ARMTTI: We don't have 16 allocatable scalar registers

This caused an regression on libquantum after enabling the new loop vectorizer
unroll heuristics.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200616 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/ARM/ARMTargetTransformInfo.cpp b/lib/Target/ARM/ARMTargetTransformInfo.cpp
index 25caa11b9250b194e9a891fdb088443abb524fa1..9aacee44c04eaf13d53be6ddd9b062acca4657da 100644 (file)
--- a/lib/Target/ARM/ARMTargetTransformInfo.cpp
+++ b/lib/Target/ARM/ARMTargetTransformInfo.cpp
@@ -95,7 +95,7 @@ public:
 
     if (ST->isThumb1Only())
       return 8;
-    return 16;
+    return 13;
   }
 
   unsigned getRegisterBitWidth(bool Vector) const {

diff --git a/test/Transforms/LoopVectorize/ARM/arm-unroll.ll b/test/Transforms/LoopVectorize/ARM/arm-unroll.ll
index 39363ab2d802b943f3902f5c86d69f6759ceb57c..0b87e0e74055d2a86d4e82ae8c2ebb5069462580 100644 (file)
--- a/test/Transforms/LoopVectorize/ARM/arm-unroll.ll
+++ b/test/Transforms/LoopVectorize/ARM/arm-unroll.ll
@@ -1,5 +1,6 @@
 ; RUN: opt < %s  -loop-vectorize -mtriple=thumbv7-apple-ios3.0.0 -S | FileCheck %s
 ; RUN: opt < %s  -loop-vectorize -mtriple=thumbv7-apple-ios3.0.0 -mcpu=swift -S | FileCheck %s --check-prefix=SWIFT
+; RUN: opt < %s  -loop-vectorize -force-vector-width=1 -mtriple=thumbv7-apple-ios3.0.0 -mcpu=swift -S | FileCheck %s --check-prefix=SWIFTUNROLL
 
 target datalayout = "e-p:32:32:32-i1:8:32-i8:8:32-i16:16:32-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:32:64-v128:32:128-a0:0:32-n32-S32"
 target triple = "thumbv7-apple-ios3.0.0"
@@ -30,3 +31,38 @@ define i32 @foo(i32* nocapture %A, i32 %n) nounwind readonly ssp {
   %sum.0.lcssa = phi i32 [ 0, %0 ], [ %4, %.lr.ph ]
   ret i32 %sum.0.lcssa
 }
+
+; Verify the register limit. On arm we don't have 16 allocatable registers.
+;SWIFTUNROLL-LABEL: @register_limit(
+;SWIFTUNROLL: load i32
+;SWIFTUNROLL-NOT: load i32
+define i32 @register_limit(i32* nocapture %A, i32 %n) {
+  %1 = icmp sgt i32 %n, 0
+  br i1 %1, label %.lr.ph, label %._crit_edge
+
+.lr.ph:
+  %i.02 = phi i32 [ %5, %.lr.ph ], [ 0, %0 ]
+  %sum.01 = phi i32 [ %4, %.lr.ph ], [ 0, %0 ]
+  %sum.02 = phi i32 [ %6, %.lr.ph ], [ 0, %0 ]
+  %sum.03 = phi i32 [ %7, %.lr.ph ], [ 0, %0 ]
+  %sum.04 = phi i32 [ %8, %.lr.ph ], [ 0, %0 ]
+  %sum.05 = phi i32 [ %9, %.lr.ph ], [ 0, %0 ]
+  %2 = getelementptr inbounds i32* %A, i32 %i.02
+  %3 = load i32* %2, align 4
+  %4 = add nsw i32 %3, %sum.01
+  %5 = add nsw i32 %i.02, 1
+  %6 = add nsw i32 %3, %sum.02
+  %7 = add nsw i32 %3, %sum.03
+  %8 = add nsw i32 %3, %sum.04
+  %9 = add nsw i32 %3, %sum.05
+  %exitcond = icmp eq i32 %5, %n
+  br i1 %exitcond, label %._crit_edge, label %.lr.ph
+
+._crit_edge:                                      ; preds = %.lr.ph, %0
+  %sum.0.lcssa = phi i32 [ 0, %0 ], [ %4, %.lr.ph ]
+  %sum.1.lcssa = phi i32 [ 0, %0 ], [ %6, %.lr.ph ]
+  %sum.2.lcssa = phi i32 [ 0, %0 ], [ %7, %.lr.ph ]
+  %sum.4.lcssa = phi i32 [ 0, %0 ], [ %8, %.lr.ph ]
+  %sum.5.lcssa = phi i32 [ 0, %0 ], [ %9, %.lr.ph ]
+  ret i32 %sum.0.lcssa
+}