void *
util_make_fs_msaa_resolve(struct pipe_context *pipe,
enum tgsi_texture_type tgsi_tex, unsigned nr_samples,
enum tgsi_return_type stype)
{
struct ureg_program *ureg;
struct ureg_src sampler, coord;
struct ureg_dst out, tmp_sum, tmp_coord, tmp;
unsigned i;
ureg = ureg_create(PIPE_SHADER_FRAGMENT);
if (!ureg)
return NULL;
/* Declarations. */
sampler = ureg_DECL_sampler(ureg, 0);
ureg_DECL_sampler_view(ureg, 0, tgsi_tex, stype, stype, stype, stype);
coord = ureg_DECL_fs_input(ureg, TGSI_SEMANTIC_GENERIC, 0,
TGSI_INTERPOLATE_LINEAR);
out = ureg_DECL_output(ureg, TGSI_SEMANTIC_COLOR, 0);
tmp_sum = ureg_DECL_temporary(ureg);
tmp_coord = ureg_DECL_temporary(ureg);
tmp = ureg_DECL_temporary(ureg);
/* Instructions. */
ureg_MOV(ureg, tmp_sum, ureg_imm1f(ureg, 0));
ureg_F2U(ureg, tmp_coord, coord);
for (i = 0; i < nr_samples; i++) {
/* Read one sample. */
ureg_MOV(ureg, ureg_writemask(tmp_coord, TGSI_WRITEMASK_W),
ureg_imm1u(ureg, i));
ureg_TXF(ureg, tmp, tgsi_tex, ureg_src(tmp_coord), sampler);
if (stype == TGSI_RETURN_TYPE_UINT)
ureg_U2F(ureg, tmp, ureg_src(tmp));
else if (stype == TGSI_RETURN_TYPE_SINT)
ureg_I2F(ureg, tmp, ureg_src(tmp));
/* Add it to the sum.*/
ureg_ADD(ureg, tmp_sum, ureg_src(tmp_sum), ureg_src(tmp));
}
/* Calculate the average and return. */
ureg_MUL(ureg, tmp_sum, ureg_src(tmp_sum),
ureg_imm1f(ureg, 1.0 / nr_samples));
if (stype == TGSI_RETURN_TYPE_UINT)
ureg_F2U(ureg, out, ureg_src(tmp_sum));
else if (stype == TGSI_RETURN_TYPE_SINT)
ureg_F2I(ureg, out, ureg_src(tmp_sum));
else
ureg_MOV(ureg, out, ureg_src(tmp_sum));
ureg_END(ureg);
return ureg_create_shader_and_destroy(ureg, pipe);
}
static struct ureg_src
prepare_argument(struct st_translate *t, const unsigned argId,
const struct atifragshader_src_register *srcReg)
{
struct ureg_src src = get_source(t, srcReg->Index);
struct ureg_dst arg = get_temp(t, MAX_NUM_FRAGMENT_REGISTERS_ATI + argId);
switch (srcReg->argRep) {
case GL_NONE:
break;
case GL_RED:
src = ureg_scalar(src, TGSI_SWIZZLE_X);
break;
case GL_GREEN:
src = ureg_scalar(src, TGSI_SWIZZLE_Y);
break;
case GL_BLUE:
src = ureg_scalar(src, TGSI_SWIZZLE_Z);
break;
case GL_ALPHA:
src = ureg_scalar(src, TGSI_SWIZZLE_W);
break;
}
ureg_insn(t->ureg, TGSI_OPCODE_MOV, &arg, 1, &src, 1);
if (srcReg->argMod & GL_COMP_BIT_ATI) {
struct ureg_src modsrc[2];
modsrc[0] = ureg_imm1f(t->ureg, 1.0f);
modsrc[1] = ureg_src(arg);
ureg_insn(t->ureg, TGSI_OPCODE_SUB, &arg, 1, modsrc, 2);
}
if (srcReg->argMod & GL_BIAS_BIT_ATI) {
struct ureg_src modsrc[2];
modsrc[0] = ureg_src(arg);
modsrc[1] = ureg_imm1f(t->ureg, 0.5f);
ureg_insn(t->ureg, TGSI_OPCODE_SUB, &arg, 1, modsrc, 2);
}
if (srcReg->argMod & GL_2X_BIT_ATI) {
struct ureg_src modsrc[2];
modsrc[0] = ureg_src(arg);
modsrc[1] = ureg_src(arg);
ureg_insn(t->ureg, TGSI_OPCODE_ADD, &arg, 1, modsrc, 2);
}
if (srcReg->argMod & GL_NEGATE_BIT_ATI) {
struct ureg_src modsrc[2];
modsrc[0] = ureg_src(arg);
modsrc[1] = ureg_imm1f(t->ureg, -1.0f);
ureg_insn(t->ureg, TGSI_OPCODE_MUL, &arg, 1, modsrc, 2);
}
return ureg_src(arg);
}
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_dst
calc_position(struct vl_mc *r, struct ureg_program *shader, struct ureg_src block_scale)
{
struct ureg_src vrect, vpos;
struct ureg_dst t_vpos;
struct ureg_dst o_vpos;
vrect = ureg_DECL_vs_input(shader, VS_I_RECT);
