From: Eric Anholt <eric@anholt.net>
Date: Mon, 8 Aug 2011 22:56:11 +0000 (-0700)
Subject: i965/vs: Slightly improve the trivial reg allocator to skip unused regs.
X-Git-Url: https://git.libre-soc.org/?a=commitdiff_plain;h=7b91eefe7cbe771397684b5970f7c04313baa2f0;p=mesa.git

i965/vs: Slightly improve the trivial reg allocator to skip unused regs.

This fixes most of the regressions in the vs array test set from the
varying array indexing work, since the giant array that was originally
allocated in virtual GRF space never gets used and is only ever
read/stored from scratch space.
---

diff --git a/src/mesa/drivers/dri/i965/brw_vec4_reg_allocate.cpp b/src/mesa/drivers/dri/i965/brw_vec4_reg_allocate.cpp
index e7f6b28a536..1bfd84d76e8 100644
--- a/src/mesa/drivers/dri/i965/brw_vec4_reg_allocate.cpp
+++ b/src/mesa/drivers/dri/i965/brw_vec4_reg_allocate.cpp
@@ -41,15 +41,37 @@ vec4_visitor::reg_allocate_trivial()
 {
    int last_grf = 0;
    int hw_reg_mapping[this->virtual_grf_count];
+   bool virtual_grf_used[this->virtual_grf_count];
    int i;
    int next;
 
+   /* Calculate which virtual GRFs are actually in use after whatever
+    * optimization passes have occurred.
+    */
+   for (int i = 0; i < this->virtual_grf_count; i++) {
+      virtual_grf_used[i] = false;
+   }
+
+   foreach_iter(exec_list_iterator, iter, this->instructions) {
+      vec4_instruction *inst = (vec4_instruction *)iter.get();
+
+      if (inst->dst.file == GRF)
+	 virtual_grf_used[inst->dst.reg] = true;
+
+      for (int i = 0; i < 3; i++) {
+	 if (inst->src[i].file == GRF)
+	    virtual_grf_used[inst->src[i].reg] = true;
+      }
+   }
+
    /* Note that compressed instructions require alignment to 2 registers. */
    hw_reg_mapping[0] = this->first_non_payload_grf;
    next = hw_reg_mapping[0] + this->virtual_grf_sizes[0];
    for (i = 1; i < this->virtual_grf_count; i++) {
-      hw_reg_mapping[i] = next;
-      next += this->virtual_grf_sizes[i];
+      if (virtual_grf_used[i]) {
+	 hw_reg_mapping[i] = next;
+	 next += this->virtual_grf_sizes[i];
+      }
    }
    prog_data->total_grf = next;