static void
handle_sample_op(GLcontext * ctx, struct atifs_machine *machine,
struct atifs_instruction *inst, const struct sw_span *span,
GLuint column)
{
GLuint idx = inst->DstReg[0].Index - GL_REG_0_ATI;
GLuint swizzle = inst->DstReg[0].Swizzle;
GLuint sample_tex = inst->SrcReg[0][0].Index;
/* if we get here after passing pass one then we are starting pass two - backup the registers */
if (machine->pass == 1) {
finish_pass(machine);
machine->pass = 2;
}
if (sample_tex >= GL_TEXTURE0_ARB && sample_tex <= GL_TEXTURE7_ARB) {
sample_tex -= GL_TEXTURE0_ARB;
fetch_texel(ctx, span->array->texcoords[sample_tex][column], 0.0F,
sample_tex, machine->Registers[idx]);
}
else if (sample_tex >= GL_REG_0_ATI && sample_tex <= GL_REG_5_ATI) {
/* this is wrong... */
sample_tex -= GL_REG_0_ATI;
fetch_texel(ctx, machine->Registers[sample_tex], 0, sample_tex,
machine->Registers[idx]);
}
apply_swizzle(machine, idx, swizzle);
}
static void
handle_pass_op(struct atifs_machine *machine, struct atifs_instruction *inst,
const struct sw_span *span, GLuint column)
{
GLuint idx = inst->DstReg[0].Index - GL_REG_0_ATI;
GLuint swizzle = inst->DstReg[0].Swizzle;
GLuint pass_tex = inst->SrcReg[0][0].Index;
/* if we get here after passing pass one then we are starting pass two - backup the registers */
if (machine->pass == 1) {
finish_pass(machine);
machine->pass = 2;
}
if (pass_tex >= GL_TEXTURE0_ARB && pass_tex <= GL_TEXTURE7_ARB) {
pass_tex -= GL_TEXTURE0_ARB;
COPY_4V(machine->Registers[idx],
span->array->texcoords[pass_tex][column]);
}
else if (pass_tex >= GL_REG_0_ATI && pass_tex <= GL_REG_5_ATI
&& machine->pass == 2) {
pass_tex -= GL_REG_0_ATI;
COPY_4V(machine->Registers[idx], machine->PrevPassRegisters[pass_tex]);
}
apply_swizzle(machine, idx, swizzle);
}
/**
* Execute the given fragment shader.
* NOTE: we do everything in single-precision floating point
* \param ctx - rendering context
* \param shader - the shader to execute
* \param machine - virtual machine state
* \param span - the SWspan we're operating on
* \param column - which pixel [i] we're operating on in the span
*/
static void
execute_shader(struct gl_context *ctx, const struct ati_fragment_shader *shader,
struct atifs_machine *machine, const SWspan *span,
GLuint column)
{
GLuint pc;
struct atifs_instruction *inst;
struct atifs_setupinst *texinst;
GLint optype;
GLuint i;
GLint j, pass;
GLint dstreg;
GLfloat src[2][3][4];
GLfloat zeros[4] = { 0.0, 0.0, 0.0, 0.0 };
GLfloat ones[4] = { 1.0, 1.0, 1.0, 1.0 };
GLfloat dst[2][4], *dstp;
for (pass = 0; pass < shader->NumPasses; pass++) {
if (pass > 0)
finish_pass(machine);
for (j = 0; j < MAX_NUM_FRAGMENT_REGISTERS_ATI; j++) {
texinst = &shader->SetupInst[pass][j];
if (texinst->Opcode == ATI_FRAGMENT_SHADER_PASS_OP)
handle_pass_op(machine, texinst, span, column, j);
else if (texinst->Opcode == ATI_FRAGMENT_SHADER_SAMPLE_OP)
handle_sample_op(ctx, machine, texinst, span, column, j);
}
for (pc = 0; pc < shader->numArithInstr[pass]; pc++) {
inst = &shader->Instructions[pass][pc];
/* setup the source registers for color and alpha ops */
for (optype = 0; optype < 2; optype++) {
for (i = 0; i < inst->ArgCount[optype]; i++) {
GLint index = inst->SrcReg[optype][i].Index;
if (index >= GL_REG_0_ATI && index <= GL_REG_5_ATI)
SETUP_SRC_REG(optype, i,
machine->Registers[index - GL_REG_0_ATI]);
else if (index >= GL_CON_0_ATI && index <= GL_CON_7_ATI) {
if (shader->LocalConstDef & (1 << (index - GL_CON_0_ATI))) {
SETUP_SRC_REG(optype, i,
shader->Constants[index - GL_CON_0_ATI]);
} else {
SETUP_SRC_REG(optype, i,
ctx->ATIFragmentShader.GlobalConstants[index - GL_CON_0_ATI]);
}
}
else if (index == GL_ONE)
SETUP_SRC_REG(optype, i, ones);
else if (index == GL_ZERO)
SETUP_SRC_REG(optype, i, zeros);
else if (index == GL_PRIMARY_COLOR_EXT)
SETUP_SRC_REG(optype, i,
machine->Inputs[ATI_FS_INPUT_PRIMARY]);
else if (index == GL_SECONDARY_INTERPOLATOR_ATI)
SETUP_SRC_REG(optype, i,
machine->Inputs[ATI_FS_INPUT_SECONDARY]);
apply_src_rep(optype, inst->SrcReg[optype][i].argRep,
src[optype][i]);
apply_src_mod(optype, inst->SrcReg[optype][i].argMod,
src[optype][i]);
}
}
/* Execute the operations - color then alpha */
for (optype = 0; optype < 2; optype++) {
if (inst->Opcode[optype]) {
switch (inst->Opcode[optype]) {
case GL_ADD_ATI:
if (!optype)
for (i = 0; i < 3; i++) {
dst[optype][i] =
src[optype][0][i] + src[optype][1][i];
}
else
dst[optype][3] = src[optype][0][3] + src[optype][1][3];
break;
case GL_SUB_ATI:
if (!optype)
for (i = 0; i < 3; i++) {
dst[optype][i] =
src[optype][0][i] - src[optype][1][i];
}
else
dst[optype][3] = src[optype][0][3] - src[optype][1][3];
break;
case GL_MUL_ATI:
if (!optype)
for (i = 0; i < 3; i++) {
dst[optype][i] =
src[optype][0][i] * src[optype][1][i];
}
else
dst[optype][3] = src[optype][0][3] * src[optype][1][3];
break;
case GL_MAD_ATI:
if (!optype)
for (i = 0; i < 3; i++) {
dst[optype][i] =
src[optype][0][i] * src[optype][1][i] +
src[optype][2][i];
}
else
dst[optype][3] =
src[optype][0][3] * src[optype][1][3] +
src[optype][2][3];
break;
case GL_LERP_ATI:
if (!optype)
for (i = 0; i < 3; i++) {
dst[optype][i] =
src[optype][0][i] * src[optype][1][i] + (1 -
src
[optype]
[0][i]) *
src[optype][2][i];
}
else
dst[optype][3] =
src[optype][0][3] * src[optype][1][3] + (1 -
src[optype]
[0][3]) *
src[optype][2][3];
break;
case GL_MOV_ATI:
if (!optype)
for (i = 0; i < 3; i++) {
dst[optype][i] = src[optype][0][i];
}
else
dst[optype][3] = src[optype][0][3];
break;
case GL_CND_ATI:
if (!optype) {
for (i = 0; i < 3; i++) {
dst[optype][i] =
(src[optype][2][i] >
0.5) ? src[optype][0][i] : src[optype][1][i];
}
}
else {
dst[optype][3] =
(src[optype][2][3] >
0.5) ? src[optype][0][3] : src[optype][1][3];
}
break;
case GL_CND0_ATI:
if (!optype)
for (i = 0; i < 3; i++) {
dst[optype][i] =
(src[optype][2][i] >=
0) ? src[optype][0][i] : src[optype][1][i];
}
else {
dst[optype][3] =
(src[optype][2][3] >=
0) ? src[optype][0][3] : src[optype][1][3];
}
break;
case GL_DOT2_ADD_ATI:
{
GLfloat result;
/* DOT 2 always uses the source from the color op */
/* could save recalculation of dot products for alpha inst */
result = src[0][0][0] * src[0][1][0] +
src[0][0][1] * src[0][1][1] + src[0][2][2];
if (!optype) {
for (i = 0; i < 3; i++) {
dst[optype][i] = result;
}
}
else
dst[optype][3] = result;
}
break;
case GL_DOT3_ATI:
{
GLfloat result;
/* DOT 3 always uses the source from the color op */
result = src[0][0][0] * src[0][1][0] +
src[0][0][1] * src[0][1][1] +
src[0][0][2] * src[0][1][2];
if (!optype) {
for (i = 0; i < 3; i++) {
dst[optype][i] = result;
}
}
else
dst[optype][3] = result;
}
break;
case GL_DOT4_ATI:
{
GLfloat result;
/* DOT 4 always uses the source from the color op */
result = src[0][0][0] * src[0][1][0] +
src[0][0][1] * src[0][1][1] +
src[0][0][2] * src[0][1][2] +
src[0][0][3] * src[0][1][3];
if (!optype) {
for (i = 0; i < 3; i++) {
dst[optype][i] = result;
}
}
else
dst[optype][3] = result;
}
break;
}
}
}
/* write out the destination registers */
for (optype = 0; optype < 2; optype++) {
if (inst->Opcode[optype]) {
dstreg = inst->DstReg[optype].Index;
dstp = machine->Registers[dstreg - GL_REG_0_ATI];
if ((optype == 0) || ((inst->Opcode[1] != GL_DOT2_ADD_ATI) &&
(inst->Opcode[1] != GL_DOT3_ATI) && (inst->Opcode[1] != GL_DOT4_ATI)))
write_dst_addr(optype, inst->DstReg[optype].dstMod,
inst->DstReg[optype].dstMask, dst[optype],
dstp);
else
write_dst_addr(1, inst->DstReg[0].dstMod, 0, dst[1], dstp);
}
}
}
}
}