struct nvfx_context* nvfx;
struct pipe_shader_state pipe;
struct nvfx_vertex_program *vp;
+ struct tgsi_shader_info* info;
struct nvfx_vertex_program_exec *vpi;
if(insn.sat)
{
- assert(nvfx->is_nv4x);
- if(nvfx->is_nv4x)
+ assert(nvfx->use_nv4x);
+ if(nvfx->use_nv4x)
hw[0] |= NV40_VP_INST_SATURATE;
}
src.swz[2] = fsrc->Register.SwizzleZ;
src.swz[3] = fsrc->Register.SwizzleW;
src.indirect = 0;
+ src.indirect_reg = 0;
+ src.indirect_swz = 0;
if(fsrc->Register.Indirect) {
if(fsrc->Indirect.File == TGSI_FILE_ADDRESS &&
int mask;
int ai = -1, ci = -1, ii = -1;
int i;
-
- if (finst->Instruction.Opcode == TGSI_OPCODE_END)
- return TRUE;
+ unsigned sub_depth = 0;
for (i = 0; i < finst->Instruction.NumSrcRegs; i++) {
const struct tgsi_full_src_register *fsrc;
if(finst->Instruction.Saturate == TGSI_SAT_ZERO_ONE)
{
assert(finst->Instruction.Opcode != TGSI_OPCODE_ARL);
- if(nvfx->is_nv4x)
+ if(nvfx->use_nv4x)
sat = TRUE;
else if(dst.type != NVFXSR_TEMP)
dst = temp(vpc);
break;
case TGSI_OPCODE_RET:
- tmp = none;
- tmp.swz[0] = tmp.swz[1] = tmp.swz[2] = tmp.swz[3] = 0;
- nvfx_vp_emit(vpc, arith(0, SCA, RET, none.reg, 0, none, none, tmp));
+ if(sub_depth || !nvfx->use_vp_clipping) {
+ tmp = none;
+ tmp.swz[0] = tmp.swz[1] = tmp.swz[2] = tmp.swz[3] = 0;
+ nvfx_vp_emit(vpc, arith(0, SCA, RET, none.reg, 0, none, none, tmp));
+ } else {
+ reloc.location = vpc->vp->nr_insns;
+ reloc.target = vpc->info->num_instructions;
+ util_dynarray_append(&vpc->label_relocs, struct nvfx_relocation, reloc);
+ nvfx_vp_emit(vpc, arith(0, SCA, BRA, none.reg, 0, none, none, none));
+ }
break;
case TGSI_OPCODE_BGNSUB:
+ ++sub_depth;
+ break;
case TGSI_OPCODE_ENDSUB:
+ --sub_depth;
+ break;
case TGSI_OPCODE_ENDIF:
/* nothing to do here */
break;
nvfx_vp_emit(vpc, arith(0, SCA, BRA, none.reg, 0, none, none, none));
break;
+ case TGSI_OPCODE_END:
+ assert(!sub_depth);
+ if(nvfx->use_vp_clipping) {
+ if(idx != (vpc->info->num_instructions - 1)) {
+ reloc.location = vpc->vp->nr_insns;
+ reloc.target = vpc->info->num_instructions;
+ util_dynarray_append(&vpc->label_relocs, struct nvfx_relocation, reloc);
+ nvfx_vp_emit(vpc, arith(0, SCA, BRA, none.reg, 0, none, none, none));
+ }
+ } else {
+ if(vpc->vp->nr_insns)
+ vpc->vp->insns[vpc->vp->nr_insns - 1].data[3] |= NVFX_VP_INST_LAST;
+ nvfx_vp_emit(vpc, arith(0, VEC, NOP, none.reg, 0, none, none, none));
+ vpc->vp->insns[vpc->vp->nr_insns - 1].data[3] |= NVFX_VP_INST_LAST;
+ }
+ break;
+
default:
NOUVEAU_ERR("invalid opcode %d\n", finst->Instruction.Opcode);
return FALSE;
}
- if(finst->Instruction.Saturate == TGSI_SAT_ZERO_ONE && !nvfx->is_nv4x)
+ if(finst->Instruction.Saturate == TGSI_SAT_ZERO_ONE && !nvfx->use_nv4x)
{
if(!vpc->r_0_1.type)
vpc->r_0_1 = constant(vpc, -1, 0, 1, 0, 0);
/* hope 0xf is (0, 0, 0, 1) initialized; otherwise, we are _probably_ not required to do this */
memset(vpc->vp->generic_to_fp_input, 0x0f, sizeof(vpc->vp->generic_to_fp_input));
- for(int i = 0; i < 10; ++i) {
+ for(int i = 0; i < num_texcoords; ++i) {
if(sem_layout[i] == 0xff)
continue;
//printf("vp: GENERIC[%i] to fpreg %i\n", sem_layout[i], NVFX_FP_OP_INPUT_SRC_TC(0) + i);
}
vpc->vp->sprite_fp_input = -1;
- for(int i = 0; i < 10; ++i)
+ for(int i = 0; i < num_texcoords; ++i)
{
if(sem_layout[i] == 0xff)
{
DEBUG_GET_ONCE_BOOL_OPTION(nvfx_dump_vp, "NVFX_DUMP_VP", FALSE)
static struct nvfx_vertex_program*
-nvfx_vertprog_translate(struct nvfx_context *nvfx, const struct pipe_shader_state* vps)
+nvfx_vertprog_translate(struct nvfx_context *nvfx, const struct pipe_shader_state* vps, struct tgsi_shader_info* info)
{
struct tgsi_parse_context parse;
struct nvfx_vertex_program* vp = NULL;
vpc->nvfx = nvfx;
vpc->vp = vp;
vpc->pipe = *vps;
+ vpc->info = info;
{
// TODO: use a 64-bit atomic here!
nvfx_vp_emit(vpc, arith(0, VEC, DP4, cdst, mask, htmp, ceqn, none));
}
}
- else
- {
- if(vp->nr_insns)
- vp->insns[vp->nr_insns - 1].data[3] |= NVFX_VP_INST_LAST;
-
- nvfx_vp_emit(vpc, arith(0, VEC, NOP, none.reg, 0, none, none, none));
- vp->insns[vp->nr_insns - 1].data[3] |= NVFX_VP_INST_LAST;
- }
if(debug_get_option_nvfx_dump_vp())
{
{
struct nvfx_vertex_program* vp = NULL;
struct pipe_shader_state vps;
+ struct tgsi_shader_info info;
struct ureg_program *ureg = NULL;
unsigned num_outputs = MIN2(pvp->info.num_outputs, 16);
ureg_END( ureg );
vps.tokens = ureg_get_tokens(ureg, 0);
- vp = nvfx_vertprog_translate(nvfx, &vps);
+ tgsi_scan_shader(vps.tokens, &info);
+ vp = nvfx_vertprog_translate(nvfx, &vps, &info);
ureg_free_tokens(vps.tokens);
ureg_destroy(ureg);
{
struct nvfx_screen *screen = nvfx->screen;
struct nouveau_channel *chan = screen->base.channel;
+ struct nouveau_grobj *eng3d = screen->eng3d;
struct nvfx_pipe_vertex_program *pvp = nvfx->vertprog;
struct nvfx_vertex_program* vp;
struct pipe_resource *constbuf;
vp = pvp->vp;
if(!vp) {
- vp = nvfx_vertprog_translate(nvfx, &pvp->pipe);
+ vp = nvfx_vertprog_translate(nvfx, &pvp->pipe, &pvp->info);
if(!vp)
vp = NVFX_VP_FAILED;
pvp->vp = vp;
* WAIT_RING(chan, 512 * 6);
for (i = 0; i < 512; i++) {
float v[4] = {0.1, 0,2, 0.3, 0.4};
- OUT_RING(chan, RING_3D(NV34TCL_VP_UPLOAD_CONST_ID, 5));
+ OUT_RING(chan, RING_3D(NV30_3D_VP_UPLOAD_CONST_ID, 5));
OUT_RING(chan, i);
OUT_RINGp(chan, (uint32_t *)v, 4);
printf("frob %i\n", i);
}
*/
- WAIT_RING(chan, 6 * vp->nr_consts);
for (i = nvfx->use_vp_clipping ? 6 : 0; i < vp->nr_consts; i++) {
struct nvfx_vertex_program_data *vpd = &vp->consts[i];
//printf("upload into %i + %i: %f %f %f %f\n", vp->data->start, i, vpd->value[0], vpd->value[1], vpd->value[2], vpd->value[3]);
- OUT_RING(chan, RING_3D(NV34TCL_VP_UPLOAD_CONST_ID, 5));
+ BEGIN_RING(chan, eng3d, NV30_3D_VP_UPLOAD_CONST_ID, 5);
OUT_RING(chan, i + vp->data->start);
OUT_RINGp(chan, (uint32_t *)vpd->value, 4);
}
/* Upload vtxprog */
if (upload_code) {
- WAIT_RING(chan, 2 + 5 * vp->nr_insns);
- OUT_RING(chan, RING_3D(NV34TCL_VP_UPLOAD_FROM_ID, 1));
+ BEGIN_RING(chan, eng3d, NV30_3D_VP_UPLOAD_FROM_ID, 1);
OUT_RING(chan, vp->exec->start);
for (i = 0; i < vp->nr_insns; i++) {
- OUT_RING(chan, RING_3D(NV34TCL_VP_UPLOAD_INST(0), 4));
+ BEGIN_RING(chan, eng3d, NV30_3D_VP_UPLOAD_INST(0), 4);
//printf("%08x %08x %08x %08x\n", vp->insns[i].data[0], vp->insns[i].data[1], vp->insns[i].data[2], vp->insns[i].data[3]);
OUT_RINGp(chan, vp->insns[i].data, 4);
}
if(nvfx->dirty & (NVFX_NEW_VERTPROG))
{
- WAIT_RING(chan, 6);
- OUT_RING(chan, RING_3D(NV34TCL_VP_START_FROM_ID, 1));
+ BEGIN_RING(chan, eng3d, NV30_3D_VP_START_FROM_ID, 1);
OUT_RING(chan, vp->exec->start);
if(nvfx->is_nv4x) {
- OUT_RING(chan, RING_3D(NV40TCL_VP_ATTRIB_EN, 1));
+ BEGIN_RING(chan, eng3d, NV40_3D_VP_ATTRIB_EN, 1);
OUT_RING(chan, vp->ir);
}
}
static void *
nvfx_vp_state_create(struct pipe_context *pipe, const struct pipe_shader_state *cso)
{
- struct nvfx_context *nvfx = nvfx_context(pipe);
struct nvfx_pipe_vertex_program *pvp;
pvp = CALLOC(1, sizeof(struct nvfx_pipe_vertex_program));