#include "pipe/tgsi/exec/tgsi_core.h"
+
+#define DBG 0
+
+
static INLINE unsigned
-compute_clipmask(float cx, float cy, float cz, float cw)
+compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr)
{
unsigned mask = 0;
+ unsigned i;
- if (-cx + cw < 0) mask |= CLIP_RIGHT_BIT;
- if ( cx + cw < 0) mask |= CLIP_LEFT_BIT;
- if (-cy + cw < 0) mask |= CLIP_TOP_BIT;
- if ( cy + cw < 0) mask |= CLIP_BOTTOM_BIT;
- if (-cz + cw < 0) mask |= CLIP_FAR_BIT;
- if ( cz + cw < 0) mask |= CLIP_NEAR_BIT;
+ for (i = 0; i < nr; i++) {
+ if (dot4(clip, plane[i]) < 0)
+ mask |= (1<<i);
+ }
return mask;
}
-
-
-#if !defined(XSTDCALL)
-#if defined(WIN32)
-#define XSTDCALL __stdcall
-#else
-#define XSTDCALL
-#endif
-#endif
-
typedef void (XSTDCALL *codegen_function) (
const struct tgsi_exec_vector *input,
struct tgsi_exec_vector *output,
unsigned elts[4], unsigned count,
struct vertex_header *vOut[])
{
- struct tgsi_exec_machine machine;
+ struct tgsi_exec_machine *machine = &draw->machine;
unsigned int j;
ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX);
assert(draw->vertex_shader->state->output_semantic_name[0]
== TGSI_SEMANTIC_POSITION);
-#ifdef DEBUG
- memset( &machine, 0, sizeof( machine ) );
-#endif
-
- /* init machine state */
- tgsi_exec_machine_init(&machine,
- draw->vertex_shader->state->tokens,
- PIPE_MAX_SAMPLERS,
- NULL /*samplers*/ );
+ fprintf(stderr, "------ run_vertex\n");
/* Consts does not require 16 byte alignment. */
- machine.Consts = (float (*)[4]) draw->mapped_constants;
+ machine->Consts = (float (*)[4]) draw->mapped_constants;
- machine.Inputs = ALIGN16_ASSIGN(inputs);
- machine.Outputs = ALIGN16_ASSIGN(outputs);
+ machine->Inputs = ALIGN16_ASSIGN(inputs);
+ machine->Outputs = ALIGN16_ASSIGN(outputs);
- draw_vertex_fetch( draw, &machine, elts, count );
+ draw_vertex_fetch( draw, machine, elts, count );
/* run shader */
if( draw->vertex_shader->state->executable != NULL ) {
+ /* SSE */
codegen_function func = (codegen_function) draw->vertex_shader->state->executable;
func(
- machine.Inputs,
- machine.Outputs,
- machine.Consts,
- machine.Temps );
+ machine->Inputs,
+ machine->Outputs,
+ machine->Consts,
+ machine->Temps );
}
else {
- tgsi_exec_machine_run( &machine );
+ /* interpreter */
+ tgsi_exec_machine_run( machine );
}
unsigned slot;
float x, y, z, w;
- /* Handle attr[0] (position) specially: */
- x = vOut[j]->clip[0] = machine.Outputs[0].xyzw[0].f[j];
- y = vOut[j]->clip[1] = machine.Outputs[0].xyzw[1].f[j];
- z = vOut[j]->clip[2] = machine.Outputs[0].xyzw[2].f[j];
- w = vOut[j]->clip[3] = machine.Outputs[0].xyzw[3].f[j];
+ /* Handle attr[0] (position) specially:
+ *
+ * XXX: Computing the clipmask should be done in the vertex
+ * program as a set of DP4 instructions appended to the
+ * user-provided code.
+ */
+ x = vOut[j]->clip[0] = machine->Outputs[0].xyzw[0].f[j];
+ y = vOut[j]->clip[1] = machine->Outputs[0].xyzw[1].f[j];
+ z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j];
+ w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j];
+ printf("output %d: %f %f %f %f\n", 0, x, y, z, w);
- vOut[j]->clipmask = compute_clipmask(x, y, z, w) | draw->user_clipmask;
+ vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes);
vOut[j]->edgeflag = 1;
/* divide by w */
vOut[j]->data[0][2] = z * scale[2] + trans[2];
vOut[j]->data[0][3] = w;
+#if DBG
+ printf("output[%d]win: %f %f %f %f\n", x, y, z, w);
+#endif
/* Remaining attributes are packed into sequential post-transform
* vertex attrib slots.
* Skip 0 since we just did it above.
* Subtract two because of the VERTEX_HEADER, CLIP_POS attribs.
*/
for (slot = 1; slot < draw->vertex_info.num_attribs - 2; slot++) {
- vOut[j]->data[slot][0] = machine.Outputs[slot].xyzw[0].f[j];
- vOut[j]->data[slot][1] = machine.Outputs[slot].xyzw[1].f[j];
- vOut[j]->data[slot][2] = machine.Outputs[slot].xyzw[2].f[j];
- vOut[j]->data[slot][3] = machine.Outputs[slot].xyzw[3].f[j];
- /*
- printf("output %d: %f %f %f %f\n", slot,
- vOut[j]->data[slot][0],
- vOut[j]->data[slot][1],
- vOut[j]->data[slot][2],
- vOut[j]->data[slot][3]);
- */
+ vOut[j]->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j];
+ vOut[j]->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j];
+ vOut[j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j];
+ vOut[j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j];
+#if DBG
+ printf("output[%d][%d]: %f %f %f %f\n", j, slot,
+#endif
}
} /* loop over vertices */
}
{
unsigned i, j;
-// fprintf(stderr, " q(%d) ", draw->vs.queue_nr );
+ fprintf(stderr, "XX q(%d) ", draw->vs.queue_nr );
+
+ if (draw->vertex_shader->state->llvm_prog) {
+ draw_vertex_shader_queue_flush_llvm(draw);
+ return;
+ }
/* run vertex shader on vertex cache entries, four per invokation */
for (i = 0; i < draw->vs.queue_nr; i += 4) {
vs->state = shader;
#if defined(__i386__) || defined(__386__)
- x86_init_func(&vs->sse2_program);
+ if (draw->use_sse) {
+ /* cast-away const */
+ struct pipe_shader_state *sh = (struct pipe_shader_state *) shader;
+
+ x86_init_func( &sh->sse2_program );
- tgsi_emit_sse2(shader->tokens, &vs->sse2_program);
+ tgsi_emit_sse2( sh->tokens, &sh->sse2_program );
- ((struct pipe_shader_state*)(vs->state))->executable =
- x86_get_func(&vs->sse2_program);
+ sh->executable = x86_get_func( &sh->sse2_program );
+ }
#endif
return vs;
{
draw_flush(draw);
draw->vertex_shader = (struct draw_vertex_shader*)(vcso);
+
+ /* specify the fragment program to interpret/execute */
+ tgsi_exec_machine_init(&draw->machine,
+ draw->vertex_shader->state->tokens,
+ PIPE_MAX_SAMPLERS,
+ NULL /*samplers*/ );
}
void draw_delete_vertex_shader(struct draw_context *draw,
void *vcso)
{
struct draw_vertex_shader *vs = (struct draw_vertex_shader*)(vcso);
+
#if defined(__i386__) || defined(__386__)
- x86_release_func(&vs->sse2_program);
+ x86_release_func((struct x86_function *) &vs->state->sse2_program);
#endif
+
+ free((void *) vs->state);
free(vcso);
}