We need our reads to provide well-defined results for all enabled
channels even for helper invocations, which means that we should not
use a pixel mask with them.
---
src/mesa/drivers/dri/i965/brw_eu_emit.c | 4 ++--
src/mesa/drivers/dri/i965/brw_fs_visitor.cpp | 35 ++++------------------------
2 files changed, 7 insertions(+), 32 deletions(-)
diff --git a/src/mesa/drivers/dri/i965/brw_eu_emit.c
b/src/mesa/drivers/dri/i965/brw_eu_emit.c
index 4d39762..448bb49 100644
--- a/src/mesa/drivers/dri/i965/brw_eu_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_eu_emit.c
@@ -2886,11 +2886,11 @@ brw_untyped_surface_read(struct brw_codegen *p,
const unsigned sfid = (devinfo->gen >= 8 || devinfo->is_haswell ?
HSW_SFID_DATAPORT_DATA_CACHE_1 :
GEN7_SFID_DATAPORT_DATA_CACHE);
- const bool align1 = (brw_inst_access_mode(devinfo, p->current) ==
BRW_ALIGN_1);
+
struct brw_inst *insn = brw_send_indirect_surface_message(
p, sfid, dst, payload, surface, msg_length,
brw_surface_payload_size(p, num_channels, true, true),
- align1);
+ false);
brw_set_dp_untyped_surface_read_message(
p, insn, num_channels);
diff --git a/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp
b/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp
index 94d6a58..cf35556 100644
--- a/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp
@@ -1185,41 +1185,16 @@ fs_visitor::emit_untyped_surface_read(unsigned
surf_index, fs_reg dst,
{
int reg_width = dispatch_width / 8;
- fs_reg *sources = ralloc_array(mem_ctx, fs_reg, 2);
-
- sources[0] = fs_reg(GRF, alloc.allocate(1), BRW_REGISTER_TYPE_UD);
- /* Initialize the sample mask in the message header. */
- bld.exec_all()
- .MOV(sources[0], fs_reg(0u));
-
- if (stage == MESA_SHADER_FRAGMENT) {
- if (((brw_wm_prog_data*)this->prog_data)->uses_kill) {
- bld.exec_all()
- .MOV(component(sources[0], 7), brw_flag_reg(0, 1));
- } else {
- bld.exec_all()
- .MOV(component(sources[0], 7),
- retype(brw_vec1_grf(1, 7), BRW_REGISTER_TYPE_UD));
- }
- } else {
- /* The execution mask is part of the side-band information sent together
with
- * the message payload to the data port. It's implicitly ANDed with the
sample
- * mask sent in the header to compute the actual set of channels that
execute
- * the atomic operation.
- */
- assert(stage == MESA_SHADER_VERTEX || stage == MESA_SHADER_COMPUTE);
- bld.exec_all()
- .MOV(component(sources[0], 7), fs_reg(0xffffu));
- }
+ fs_reg *sources = ralloc_array(mem_ctx, fs_reg, 1);
/* Set the surface read offset. */
- sources[1] = vgrf(glsl_type::uint_type);
- bld.MOV(sources[1], offset);
+ sources[0] = vgrf(glsl_type::uint_type);
+ bld.MOV(sources[0], offset);
- int mlen = 1 + reg_width;
+ int mlen = reg_width;
fs_reg src_payload = fs_reg(GRF, alloc.allocate(mlen),
BRW_REGISTER_TYPE_UD);
- fs_inst *inst = bld.LOAD_PAYLOAD(src_payload, sources, 2, 1);
+ fs_inst *inst = bld.LOAD_PAYLOAD(src_payload, sources, 1, 0);
/* Emit the instruction. */
inst = bld.emit(SHADER_OPCODE_UNTYPED_SURFACE_READ, dst, src_payload,
--
1.9.1
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev
