bool
nv50_program_translate(struct nv50_program *prog, uint16_t chipset,
struct pipe_debug_callback *debug)
{
struct nv50_ir_prog_info *info;
int i, ret;
const uint8_t map_undef = (prog->type == PIPE_SHADER_VERTEX) ? 0x40 : 0x80;
info = CALLOC_STRUCT(nv50_ir_prog_info);
if (!info)
return false;
info->type = prog->type;
info->target = chipset;
info->bin.sourceRep = prog->pipe.type;
switch (prog->pipe.type) {
case PIPE_SHADER_IR_TGSI:
info->bin.source = (void *)prog->pipe.tokens;
break;
case PIPE_SHADER_IR_NIR:
info->bin.source = (void *)nir_shader_clone(NULL, prog->pipe.ir.nir);
break;
default:
assert(!"unsupported IR!");
return false;
}
info->bin.smemSize = prog->cp.smem_size;
info->io.auxCBSlot = 15;
info->io.ucpBase = NV50_CB_AUX_UCP_OFFSET;
info->io.genUserClip = prog->vp.clpd_nr;
if (prog->fp.alphatest)
info->io.alphaRefBase = NV50_CB_AUX_ALPHATEST_OFFSET;
info->io.suInfoBase = NV50_CB_AUX_TEX_MS_OFFSET;
info->io.sampleInfoBase = NV50_CB_AUX_SAMPLE_OFFSET;
info->io.msInfoCBSlot = 15;
info->io.msInfoBase = NV50_CB_AUX_MS_OFFSET;
info->assignSlots = nv50_program_assign_varying_slots;
prog->vp.bfc[0] = 0xff;
prog->vp.bfc[1] = 0xff;
prog->vp.edgeflag = 0xff;
prog->vp.clpd[0] = map_undef;
prog->vp.clpd[1] = map_undef;
prog->vp.psiz = map_undef;
prog->gp.has_layer = 0;
prog->gp.has_viewport = 0;
if (prog->type == PIPE_SHADER_COMPUTE)
info->prop.cp.inputOffset = 0x10;
info->driverPriv = prog;
#ifdef DEBUG
info->optLevel = debug_get_num_option("NV50_PROG_OPTIMIZE", 3);
info->dbgFlags = debug_get_num_option("NV50_PROG_DEBUG", 0);
info->omitLineNum = debug_get_num_option("NV50_PROG_DEBUG_OMIT_LINENUM", 0);
#else
info->optLevel = 3;
#endif
ret = nv50_ir_generate_code(info);
if (ret) {
NOUVEAU_ERR("shader translation failed: %i\n", ret);
goto out;
}
prog->code = info->bin.code;
prog->code_size = info->bin.codeSize;
prog->fixups = info->bin.relocData;
prog->interps = info->bin.fixupData;
prog->max_gpr = MAX2(4, (info->bin.maxGPR >> 1) + 1);
prog->tls_space = info->bin.tlsSpace;
prog->cp.smem_size = info->bin.smemSize;
prog->mul_zero_wins = info->io.mul_zero_wins;
prog->vp.need_vertex_id = info->io.vertexId < PIPE_MAX_SHADER_INPUTS;
prog->vp.clip_enable = (1 << info->io.clipDistances) - 1;
prog->vp.cull_enable =
((1 << info->io.cullDistances) - 1) << info->io.clipDistances;
prog->vp.clip_mode = 0;
for (i = 0; i < info->io.cullDistances; ++i)
prog->vp.clip_mode |= 1 << ((info->io.clipDistances + i) * 4);
if (prog->type == PIPE_SHADER_FRAGMENT) {
if (info->prop.fp.writesDepth) {
prog->fp.flags[0] |= NV50_3D_FP_CONTROL_EXPORTS_Z;
prog->fp.flags[1] = 0x11;
}
if (info->prop.fp.usesDiscard)
