[Mesa-dev] [PATCH 17/22] i965/vec4: fix register_coalesce() for partial writes

[Samuel Iglesias Gonsálvez]

From: "Juan A. Suarez Romero" <jasuarez at igalia.com>

When lowering double_to_single() we added a final mov() that puts 32-bit
values from one register in the second half of destination.
---
 src/mesa/drivers/dri/i965/brw_vec4.cpp                  | 5 +++++
 src/mesa/drivers/dri/i965/brw_vec4_copy_propagation.cpp | 3 ++-
 2 files changed, 7 insertions(+), 1 deletion(-)

diff --git a/src/mesa/drivers/dri/i965/brw_vec4.cpp b/src/mesa/drivers/dri/i965/brw_vec4.cpp
index afabc22..c654c8b 100644
--- a/src/mesa/drivers/dri/i965/brw_vec4.cpp
+++ b/src/mesa/drivers/dri/i965/brw_vec4.cpp
@@ -1223,6 +1223,11 @@ vec4_visitor::opt_register_coalesce()
                }
             }
 
+            /* Disallow partial register writes. */
+            if (scan_inst->exec_size < 8 &&
+                type_sz(scan_inst->src[0].type) == 4)
+               break;
+
             /* This doesn't handle saturation on the instruction we
              * want to coalesce away if the register types do not match.
              * But if scan_inst is a non type-converting 'mov', we can fix
diff --git a/src/mesa/drivers/dri/i965/brw_vec4_copy_propagation.cpp b/src/mesa/drivers/dri/i965/brw_vec4_copy_propagation.cpp
index e7f6f93..f5c10d1 100644
--- a/src/mesa/drivers/dri/i965/brw_vec4_copy_propagation.cpp
+++ b/src/mesa/drivers/dri/i965/brw_vec4_copy_propagation.cpp
@@ -499,7 +499,8 @@ vec4_visitor::opt_copy_propagation(bool do_constant_prop)
 
          /* We only handle register-aligned single GRF copies. */
          if (inst->size_read(i) != REG_SIZE ||
-             inst->src[i].offset % REG_SIZE)
+             inst->src[i].offset % REG_SIZE ||
+             inst->dst.offset % REG_SIZE)
             continue;
 
          const unsigned reg = (alloc.offsets[inst->src[i].nr] +
-- 
2.9.3