#include "pipe/p_state.h"
#include "pipe/p_defines.h"
-#include "pipe/p_util.h"
+#include "util/u_memory.h"
#include "pipe/p_inlines.h"
-#include "tgsi/exec/tgsi_exec.h"
-#include "tgsi/exec/tgsi_sse2.h"
+#include "tgsi/tgsi_exec.h"
+#include "tgsi/tgsi_sse2.h"
-#if defined(__i386__) || defined(__386__)
+#if defined(PIPE_ARCH_X86)
#include "rtasm/rtasm_x86sse.h"
/* Surely this should be defined somewhere in a tgsi header:
*/
-typedef void (XSTDCALL *codegen_function)(
+typedef void (PIPE_CDECL *codegen_function)(
const struct tgsi_exec_vector *input,
struct tgsi_exec_vector *output,
- float (*constant)[4],
+ const float (*constant)[4],
struct tgsi_exec_vector *temporary,
- const struct tgsi_interp_coef *coef
+ const struct tgsi_interp_coef *coef,
+ float (*immediates)[4]
//, const struct tgsi_exec_vector *quadPos
);
struct sp_fragment_shader base;
struct x86_function sse2_program;
codegen_function func;
+ float immediates[TGSI_EXEC_NUM_IMMEDIATES][4];
};
static void
-fs_sse_prepare( struct sp_fragment_shader *base,
- struct tgsi_exec_machine *machine,
- struct tgsi_sampler *samplers )
+fs_sse_prepare( const struct sp_fragment_shader *base,
+ struct tgsi_exec_machine *machine,
+ struct tgsi_sampler *samplers )
{
}
* TODO: process >1 quad at a time
*/
static unsigned
-fs_sse_run( struct sp_fragment_shader *base,
- struct tgsi_exec_machine *machine,
- struct quad_header *quad )
+fs_sse_run( const struct sp_fragment_shader *base,
+ struct tgsi_exec_machine *machine,
+ struct quad_header *quad )
{
struct sp_sse_fragment_shader *shader = (struct sp_sse_fragment_shader *) base;
/* Compute X, Y, Z, W vals for this quad -- place in temp[0] for now */
sp_setup_pos_vector(quad->posCoef,
- (float)quad->x0, (float)quad->y0,
+ (float)quad->input.x0, (float)quad->input.y0,
machine->Temps);
/* init kill mask */
- machine->Temps[TGSI_EXEC_TEMP_KILMASK_I].xyzw[TGSI_EXEC_TEMP_KILMASK_C].u[0] = 0x0;
+ tgsi_set_kill_mask(machine, 0x0);
+ tgsi_set_exec_mask(machine, 1, 1, 1, 1);
shader->func( machine->Inputs,
machine->Outputs,
machine->Consts,
machine->Temps,
- machine->InterpCoefs
+ machine->InterpCoefs,
+ shader->immediates
// , &machine->QuadPos
);
x86_init_func( &shader->sse2_program );
- if (!tgsi_emit_sse2_fs( templ->tokens, &shader->sse2_program )) {
+ if (!tgsi_emit_sse2( templ->tokens, &shader->sse2_program,
+ shader->immediates, FALSE )) {
FREE(shader);
return NULL;
}
shader->func = (codegen_function) x86_get_func( &shader->sse2_program );
- assert(shader->func);
+ if (!shader->func) {
+ x86_release_func( &shader->sse2_program );
+ FREE(shader);
+ return NULL;
+ }
shader->base.shader = *templ;
shader->base.prepare = fs_sse_prepare;