Mesa (master): i965/vec4: Relax writemask condition in CSE

[Ian Romanick]

Module: Mesa
Branch: master
Commit: 360899d4577a2431dc73b5c702d60ac6bd59ca07
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=360899d4577a2431dc73b5c702d60ac6bd59ca07

Author: Ian Romanick <ian.d.romanick at intel.com>
Date:   Fri Feb 16 17:26:11 2018 -0800

i965/vec4: Relax writemask condition in CSE

If the previously seen instruction generates more fields than the new
instruction, still allow CSE to happen.  This doesn't do much, but it
also enables a couple more shaders in the next patch.  It helped quite a
bit in another change series that I have (at least for now) abandoned.

v2: Add some extra comentary about the parameters to instructions_match.
Suggested by Ken.

No changes on Skylake, Broadwell, Iron Lake or GM45.

Ivy Bridge and Haswell had similar results. (Ivy Bridge shown)
total instructions in shared programs: 11780295 -> 11780294 (<.01%)
instructions in affected programs: 302 -> 301 (-0.33%)
helped: 1
HURT: 0

total cycles in shared programs: 257308315 -> 257308313 (<.01%)
cycles in affected programs: 2074 -> 2072 (-0.10%)
helped: 1
HURT: 0

Sandy Bridge
total instructions in shared programs: 10506687 -> 10506686 (<.01%)
instructions in affected programs: 335 -> 334 (-0.30%)
helped: 1
HURT: 0

Signed-off-by: Ian Romanick <ian.d.romanick at intel.com>
Reviewed-by: Kenneth Graunke <kenneth at whitecape.org>

---

 src/intel/compiler/brw_vec4_cse.cpp | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/src/intel/compiler/brw_vec4_cse.cpp b/src/intel/compiler/brw_vec4_cse.cpp
index 2e65ef7854..d9f08c9631 100644
--- a/src/intel/compiler/brw_vec4_cse.cpp
+++ b/src/intel/compiler/brw_vec4_cse.cpp
@@ -112,6 +112,14 @@ operands_match(const vec4_instruction *a, const vec4_instruction *b)
    }
 }
 
+/**
+ * Checks if instructions match, exactly for sources, but loosely for
+ * destination writemasks.
+ *
+ * \param 'a' is the generating expression from the AEB entry.
+ * \param 'b' is the second occurrence of the expression that we're
+ *        considering eliminating.
+ */
 static bool
 instructions_match(vec4_instruction *a, vec4_instruction *b)
 {
@@ -127,7 +135,7 @@ instructions_match(vec4_instruction *a, vec4_instruction *b)
           a->base_mrf == b->base_mrf &&
           a->header_size == b->header_size &&
           a->shadow_compare == b->shadow_compare &&
-          a->dst.writemask == b->dst.writemask &&
+          ((a->dst.writemask & b->dst.writemask) == a->dst.writemask) &&
           a->force_writemask_all == b->force_writemask_all &&
           a->size_written == b->size_written &&
           a->exec_size == b->exec_size &&