[llvm] r362928 - [ARM] Enable Unroll UpperBound

[David Green via llvm-commits]

Author: dmgreen
Date: Mon Jun 10 03:22:14 2019
New Revision: 362928

URL: http://llvm.org/viewvc/llvm-project?rev=362928&view=rev
Log:
[ARM] Enable Unroll UpperBound

This option allows loops with small max trip counts to be fully unrolled. This
can help with code like the remainder loops from manually unrolled loops like
those that appear in the cmsis dsp library. We would apparently previously
runtime unroll them with the default unroll count (4).

Differential Revision: https://reviews.llvm.org/D63064

Added:
    llvm/trunk/test/Transforms/LoopUnroll/ARM/upperbound.ll
Modified:
    llvm/trunk/lib/Target/ARM/ARMTargetTransformInfo.cpp

Modified: llvm/trunk/lib/Target/ARM/ARMTargetTransformInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMTargetTransformInfo.cpp?rev=362928&r1=362927&r2=362928&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMTargetTransformInfo.cpp (original)
+++ llvm/trunk/lib/Target/ARM/ARMTargetTransformInfo.cpp Mon Jun 10 03:22:14 2019
@@ -683,6 +683,7 @@ void ARMTTIImpl::getUnrollingPreferences
 
   UP.Partial = true;
   UP.Runtime = true;
+  UP.UpperBound = true;
   UP.UnrollRemainder = true;
   UP.DefaultUnrollRuntimeCount = 4;
   UP.UnrollAndJam = true;

Added: llvm/trunk/test/Transforms/LoopUnroll/ARM/upperbound.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/LoopUnroll/ARM/upperbound.ll?rev=362928&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/LoopUnroll/ARM/upperbound.ll (added)
+++ llvm/trunk/test/Transforms/LoopUnroll/ARM/upperbound.ll Mon Jun 10 03:22:14 2019
@@ -0,0 +1,38 @@
+; RUN: opt -loop-unroll -S -mtriple arm-none-eabi -mcpu=cortex-m7 %s | FileCheck %s -check-prefix=UNROLL
+
+; This test is meant to check that this loop is unrolled into three iterations.
+
+; UNROLL-LABEL: @test
+; UNROLL: load i32, i32*
+; UNROLL: load i32, i32*
+; UNROLL: load i32, i32*
+; UNROLL-NOT: load i32, i32*
+
+define void @test(i32* %x, i32 %n) {
+entry:
+  %sub = add nsw i32 %n, -1
+  %rem = srem i32 %sub, 4
+  %cmp7 = icmp sgt i32 %rem, 0
+  br i1 %cmp7, label %while.body, label %while.end
+
+while.body:                                       ; preds = %entry, %if.end
+  %x.addr.09 = phi i32* [ %incdec.ptr, %if.end ], [ %x, %entry ]
+  %n.addr.08 = phi i32 [ %dec, %if.end ], [ %rem, %entry ]
+  %0 = load i32, i32* %x.addr.09, align 4
+  %cmp1 = icmp slt i32 %0, 10
+  br i1 %cmp1, label %if.then, label %if.end
+
+if.then:                                          ; preds = %while.body
+  store i32 0, i32* %x.addr.09, align 4
+  br label %if.end
+
+if.end:                                           ; preds = %if.then, %while.body
+  %incdec.ptr = getelementptr inbounds i32, i32* %x.addr.09, i32 1
+  %dec = add nsw i32 %n.addr.08, -1
+  %cmp = icmp sgt i32 %dec, 0
+  br i1 %cmp, label %while.body, label %while.end
+
+while.end:                                        ; preds = %if.end, %entry
+  ret void
+}
+