Clean up spill weight computation. Also some changes to give loop induction

variable increment / decrement slighter high priority. 

This has major impact on some micro-benchmarks. On MultiSource/Applications
and spec tests, it's a minor win. It also reduce 256.bzip instruction count
by 8%, 55 on 164.gzip on i386 / Darwin.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82485 91177308-0d34-0410-b5e6-96231b3b80d8

7 files changed, 44 insertions, 8 deletions

diff --git a/test/CodeGen/ARM/2008-11-19-ScavengerAssert.ll b/test/CodeGen/ARM/2008-11-19-ScavengerAssert.ll
index 311c855742..221a168cba 100644
--- a/test/CodeGen/ARM/2008-11-19-ScavengerAssert.ll
+++ b/test/CodeGen/ARM/2008-11-19-ScavengerAssert.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -mtriple=arm-apple-darwin9 -stats |& grep asm-printer | grep 164
+; RUN: llc < %s -mtriple=arm-apple-darwin9 -stats |& grep asm-printer | grep 161
 
 	%"struct.Adv5::Ekin<3>" = type <{ i8 }>
 	%"struct.Adv5::X::Energyflux<3>" = type { double }
diff --git a/test/CodeGen/ARM/remat.ll b/test/CodeGen/ARM/remat.ll
index 21d117aaf2..50da997ed4 100644
--- a/test/CodeGen/ARM/remat.ll
+++ b/test/CodeGen/ARM/remat.ll
@@ -1,5 +1,5 @@
 ; RUN: llc < %s -mtriple=arm-apple-darwin 
-; RUN: llc < %s -mtriple=arm-apple-darwin -stats -info-output-file - | grep "Number of re-materialization" | grep 2
+; RUN: llc < %s -mtriple=arm-apple-darwin -stats -info-output-file - | grep "Number of re-materialization" | grep 5
 
 	%struct.CONTENTBOX = type { i32, i32, i32, i32, i32 }
 	%struct.LOCBOX = type { i32, i32, i32, i32 }
diff --git a/test/CodeGen/X86/2008-02-22-ReMatBug.ll b/test/CodeGen/X86/2008-02-22-ReMatBug.ll
index 95df19ad1f..a91ac27f98 100644
--- a/test/CodeGen/X86/2008-02-22-ReMatBug.ll
+++ b/test/CodeGen/X86/2008-02-22-ReMatBug.ll
@@ -1,5 +1,4 @@
-; RUN: llc < %s -march=x86 -stats |& grep {Number of re-materialization} | grep 3
-; RUN: llc < %s -march=x86 -stats |& grep {Number of dead spill slots removed}
+; RUN: llc < %s -march=x86 -stats |& grep {Number of re-materialization} | grep 2
 ; rdar://5761454
 
 	%struct.quad_struct = type { i32, i32, %struct.quad_struct*, %struct.quad_struct*, %struct.quad_struct*, %struct.quad_struct*, %struct.quad_struct* }
diff --git a/test/CodeGen/X86/2008-07-11-SpillerBug.ll b/test/CodeGen/X86/2008-07-11-SpillerBug.ll
index 6f3019e78c..f75e605168 100644
--- a/test/CodeGen/X86/2008-07-11-SpillerBug.ll
+++ b/test/CodeGen/X86/2008-07-11-SpillerBug.ll
@@ -2,8 +2,9 @@
 ; PR2536
 
 
-; CHECK: movw %ax
+; CHECK: movw %cx
 ; CHECK-NEXT: andl    $65534, %
+; CHECK-NEXT: movl %
 ; CHECK-NEXT: movl $17
 
 @g_5 = external global i16		; <i16*> [#uses=2]
diff --git a/test/CodeGen/X86/2009-04-20-LinearScanOpt.ll b/test/CodeGen/X86/2009-04-20-LinearScanOpt.ll
index 9f11c4e58a..4d25b0f983 100644
--- a/test/CodeGen/X86/2009-04-20-LinearScanOpt.ll
+++ b/test/CodeGen/X86/2009-04-20-LinearScanOpt.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -mtriple=x86_64-apple-darwin10.0 -relocation-model=pic -disable-fp-elim -stats |& grep {Number of registers downgraded}
+; RUN: llc < %s -mtriple=x86_64-apple-darwin10.0 -relocation-model=pic -disable-fp-elim -stats |& grep asm-printer | grep 84
 ; rdar://6802189
 
 ; Test if linearscan is unfavoring registers for allocation to allow more reuse
diff --git a/test/CodeGen/X86/2009-09-21-NoSpillLoopCount.ll b/test/CodeGen/X86/2009-09-21-NoSpillLoopCount.ll
new file mode 100644
index 0000000000..80b883582c
--- /dev/null
+++ b/test/CodeGen/X86/2009-09-21-NoSpillLoopCount.ll
@@ -0,0 +1,36 @@
+; RUN: llc < %s -mtriple=i386-apple-darwin10.0 -relocation-model=pic | FileCheck %s
+
+define void @dot(i16* nocapture %A, i32 %As, i16* nocapture %B, i32 %Bs, i16* nocapture %C, i32 %N) nounwind ssp {
+; CHECK: dot:
+; CHECK: decl %
+; CHECK-NEXT: jne
+entry:
+	%0 = icmp sgt i32 %N, 0		; <i1> [#uses=1]
+	br i1 %0, label %bb, label %bb2
+
+bb:		; preds = %bb, %entry
+	%i.03 = phi i32 [ 0, %entry ], [ %indvar.next, %bb ]		; <i32> [#uses=3]
+	%sum.04 = phi i32 [ 0, %entry ], [ %10, %bb ]		; <i32> [#uses=1]
+	%1 = mul i32 %i.03, %As		; <i32> [#uses=1]
+	%2 = getelementptr i16* %A, i32 %1		; <i16*> [#uses=1]
+	%3 = load i16* %2, align 2		; <i16> [#uses=1]
+	%4 = sext i16 %3 to i32		; <i32> [#uses=1]
+	%5 = mul i32 %i.03, %Bs		; <i32> [#uses=1]
+	%6 = getelementptr i16* %B, i32 %5		; <i16*> [#uses=1]
+	%7 = load i16* %6, align 2		; <i16> [#uses=1]
+	%8 = sext i16 %7 to i32		; <i32> [#uses=1]
+	%9 = mul i32 %8, %4		; <i32> [#uses=1]
+	%10 = add i32 %9, %sum.04		; <i32> [#uses=2]
+	%indvar.next = add i32 %i.03, 1		; <i32> [#uses=2]
+	%exitcond = icmp eq i32 %indvar.next, %N		; <i1> [#uses=1]
+	br i1 %exitcond, label %bb1.bb2_crit_edge, label %bb
+
+bb1.bb2_crit_edge:		; preds = %bb
+	%phitmp = trunc i32 %10 to i16		; <i16> [#uses=1]
+	br label %bb2
+
+bb2:		; preds = %entry, %bb1.bb2_crit_edge
+	%sum.0.lcssa = phi i16 [ %phitmp, %bb1.bb2_crit_edge ], [ 0, %entry ]		; <i16> [#uses=1]
+	store i16 %sum.0.lcssa, i16* %C, align 2
+	ret void
+}
diff --git a/test/CodeGen/X86/stack-color-with-reg.ll b/test/CodeGen/X86/stack-color-with-reg.ll
index a8ae8e8168..672f77eef0 100644
--- a/test/CodeGen/X86/stack-color-with-reg.ll
+++ b/test/CodeGen/X86/stack-color-with-reg.ll
@@ -1,6 +1,6 @@
 ; RUN: llc < %s -mtriple=x86_64-apple-darwin10 -relocation-model=pic -disable-fp-elim -color-ss-with-regs -stats -info-output-file - > %t
-; RUN:   grep stackcoloring %t | grep "stack slot refs replaced with reg refs"  | grep 8
-; RUN:   grep asm-printer %t   | grep 182
+; RUN:   grep stackcoloring %t | grep "stack slot refs replaced with reg refs"  | grep 5
+; RUN:   grep asm-printer %t   | grep 179
 
 	type { [62 x %struct.Bitvec*] }		; type %0
 	type { i8* }		; type %1