i965/fs: Do more register coalescing by using the interference graph.

By using the live variables code for determining interference, we can
handle coalescing in the presence of control flow, which the other
register coalescing path couldn't.

Total instructions: 207184 -> 206990
74/1246 programs affected (5.9%)
33993 -> 33799 instructions in affected programs (0.6% reduction)

There is a newerth shader that loses out, because of some extra MOVs
that now get their dead-code nature obscured by coalescing.  This
should be fixed by doing better at dead code elimination.

diff --git a/src/mesa/drivers/dri/i965/brw_fs.cpp b/src/mesa/drivers/dri/i965/brw_fs.cpp
index 7176d40..313e720 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_fs.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs.cpp
@@ -1218,6 +1218,66 @@ fs_visitor::dead_code_eliminate()
    return progress;
 }
 
+/**
+ * Implements a second type of register coalescing: This one checks if
+ * the two regs involved in a raw move don't interfere, in which case
+ * they can both by stored in the same place and the MOV removed.
+ */
+bool
+fs_visitor::register_coalesce_2()
+{
+   bool progress = false;
+
+   calculate_live_intervals();
+
+   foreach_list_safe(node, &this->instructions) {
+      fs_inst *inst = (fs_inst *)node;
+
+      if (inst->opcode != BRW_OPCODE_MOV ||
+         inst->predicated ||
+         inst->saturate ||
+         inst->src[0].file != GRF ||
+         inst->src[0].negate ||
+         inst->src[0].abs ||
+         inst->src[0].smear != -1 ||
+         inst->dst.file != GRF ||
+         inst->dst.type != inst->src[0].type ||
+         virtual_grf_sizes[inst->src[0].reg] != 1 ||
+         virtual_grf_interferes(inst->dst.reg, inst->src[0].reg)) {
+        continue;
+      }
+
+      int reg_from = inst->src[0].reg;
+      assert(inst->src[0].reg_offset == 0);
+      int reg_to = inst->dst.reg;
+      int reg_to_offset = inst->dst.reg_offset;
+
+      foreach_list_safe(node, &this->instructions) {
+        fs_inst *scan_inst = (fs_inst *)node;
+
+        if (scan_inst->dst.file == GRF &&
+            scan_inst->dst.reg == reg_from) {
+           scan_inst->dst.reg = reg_to;
+           scan_inst->dst.reg_offset = reg_to_offset;
+        }
+        for (int i = 0; i < 3; i++) {
+           if (scan_inst->src[i].file == GRF &&
+               scan_inst->src[i].reg == reg_from) {
+              scan_inst->src[i].reg = reg_to;
+              scan_inst->src[i].reg_offset = reg_to_offset;
+           }
+        }
+      }
+
+      inst->remove();
+      live_intervals_valid = false;
+      progress = true;
+      continue;
+   }
+
+   return progress;
+}
+
 bool
 fs_visitor::register_coalesce()
 {
@@ -1684,6 +1744,7 @@ fs_visitor::run()
         progress = opt_cse() || progress;
         progress = opt_copy_propagate() || progress;
         progress = register_coalesce() || progress;
+        progress = register_coalesce_2() || progress;
         progress = compute_to_mrf() || progress;
         progress = dead_code_eliminate() || progress;
       } while (progress);

diff --git a/src/mesa/drivers/dri/i965/brw_fs.h b/src/mesa/drivers/dri/i965/brw_fs.h
index f04c6fc..d7fd9a4 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_fs.h
+++ b/src/mesa/drivers/dri/i965/brw_fs.h
@@ -495,6 +495,7 @@ public:
    bool opt_copy_propagate_local(void *mem_ctx, fs_bblock *block,
                                 exec_list *acp);
    bool register_coalesce();
+   bool register_coalesce_2();
    bool compute_to_mrf();
    bool dead_code_eliminate();
    bool remove_dead_constants();