent. We swizzle that into the W position. This
* negation makes the back-face specular term positive again.
*/
dots = negate(swizzle(dots,X,Y,W,Z));
if (!is_undef(att)) {
emit_op1(p, OPCODE_LIT, lit, 0, dots);
emit_op2(p, OPCODE_MUL, lit, 0, lit, att);
emit_op3(p, OPCODE_MAD, _bfc0, 0, swizzle1(lit,X), ambient, _bfc0);
}
else if (!p->state->material_shininess_is_zero) {
emit_op1(p, OPCODE_LIT, lit, 0, dots);
emit_op2(p, OPCODE_ADD, _bfc0, 0, ambient, _bfc0); /**/
}
else {
emit_degenerate_lit(p, lit, dots);
emit_op2(p, OPCODE_ADD, _bfc0, 0, ambient, _bfc0);
}
emit_op3(p, OPCODE_MAD, res0, mask0, swizzle1(lit,Y), diffuse, _bfc0);
emit_op3(p, OPCODE_MAD, res1, mask1, swizzle1(lit,Z), specular, _bfc1);
/* restore dots to its original state for subsequent lights
* by negating and swizzling again.
*/
dots = negate(swizzle(dots,X,Y,W,Z));
release_temp(p, ambient);
release_temp(p, diffuse);
release_temp(p, specular);
}
release_temp(p, half);
release_temp(p, VPpli);
release_temp(p, att);
}
}
release_temps( p );
}
/**
* Some TEX instructions require extra code, cube map coordinate
* normalization, or coordinate scaling for RECT textures, etc.
* This function emits those extra instructions and the TEX
* instruction itself.
*/
static void precalc_tex( struct brw_wm_compile *c,
struct brw_fp_dst dst,
unsigned target,
unsigned unit,
struct brw_fp_src src0,
struct brw_fp_src sampler )
{
struct brw_fp_src coord;
struct brw_fp_dst tmp = dst_undef();
assert(unit < BRW_MAX_TEX_UNIT);
/* Cubemap: find longest component of coord vector and normalize
* it.
*/
if (target == TGSI_TEXTURE_CUBE) {
struct brw_fp_src tmpsrc;
tmp = get_temp(c);
tmpsrc = src_reg_from_dst(tmp);
/* tmp = abs(src0) */
emit_op1(c,
TGSI_OPCODE_MOV,
tmp,
src_abs(src0));
/* tmp.X = MAX(tmp.X, tmp.Y) */
emit_op2(c, TGSI_OPCODE_MAX,
dst_mask(tmp, BRW_WRITEMASK_X),
src_scalar(tmpsrc, X),
src_scalar(tmpsrc, Y));
/* tmp.X = MAX(tmp.X, tmp.Z) */
emit_op2(c, TGSI_OPCODE_MAX,
dst_mask(tmp, BRW_WRITEMASK_X),
tmpsrc,
src_scalar(tmpsrc, Z));
/* tmp.X = 1 / tmp.X */
emit_op1(c, TGSI_OPCODE_RCP,
dst_mask(tmp, BRW_WRITEMASK_X),
tmpsrc);
/* tmp = src0 * tmp.xxxx */
emit_op2(c, TGSI_OPCODE_MUL,
tmp,
src0,
src_scalar(tmpsrc, X));
coord = tmpsrc;
}
else if (target == TGSI_TEXTURE_RECT ||
target == TGSI_TEXTURE_SHADOWRECT) {
/* XXX: need a mechanism for internally generated constants.
*/
coord = src0;
}
else {
coord = src0;
}
/* Need to emit YUV texture conversions by hand. Probably need to
* do this here - the alternative is in brw_wm_emit.c, but the
* conversion requires allocating a temporary variable which we
* don't have the facility to do that late in the compilation.
*/
if (c->key.yuvtex_mask & (1 << unit)) {
/* convert ycbcr to RGBA */
GLboolean swap_uv = c->key.yuvtex_swap_mask & (1<<unit);
struct brw_fp_dst tmp = get_temp(c);
struct brw_fp_src tmpsrc = src_reg_from_dst(tmp);
struct brw_fp_src C0 = src_imm4f( c, -.5, -.0625, -.5, 1.164 );
struct brw_fp_src C1 = src_imm4f( c, 1.596, -0.813, 2.018, -.391 );
/* tmp = TEX ...
*/
emit_tex_op(c,
TGSI_OPCODE_TEX,
dst_saturate(tmp, dst.saturate),
unit,
target,
sampler.index,
coord,
src_undef(),
src_undef());
/* tmp.xyz = ADD TMP, C0
*/
emit_op2(c, TGSI_OPCODE_ADD,
dst_mask(tmp, BRW_WRITEMASK_XYZ),
tmpsrc,
C0);
/* YUV.y = MUL YUV.y, C0.w
*/
emit_op2(c, TGSI_OPCODE_MUL,
dst_mask(tmp, BRW_WRITEMASK_Y),
tmpsrc,
src_scalar(C0, W));
/*
* if (UV swaped)
* RGB.xyz = MAD YUV.zzx, C1, YUV.y
* else
* RGB.xyz = MAD YUV.xxz, C1, YUV.y
*/
emit_op3(c, TGSI_OPCODE_MAD,
dst_mask(dst, BRW_WRITEMASK_XYZ),
( swap_uv ?
src_swizzle(tmpsrc, Z,Z,X,X) :
src_swizzle(tmpsrc, X,X,Z,Z)),
C1,
src_scalar(tmpsrc, Y));
/* RGB.y = MAD YUV.z, C1.w, RGB.y
*/
emit_op3(c,
TGSI_OPCODE_MAD,
dst_mask(dst, BRW_WRITEMASK_Y),
src_scalar(tmpsrc, Z),
src_scalar(C1, W),
src_scalar(src_reg_from_dst(dst), Y));
release_temp(c, tmp);
}
else {
/* ordinary RGBA tex instruction */
emit_tex_op(c,
TGSI_OPCODE_TEX,
dst,
unit,
target,
sampler.index,
coord,
src_undef(),
src_undef());
}
/* XXX: add GL_EXT_texture_swizzle support to gallium -- by
* generating shader variants in mesa state tracker.
*/
/* Release this temp if we ended up allocating it:
*/
if (!dst_is_undef(tmp))
release_temp(c, tmp);
}
/**
* Some TEX instructions require extra code, cube map coordinate
* normalization, or coordinate scaling for RECT textures, etc.
* This function emits those extra instructions and the TEX
* instruction itself.
*/
static void precalc_tex( struct brw_wm_compile *c,
const struct prog_instruction *inst )
{
struct prog_src_register coord;
struct prog_dst_register tmpcoord;
const GLuint unit = c->fp->program.Base.SamplerUnits[inst->TexSrcUnit];
assert(unit < BRW_MAX_TEX_UNIT);
if (inst->TexSrcTarget == TEXTURE_CUBE_INDEX) {
struct prog_instruction *out;
struct prog_dst_register tmp0 = get_temp(c);
struct prog_src_register tmp0src = src_reg_from_dst(tmp0);
struct prog_dst_register tmp1 = get_temp(c);
struct prog_src_register tmp1src = src_reg_from_dst(tmp1);
struct prog_src_register src0 = inst->SrcReg[0];
/* find longest component of coord vector and normalize it */
tmpcoord = get_temp(c);
coord = src_reg_from_dst(tmpcoord);
/* tmpcoord = src0 (i.e.: coord = src0) */
out = emit_op(c, OPCODE_MOV,
tmpcoord,
0,
src0,
src_undef(),
src_undef());
out->SrcReg[0].Negate = NEGATE_NONE;
out->SrcReg[0].Abs = 1;
/* tmp0 = MAX(coord.X, coord.Y) */
emit_op(c, OPCODE_MAX,
tmp0,
0,
src_swizzle1(coord, X),
src_swizzle1(coord, Y),
src_undef());
/* tmp1 = MAX(tmp0, coord.Z) */
emit_op(c, OPCODE_MAX,
tmp1,
0,
tmp0src,
src_swizzle1(coord, Z),
src_undef());
/* tmp0 = 1 / tmp1 */
emit_op(c, OPCODE_RCP,
dst_mask(tmp0, WRITEMASK_X),
0,
tmp1src,
src_undef(),
src_undef());
/* tmpCoord = src0 * tmp0 */
emit_op(c, OPCODE_MUL,
tmpcoord,
0,
src0,
src_swizzle1(tmp0src, SWIZZLE_X),
src_undef());
release_temp(c, tmp0);
release_temp(c, tmp1);
}
else if (inst->TexSrcTarget == TEXTURE_RECT_INDEX) {
struct prog_src_register scale =
search_or_add_param5( c,
STATE_INTERNAL,
STATE_TEXRECT_SCALE,
unit,
0,0 );
tmpcoord = get_temp(c);
/* coord.xy = MUL inst->SrcReg[0], { 1/width, 1/height }
*/
emit_op(c,
OPCODE_MUL,
tmpcoord,
0,
inst->SrcReg[0],
src_swizzle(scale,
SWIZZLE_X,
SWIZZLE_Y,
SWIZZLE_ONE,
SWIZZLE_ONE),
src_undef());
coord = src_reg_from_dst(tmpcoord);
}
else {
coord = inst->SrcReg[0];
}
/* Need to emit YUV texture conversions by hand. Probably need to
* do this here - the alternative is in brw_wm_emit.c, but the
* conversion requires allocating a temporary variable which we
* don't have the facility to do that late in the compilation.
*/
if (c->key.yuvtex_mask & (1 << unit)) {
/* convert ycbcr to RGBA */
GLboolean swap_uv = c->key.yuvtex_swap_mask & (1<<unit);
/*
CONST C0 = { -.5, -.0625, -.5, 1.164 }
CONST C1 = { 1.596, -0.813, 2.018, -.391 }
UYV = TEX ...
UYV.xyz = ADD UYV, C0
UYV.y = MUL UYV.y, C0.w
if (UV swaped)
RGB.xyz = MAD UYV.zzx, C1, UYV.y
else
RGB.xyz = MAD UYV.xxz, C1, UYV.y
RGB.y = MAD UYV.z, C1.w, RGB.y
*/
struct prog_dst_register dst = inst->DstReg;
struct prog_dst_register tmp = get_temp(c);
struct prog_src_register tmpsrc = src_reg_from_dst(tmp);
struct prog_src_register C0 = search_or_add_const4f( c, -.5, -.0625, -.5, 1.164 );
struct prog_src_register C1 = search_or_add_const4f( c, 1.596, -0.813, 2.018, -.391 );
/* tmp = TEX ...
*/
emit_tex_op(c,
OPCODE_TEX,
tmp,
inst->SaturateMode,
unit,
inst->TexSrcTarget,
inst->TexShadow,
coord,
src_undef(),
src_undef());
/* tmp.xyz = ADD TMP, C0
*/
emit_op(c,
OPCODE_ADD,
dst_mask(tmp, WRITEMASK_XYZ),
0,
tmpsrc,
C0,
src_undef());
/* YUV.y = MUL YUV.y, C0.w
*/
emit_op(c,
OPCODE_MUL,
dst_mask(tmp, WRITEMASK_Y),
0,
tmpsrc,
src_swizzle1(C0, W),
src_undef());
/*
* if (UV swaped)
* RGB.xyz = MAD YUV.zzx, C1, YUV.y
* else
* RGB.xyz = MAD YUV.xxz, C1, YUV.y
*/
emit_op(c,
OPCODE_MAD,
dst_mask(dst, WRITEMASK_XYZ),
0,
swap_uv?src_swizzle(tmpsrc, Z,Z,X,X):src_swizzle(tmpsrc, X,X,Z,Z),
C1,
src_swizzle1(tmpsrc, Y));
/* RGB.y = MAD YUV.z, C1.w, RGB.y
*/
emit_op(c,
OPCODE_MAD,
dst_mask(dst, WRITEMASK_Y),
0,
src_swizzle1(tmpsrc, Z),
src_swizzle1(C1, W),
src_swizzle1(src_reg_from_dst(dst), Y));
release_temp(c, tmp);
}
else {
/* ordinary RGBA tex instruction */
emit_tex_op(c,
OPCODE_TEX,
inst->DstReg,
inst->SaturateMode,
unit,
inst->TexSrcTarget,
inst->TexShadow,
coord,
src_undef(),
src_undef());
}
/* For GL_EXT_texture_swizzle: */
if (c->key.tex_swizzles[unit] != SWIZZLE_NOOP) {
/* swizzle the result of the TEX instruction */
struct prog_src_register tmpsrc = src_reg_from_dst(inst->DstReg);
emit_op(c, OPCODE_SWZ,
inst->DstReg,
SATURATE_OFF, /* saturate already done above */
src_swizzle4(tmpsrc, c->key.tex_swizzles[unit]),
src_undef(),
src_undef());
}
if ((inst->TexSrcTarget == TEXTURE_RECT_INDEX) ||
(inst->TexSrcTarget == TEXTURE_CUBE_INDEX))
release_temp(c, tmpcoord);
}
static lua_Number ljc_relational( lua_Number st, lua_Number sv
, lua_Number tt, lua_Number tv
, int op ) {
assert( !( st == LUA_TNUMBER && tt == LUA_TNUMBER ) );
struct TValue s = { .t = st, .v = (union Value)sv };
struct TValue t = { .t = tt, .v = (union Value)tv };
switch( op ){
case REL_LT:
return do_lt( &s, &t );
case REL_LEQ:
return do_leq( &s, &t );
case REL_EQ:
return do_eq( &s, &t );
default:
assert( false );
}
}
typedef void (*arch_rel)( struct emitter*, struct machine*
, operand, operand, label );
static void emit_relational( struct emitter *me, struct machine_ops *mop
, struct frame* f
, loperand s, loperand t
, arch_rel ar, int op
, bool expect ){
vreg_operand os = loperand_to_operand( f, s ),
ot = loperand_to_operand( f, t );
unsigned int pc = me->ops->pc( me ) + 2;
label l = LBL_PC( pc );
// determine if coercion is required
operand tag = OP_TARGETREG( acquire_temp( mop, me, f->m ) );
mop->bor( me, f->m, tag, os.type, ot.type );
mop->beq( me, f->m, tag, OP_TARGETIMMED( 0 ), LBL_NEXT( 0 ) );
// do coercion
mop->call_static_cfn( me, f, (uintptr_t)&ljc_relational
, &tag, 5, os.type, os.value
, ot.type, ot.value
, OP_TARGETIMMED( op ) );
mop->beq( me, f->m, tag, OP_TARGETIMMED( expect ), l );
mop->b( me, f->m, LBL_NEXT( 1 ) );
// do primitive relational
me->ops->label_local( me, 0 );
ar( me, f->m, os.value, ot.value, l );
me->ops->label_local( me, 1 );
release_temp( mop, me, f->m );
return;
}
void emit_jmp( struct emitter** mce, struct machine_ops* mop
, struct frame *f
, loperand a
, int offset ){
assert( a.islocal );
// if not zero then any upvalues below the vreg need to be closed.
if( a.index > 0 ){
vreg_operand op = vreg_to_operand( f, a.index + 1, true );
operand base = OP_TARGETREG( acquire_temp( mop, REF, f->m ) );
address_of( mop, REF, f->m, base, op.type );
mop->call_static_cfn( REF, f, (uintptr_t)&closure_close, NULL
, 1
, base );
release_temp( mop, REF, f->m );
}
unsigned int pc = (int)REF->ops->pc( REF ) + offset + 1;
mop->b( REF, f->m, LBL_PC( pc ) );
}
static void precalc_tex( struct brw_wm_compile *c,
const struct prog_instruction *inst )
{
struct prog_src_register coord;
struct prog_dst_register tmpcoord;
if (inst->TexSrcTarget == TEXTURE_RECT_INDEX) {
struct prog_src_register scale =
search_or_add_param5( c,
STATE_INTERNAL,
STATE_TEXRECT_SCALE,
inst->TexSrcUnit,
0,0 );
tmpcoord = get_temp(c);
/* coord.xy = MUL inst->SrcReg[0], { 1/width, 1/height }
*/
emit_op(c,
OPCODE_MUL,
tmpcoord,
0, 0, 0,
inst->SrcReg[0],
scale,
src_undef());
coord = src_reg_from_dst(tmpcoord);
}
else {
coord = inst->SrcReg[0];
}
/* Need to emit YUV texture conversions by hand. Probably need to
* do this here - the alternative is in brw_wm_emit.c, but the
* conversion requires allocating a temporary variable which we
* don't have the facility to do that late in the compilation.
*/
if (!(c->key.yuvtex_mask & (1<<inst->TexSrcUnit))) {
emit_op(c,
OPCODE_TEX,
inst->DstReg,
inst->SaturateMode,
inst->TexSrcUnit,
inst->TexSrcTarget,
coord,
src_undef(),
src_undef());
}
else {
/*
CONST C0 = { -.5, -.0625, -.5, 1.164 }
CONST C1 = { 1.596, -0.813, 2.018, -.391 }
UYV = TEX ...
UYV.xyz = ADD UYV, C0
UYV.y = MUL UYV.y, C0.w
RGB.xyz = MAD UYV.xxz, C1, UYV.y
RGB.y = MAD UYV.z, C1.w, RGB.y
*/
struct prog_dst_register dst = inst->DstReg;
struct prog_src_register src0 = inst->SrcReg[0];
struct prog_dst_register tmp = get_temp(c);
struct prog_src_register tmpsrc = src_reg_from_dst(tmp);
struct prog_src_register C0 = search_or_add_const4f( c, -.5, -.0625, -.5, 1.164 );
struct prog_src_register C1 = search_or_add_const4f( c, 1.596, -0.813, 2.018, -.391 );
/* tmp = TEX ...
*/
emit_op(c,
OPCODE_TEX,
tmp,
inst->SaturateMode,
inst->TexSrcUnit,
inst->TexSrcTarget,
src0,
src_undef(),
src_undef());
/* tmp.xyz = ADD TMP, C0
*/
emit_op(c,
OPCODE_ADD,
dst_mask(tmp, WRITEMASK_XYZ),
0, 0, 0,
tmpsrc,
C0,
src_undef());
/* YUV.y = MUL YUV.y, C0.w
*/
emit_op(c,
OPCODE_MUL,
dst_mask(tmp, WRITEMASK_Y),
0, 0, 0,
tmpsrc,
src_swizzle1(C0, W),
src_undef());
/* RGB.xyz = MAD YUV.xxz, C1, YUV.y
*/
emit_op(c,
OPCODE_MAD,
dst_mask(dst, WRITEMASK_XYZ),
0, 0, 0,
src_swizzle(tmpsrc, X,X,Z,Z),
C1,
src_swizzle1(tmpsrc, Y));
/* RGB.y = MAD YUV.z, C1.w, RGB.y
*/
emit_op(c,
OPCODE_MAD,
dst_mask(dst, WRITEMASK_Y),
0, 0, 0,
src_swizzle1(tmpsrc, Z),
src_swizzle1(C1, W),
src_swizzle1(src_reg_from_dst(dst), Y));
release_temp(c, tmp);
}
if (inst->TexSrcTarget == GL_TEXTURE_RECTANGLE_NV)
release_temp(c, tmpcoord);
}
/* Need to add some addtional parameters to allow lighting in object
* space - STATE_SPOT_DIRECTION and STATE_HALF_VECTOR implicitly assume eye
* space lighting.
*/
static void build_lighting( struct tnl_program *p )
{
const GLboolean twoside = p->state->light_twoside;
const GLboolean separate = p->state->separate_specular;
GLuint nr_lights = 0, count = 0;
struct ureg normal = get_eye_normal(p);
struct ureg lit = get_temp(p);
struct ureg dots = get_temp(p);
struct ureg _col0 = undef, _col1 = undef;
struct ureg _bfc0 = undef, _bfc1 = undef;
GLuint i;
for (i = 0; i < MAX_LIGHTS; i++)
if (p->state->unit[i].light_enabled)
nr_lights++;
set_material_flags(p);
{
struct ureg shininess = get_material(p, 0, STATE_SHININESS);
emit_op1(p, OPCODE_MOV, dots, WRITEMASK_W, swizzle1(shininess,X));
release_temp(p, shininess);
_col0 = make_temp(p, get_scenecolor(p, 0));
if (separate)
_col1 = make_temp(p, get_identity_param(p));
else
_col1 = _col0;
}
if (twoside) {
struct ureg shininess = get_material(p, 1, STATE_SHININESS);
emit_op1(p, OPCODE_MOV, dots, WRITEMASK_Z,
ureg_negate(swizzle1(shininess,X)));
release_temp(p, shininess);
_bfc0 = make_temp(p, get_scenecolor(p, 1));
if (separate)
_bfc1 = make_temp(p, get_identity_param(p));
else
_bfc1 = _bfc0;
}
/* If no lights, still need to emit the scenecolor.
*/
/* KW: changed to do this always - v1.17 "Fix lighting alpha result"?
*/
if (p->state->fragprog_inputs_read & FRAG_BIT_COL0)
{
struct ureg res0 = register_output( p, VERT_RESULT_COL0 );
emit_op1(p, OPCODE_MOV, res0, 0, _col0);
if (twoside) {
struct ureg res0 = register_output( p, VERT_RESULT_BFC0 );
emit_op1(p, OPCODE_MOV, res0, 0, _bfc0);
}
}
if (separate && (p->state->fragprog_inputs_read & FRAG_BIT_COL1)) {
struct ureg res1 = register_output( p, VERT_RESULT_COL1 );
emit_op1(p, OPCODE_MOV, res1, 0, _col1);
if (twoside) {
struct ureg res1 = register_output( p, VERT_RESULT_BFC1 );
emit_op1(p, OPCODE_MOV, res1, 0, _bfc1);
}
}
if (nr_lights == 0) {
release_temps(p);
return;
}
for (i = 0; i < MAX_LIGHTS; i++) {
if (p->state->unit[i].light_enabled) {
struct ureg half = undef;
struct ureg att = undef, VPpli = undef;
count++;
if (p->state->unit[i].light_eyepos3_is_zero) {
/* Can used precomputed constants in this case.
* Attenuation never applies to infinite lights.
*/
VPpli = register_param3(p, STATE_LIGHT, i,
STATE_POSITION_NORMALIZED);
if (p->state->light_local_viewer) {
struct ureg eye_hat = get_eye_position_normalized(p);
half = get_temp(p);
emit_op2(p, OPCODE_SUB, half, 0, VPpli, eye_hat);
emit_normalize_vec3(p, half, half);
} else {
half = register_param3(p, STATE_LIGHT, i, STATE_HALF_VECTOR);
}
}
else {
struct ureg Ppli = register_param3(p, STATE_LIGHT, i,
STATE_POSITION);
struct ureg V = get_eye_position(p);
struct ureg dist = get_temp(p);
VPpli = get_temp(p);
half = get_temp(p);
/* Calulate VPpli vector
*/
emit_op2(p, OPCODE_SUB, VPpli, 0, Ppli, V);
/* Normalize VPpli. The dist value also used in
* attenuation below.