vpos = ureg_DECL_vs_input(shader, VS_I_VPOS);
t_vpos = ureg_DECL_temporary(shader);
o_vpos = ureg_DECL_output(shader, TGSI_SEMANTIC_POSITION, VS_O_VPOS);
/*
* block_scale = (VL_MACROBLOCK_WIDTH, VL_MACROBLOCK_HEIGHT) / (dst.width, dst.height)
*
* t_vpos = (vpos + vrect) * block_scale
* o_vpos.xy = t_vpos
* o_vpos.zw = vpos
*/
ureg_ADD(shader, ureg_writemask(t_vpos, TGSI_WRITEMASK_XY), vpos, vrect);
ureg_MUL(shader, ureg_writemask(t_vpos, TGSI_WRITEMASK_XY), ureg_src(t_vpos), block_scale);
ureg_MOV(shader, ureg_writemask(o_vpos, TGSI_WRITEMASK_XY), ureg_src(t_vpos));
ureg_MOV(shader, ureg_writemask(o_vpos, TGSI_WRITEMASK_ZW), ureg_imm1f(shader, 1.0f));
return t_vpos;
}
/* These instructions need special treatment */
static void
emit_special_inst(struct st_translate *t, const struct instruction_desc *desc,
struct ureg_dst *dst, struct ureg_src *args, unsigned argcount)
{
struct ureg_dst tmp[1];
struct ureg_src src[3];
if (!strcmp(desc->name, "CND")) {
tmp[0] = get_temp(t, MAX_NUM_FRAGMENT_REGISTERS_ATI + 2); /* re-purpose a3 */
src[0] = ureg_imm1f(t->ureg, 0.5f);
src[1] = args[2];
ureg_insn(t->ureg, TGSI_OPCODE_SUB, tmp, 1, src, 2);
src[0] = ureg_src(tmp[0]);
src[1] = args[0];
src[2] = args[1];
ureg_insn(t->ureg, TGSI_OPCODE_CMP, dst, 1, src, 3);
} else if (!strcmp(desc->name, "CND0")) {
src[0] = args[2];
src[1] = args[1];
src[2] = args[0];
ureg_insn(t->ureg, TGSI_OPCODE_CMP, dst, 1, src, 3);
} else if (!strcmp(desc->name, "DOT2_ADD")) {
/* note: DP2A is not implemented in most pipe drivers */
tmp[0] = get_temp(t, MAX_NUM_FRAGMENT_REGISTERS_ATI); /* re-purpose a1 */
src[0] = args[0];
src[1] = args[1];
ureg_insn(t->ureg, TGSI_OPCODE_DP2, tmp, 1, src, 2);
src[0] = ureg_src(tmp[0]);
src[1] = ureg_scalar(args[2], TGSI_SWIZZLE_Z);
ureg_insn(t->ureg, TGSI_OPCODE_ADD, dst, 1, src, 2);
}
}
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);
}
static void
increment_addr(struct ureg_program *shader, struct ureg_dst daddr[2],
struct ureg_src saddr[2], bool right_side, bool transposed,
int pos, float size)
{
unsigned wm_start = (right_side == transposed) ? TGSI_WRITEMASK_X : TGSI_WRITEMASK_Y;
unsigned wm_tc = (right_side == transposed) ? TGSI_WRITEMASK_Y : TGSI_WRITEMASK_X;
/*
* daddr[0..1].(start) = saddr[0..1].(start)
* daddr[0..1].(tc) = saddr[0..1].(tc)
*/
ureg_MOV(shader, ureg_writemask(daddr[0], wm_start), saddr[0]);
ureg_ADD(shader, ureg_writemask(daddr[0], wm_tc), saddr[0], ureg_imm1f(shader, pos / size));
ureg_MOV(shader, ureg_writemask(daddr[1], wm_start), saddr[1]);
ureg_ADD(shader, ureg_writemask(daddr[1], wm_tc), saddr[1], ureg_imm1f(shader, pos / size));
}
static struct ureg_dst
calc_line(struct ureg_program *shader)
{
struct ureg_dst tmp;
struct ureg_src pos;
tmp = ureg_DECL_temporary(shader);
pos = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_POSITION, VS_O_VPOS, TGSI_INTERPOLATE_LINEAR);
/*
* tmp.y = fraction(pos.y / 2) >= 0.5 ? 1 : 0
*/
ureg_MUL(shader, ureg_writemask(tmp, TGSI_WRITEMASK_Y), pos, ureg_imm1f(shader, 0.5f));
ureg_FRC(shader, ureg_writemask(tmp, TGSI_WRITEMASK_Y), ureg_src(tmp));
ureg_SGE(shader, ureg_writemask(tmp, TGSI_WRITEMASK_Y), ureg_src(tmp), ureg_imm1f(shader, 0.5f));
return tmp;
}
static void *
create_frag_shader_video_buffer(struct vl_compositor *c)
{
