The same effect can be achieved using a combination of ::stride and
::subreg_offset. Remove the less flexible ::smear to keep the data
members of fs_reg orthogonal.
---
src/mesa/drivers/dri/i965/brw_fs.cpp | 23 +++++++++++++---------
src/mesa/drivers/dri/i965/brw_fs.h | 4 +++-
.../drivers/dri/i965/brw_fs_copy_propagation.cpp | 4 +---
src/mesa/drivers/dri/i965/brw_fs_generator.cpp | 4 ++--
.../drivers/dri/i965/brw_fs_live_variables.cpp | 2 +-
src/mesa/drivers/dri/i965/brw_fs_reg_allocate.cpp | 4 ++--
src/mesa/drivers/dri/i965/brw_fs_visitor.cpp | 14 ++++++-------
7 files changed, 30 insertions(+), 25 deletions(-)
diff --git a/src/mesa/drivers/dri/i965/brw_fs.cpp
b/src/mesa/drivers/dri/i965/brw_fs.cpp
index e4cee33..e6e32fc 100644
--- a/src/mesa/drivers/dri/i965/brw_fs.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs.cpp
@@ -379,7 +379,6 @@ void
fs_reg::init()
{
memset(this, 0, sizeof(*this));
- this->smear = -1;
stride = 1;
}
@@ -445,7 +444,6 @@ fs_reg::equals(const fs_reg &r) const
!reladdr && !r.reladdr &&
memcmp(&fixed_hw_reg, &r.fixed_hw_reg,
sizeof(fixed_hw_reg)) == 0 &&
- smear == r.smear &&
stride == r.stride &&
imm.u == r.imm.u);
}
@@ -468,6 +466,15 @@ fs_reg::apply_stride(unsigned stride)
return *this;
}
+fs_reg &
+fs_reg::set_smear(unsigned subreg)
+{
+ assert(file != HW_REG && file != IMM);
+ subreg_offset = subreg * type_sz(type);
+ stride = 0;
+ return *this;
+}
+
bool
fs_reg::is_contiguous() const
{
@@ -545,7 +552,7 @@ fs_visitor::get_timestamp()
* else that might disrupt timing) by setting smear to 2 and checking if
* that field is != 0.
*/
- dst.smear = 0;
+ dst.set_smear(0);
return dst;
}
@@ -580,7 +587,7 @@ fs_visitor::emit_shader_time_end()
* were the only two timestamp reads that happened).
*/
fs_reg reset = shader_end_time;
- reset.smear = 2;
+ reset.set_smear(2);
fs_inst *test = emit(AND(reg_null_d, reset, fs_reg(1u)));
test->conditional_mod = BRW_CONDITIONAL_Z;
emit(IF(BRW_PREDICATE_NORMAL));
@@ -1934,7 +1941,7 @@ fs_visitor::setup_pull_constants()
inst->src[i].file = GRF;
inst->src[i].reg = dst.reg;
inst->src[i].reg_offset = 0;
- inst->src[i].smear = pull_index & 3;
+ inst->src[i].set_smear(pull_index & 3);
}
}
}
@@ -2263,8 +2270,7 @@ fs_visitor::register_coalesce_2()
inst->src[0].file != GRF ||
inst->src[0].negate ||
inst->src[0].abs ||
- inst->src[0].smear != -1 ||
- !inst->src[0].is_contiguous() ||
+ !inst->src[0].is_contiguous() ||
inst->dst.file != GRF ||
inst->dst.type != inst->src[0].type ||
virtual_grf_sizes[inst->src[0].reg] != 1) {
@@ -2356,7 +2362,6 @@ fs_visitor::register_coalesce()
bool has_source_modifiers = (inst->src[0].abs ||
inst->src[0].negate ||
- inst->src[0].smear != -1 ||
!inst->src[0].is_contiguous() ||
inst->src[0].file == UNIFORM);
@@ -2480,7 +2485,7 @@ fs_visitor::compute_to_mrf()
inst->dst.file != MRF || inst->src[0].file != GRF ||
inst->dst.type != inst->src[0].type ||
inst->src[0].abs || inst->src[0].negate ||
- inst->src[0].smear != -1 || !inst->src[0].is_contiguous() ||
+ !inst->src[0].is_contiguous() ||
inst->src[0].subreg_offset)
continue;
diff --git a/src/mesa/drivers/dri/i965/brw_fs.h
b/src/mesa/drivers/dri/i965/brw_fs.h
index b0ce812..e36943c 100644
--- a/src/mesa/drivers/dri/i965/brw_fs.h
+++ b/src/mesa/drivers/dri/i965/brw_fs.h
@@ -81,10 +81,12 @@ public:
fs_reg retype(uint32_t type);
fs_reg &apply_stride(unsigned stride);
+ /** Smear a channel of the reg to all channels. */
+ fs_reg &set_smear(unsigned subreg);
+
bool negate;
bool abs;
bool sechalf;
- int smear; /* -1, or a channel of the reg to smear to all channels. */
int subreg_offset; /**< Offset in bytes from the start of the register. */
int stride; /**< Register region horizontal stride */
diff --git a/src/mesa/drivers/dri/i965/brw_fs_copy_propagation.cpp
b/src/mesa/drivers/dri/i965/brw_fs_copy_propagation.cpp
index 2f2d6b6..6b67d39 100644
--- a/src/mesa/drivers/dri/i965/brw_fs_copy_propagation.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs_copy_propagation.cpp
@@ -298,7 +298,7 @@ fs_visitor::try_copy_propagate(fs_inst *inst, int arg,
acp_entry *entry)
bool has_source_modifiers = entry->src.abs || entry->src.negate;
if ((has_source_modifiers || entry->src.file == UNIFORM ||
- entry->src.smear != -1 || !entry->src.is_contiguous()) &&
+ !entry->src.is_contiguous()) &&
!can_do_source_mods(inst))
return false;
@@ -311,8 +311,6 @@ fs_visitor::try_copy_propagate(fs_inst *inst, int arg,
acp_entry *entry)
inst->src[arg].file = entry->src.file;
inst->src[arg].reg = entry->src.reg;
inst->src[arg].reg_offset = entry->src.reg_offset;
- if (entry->src.smear != -1)
- inst->src[arg].smear = entry->src.smear;
inst->src[arg].subreg_offset = entry->src.subreg_offset;
inst->src[arg].stride *= entry->src.stride;
diff --git a/src/mesa/drivers/dri/i965/brw_fs_generator.cpp
b/src/mesa/drivers/dri/i965/brw_fs_generator.cpp
index 66321bd..3972fda 100644
--- a/src/mesa/drivers/dri/i965/brw_fs_generator.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs_generator.cpp
@@ -988,8 +988,8 @@ brw_reg_from_fs_reg(fs_reg *reg)
switch (reg->file) {
case GRF:
case MRF:
- if (reg->stride == 0 || reg->smear >= 0) {
- brw_reg = brw_vec1_reg(brw_file_from_reg(reg), reg->reg, reg->smear);
+ if (reg->stride == 0) {
+ brw_reg = brw_vec1_reg(brw_file_from_reg(reg), reg->reg, 0);
} else {
brw_reg = brw_vec8_reg(brw_file_from_reg(reg), reg->reg, 0);
brw_reg = stride(brw_reg, 8 * reg->stride, 8, reg->stride);
diff --git a/src/mesa/drivers/dri/i965/brw_fs_live_variables.cpp
b/src/mesa/drivers/dri/i965/brw_fs_live_variables.cpp
index 1a2d398..46fdd74 100644
--- a/src/mesa/drivers/dri/i965/brw_fs_live_variables.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs_live_variables.cpp
@@ -84,7 +84,7 @@ fs_live_variables::setup_one_read(bblock_t *block, fs_inst
*inst,
* would get stomped by the first decode as well.
*/
int end_ip = ip;
- if (v->dispatch_width == 16 && (reg.smear != -1 || reg.stride == 0 ||
+ if (v->dispatch_width == 16 && (reg.stride == 0 ||
(v->pixel_x.reg == reg.reg ||
v->pixel_y.reg == reg.reg))) {
end_ip++;
diff --git a/src/mesa/drivers/dri/i965/brw_fs_reg_allocate.cpp
b/src/mesa/drivers/dri/i965/brw_fs_reg_allocate.cpp
index 3290004..c04944e 100644
--- a/src/mesa/drivers/dri/i965/brw_fs_reg_allocate.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs_reg_allocate.cpp
@@ -591,7 +591,7 @@ fs_visitor::choose_spill_reg(struct ra_graph *g)
* loading pull constants, so spilling them is unlikely to reduce
* register pressure anyhow.
*/
- if (inst->src[i].smear >= 0 || !inst->src[i].is_contiguous()) {
+ if (!inst->src[i].is_contiguous()) {
no_spill[inst->src[i].reg] = true;
}
}
@@ -600,7 +600,7 @@ fs_visitor::choose_spill_reg(struct ra_graph *g)
if (inst->dst.file == GRF) {
spill_costs[inst->dst.reg] += inst->regs_written * loop_scale;
- if (inst->dst.smear >= 0 || !inst->dst.is_contiguous()) {
+ if (!inst->dst.is_contiguous()) {
no_spill[inst->dst.reg] = true;
}
}
diff --git a/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp
b/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp
index 3f065e7..8a76094 100644
--- a/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp
@@ -726,8 +726,14 @@ fs_visitor::visit(ir_expression *ir)
emit(fs_inst(FS_OPCODE_UNIFORM_PULL_CONSTANT_LOAD,
packed_consts, surf_index, const_offset_reg));
- packed_consts.smear = const_offset->value.u[0] % 16 / 4;
for (int i = 0; i < ir->type->vector_elements; i++) {
+ packed_consts.set_smear(const_offset->value.u[0] % 16 / 4 + i);
+
+ /* The std140 packing rules don't allow vectors to cross 16-byte
+ * boundaries, and a reg is 32 bytes.
+ */
+ assert(packed_consts.subreg_offset < 32);
+
/* UBO bools are any nonzero value. We consider bools to be
* values with the low bit set to 1. Convert them using CMP.
*/
@@ -737,13 +743,7 @@ fs_visitor::visit(ir_expression *ir)
emit(MOV(result, packed_consts));
}
- packed_consts.smear++;
result.reg_offset++;
-
- /* The std140 packing rules don't allow vectors to cross 16-byte
- * boundaries, and a reg is 32 bytes.
- */
- assert(packed_consts.smear < 8);
}
} else {
/* Turn the byte offset into a dword offset. */
--
1.8.3.4
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev
