elem_types[DRAW_JIT_DVBUFFER_SIZE] = int32_type;
dvbuffer_type = LLVMStructTypeInContext(gallivm->context, elem_types,
- Elements(elem_types), 0);
+ ARRAY_SIZE(elem_types), 0);
(void) target; /* silence unused var warning for non-debug build */
LP_CHECK_MEMBER_OFFSET(struct draw_vertex_buffer, map,
LLVMArrayType(int32_type, PIPE_MAX_TEXTURE_LEVELS);
texture_type = LLVMStructTypeInContext(gallivm->context, elem_types,
- Elements(elem_types), 0);
+ ARRAY_SIZE(elem_types), 0);
(void) target; /* silence unused var warning for non-debug build */
LP_CHECK_MEMBER_OFFSET(struct draw_jit_texture, width,
LLVMArrayType(LLVMFloatTypeInContext(gallivm->context), 4);
sampler_type = LLVMStructTypeInContext(gallivm->context, elem_types,
- Elements(elem_types), 0);
+ ARRAY_SIZE(elem_types), 0);
(void) target; /* silence unused var warning for non-debug build */
LP_CHECK_MEMBER_OFFSET(struct draw_jit_sampler, min_lod,
elem_types[5] = LLVMArrayType(sampler_type,
PIPE_MAX_SAMPLERS); /* samplers */
context_type = LLVMStructTypeInContext(gallivm->context, elem_types,
- Elements(elem_types), 0);
+ ARRAY_SIZE(elem_types), 0);
(void) target; /* silence unused var warning for non-debug build */
LP_CHECK_MEMBER_OFFSET(struct draw_jit_context, vs_constants,
vector_length), 0);
context_type = LLVMStructTypeInContext(gallivm->context, elem_types,
- Elements(elem_types), 0);
+ ARRAY_SIZE(elem_types), 0);
(void) target; /* silence unused var warning for non-debug build */
LP_CHECK_MEMBER_OFFSET(struct draw_gs_jit_context, constants,
elem_types[3] = LLVMPointerType(LLVMInt8TypeInContext(gallivm->context), 0);
vb_type = LLVMStructTypeInContext(gallivm->context, elem_types,
- Elements(elem_types), 0);
+ ARRAY_SIZE(elem_types), 0);
(void) target; /* silence unused var warning for non-debug build */
LP_CHECK_MEMBER_OFFSET(struct pipe_vertex_buffer, stride,
elem_types[DRAW_JIT_VERTEX_DATA] = LLVMArrayType(elem_types[1], data_elems);
vertex_header = LLVMStructTypeInContext(gallivm->context, elem_types,
- Elements(elem_types), 0);
+ ARRAY_SIZE(elem_types), 0);
/* these are bit-fields and we can't take address of them
LP_CHECK_MEMBER_OFFSET(struct vertex_header, clipmask,
#endif
/* Unaligned store due to the vertex header */
- lp_set_store_alignment(LLVMBuildStore(builder, value, data_ptr), sizeof(float));
+ LLVMSetAlignment(LLVMBuildStore(builder, value, data_ptr), sizeof(float));
}
/**
LLVMValueRef* aos,
int attrib,
int num_outputs,
- LLVMValueRef clipmask)
+ LLVMValueRef clipmask,
+ boolean need_edgeflag)
{
LLVMBuilderRef builder = gallivm->builder;
LLVMValueRef attr_index = lp_build_const_int32(gallivm, attrib);
*/
assert(DRAW_TOTAL_CLIP_PLANES==14);
/* initialize vertex id:16 = 0xffff, pad:1 = 0, edgeflag:1 = 1 */
- vertex_id_pad_edgeflag = (0xffff << 16) | (1 << DRAW_TOTAL_CLIP_PLANES);
- val = lp_build_const_int_vec(gallivm, lp_int_type(soa_type), vertex_id_pad_edgeflag);
+ if (!need_edgeflag) {
+ vertex_id_pad_edgeflag = (0xffff << 16) | (1 << DRAW_TOTAL_CLIP_PLANES);
+ }
+ else {
+ vertex_id_pad_edgeflag = (0xffff << 16);
+ }
+ val = lp_build_const_int_vec(gallivm, lp_int_type(soa_type),
+ vertex_id_pad_edgeflag);
/* OR with the clipmask */
cliptmp = LLVMBuildOr(builder, val, clipmask, "");
for (i = 0; i < vector_length; i++) {
LLVMValueRef clipmask,
int num_outputs,
struct lp_type soa_type,
- boolean have_clipdist)
+ boolean need_edgeflag)
{
LLVMBuilderRef builder = gallivm->builder;
unsigned chan, attrib, i;
aos,
attrib,
num_outputs,
- clipmask);
+ clipmask,
+ need_edgeflag);
}
#if DEBUG_STORE
lp_build_printf(gallivm, " # storing end\n");
clip_ptr = LLVMBuildPointerCast(builder, clip_ptr, clip_ptr_type, "");
/* Unaligned store */
- lp_set_store_alignment(LLVMBuildStore(builder, aos[j], clip_ptr), sizeof(float));
+ LLVMSetAlignment(LLVMBuildStore(builder, aos[j], clip_ptr), sizeof(float));
}
}
/* divide by w */
out = LLVMBuildFMul(builder, out, out3, "");
- /* mult by scale */
- out = LLVMBuildFMul(builder, out, scale, "");
- /* add translation */
- out = LLVMBuildFAdd(builder, out, trans, "");
+ /* mult by scale, add translation */
+ out = lp_build_fmuladd(builder, out, scale, trans);
/* store transformed outputs */
LLVMBuildStore(builder, out, outputs[pos][i]);
struct gallivm_state *gallivm,
struct lp_type vs_type,
LLVMValueRef (*outputs)[TGSI_NUM_CHANNELS],
- boolean clip_xy,
- boolean clip_z,
- boolean clip_user,
- boolean clip_halfz,
- unsigned ucp_enable,
+ struct draw_llvm_variant_key *key,
LLVMValueRef context_ptr,
boolean *have_clipdist)
{
const unsigned pos = llvm->draw->vs.position_output;
const unsigned cv = llvm->draw->vs.clipvertex_output;
int num_written_clipdistance = llvm->draw->vs.vertex_shader->info.num_written_clipdistance;
- bool have_cd = false;
+ boolean have_cd = false;
+ boolean clip_user = key->clip_user;
+ unsigned ucp_enable = key->ucp_enable;
unsigned cd[2];
- cd[0] = llvm->draw->vs.clipdistance_output[0];
- cd[1] = llvm->draw->vs.clipdistance_output[1];
+ cd[0] = llvm->draw->vs.ccdistance_output[0];
+ cd[1] = llvm->draw->vs.ccdistance_output[1];
if (cd[0] != pos || cd[1] != pos)
have_cd = true;
cv_w = pos_w;
}
+ /*
+ * Be careful with the comparisons and NaNs (using llvm's unordered
+ * comparisons here).
+ */
/* Cliptest, for hardwired planes */
- if (clip_xy) {
+ /*
+ * XXX should take guardband into account (currently not in key).
+ * Otherwise might run the draw pipeline stages for nothing.
+ */
+ if (key->clip_xy) {
/* plane 1 */
test = lp_build_compare(gallivm, f32_type, PIPE_FUNC_GREATER, pos_x , pos_w);
temp = shift;
mask = LLVMBuildOr(builder, mask, test, "");
}
- if (clip_z) {
+ if (key->clip_z) {
temp = lp_build_const_int_vec(gallivm, i32_type, 16);
- if (clip_halfz) {
+ if (key->clip_halfz) {
/* plane 5 */
test = lp_build_compare(gallivm, f32_type, PIPE_FUNC_GREATER, zero, pos_z);
test = LLVMBuildAnd(builder, test, temp, "");
plane_ptr = LLVMBuildGEP(builder, planes_ptr, indices, 3, "");
plane1 = LLVMBuildLoad(builder, plane_ptr, "plane_y");
planes = lp_build_broadcast(gallivm, vs_type_llvm, plane1);
- test = LLVMBuildFMul(builder, planes, cv_y, "");
- sum = LLVMBuildFAdd(builder, sum, test, "");
+ sum = lp_build_fmuladd(builder, planes, cv_y, sum);
indices[2] = lp_build_const_int32(gallivm, 2);
plane_ptr = LLVMBuildGEP(builder, planes_ptr, indices, 3, "");
plane1 = LLVMBuildLoad(builder, plane_ptr, "plane_z");
planes = lp_build_broadcast(gallivm, vs_type_llvm, plane1);
- test = LLVMBuildFMul(builder, planes, cv_z, "");
- sum = LLVMBuildFAdd(builder, sum, test, "");
+ sum = lp_build_fmuladd(builder, planes, cv_z, sum);
indices[2] = lp_build_const_int32(gallivm, 3);
plane_ptr = LLVMBuildGEP(builder, planes_ptr, indices, 3, "");
plane1 = LLVMBuildLoad(builder, plane_ptr, "plane_w");
planes = lp_build_broadcast(gallivm, vs_type_llvm, plane1);
- test = LLVMBuildFMul(builder, planes, cv_w, "");
- sum = LLVMBuildFAdd(builder, sum, test, "");
+ sum = lp_build_fmuladd(builder, planes, cv_w, sum);
test = lp_build_compare(gallivm, f32_type, PIPE_FUNC_GREATER, zero, sum);
temp = lp_build_const_int_vec(gallivm, i32_type, 1LL << plane_idx);
}
}
}
+ if (key->need_edgeflags) {
+ /*
+ * This isn't really part of clipmask but stored the same in vertex
+ * header later, so do it here.
+ */
+ unsigned edge_attr = llvm->draw->vs.edgeflag_output;
+ LLVMValueRef one = lp_build_const_vec(gallivm, f32_type, 1.0);
+ LLVMValueRef edgeflag = LLVMBuildLoad(builder, outputs[edge_attr][0], "");
+ test = lp_build_compare(gallivm, f32_type, PIPE_FUNC_EQUAL, one, edgeflag);
+ temp = lp_build_const_int_vec(gallivm, i32_type,
+ 1LL << DRAW_TOTAL_CLIP_PLANES);
+ test = LLVMBuildAnd(builder, test, temp, "");
+ mask = LLVMBuildOr(builder, mask, test, "");
+ }
return mask;
}
static LLVMValueRef
clipmask_booli32(struct gallivm_state *gallivm,
const struct lp_type vs_type,
- LLVMValueRef clipmask_bool_ptr)
+ LLVMValueRef clipmask_bool_ptr,
+ boolean edgeflag_in_clipmask)
{
LLVMBuilderRef builder = gallivm->builder;
LLVMTypeRef int32_type = LLVMInt32TypeInContext(gallivm->context);
int i;
/*
- * Can do this with log2(vector length) pack instructions and one extract
- * (as we don't actually need a or) with sse2 which would be way better.
+ * We need to invert the edgeflag bit from the clipmask here
+ * (because the result is really if we want to run the pipeline or not
+ * and we (may) need it if edgeflag was 0).
+ */
+ if (edgeflag_in_clipmask) {
+ struct lp_type i32_type = lp_int_type(vs_type);
+ LLVMValueRef edge = lp_build_const_int_vec(gallivm, i32_type,
+ 1LL << DRAW_TOTAL_CLIP_PLANES);
+ clipmask_bool = LLVMBuildXor(builder, clipmask_bool, edge, "");
+ }
+ /*
+ * Could do much better with just cmp/movmskps.
*/
for (i=0; i < vs_type.length; i++) {
temp = LLVMBuildExtractElement(builder, clipmask_bool,
LLVMTypeRef int32_type = LLVMInt32TypeInContext(context);
LLVMTypeRef arg_types[11];
unsigned num_arg_types =
- elts ? Elements(arg_types) : Elements(arg_types) - 1;
+ elts ? ARRAY_SIZE(arg_types) : ARRAY_SIZE(arg_types) - 1;
LLVMTypeRef func_type;
LLVMValueRef context_ptr;
LLVMBasicBlockRef block;
const boolean bypass_viewport = key->has_gs || key->bypass_viewport ||
llvm->draw->vs.vertex_shader->info.writes_viewport_index;
const boolean enable_cliptest = !key->has_gs && (key->clip_xy ||
- key->clip_z ||
- key->clip_user);
+ key->clip_z ||
+ key->clip_user ||
+ key->need_edgeflags);
LLVMValueRef variant_func;
const unsigned pos = llvm->draw->vs.position_output;
const unsigned cv = llvm->draw->vs.clipvertex_output;
context_ptr = LLVMGetParam(variant_func, 0);
io_ptr = LLVMGetParam(variant_func, 1);
vbuffers_ptr = LLVMGetParam(variant_func, 2);
+ /*
+ * XXX: stride is actually unused. The stride we use is strictly calculated
+ * from the number of outputs (including the draw_extra outputs).
+ * Should probably fix some day (we need a new vs just because of extra
+ * outputs which the generated vs won't touch).
+ */
stride = LLVMGetParam(variant_func, 5 + (elts ? 1 : 0));
vb_ptr = LLVMGetParam(variant_func, 6 + (elts ? 1 : 0));
system_values.instance_id = LLVMGetParam(variant_func, 7 + (elts ? 1 : 0));
gallivm,
vs_type,
outputs,
- key->clip_xy,
- key->clip_z,
- key->clip_user,
- key->clip_halfz,
- key->ucp_enable,
+ key,
context_ptr, &have_clipdist);
temp = LLVMBuildOr(builder, clipmask, temp, "");
/* store temporary clipping boolean value */
*/
convert_to_aos(gallivm, io, NULL, outputs, clipmask,
vs_info->num_outputs, vs_type,
- have_clipdist);
+ enable_cliptest && key->need_edgeflags);
}
lp_build_loop_end_cond(&lp_loop, count, step, LLVMIntUGE);
sampler->destroy(sampler);
/* return clipping boolean value for function */
- ret = clipmask_booli32(gallivm, vs_type, clipmask_bool_ptr);
+ ret = clipmask_booli32(gallivm, vs_type, clipmask_bool_ptr,
+ enable_cliptest && key->need_edgeflags);
LLVMBuildRet(builder, ret);
key = (struct draw_llvm_variant_key *)store;
+ memset(key, 0, offsetof(struct draw_llvm_variant_key, vertex_element[0]));
+
key->clamp_vertex_color = llvm->draw->rasterizer->clamp_vertex_color; /**/
/* Presumably all variants of the shader should have the same
key->clip_user = llvm->draw->clip_user;
key->bypass_viewport = llvm->draw->bypass_viewport;
key->clip_halfz = llvm->draw->rasterizer->clip_halfz;
+ /* XXX assumes edgeflag output not at 0 */
key->need_edgeflags = (llvm->draw->vs.edgeflag_output ? TRUE : FALSE);
key->ucp_enable = llvm->draw->rasterizer->clip_plane_enable;
key->has_gs = llvm->draw->gs.geometry_shader != NULL;
key->num_outputs = draw_total_vs_outputs(llvm->draw);
- key->pad1 = 0;
/* All variants of this shader will have the same value for
* nr_samplers. Not yet trying to compact away holes in the
shader_stage == PIPE_SHADER_GEOMETRY);
if (shader_stage == PIPE_SHADER_VERTEX) {
- assert(sview_idx < Elements(draw->llvm->jit_context.textures));
+ assert(sview_idx < ARRAY_SIZE(draw->llvm->jit_context.textures));
jit_tex = &draw->llvm->jit_context.textures[sview_idx];
} else if (shader_stage == PIPE_SHADER_GEOMETRY) {
- assert(sview_idx < Elements(draw->llvm->gs_jit_context.textures));
+ assert(sview_idx < ARRAY_SIZE(draw->llvm->gs_jit_context.textures));
jit_tex = &draw->llvm->gs_jit_context.textures[sview_idx];
} else {
LLVMVectorType(int32_type, vector_length), 0); /* prim_id_ptr */
arg_types[6] = int32_type;
- func_type = LLVMFunctionType(int32_type, arg_types, Elements(arg_types), 0);
+ func_type = LLVMFunctionType(int32_type, arg_types, ARRAY_SIZE(arg_types), 0);
variant_func = LLVMAddFunction(gallivm->module, func_name, func_type);
LLVMSetFunctionCallConv(variant_func, LLVMCCallConv);
- for (i = 0; i < Elements(arg_types); ++i)
+ for (i = 0; i < ARRAY_SIZE(arg_types); ++i)
if (LLVMGetTypeKind(arg_types[i]) == LLVMPointerTypeKind)
LLVMAddAttribute(LLVMGetParam(variant_func, i),
LLVMNoAliasAttribute);
key = (struct draw_gs_llvm_variant_key *)store;
+ memset(key, 0, offsetof(struct draw_gs_llvm_variant_key, samplers[0]));
+
key->num_outputs = draw_total_gs_outputs(llvm->draw);
/* All variants of this shader will have the same value for
projects / mesa.git / blobdiff