struct ureg_program *shader;
struct ureg_src tc;
struct ureg_src csc[3];
struct ureg_src sampler[3];
struct ureg_dst texel;
struct ureg_dst fragment;
unsigned i;
shader = ureg_create(TGSI_PROCESSOR_FRAGMENT);
if (!shader)
return false;
tc = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_VTEX, TGSI_INTERPOLATE_LINEAR);
for (i = 0; i < 3; ++i) {
csc[i] = ureg_DECL_constant(shader, i);
sampler[i] = ureg_DECL_sampler(shader, i);
}
texel = ureg_DECL_temporary(shader);
fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
/*
* texel.xyz = tex(tc, sampler[i])
* fragment = csc * texel
*/
for (i = 0; i < 3; ++i)
ureg_TEX(shader, ureg_writemask(texel, TGSI_WRITEMASK_X << i), TGSI_TEXTURE_3D, tc, sampler[i]);
ureg_MOV(shader, ureg_writemask(texel, TGSI_WRITEMASK_W), ureg_imm1f(shader, 1.0f));
for (i = 0; i < 3; ++i)
ureg_DP4(shader, ureg_writemask(fragment, TGSI_WRITEMASK_X << i), csc[i], ureg_src(texel));
ureg_MOV(shader, ureg_writemask(fragment, TGSI_WRITEMASK_W), ureg_imm1f(shader, 1.0f));
ureg_release_temporary(shader, texel);
ureg_END(shader);
return ureg_create_shader_and_destroy(shader, c->pipe);
}
/**
* Compile one arithmetic operation COLOR&ALPHA pair into TGSI instructions.
*/
static void
compile_instruction(struct st_translate *t,
const struct atifs_instruction *inst)
{
unsigned optype;
for (optype = 0; optype < 2; optype++) { /* color, alpha */
const struct instruction_desc *desc;
struct ureg_dst dst[1];
struct ureg_src args[3]; /* arguments for the main operation */
unsigned arg;
unsigned dstreg = inst->DstReg[optype].Index - GL_REG_0_ATI;
if (!inst->Opcode[optype])
continue;
desc = &inst_desc[inst->Opcode[optype] - GL_MOV_ATI];
/* prepare the arguments */
for (arg = 0; arg < desc->arg_count; arg++) {
if (arg >= inst->ArgCount[optype]) {
_mesa_warning(0, "Using 0 for missing argument %d of %s\n",
arg, desc->name);
args[arg] = ureg_imm1f(t->ureg, 0.0f);
} else {
args[arg] = prepare_argument(t, arg,
&inst->SrcReg[optype][arg]);
}
}
/* prepare dst */
dst[0] = get_temp(t, dstreg);
if (optype) {
dst[0] = ureg_writemask(dst[0], TGSI_WRITEMASK_W);
} else {
GLuint dstMask = inst->DstReg[optype].dstMask;
if (dstMask == GL_NONE) {
dst[0] = ureg_writemask(dst[0], TGSI_WRITEMASK_XYZ);
} else {
dst[0] = ureg_writemask(dst[0], dstMask); /* the enum values match */
}
}
/* emit the main instruction */
emit_arith_inst(t, desc, dst, args, arg);
emit_dstmod(t, *dst, inst->DstReg[optype].dstMod);
t->regs_written[t->current_pass][dstreg] = true;
}
}
void
vl_idct_stage2_vert_shader(struct vl_idct *idct, struct ureg_program *shader,
unsigned first_output, struct ureg_dst tex)
{
struct ureg_src vrect, vpos;
struct ureg_src scale;
struct ureg_dst t_start;
struct ureg_dst o_l_addr[2], o_r_addr[2];
vrect = ureg_DECL_vs_input(shader, VS_I_RECT);
vpos = ureg_DECL_vs_input(shader, VS_I_VPOS);
t_start = ureg_DECL_temporary(shader);
--first_output;
o_l_addr[0] = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, first_output + VS_O_L_ADDR0);
o_l_addr[1] = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, first_output + VS_O_L_ADDR1);
o_r_addr[0] = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, first_output + VS_O_R_ADDR0);
o_r_addr[1] = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, first_output + VS_O_R_ADDR1);
scale = ureg_imm2f(shader,
(float)VL_BLOCK_WIDTH / idct->buffer_width,
(float)VL_BLOCK_HEIGHT / idct->buffer_height);
ureg_MUL(shader, ureg_writemask(tex, TGSI_WRITEMASK_Z),
ureg_scalar(vrect, TGSI_SWIZZLE_X),
ureg_imm1f(shader, VL_BLOCK_WIDTH / idct->nr_of_render_targets));
ureg_MUL(shader, ureg_writemask(t_start, TGSI_WRITEMASK_XY), vpos, scale);
calc_addr(shader, o_l_addr, vrect, ureg_imm1f(shader, 0.0f), false, false, VL_BLOCK_WIDTH / 4);
calc_addr(shader, o_r_addr, ureg_src(tex), ureg_src(t_start), true, false, idct->buffer_height / 4);
ureg_MOV(shader, ureg_writemask(o_r_addr[0], TGSI_WRITEMASK_Z), ureg_src(tex));
ureg_MOV(shader, ureg_writemask(o_r_addr[1], TGSI_WRITEMASK_Z), ureg_src(tex));
}
static struct ureg_src
get_source(struct st_translate *t, GLuint src_type)
{
if (src_type >= GL_REG_0_ATI && src_type <= GL_REG_5_ATI) {
if (t->regs_written[t->current_pass][src_type - GL_REG_0_ATI]) {
return ureg_src(get_temp(t, src_type - GL_REG_0_ATI));
} else {
return ureg_imm1f(t->ureg, 0.0f);
}
} else if (src_type >= GL_CON_0_ATI && src_type <= GL_CON_7_ATI) {
return t->constants[src_type - GL_CON_0_ATI];
} else if (src_type == GL_ZERO) {
return ureg_imm1f(t->ureg, 0.0f);
} else if (src_type == GL_ONE) {
return ureg_imm1f(t->ureg, 1.0f);
} else if (src_type == GL_PRIMARY_COLOR_ARB) {
return t->inputs[t->inputMapping[VARYING_SLOT_COL0]];
} else if (src_type == GL_SECONDARY_INTERPOLATOR_ATI) {
return t->inputs[t->inputMapping[VARYING_SLOT_COL1]];
} else {
/* frontend prevents this */
unreachable("unknown source");
}
}
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);
}
/**
* Compile one setup instruction to TGSI instructions.
*/
static void
compile_setupinst(struct st_translate *t,
const unsigned r,
const struct atifs_setupinst *texinst)
{
struct ureg_dst dst[1];
struct ureg_src src[2];
if (!texinst->Opcode)
return;
dst[0] = get_temp(t, r);
GLuint pass_tex = texinst->src;
if (pass_tex >= GL_TEXTURE0_ARB && pass_tex <= GL_TEXTURE7_ARB) {
unsigned attr = pass_tex - GL_TEXTURE0_ARB + VARYING_SLOT_TEX0;
src[0] = t->inputs[t->inputMapping[attr]];
} else if (pass_tex >= GL_REG_0_ATI && pass_tex <= GL_REG_5_ATI) {
unsigned reg = pass_tex - GL_REG_0_ATI;
/* the frontend already validated that REG is only allowed in second pass */
if (t->regs_written[0][reg]) {
src[0] = ureg_src(t->temps[reg]);
} else {
src[0] = ureg_imm1f(t->ureg, 0.0f);
}
}
src[0] = apply_swizzle(t, src[0], texinst->swizzle);
if (texinst->Opcode == ATI_FRAGMENT_SHADER_SAMPLE_OP) {
/* by default texture and sampler indexes are the same */
src[1] = t->samplers[r];
/* the texture target is still unknown, it will be fixed in the draw call */
ureg_tex_insn(t->ureg, TGSI_OPCODE_TEX, dst, 1, TGSI_TEXTURE_2D,
NULL, 0, src, 2);
} else if (texinst->Opcode == ATI_FRAGMENT_SHADER_PASS_OP) {
ureg_insn(t->ureg, TGSI_OPCODE_MOV, dst, 1, src, 1);
}
t->regs_written[t->current_pass][r] = true;
}
static void
emit_dstmod(struct st_translate *t,
struct ureg_dst dst, GLuint dstMod)
{
float imm;
struct ureg_src src[3];
GLuint scale = dstMod & ~GL_SATURATE_BIT_ATI;
if (dstMod == GL_NONE) {
return;
}
switch (scale) {
case GL_2X_BIT_ATI:
imm = 2.0f;
break;
case GL_4X_BIT_ATI:
imm = 4.0f;
break;
case GL_8X_BIT_ATI:
imm = 8.0f;
break;
case GL_HALF_BIT_ATI:
imm = 0.5f;
break;
case GL_QUARTER_BIT_ATI:
imm = 0.25f;
break;
case GL_EIGHTH_BIT_ATI:
imm = 0.125f;
break;
default:
imm = 1.0f;
}
src[0] = ureg_src(dst);
src[1] = ureg_imm1f(t->ureg, imm);
if (dstMod & GL_SATURATE_BIT_ATI) {
dst = ureg_saturate(dst);
}
ureg_insn(t->ureg, TGSI_OPCODE_MUL, &dst, 1, src, 2);
}
static void
calc_addr(struct ureg_program *shader, struct ureg_dst addr[2],
struct ureg_src tc, struct ureg_src start, bool right_side,
bool transposed, float size)
{
unsigned wm_start = (right_side == transposed) ? TGSI_WRITEMASK_X : TGSI_WRITEMASK_Y;
unsigned sw_start = right_side ? TGSI_SWIZZLE_Y : TGSI_SWIZZLE_X;
unsigned wm_tc = (right_side == transposed) ? TGSI_WRITEMASK_Y : TGSI_WRITEMASK_X;
unsigned sw_tc = right_side ? TGSI_SWIZZLE_X : TGSI_SWIZZLE_Y;
/*
* addr[0..1].(start) = right_side ? start.x : tc.x
* addr[0..1].(tc) = right_side ? tc.y : start.y
* addr[0..1].z = tc.z
* addr[1].(start) += 1.0f / scale
*/
ureg_MOV(shader, ureg_writemask(addr[0], wm_start), ureg_scalar(start, sw_start));
ureg_MOV(shader, ureg_writemask(addr[0], wm_tc), ureg_scalar(tc, sw_tc));
ureg_ADD(shader, ureg_writemask(addr[1], wm_start), ureg_scalar(start, sw_start), ureg_imm1f(shader, 1.0f / size));
ureg_MOV(shader, ureg_writemask(addr[1], wm_tc), ureg_scalar(tc, sw_tc));
}
static void
compile_instruction(
struct gl_context *ctx,
struct st_translate *t,
const struct prog_instruction *inst,
boolean clamp_dst_color_output)
{
struct ureg_program *ureg = t->ureg;
GLuint i;
struct ureg_dst dst[1] = { { 0 } };
struct ureg_src src[4];
unsigned num_dst;
unsigned num_src;
num_dst = _mesa_num_inst_dst_regs( inst->Opcode );
num_src = _mesa_num_inst_src_regs( inst->Opcode );
if (num_dst)
dst[0] = translate_dst( t,
&inst->DstReg,
inst->Saturate,
clamp_dst_color_output);
for (i = 0; i < num_src; i++)
src[i] = translate_src( t, &inst->SrcReg[i] );
switch( inst->Opcode ) {
case OPCODE_SWZ:
emit_swz( t, dst[0], &inst->SrcReg[0] );
return;
case OPCODE_BGNLOOP:
case OPCODE_CAL:
case OPCODE_ELSE:
case OPCODE_ENDLOOP:
debug_assert(num_dst == 0);
ureg_label_insn( ureg,
translate_opcode( inst->Opcode ),
src, num_src,
get_label( t, inst->BranchTarget ));
return;
case OPCODE_IF:
debug_assert(num_dst == 0);
ureg_label_insn( ureg,
ctx->Const.NativeIntegers ? TGSI_OPCODE_UIF : TGSI_OPCODE_IF,
src, num_src,
get_label( t, inst->BranchTarget ));
return;
case OPCODE_TEX:
case OPCODE_TXB:
case OPCODE_TXD:
case OPCODE_TXL:
case OPCODE_TXP:
src[num_src++] = t->samplers[inst->TexSrcUnit];
ureg_tex_insn( ureg,
translate_opcode( inst->Opcode ),
dst, num_dst,
st_translate_texture_target( inst->TexSrcTarget,
inst->TexShadow ),
NULL, 0,
src, num_src );
return;
case OPCODE_SCS:
dst[0] = ureg_writemask(dst[0], TGSI_WRITEMASK_XY );
ureg_insn( ureg,
translate_opcode( inst->Opcode ),
dst, num_dst,
src, num_src );
break;
case OPCODE_XPD:
dst[0] = ureg_writemask(dst[0], TGSI_WRITEMASK_XYZ );
ureg_insn( ureg,
translate_opcode( inst->Opcode ),
dst, num_dst,
src, num_src );
break;
case OPCODE_NOISE1:
case OPCODE_NOISE2:
case OPCODE_NOISE3:
case OPCODE_NOISE4:
/* At some point, a motivated person could add a better
* implementation of noise. Currently not even the nvidia
* binary drivers do anything more than this. In any case, the
* place to do this is in the GL state tracker, not the poor
* driver.
*/
ureg_MOV( ureg, dst[0], ureg_imm1f(ureg, 0.5) );
break;
case OPCODE_DDY:
emit_ddy( t, dst[0], &inst->SrcReg[0] );
break;
case OPCODE_RSQ:
ureg_RSQ( ureg, dst[0], ureg_abs(src[0]) );
break;
default:
ureg_insn( ureg,
translate_opcode( inst->Opcode ),
dst, num_dst,
src, num_src );
break;
}
}
static void *
create_frag_shader_weave(struct vl_compositor *c)
{
struct ureg_program *shader;
struct ureg_src i_tc[2];
struct ureg_src csc[3];
struct ureg_src sampler[3];
struct ureg_dst t_tc[2];
struct ureg_dst t_texel[2];
struct ureg_dst o_fragment;
unsigned i, j;
shader = ureg_create(TGSI_PROCESSOR_FRAGMENT);
if (!shader)
return false;
i_tc[0] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_VTOP, TGSI_INTERPOLATE_LINEAR);
i_tc[1] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_VBOTTOM, TGSI_INTERPOLATE_LINEAR);
for (i = 0; i < 3; ++i) {
csc[i] = ureg_DECL_constant(shader, i);
sampler[i] = ureg_DECL_sampler(shader, i);
}
for (i = 0; i < 2; ++i) {
t_tc[i] = ureg_DECL_temporary(shader);
t_texel[i] = ureg_DECL_temporary(shader);
}
o_fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
/* calculate the texture offsets
* t_tc.x = i_tc.x
* t_tc.y = (round(i_tc.y) + 0.5) / height * 2
*/
for (i = 0; i < 2; ++i) {
ureg_MOV(shader, ureg_writemask(t_tc[i], TGSI_WRITEMASK_X), i_tc[i]);
ureg_ROUND(shader, ureg_writemask(t_tc[i], TGSI_WRITEMASK_YZ), i_tc[i]);
ureg_MOV(shader, ureg_writemask(t_tc[i], TGSI_WRITEMASK_W),
ureg_imm1f(shader, i ? 0.75f : 0.25f));
ureg_ADD(shader, ureg_writemask(t_tc[i], TGSI_WRITEMASK_YZ),
ureg_src(t_tc[i]), ureg_imm1f(shader, 0.5f));
ureg_MUL(shader, ureg_writemask(t_tc[i], TGSI_WRITEMASK_Y),
ureg_src(t_tc[i]), ureg_scalar(i_tc[0], TGSI_SWIZZLE_W));
ureg_MUL(shader, ureg_writemask(t_tc[i], TGSI_WRITEMASK_Z),
ureg_src(t_tc[i]), ureg_scalar(i_tc[1], TGSI_SWIZZLE_W));
}
/* fetch the texels
* texel[0..1].x = tex(t_tc[0..1][0])
* texel[0..1].y = tex(t_tc[0..1][1])
* texel[0..1].z = tex(t_tc[0..1][2])
*/
for (i = 0; i < 2; ++i)
for (j = 0; j < 3; ++j) {
struct ureg_src src = ureg_swizzle(ureg_src(t_tc[i]),
TGSI_SWIZZLE_X, j ? TGSI_SWIZZLE_Z : TGSI_SWIZZLE_Y, TGSI_SWIZZLE_W, TGSI_SWIZZLE_W);
ureg_TEX(shader, ureg_writemask(t_texel[i], TGSI_WRITEMASK_X << j),
TGSI_TEXTURE_3D, src, sampler[j]);
}
/* calculate linear interpolation factor
* factor = |round(i_tc.y) - i_tc.y| * 2
*/
ureg_ROUND(shader, ureg_writemask(t_tc[0], TGSI_WRITEMASK_YZ), i_tc[0]);
ureg_ADD(shader, ureg_writemask(t_tc[0], TGSI_WRITEMASK_YZ),
ureg_src(t_tc[0]), ureg_negate(i_tc[0]));
ureg_MUL(shader, ureg_writemask(t_tc[0], TGSI_WRITEMASK_XY),
ureg_abs(ureg_src(t_tc[0])), ureg_imm1f(shader, 2.0f));
ureg_LRP(shader, t_texel[0], ureg_swizzle(ureg_src(t_tc[0]),
TGSI_SWIZZLE_Y, TGSI_SWIZZLE_Z, TGSI_SWIZZLE_Z, TGSI_SWIZZLE_Z),
ureg_src(t_texel[1]), ureg_src(t_texel[0]));
/* and finally do colour space transformation
* fragment = csc * texel
*/
ureg_MOV(shader, ureg_writemask(t_texel[0], TGSI_WRITEMASK_W), ureg_imm1f(shader, 1.0f));
for (i = 0; i < 3; ++i)
ureg_DP4(shader, ureg_writemask(o_fragment, TGSI_WRITEMASK_X << i), csc[i], ureg_src(t_texel[0]));
ureg_MOV(shader, ureg_writemask(o_fragment, TGSI_WRITEMASK_W), ureg_imm1f(shader, 1.0f));
for (i = 0; i < 2; ++i) {
ureg_release_temporary(shader, t_texel[i]);
ureg_release_temporary(shader, t_tc[i]);
}
ureg_END(shader);
return ureg_create_shader_and_destroy(shader, c->pipe);
}
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_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);
}