*/
emit_op2(p, OPCODE_DP3, dist, 0, VPpli, VPpli);
emit_op1(p, OPCODE_RSQ, dist, 0, dist);
emit_op2(p, OPCODE_MUL, VPpli, 0, VPpli, dist);
/* Calculate attenuation:
*/
if (!p->state->unit[i].light_spotcutoff_is_180 ||
p->state->unit[i].light_attenuated) {
att = calculate_light_attenuation(p, i, VPpli, dist);
}
/* Calculate viewer direction, or use infinite viewer:
*/
if (p->state->light_local_viewer) {
struct ureg eye_hat = get_eye_position_normalized(p);
emit_op2(p, OPCODE_SUB, half, 0, VPpli, eye_hat);
}
else {
struct ureg z_dir = swizzle(get_identity_param(p),X,Y,W,Z);
emit_op2(p, OPCODE_ADD, half, 0, VPpli, z_dir);
}
emit_normalize_vec3(p, half, half);
release_temp(p, dist);
}
/* Calculate dot products:
*/
emit_op2(p, OPCODE_DP3, dots, WRITEMASK_X, normal, VPpli);
emit_op2(p, OPCODE_DP3, dots, WRITEMASK_Y, normal, half);
/* Front face lighting:
*/
{
struct ureg ambient = get_lightprod(p, i, 0, STATE_AMBIENT);
struct ureg diffuse = get_lightprod(p, i, 0, STATE_DIFFUSE);
struct ureg specular = get_lightprod(p, i, 0, STATE_SPECULAR);
struct ureg res0, res1;
GLuint mask0, mask1;
emit_op1(p, OPCODE_LIT, lit, 0, dots);
if (!is_undef(att))
emit_op2(p, OPCODE_MUL, lit, 0, lit, att);
mask0 = 0;
mask1 = 0;
res0 = _col0;
res1 = _col1;
if (count == nr_lights) {
if (separate) {
mask0 = WRITEMASK_XYZ;
mask1 = WRITEMASK_XYZ;
if (p->state->fragprog_inputs_read & FRAG_BIT_COL0)
res0 = register_output( p, VERT_RESULT_COL0 );
if (p->state->fragprog_inputs_read & FRAG_BIT_COL1)
res1 = register_output( p, VERT_RESULT_COL1 );
}
else {
mask1 = WRITEMASK_XYZ;
if (p->state->fragprog_inputs_read & FRAG_BIT_COL0)
res1 = register_output( p, VERT_RESULT_COL0 );
}
}
emit_op3(p, OPCODE_MAD, _col0, 0, swizzle1(lit,X), ambient, _col0);
emit_op3(p, OPCODE_MAD, res0, mask0, swizzle1(lit,Y), diffuse, _col0);
emit_op3(p, OPCODE_MAD, res1, mask1, swizzle1(lit,Z), specular, _col1);
release_temp(p, ambient);
release_temp(p, diffuse);
release_temp(p, specular);
}
/* Back face lighting:
*/
if (twoside) {
struct ureg ambient = get_lightprod(p, i, 1, STATE_AMBIENT);
struct ureg diffuse = get_lightprod(p, i, 1, STATE_DIFFUSE);
struct ureg specular = get_lightprod(p, i, 1, STATE_SPECULAR);
struct ureg res0, res1;
GLuint mask0, mask1;
emit_op1(p, OPCODE_LIT, lit, 0, ureg_negate(swizzle(dots,X,Y,W,Z)));
if (!is_undef(att))
emit_op2(p, OPCODE_MUL, lit, 0, lit, att);
mask0 = 0;
mask1 = 0;
res0 = _bfc0;
res1 = _bfc1;
if (count == nr_lights) {
if (separate) {
mask0 = WRITEMASK_XYZ;
mask1 = WRITEMASK_XYZ;
if (p->state->fragprog_inputs_read & FRAG_BIT_COL0)
res0 = register_output( p, VERT_RESULT_BFC0 );
if (p->state->fragprog_inputs_read & FRAG_BIT_COL1)
res1 = register_output( p, VERT_RESULT_BFC1 );
}
else {
mask1 = WRITEMASK_XYZ;
if (p->state->fragprog_inputs_read & FRAG_BIT_COL0)
res1 = register_output( p, VERT_RESULT_BFC0 );
}
}
emit_op3(p, OPCODE_MAD, _bfc0, 0, swizzle1(lit,X), ambient, _bfc0);
emit_op3(p, OPCODE_MAD, res0, mask0, swizzle1(lit,Y), diffuse, _bfc0);
emit_op3(p, OPCODE_MAD, res1, mask1, swizzle1(lit,Z), specular, _bfc1);
release_temp(p, ambient);
release_temp(p, diffuse);
release_temp(p, specular);
}
release_temp(p, half);
release_temp(p, VPpli);
release_temp(p, att);
}
}
release_temps( p );
}