prog->fp.flags[0] |= NV50_3D_FP_CONTROL_USES_KIL;
} else
if (prog->type == PIPE_SHADER_GEOMETRY) {
switch (info->prop.gp.outputPrim) {
case PIPE_PRIM_LINE_STRIP:
prog->gp.prim_type = NV50_3D_GP_OUTPUT_PRIMITIVE_TYPE_LINE_STRIP;
break;
case PIPE_PRIM_TRIANGLE_STRIP:
prog->gp.prim_type = NV50_3D_GP_OUTPUT_PRIMITIVE_TYPE_TRIANGLE_STRIP;
break;
case PIPE_PRIM_POINTS:
default:
assert(info->prop.gp.outputPrim == PIPE_PRIM_POINTS);
prog->gp.prim_type = NV50_3D_GP_OUTPUT_PRIMITIVE_TYPE_POINTS;
break;
}
prog->gp.vert_count = CLAMP(info->prop.gp.maxVertices, 1, 1024);
}
if (prog->type == PIPE_SHADER_COMPUTE) {
prog->cp.syms = info->bin.syms;
prog->cp.num_syms = info->bin.numSyms;
} else {
FREE(info->bin.syms);
}
if (prog->pipe.stream_output.num_outputs)
prog->so = nv50_program_create_strmout_state(info,
&prog->pipe.stream_output);
pipe_debug_message(debug, SHADER_INFO,
"type: %d, local: %d, shared: %d, gpr: %d, inst: %d, bytes: %d",
prog->type, info->bin.tlsSpace, info->bin.smemSize,
prog->max_gpr, info->bin.instructions,
info->bin.codeSize);
out:
if (info->bin.sourceRep == PIPE_SHADER_IR_NIR)
ralloc_free((void *)info->bin.source);
FREE(info);
return !ret;
}
boolean
nv50_program_translate(struct nv50_program *prog, uint16_t chipset)
{
struct nv50_ir_prog_info *info;
int ret;
const uint8_t map_undef = (prog->type == PIPE_SHADER_VERTEX) ? 0x40 : 0x80;
info = CALLOC_STRUCT(nv50_ir_prog_info);
if (!info)
return FALSE;
info->type = prog->type;
info->target = chipset;
info->bin.sourceRep = NV50_PROGRAM_IR_TGSI;
info->bin.source = (void *)prog->pipe.tokens;
info->io.ucpCBSlot = 15;
info->io.ucpBase = 0;
info->io.genUserClip = prog->vp.clpd_nr;
info->assignSlots = nv50_program_assign_varying_slots;
prog->vp.bfc[0] = 0xff;
prog->vp.bfc[1] = 0xff;
prog->vp.edgeflag = 0xff;
prog->vp.clpd[0] = map_undef;
prog->vp.clpd[1] = map_undef;
prog->vp.psiz = map_undef;
prog->gp.primid = 0x80;
info->driverPriv = prog;
#ifdef DEBUG
info->optLevel = debug_get_num_option("NV50_PROG_OPTIMIZE", 3);
info->dbgFlags = debug_get_num_option("NV50_PROG_DEBUG", 0);
#else
info->optLevel = 3;
#endif
ret = nv50_ir_generate_code(info);
if (ret) {
NOUVEAU_ERR("shader translation failed: %i\n", ret);
goto out;
}
FREE(info->bin.syms);
prog->code = info->bin.code;
prog->code_size = info->bin.codeSize;
prog->fixups = info->bin.relocData;
prog->max_gpr = MAX2(4, (info->bin.maxGPR >> 1) + 1);
prog->tls_space = info->bin.tlsSpace;
if (prog->type == PIPE_SHADER_FRAGMENT) {
if (info->prop.fp.writesDepth) {
prog->fp.flags[0] |= NV50_3D_FP_CONTROL_EXPORTS_Z;
prog->fp.flags[1] = 0x11;
}
if (info->prop.fp.usesDiscard)
prog->fp.flags[0] |= NV50_3D_FP_CONTROL_USES_KIL;
}
if (prog->pipe.stream_output.num_outputs)
prog->so = nv50_program_create_strmout_state(info,
&prog->pipe.stream_output);
out:
FREE(info);
return !ret;
}
bool
nv50_program_translate(struct nv50_program *prog, uint16_t chipset)
{
struct nv50_ir_prog_info *info;
int ret;
const uint8_t map_undef = (prog->type == PIPE_SHADER_VERTEX) ? 0x40 : 0x80;
info = CALLOC_STRUCT(nv50_ir_prog_info);
if (!info)
return false;
info->type = prog->type;
info->target = chipset;
info->bin.sourceRep = NV50_PROGRAM_IR_TGSI;
info->bin.source = (void *)prog->pipe.tokens;
info->io.ucpCBSlot = 15;
info->io.ucpBase = NV50_CB_AUX_UCP_OFFSET;
info->io.genUserClip = prog->vp.clpd_nr;
info->io.sampleInterp = prog->fp.sample_interp;
info->io.resInfoCBSlot = 15;
info->io.suInfoBase = NV50_CB_AUX_TEX_MS_OFFSET;
info->io.sampleInfoBase = NV50_CB_AUX_SAMPLE_OFFSET;
info->io.msInfoCBSlot = 15;
info->io.msInfoBase = NV50_CB_AUX_MS_OFFSET;
info->assignSlots = nv50_program_assign_varying_slots;
prog->vp.bfc[0] = 0xff;
prog->vp.bfc[1] = 0xff;
prog->vp.edgeflag = 0xff;
prog->vp.clpd[0] = map_undef;
prog->vp.clpd[1] = map_undef;
prog->vp.psiz = map_undef;
prog->gp.has_layer = 0;
prog->gp.has_viewport = 0;
info->driverPriv = prog;
#ifdef DEBUG
info->optLevel = debug_get_num_option("NV50_PROG_OPTIMIZE", 3);
info->dbgFlags = debug_get_num_option("NV50_PROG_DEBUG", 0);
#else
info->optLevel = 3;
#endif
ret = nv50_ir_generate_code(info);
if (ret) {
NOUVEAU_ERR("shader translation failed: %i\n", ret);
goto out;
}
FREE(info->bin.syms);
prog->code = info->bin.code;
prog->code_size = info->bin.codeSize;
prog->fixups = info->bin.relocData;
prog->max_gpr = MAX2(4, (info->bin.maxGPR >> 1) + 1);
prog->tls_space = info->bin.tlsSpace;
if (prog->type == PIPE_SHADER_FRAGMENT) {
if (info->prop.fp.writesDepth) {
prog->fp.flags[0] |= NV50_3D_FP_CONTROL_EXPORTS_Z;
prog->fp.flags[1] = 0x11;
}
if (info->prop.fp.usesDiscard)
prog->fp.flags[0] |= NV50_3D_FP_CONTROL_USES_KIL;
} else
if (prog->type == PIPE_SHADER_GEOMETRY) {
switch (info->prop.gp.outputPrim) {
case PIPE_PRIM_LINE_STRIP:
prog->gp.prim_type = NV50_3D_GP_OUTPUT_PRIMITIVE_TYPE_LINE_STRIP;
break;
case PIPE_PRIM_TRIANGLE_STRIP:
prog->gp.prim_type = NV50_3D_GP_OUTPUT_PRIMITIVE_TYPE_TRIANGLE_STRIP;
break;
case PIPE_PRIM_POINTS:
default:
assert(info->prop.gp.outputPrim == PIPE_PRIM_POINTS);
prog->gp.prim_type = NV50_3D_GP_OUTPUT_PRIMITIVE_TYPE_POINTS;
break;
}
prog->gp.vert_count = info->prop.gp.maxVertices;
}
if (prog->pipe.stream_output.num_outputs)
prog->so = nv50_program_create_strmout_state(info,
&prog->pipe.stream_output);
out:
FREE(info);
return !ret;
}
