* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
*/
-#include "glheader.h"
-#include "colormac.h"
-#include "context.h"
-#include "atifragshader.h"
-#include "macros.h"
-#include "program.h"
-
-#include "s_atifragshader.h"
-#include "s_nvfragprog.h"
-#include "s_span.h"
-#include "s_texture.h"
+#include "main/glheader.h"
+#include "main/colormac.h"
+#include "main/context.h"
+#include "main/macros.h"
+#include "shader/program.h"
+#include "shader/atifragshader.h"
+#include "swrast/s_atifragshader.h"
+
+
+/**
+ * State for executing ATI fragment shader.
+ */
+struct atifs_machine
+{
+ GLfloat Registers[6][4]; /** six temporary registers */
+ GLfloat PrevPassRegisters[6][4];
+ GLfloat Inputs[2][4]; /** Primary, secondary input colors */
+};
+
+
/**
* Fetch a texel.
SWcontext *swrast = SWRAST_CONTEXT(ctx);
/* XXX use a float-valued TextureSample routine here!!! */
- swrast->TextureSample[unit] (ctx, unit, ctx->Texture.Unit[unit]._Current,
- 1, (const GLfloat(*)[4]) texcoord,
- &lambda, &rgba);
+ swrast->TextureSample[unit](ctx, ctx->Texture.Unit[unit]._Current,
+ 1, (const GLfloat(*)[4]) texcoord,
+ &lambda, &rgba);
color[0] = CHAN_TO_FLOAT(rgba[0]);
color[1] = CHAN_TO_FLOAT(rgba[1]);
color[2] = CHAN_TO_FLOAT(rgba[2]);
return;
start = optype ? 3 : 0;
- end = optype ? 4 : 3;
+ end = 4;
for (i = start; i < end; i++) {
switch (rep) {
return;
start = optype ? 3 : 0;
- end = optype ? 4 : 3;
+ end = 4;
for (i = start; i < end; i++) {
if (mod & GL_COMP_BIT_ATI)
}
}
-/**
- * Execute the given fragment shader
- * NOTE: we do everything in single-precision floating point; we don't
- * currently observe the single/half/fixed-precision qualifiers.
- * \param ctx - rendering context
- * \param program - the fragment program to execute
- * \param machine - machine state (register file)
- * \param maxInst - max number of instructions to execute
- * \return GL_TRUE if program completed or GL_FALSE if program executed KIL.
- */
-
struct ati_fs_opcode_st ati_fs_opcodes[] = {
{GL_ADD_ATI, 2},
{GL_SUB_ATI, 2},
static void
handle_pass_op(struct atifs_machine *machine, struct atifs_setupinst *texinst,
- const struct sw_span *span, GLuint column, GLuint idx)
+ const SWspan *span, GLuint column, GLuint idx)
{
GLuint swizzle = texinst->swizzle;
GLuint pass_tex = texinst->src;
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]);
+ span->array->attribs[FRAG_ATTRIB_TEX0 + pass_tex][column]);
}
else if (pass_tex >= GL_REG_0_ATI && pass_tex <= GL_REG_5_ATI) {
pass_tex -= GL_REG_0_ATI;
static void
handle_sample_op(GLcontext * ctx, struct atifs_machine *machine,
- struct atifs_setupinst *texinst, const struct sw_span *span,
+ struct atifs_setupinst *texinst, const SWspan *span,
GLuint column, GLuint idx)
{
/* sample from unit idx using texinst->src as coords */
if (coord_source >= GL_TEXTURE0_ARB && coord_source <= GL_TEXTURE7_ARB) {
coord_source -= GL_TEXTURE0_ARB;
- COPY_4V(tex_coords, span->array->texcoords[coord_source][column]);
+ COPY_4V(tex_coords,
+ span->array->attribs[FRAG_ATTRIB_TEX0 + coord_source][column]);
}
else if (coord_source >= GL_REG_0_ATI && coord_source <= GL_REG_5_ATI) {
coord_source -= GL_REG_0_ATI;
fetch_texel(ctx, tex_coords, 0.0F, idx, machine->Registers[idx]);
}
-#define SETUP_SRC_REG(optype, i, x) do { \
- if (optype) \
- src[optype][i][3] = x[3]; \
- else \
- COPY_3V(src[optype][i], x); \
- } while (0)
-
-static GLboolean
-execute_shader(GLcontext * ctx,
- const struct ati_fragment_shader *shader, GLuint maxInst,
- struct atifs_machine *machine, const struct sw_span *span,
- GLuint column)
+#define SETUP_SRC_REG(optype, i, x) \
+do { \
+ COPY_4V(src[optype][i], x); \
+} while (0)
+
+
+
+/**
+ * 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(GLcontext *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;
- GLint i, j, pass;
+ GLuint i;
+ GLint j, pass;
GLint dstreg;
GLfloat src[2][3][4];
GLfloat zeros[4] = { 0.0, 0.0, 0.0, 0.0 };
/* setup the source registers for color and alpha ops */
for (optype = 0; optype < 2; optype++) {
- for (i = 0; i < inst->ArgCount[optype]; i++) {
+ 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))) {
+ 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]);
+ ctx->ATIFragmentShader.GlobalConstants[index - GL_CON_0_ATI]);
}
}
else if (index == GL_ONE)
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) {
}
else
dst[optype][3] = result;
-
}
break;
case GL_DOT3_ATI:
GLfloat result;
/* DOT 4 always uses the source from the color op */
- result = src[optype][0][0] * src[0][1][0] +
+ 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];
dstreg = inst->DstReg[optype].Index;
dstp = machine->Registers[dstreg - GL_REG_0_ATI];
- write_dst_addr(optype, inst->DstReg[optype].dstMod,
+ 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);
}
}
}
}
- return GL_TRUE;
}
+
+/**
+ * Init fragment shader virtual machine state.
+ */
static void
init_machine(GLcontext * ctx, struct atifs_machine *machine,
const struct ati_fragment_shader *shader,
- const struct sw_span *span, GLuint col)
+ const SWspan *span, GLuint col)
{
+ GLfloat (*inputs)[4] = machine->Inputs;
GLint i, j;
for (i = 0; i < 6; i++) {
for (j = 0; j < 4; j++)
- ctx->ATIFragmentShader.Machine.Registers[i][j] = 0.0;
-
+ machine->Registers[i][j] = 0.0;
}
- ctx->ATIFragmentShader.Machine.Inputs[ATI_FS_INPUT_PRIMARY][0] =
- CHAN_TO_FLOAT(span->array->rgba[col][0]);
- ctx->ATIFragmentShader.Machine.Inputs[ATI_FS_INPUT_PRIMARY][1] =
- CHAN_TO_FLOAT(span->array->rgba[col][1]);
- ctx->ATIFragmentShader.Machine.Inputs[ATI_FS_INPUT_PRIMARY][2] =
- CHAN_TO_FLOAT(span->array->rgba[col][2]);
- ctx->ATIFragmentShader.Machine.Inputs[ATI_FS_INPUT_PRIMARY][3] =
- CHAN_TO_FLOAT(span->array->rgba[col][3]);
-
- ctx->ATIFragmentShader.Machine.Inputs[ATI_FS_INPUT_SECONDARY][0] =
- CHAN_TO_FLOAT(span->array->spec[col][0]);
- ctx->ATIFragmentShader.Machine.Inputs[ATI_FS_INPUT_SECONDARY][1] =
- CHAN_TO_FLOAT(span->array->spec[col][1]);
- ctx->ATIFragmentShader.Machine.Inputs[ATI_FS_INPUT_SECONDARY][2] =
- CHAN_TO_FLOAT(span->array->spec[col][2]);
- ctx->ATIFragmentShader.Machine.Inputs[ATI_FS_INPUT_SECONDARY][3] =
- CHAN_TO_FLOAT(span->array->spec[col][3]);
-
- ctx->ATIFragmentShader.Machine.pass = 0;
+ COPY_4V(inputs[ATI_FS_INPUT_PRIMARY], span->array->attribs[FRAG_ATTRIB_COL0][col]);
+ COPY_4V(inputs[ATI_FS_INPUT_SECONDARY], span->array->attribs[FRAG_ATTRIB_COL1][col]);
}
/**
- * Execute the current fragment program, operating on the given span.
+ * Execute the current ATI shader program, operating on the given span.
*/
void
-_swrast_exec_fragment_shader(GLcontext * ctx, struct sw_span *span)
+_swrast_exec_fragment_shader(GLcontext * ctx, SWspan *span)
{
const struct ati_fragment_shader *shader = ctx->ATIFragmentShader.Current;
+ struct atifs_machine machine;
GLuint i;
+ /* incoming colors should be floats */
+ ASSERT(span->array->ChanType == GL_FLOAT);
+
ctx->_CurrentProgram = GL_FRAGMENT_SHADER_ATI;
for (i = 0; i < span->end; i++) {
if (span->array->mask[i]) {
- init_machine(ctx, &ctx->ATIFragmentShader.Machine,
- ctx->ATIFragmentShader.Current, span, i);
+ init_machine(ctx, &machine, shader, span, i);
- if (execute_shader(ctx, shader, ~0,
- &ctx->ATIFragmentShader.Machine, span, i)) {
- span->array->mask[i] = GL_FALSE;
- }
+ execute_shader(ctx, shader, &machine, span, i);
+ /* store result color */
{
- const GLfloat *colOut =
- ctx->ATIFragmentShader.Machine.Registers[0];
-
- /*fprintf(stderr,"outputs %f %f %f %f\n", colOut[0], colOut[1], colOut[2], colOut[3]); */
- UNCLAMPED_FLOAT_TO_CHAN(span->array->rgba[i][RCOMP], colOut[0]);
- UNCLAMPED_FLOAT_TO_CHAN(span->array->rgba[i][GCOMP], colOut[1]);
- UNCLAMPED_FLOAT_TO_CHAN(span->array->rgba[i][BCOMP], colOut[2]);
- UNCLAMPED_FLOAT_TO_CHAN(span->array->rgba[i][ACOMP], colOut[3]);
+ const GLfloat *colOut = machine.Registers[0];
+ /*fprintf(stderr,"outputs %f %f %f %f\n",
+ colOut[0], colOut[1], colOut[2], colOut[3]); */
+ COPY_4V(span->array->attribs[FRAG_ATTRIB_COL0][i], colOut);
}
}
-
}
-
ctx->_CurrentProgram = 0;
-
}