v2: less invasive changes
---
src/mesa/drivers/dri/i965/brw_fs.cpp | 4 +-
src/mesa/drivers/dri/i965/brw_vec4.h | 5 ++-
src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp | 60 +++++++++++++++--------
3 files changed, 45 insertions(+), 24 deletions(-)
diff --git a/src/mesa/drivers/dri/i965/brw_fs.cpp
b/src/mesa/drivers/dri/i965/brw_fs.cpp
index bf59da3..55cbf9e 100644
--- a/src/mesa/drivers/dri/i965/brw_fs.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs.cpp
@@ -462,7 +462,7 @@ fs_visitor::emit_general_interpolation(ir_variable *ir)
* field of the setup reg.
*/
for (unsigned int k = 0; k < type->vector_elements; k++) {
- struct brw_reg interp = interp_reg(location, k);
+ struct brw_reg interp = interp_reg(location, k +
ir->horizontal_location);
interp = suboffset(interp, 3);
interp.type = reg->type;
emit(FS_OPCODE_CINTERP, attr, fs_reg(interp));
@@ -482,7 +482,7 @@ fs_visitor::emit_general_interpolation(ir_variable *ir)
k == 3 && !(c->key.proj_attrib_mask & (1 << location))) {
emit(BRW_OPCODE_MOV, attr, fs_reg(1.0f));
} else {
- struct brw_reg interp = interp_reg(location, k);
+ struct brw_reg interp = interp_reg(location, k +
ir->horizontal_location);
brw_wm_barycentric_interp_mode barycoord_mode;
if (interpolation_mode == INTERP_QUALIFIER_SMOOTH)
barycoord_mode = BRW_WM_PERSPECTIVE_PIXEL_BARYCENTRIC;
diff --git a/src/mesa/drivers/dri/i965/brw_vec4.h
b/src/mesa/drivers/dri/i965/brw_vec4.h
index bc8b392..c7b2665 100644
--- a/src/mesa/drivers/dri/i965/brw_vec4.h
+++ b/src/mesa/drivers/dri/i965/brw_vec4.h
@@ -387,7 +387,10 @@ public:
/* Regs for vertex results. Generated at ir_variable visiting time
* for the ir->location's used.
*/
- dst_reg output_reg[BRW_VERT_RESULT_MAX];
+ dst_reg output_reg[4 * BRW_VERT_RESULT_MAX];
+ unsigned horizontal_shift[4 * BRW_VERT_RESULT_MAX];
+ unsigned components[4 * BRW_VERT_RESULT_MAX];
+ unsigned output_per_ref[BRW_VERT_RESULT_MAX];
const char *output_reg_annotation[BRW_VERT_RESULT_MAX];
int uniform_size[MAX_UNIFORMS];
int uniform_vector_size[MAX_UNIFORMS];
diff --git a/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp
b/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp
index f9a08a0..776b93f 100644
--- a/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp
+++ b/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp
@@ -857,10 +857,14 @@ vec4_visitor::visit(ir_variable *ir)
reg = new(mem_ctx) dst_reg(this, ir->type);
for (int i = 0; i < type_size(ir->type); i++) {
- output_reg[ir->location + i] = *reg;
- output_reg[ir->location + i].reg_offset = i;
- output_reg[ir->location + i].type =
+ unsigned location_and_lateral_offset = 4 * ir->location + 4 * i +
output_per_ref[ir->location + i];
+ output_per_ref[ir->location + i] += 1;
+ output_reg[location_and_lateral_offset] = *reg;
+ output_reg[location_and_lateral_offset].reg_offset = i;
+ output_reg[location_and_lateral_offset].type =
brw_type_for_base_type(ir->type->get_scalar_type());
+ horizontal_shift[location_and_lateral_offset] = ir->horizontal_location;
+ components[location_and_lateral_offset] =
(ir->type->is_array())?ir->type->fields.array->vector_elements:ir->type->vector_elements;
output_reg_annotation[ir->location + i] = ir->name;
}
break;
@@ -2020,11 +2024,11 @@ void
vec4_visitor::emit_ndc_computation()
{
/* Get the position */
- src_reg pos = src_reg(output_reg[VERT_RESULT_HPOS]);
+ src_reg pos = src_reg(output_reg[4 * VERT_RESULT_HPOS]);
/* Build ndc coords, which are (x/w, y/w, z/w, 1/w) */
dst_reg ndc = dst_reg(this, glsl_type::vec4_type);
- output_reg[BRW_VERT_RESULT_NDC] = ndc;
+ output_reg[4 * BRW_VERT_RESULT_NDC] = ndc;
current_annotation = "NDC";
dst_reg ndc_w = ndc;
