commit 4ef3e45d7c6ec3c482a1a48f4df470811aa3cf0a
author buzbee <buzbee@google.com> Fri Dec 14 13:35:28 2012 -0800
committer buzbee <buzbee@google.com> Wed Feb 06 12:16:33 2013 -0800
tree 5d91ff23708048a2b453a5917dab82be1b545f91
parent f84f99fe3a944310fdfb6f17836079ac5c48c799

Compiler constant handling rework

In preparation for de-optimization, reworked the constant
handling mechanism.  Also took advantage of knowledge of
constant operands (particularly for long operations).

Significant performance improvements for Mandelbrot
(~60 seconds to ~34 seconds).  Minor improvements in other
benchmarks.

The new constant handling breaks two of the existing
optimization passes: "Skip Large Method" and "Load/Store
Elimization."

I don't intend to update the large method optimization
because it will be superceeded by the upcoming interpreter/
fingerprinting mechanism.  Leaving the code in place for
now in order to compare compile-time improvements with
fingerprinting/interpret.  All related code will be deleted
when that is complete.

The load/store elimination pass needs some rework to handle
uses of multiple-register loads and stores.  It will be
updated & restored in a future CL.

Change-Id: Ia979abaf51b8ae81bbb0428031cbcea854625fac

src/compiler/codegen/gen_loadstore.cc

diff --git a/src/compiler/codegen/gen_loadstore.cc b/src/compiler/codegen/gen_loadstore.cc
index b183f9e..c8f9c51 100644
--- a/src/compiler/codegen/gen_loadstore.cc
+++ b/src/compiler/codegen/gen_loadstore.cc
@@ -59,12 +59,20 @@
           return;
         }
       }
+      int temp_reg = zero_reg;
+      if (temp_reg == INVALID_REG) {
+        temp_reg = AllocTemp(cu);
+        cu->cg->LoadConstant(cu, temp_reg, 0);
+      }
       if (cu->promotion_map[pmap_index].core_location == kLocPhysReg) {
         // Promoted - just copy in a zero
-        OpRegCopy(cu, cu->promotion_map[pmap_index].core_reg, zero_reg);
+        OpRegCopy(cu, cu->promotion_map[pmap_index].core_reg, temp_reg);
       } else {
         // Lives in the frame, need to store.
-        StoreBaseDisp(cu, TargetReg(kSp), SRegOffset(cu, rl_dest.s_reg_low), zero_reg, kWord);
+        StoreBaseDisp(cu, TargetReg(kSp), SRegOffset(cu, rl_dest.s_reg_low), temp_reg, kWord);
+      }
+      if (zero_reg == INVALID_REG) {
+        FreeTemp(cu, temp_reg);
       }
     }
   }
@@ -92,14 +100,12 @@
   rl_src = UpdateLoc(cu, rl_src);
   if (rl_src.location == kLocPhysReg) {
     OpRegCopy(cu, r_dest, rl_src.low_reg);
+  } else if (IsInexpensiveConstant(cu, rl_src)) {
+    LoadConstantNoClobber(cu, r_dest, ConstantValue(cu, rl_src));
   } else {
     DCHECK((rl_src.location == kLocDalvikFrame) ||
            (rl_src.location == kLocCompilerTemp));
-    if (rl_src.is_const && InexpensiveConstant(r_dest, cu->constant_values[rl_src.orig_sreg])) {
-      LoadConstantNoClobber(cu, r_dest, cu->constant_values[rl_src.orig_sreg]);
-    } else {
-      LoadWordDisp(cu, TargetReg(kSp), SRegOffset(cu, rl_src.s_reg_low), r_dest);
-    }
+    LoadWordDisp(cu, TargetReg(kSp), SRegOffset(cu, rl_src.s_reg_low), r_dest);
   }
 }
 
@@ -126,6 +132,8 @@
   rl_src = UpdateLocWide(cu, rl_src);
   if (rl_src.location == kLocPhysReg) {
     OpRegCopyWide(cu, reg_lo, reg_hi, rl_src.low_reg, rl_src.high_reg);
+  } else if (IsInexpensiveConstant(cu, rl_src)) {
+    LoadConstantWide(cu, reg_lo, reg_hi, ConstantValueWide(cu, rl_src));
   } else {
     DCHECK((rl_src.location == kLocDalvikFrame) ||
            (rl_src.location == kLocCompilerTemp));
@@ -152,9 +160,7 @@
 RegLocation Codegen::LoadValue(CompilationUnit* cu, RegLocation rl_src, RegisterClass op_kind)
 {
   rl_src = EvalLoc(cu, rl_src, op_kind, false);
-  if (rl_src.location != kLocPhysReg) {
-    DCHECK((rl_src.location == kLocDalvikFrame) ||
-           (rl_src.location == kLocCompilerTemp));
+  if (IsInexpensiveConstant(cu, rl_src) || rl_src.location != kLocPhysReg) {
     LoadValueDirect(cu, rl_src, rl_src.low_reg);
     rl_src.location = kLocPhysReg;
     MarkLive(cu, rl_src.low_reg, rl_src.s_reg_low);
@@ -222,14 +228,11 @@
 {
   DCHECK(rl_src.wide);
   rl_src = EvalLoc(cu, rl_src, op_kind, false);
-  if (rl_src.location != kLocPhysReg) {
-    DCHECK((rl_src.location == kLocDalvikFrame) ||
-        (rl_src.location == kLocCompilerTemp));
+  if (IsInexpensiveConstant(cu, rl_src) || rl_src.location != kLocPhysReg) {
     LoadValueDirectWide(cu, rl_src, rl_src.low_reg, rl_src.high_reg);
     rl_src.location = kLocPhysReg;
     MarkLive(cu, rl_src.low_reg, rl_src.s_reg_low);
-    MarkLive(cu, rl_src.high_reg,
-                GetSRegHi(rl_src.s_reg_low));
+    MarkLive(cu, rl_src.high_reg, GetSRegHi(rl_src.s_reg_low));
   }
   return rl_src;
 }