[Mesa-dev] [PATCH 2/2] i965: Use is_scheduling_barrier instead of schedule_node::is_barier.

[Kenneth Graunke]

Commit a73116ecc60414ade89802150b tried to make add_barrier_deps()
walk to the next barrier, and stop.  To accomplish that, it added an
is_barrier flag.  Unfortunately, this only works half of the time.

The issue is that add_barrier_deps() walks both backward (to the
previous barrier), and forward (to the next barrier).  It also sets
is_barrier.  Assuming that we're processing instructions in forward
order, this means that is_barrier will be set for previous instructions,
but not future ones.  So we'll never see it, and walk further than we
need to.

dEQP-GLES31.functional.ssbo.layout.random.all_shared_buffer.23
now compiles its shaders in 3.6 seconds instead of 3.3 minutes.
---
 src/intel/compiler/brw_schedule_instructions.cpp | 32 ++++++++----------------
 1 file changed, 10 insertions(+), 22 deletions(-)

diff --git a/src/intel/compiler/brw_schedule_instructions.cpp b/src/intel/compiler/brw_schedule_instructions.cpp
index b3f7e877c80..a1e825c661c 100644
--- a/src/intel/compiler/brw_schedule_instructions.cpp
+++ b/src/intel/compiler/brw_schedule_instructions.cpp
@@ -94,8 +94,6 @@ public:
     * successors is an exit node.
     */
    schedule_node *exit;
-
-   bool is_barrier;
 };
 
 /**
@@ -800,7 +798,6 @@ schedule_node::schedule_node(backend_instruction *inst,
    this->cand_generation = 0;
    this->delay = 0;
    this->exit = NULL;
-   this->is_barrier = false;
 
    /* We can't measure Gen6 timings directly but expect them to be much
     * closer to Gen7 than Gen4.
@@ -921,6 +918,14 @@ instruction_scheduler::add_dep(schedule_node *before, schedule_node *after)
    add_dep(before, after, before->latency);
 }
 
+static bool
+is_scheduling_barrier(const backend_instruction *inst)
+{
+   return inst->opcode == FS_OPCODE_PLACEHOLDER_HALT ||
+          inst->is_control_flow() ||
+          inst->has_side_effects();
+}
+
 /**
  * Sometimes we really want this node to execute after everything that
  * was before it and before everything that followed it.  This adds
@@ -932,12 +937,10 @@ instruction_scheduler::add_barrier_deps(schedule_node *n)
    schedule_node *prev = (schedule_node *)n->prev;
    schedule_node *next = (schedule_node *)n->next;
 
-   n->is_barrier = true;
-
    if (prev) {
       while (!prev->is_head_sentinel()) {
          add_dep(prev, n, 0);
-         if (prev->is_barrier)
+         if (is_scheduling_barrier(prev->inst))
             break;
          prev = (schedule_node *)prev->prev;
       }
@@ -946,7 +949,7 @@ instruction_scheduler::add_barrier_deps(schedule_node *n)
    if (next) {
       while (!next->is_tail_sentinel()) {
          add_dep(n, next, 0);
-         if (next->is_barrier)
+         if (is_scheduling_barrier(next->inst))
             break;
          next = (schedule_node *)next->next;
       }
@@ -962,14 +965,6 @@ fs_instruction_scheduler::is_compressed(fs_inst *inst)
    return inst->exec_size == 16;
 }
 
-static bool
-is_scheduling_barrier(const fs_inst *inst)
-{
-   return inst->opcode == FS_OPCODE_PLACEHOLDER_HALT ||
-          inst->is_control_flow() ||
-          inst->has_side_effects();
-}
-
 void
 fs_instruction_scheduler::calculate_deps()
 {
@@ -1233,13 +1228,6 @@ fs_instruction_scheduler::calculate_deps()
    }
 }
 
-static bool
-is_scheduling_barrier(const vec4_instruction *inst)
-{
-   return inst->is_control_flow() ||
-          inst->has_side_effects();
-}
-
 void
 vec4_instruction_scheduler::calculate_deps()
 {
-- 
2.14.2