@@ -2053,7 +2057,7 @@ vec4_visitor::emit_psiz_and_flags(struct brw_reg reg)
emit(MOV(header1, 0u));
if (c->prog_data.outputs_written & BITFIELD64_BIT(VERT_RESULT_PSIZ)) {
- src_reg psiz = src_reg(output_reg[VERT_RESULT_PSIZ]);
+ src_reg psiz = src_reg(output_reg[4 * VERT_RESULT_PSIZ]);
current_annotation = "Point size";
emit(MUL(header1_w, psiz, src_reg((float)(1 << 11))));
@@ -2064,7 +2068,7 @@ vec4_visitor::emit_psiz_and_flags(struct brw_reg reg)
for (i = 0; i < c->key.nr_userclip_plane_consts; i++) {
vec4_instruction *inst;
- inst = emit(DP4(dst_null_f(), src_reg(output_reg[VERT_RESULT_HPOS]),
+ inst = emit(DP4(dst_null_f(), src_reg(output_reg[4 * VERT_RESULT_HPOS]),
src_reg(this->userplane[i])));
inst->conditional_mod = BRW_CONDITIONAL_L;
@@ -2103,7 +2107,7 @@ vec4_visitor::emit_psiz_and_flags(struct brw_reg reg)
emit(MOV(retype(reg, BRW_REGISTER_TYPE_D), src_reg(0)));
if (c->prog_data.outputs_written & BITFIELD64_BIT(VERT_RESULT_PSIZ)) {
emit(MOV(brw_writemask(reg, WRITEMASK_W),
- src_reg(output_reg[VERT_RESULT_PSIZ])));
+ src_reg(output_reg[4 * VERT_RESULT_PSIZ])));
}
}
}
@@ -2140,7 +2144,7 @@ vec4_visitor::emit_clip_distances(struct brw_reg reg, int
offset)
for (int i = 0; i + offset < c->key.nr_userclip_plane_consts && i < 4;
++i) {
emit(DP4(dst_reg(brw_writemask(reg, 1 << i)),
- src_reg(output_reg[clip_vertex]),
+ src_reg(output_reg[4 * clip_vertex]),
src_reg(this->userplane[i + offset])));
}
}
@@ -2149,17 +2153,30 @@ void
vec4_visitor::emit_generic_urb_slot(dst_reg reg, int vert_result)
{
assert (vert_result < VERT_RESULT_MAX);
- reg.type = output_reg[vert_result].type;
+ reg.type = output_reg[4 * vert_result].type;
current_annotation = output_reg_annotation[vert_result];
- /* Copy the register, saturating if necessary */
- vec4_instruction *inst = emit(MOV(reg,
- src_reg(output_reg[vert_result])));
- if ((vert_result == VERT_RESULT_COL0 ||
- vert_result == VERT_RESULT_COL1 ||
- vert_result == VERT_RESULT_BFC0 ||
- vert_result == VERT_RESULT_BFC1) &&
- c->key.clamp_vertex_color) {
- inst->saturate = true;
+ for (unsigned i = 0; i < output_per_ref[vert_result]; i++) {
+ unsigned components_count = components[4 * vert_result + i], hs =
horizontal_shift[4 * vert_result + i];
+ reg.writemask = ((1 << components_count) - 1 ) << hs;
+ src_reg src_gpr = src_reg(output_reg[4 * vert_result + i]);
+ unsigned swz = src_gpr.swizzle;
+ src_gpr.swizzle = BRW_SWIZZLE4(
+ BRW_GET_SWZ(swz,(0 + hs) % 4),
+ BRW_GET_SWZ(swz,(1 + hs) % 4),
+ BRW_GET_SWZ(swz,(2 + hs) % 4),
+ BRW_GET_SWZ(swz,(3 + hs) % 4)
+ );
+
+
+ /* Copy the register, saturating if necessary */
+ vec4_instruction *inst = emit(MOV(reg, src_gpr));
+ if ((vert_result == VERT_RESULT_COL0 ||
+ vert_result == VERT_RESULT_COL1 ||
+ vert_result == VERT_RESULT_BFC0 ||
+ vert_result == VERT_RESULT_BFC1) &&
+ c->key.clamp_vertex_color) {
+ inst->saturate = true;
+ }
}
}
@@ -2178,12 +2195,12 @@ vec4_visitor::emit_urb_slot(int mrf, int vert_result)
break;
case BRW_VERT_RESULT_NDC:
current_annotation = "NDC";
- emit(MOV(reg, src_reg(output_reg[BRW_VERT_RESULT_NDC])));
+ emit(MOV(reg, src_reg(output_reg[4 *BRW_VERT_RESULT_NDC])));
break;
case BRW_VERT_RESULT_HPOS_DUPLICATE:
case VERT_RESULT_HPOS:
current_annotation = "gl_Position";
- emit(MOV(reg, src_reg(output_reg[VERT_RESULT_HPOS])));
+ emit(MOV(reg, src_reg(output_reg[4 * VERT_RESULT_HPOS])));
break;
case VERT_RESULT_CLIP_DIST0:
case VERT_RESULT_CLIP_DIST1:
@@ -2615,6 +2632,7 @@ vec4_visitor::vec4_visitor(struct brw_vs_compile *c,
this->max_grf = intel->gen >= 7 ? GEN7_MRF_HACK_START : BRW_MAX_GRF;
this->uniforms = 0;
+ memset(output_per_ref, 0, sizeof(output_per_ref));
}
vec4_visitor::~vec4_visitor()
--
1.7.7
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev