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);
}
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);
}
void *
util_make_fs_msaa_resolve_bilinear(struct pipe_context *pipe,
enum tgsi_texture_type tgsi_tex,
unsigned nr_samples,
enum tgsi_return_type stype)
{
struct ureg_program *ureg;
struct ureg_src sampler, coord;
struct ureg_dst out, tmp, top, bottom;
struct ureg_dst tmp_coord[4], tmp_sum[4];
unsigned i, c;
ureg = ureg_create(PIPE_SHADER_FRAGMENT);
if (!ureg)
return NULL;
/* Declarations. */
sampler = ureg_DECL_sampler(ureg, 0);
ureg_DECL_sampler_view(ureg, 0, tgsi_tex, stype, stype, stype, stype);
coord = ureg_DECL_fs_input(ureg, TGSI_SEMANTIC_GENERIC, 0,
TGSI_INTERPOLATE_LINEAR);
out = ureg_DECL_output(ureg, TGSI_SEMANTIC_COLOR, 0);
for (c = 0; c < 4; c++)
tmp_sum[c] = ureg_DECL_temporary(ureg);
for (c = 0; c < 4; c++)
tmp_coord[c] = ureg_DECL_temporary(ureg);
tmp = ureg_DECL_temporary(ureg);
top = ureg_DECL_temporary(ureg);
bottom = ureg_DECL_temporary(ureg);
/* Instructions. */
for (c = 0; c < 4; c++)
ureg_MOV(ureg, tmp_sum[c], ureg_imm1f(ureg, 0));
/* Get 4 texture coordinates for the bilinear filter. */
ureg_F2U(ureg, tmp_coord[0], coord); /* top-left */
ureg_UADD(ureg, tmp_coord[1], ureg_src(tmp_coord[0]),
ureg_imm4u(ureg, 1, 0, 0, 0)); /* top-right */
ureg_UADD(ureg, tmp_coord[2], ureg_src(tmp_coord[0]),
ureg_imm4u(ureg, 0, 1, 0, 0)); /* bottom-left */
ureg_UADD(ureg, tmp_coord[3], ureg_src(tmp_coord[0]),
ureg_imm4u(ureg, 1, 1, 0, 0)); /* bottom-right */
for (i = 0; i < nr_samples; i++) {
for (c = 0; c < 4; c++) {
/* Read one sample. */
ureg_MOV(ureg, ureg_writemask(tmp_coord[c], TGSI_WRITEMASK_W),
ureg_imm1u(ureg, i));
ureg_TXF(ureg, tmp, tgsi_tex, ureg_src(tmp_coord[c]), sampler);
if (stype == TGSI_RETURN_TYPE_UINT)
ureg_U2F(ureg, tmp, ureg_src(tmp));
else if (stype == TGSI_RETURN_TYPE_SINT)
ureg_I2F(ureg, tmp, ureg_src(tmp));
/* Add it to the sum.*/
ureg_ADD(ureg, tmp_sum[c], ureg_src(tmp_sum[c]), ureg_src(tmp));
}
}
/* Calculate the average. */
for (c = 0; c < 4; c++)
ureg_MUL(ureg, tmp_sum[c], ureg_src(tmp_sum[c]),
ureg_imm1f(ureg, 1.0 / nr_samples));
/* Take the 4 average values and apply a standard bilinear filter. */
ureg_FRC(ureg, tmp, coord);
ureg_LRP(ureg, top,
ureg_scalar(ureg_src(tmp), 0),
ureg_src(tmp_sum[1]),
ureg_src(tmp_sum[0]));
ureg_LRP(ureg, bottom,
ureg_scalar(ureg_src(tmp), 0),
ureg_src(tmp_sum[3]),
ureg_src(tmp_sum[2]));
ureg_LRP(ureg, tmp,
ureg_scalar(ureg_src(tmp), 1),
ureg_src(bottom),
ureg_src(top));
/* Convert to the texture format and return. */
if (stype == TGSI_RETURN_TYPE_UINT)
ureg_F2U(ureg, out, ureg_src(tmp));
else if (stype == TGSI_RETURN_TYPE_SINT)
ureg_F2I(ureg, out, ureg_src(tmp));
else
ureg_MOV(ureg, out, ureg_src(tmp));
ureg_END(ureg);
return ureg_create_shader_and_destroy(ureg, pipe);
}
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_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_frag_shader(struct vl_zscan *zscan)
{
struct ureg_program *shader;
struct ureg_src *vtex;
struct ureg_src samp_src, samp_scan, samp_quant;
struct ureg_dst *tmp;
struct ureg_dst quant, fragment;
unsigned i;
shader = ureg_create(PIPE_SHADER_FRAGMENT);
if (!shader)
return NULL;
vtex = MALLOC(zscan->num_channels * sizeof(struct ureg_src));
tmp = MALLOC(zscan->num_channels * sizeof(struct ureg_dst));
for (i = 0; i < zscan->num_channels; ++i)
vtex[i] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_VTEX + i, TGSI_INTERPOLATE_LINEAR);
samp_src = ureg_DECL_sampler(shader, 0);
samp_scan = ureg_DECL_sampler(shader, 1);
samp_quant = ureg_DECL_sampler(shader, 2);
for (i = 0; i < zscan->num_channels; ++i)
tmp[i] = ureg_DECL_temporary(shader);
quant = ureg_DECL_temporary(shader);
fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
/*
* tmp.x = tex(vtex, 1)
* tmp.y = vtex.z
* fragment = tex(tmp, 0) * quant
*/
for (i = 0; i < zscan->num_channels; ++i)
ureg_TEX(shader, ureg_writemask(tmp[i], TGSI_WRITEMASK_X), TGSI_TEXTURE_2D, vtex[i], samp_scan);
for (i = 0; i < zscan->num_channels; ++i)
ureg_MOV(shader, ureg_writemask(tmp[i], TGSI_WRITEMASK_Y), ureg_scalar(vtex[i], TGSI_SWIZZLE_W));
for (i = 0; i < zscan->num_channels; ++i) {
ureg_TEX(shader, ureg_writemask(tmp[0], TGSI_WRITEMASK_X << i), TGSI_TEXTURE_2D, ureg_src(tmp[i]), samp_src);
ureg_TEX(shader, ureg_writemask(quant, TGSI_WRITEMASK_X << i), TGSI_TEXTURE_3D, vtex[i], samp_quant);
}
ureg_MUL(shader, quant, ureg_src(quant), ureg_imm1f(shader, 16.0f));
ureg_MUL(shader, fragment, ureg_src(tmp[0]), ureg_src(quant));
for (i = 0; i < zscan->num_channels; ++i)
ureg_release_temporary(shader, tmp[i]);
ureg_END(shader);
FREE(vtex);
FREE(tmp);
return ureg_create_shader_and_destroy(shader, zscan->pipe);
}
static void *
create_frag_shader(struct vl_median_filter *filter,
struct vertex2f *offsets,
unsigned num_offsets)
{
struct pipe_screen *screen = filter->pipe->screen;
struct ureg_program *shader;
struct ureg_src i_vtex;
struct ureg_src sampler;
struct ureg_dst *t_array = MALLOC(sizeof(struct ureg_dst) * num_offsets);
struct ureg_dst o_fragment;
const unsigned median = num_offsets >> 1;
unsigned i, j;
assert(num_offsets & 1); /* we need an odd number of offsets */
if (!(num_offsets & 1)) { /* yeah, we REALLY need an odd number of offsets!!! */
FREE(t_array);
return NULL;
}
if (num_offsets > screen->get_shader_param(
screen, PIPE_SHADER_FRAGMENT, PIPE_SHADER_CAP_MAX_TEMPS)) {
FREE(t_array);
return NULL;
}
shader = ureg_create(PIPE_SHADER_FRAGMENT);
if (!shader) {
FREE(t_array);
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);
for (i = 0; i < num_offsets; ++i)
t_array[i] = ureg_DECL_temporary(shader);
o_fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
/*
* t_array[0..*] = vtex + offset[0..*]
* t_array[0..*] = tex(t_array[0..*], sampler)
* result = partial_bubblesort(t_array)[mid]
*/
for (i = 0; i < num_offsets; ++i) {
if (!is_vec_zero(offsets[i])) {
ureg_ADD(shader, ureg_writemask(t_array[i], TGSI_WRITEMASK_XY),
i_vtex, ureg_imm2f(shader, offsets[i].x, offsets[i].y));
ureg_MOV(shader, ureg_writemask(t_array[i], TGSI_WRITEMASK_ZW),
ureg_imm1f(shader, 0.0f));
}
}
for (i = 0; i < num_offsets; ++i) {
struct ureg_src src = is_vec_zero(offsets[i]) ? i_vtex : ureg_src(t_array[i]);
ureg_TEX(shader, t_array[i], TGSI_TEXTURE_2D, src, sampler);
}
// TODO: Couldn't this be improved even more?
for (i = 0; i <= median; ++i) {
for (j = 1; j < (num_offsets - i - 1); ++j) {
struct ureg_dst tmp = ureg_DECL_temporary(shader);
ureg_MOV(shader, tmp, ureg_src(t_array[j]));
ureg_MAX(shader, t_array[j], ureg_src(t_array[j]), ureg_src(t_array[j - 1]));
ureg_MIN(shader, t_array[j - 1], ureg_src(tmp), ureg_src(t_array[j - 1]));
ureg_release_temporary(shader, tmp);
}
if (i == median)
ureg_MAX(shader, t_array[j], ureg_src(t_array[j]), ureg_src(t_array[j - 1]));
else
ureg_MIN(shader, t_array[j - 1], ureg_src(t_array[j]), ureg_src(t_array[j - 1]));
}
ureg_MOV(shader, o_fragment, ureg_src(t_array[median]));
ureg_END(shader);
FREE(t_array);
return ureg_create_shader_and_destroy(shader, filter->pipe);
}
