From f5a0946e40f347ea8fbf919797f6737abab22158 Mon Sep 17 00:00:00 2001
From: David Green <david.green@arm.com>
Date: Mon, 10 Jun 2019 10:22:14 +0000
Subject: [PATCH] [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


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@362928 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/ARM/ARMTargetTransformInfo.cpp    |  1 +
 test/Transforms/LoopUnroll/ARM/upperbound.ll | 38 ++++++++++++++++++++
 2 files changed, 39 insertions(+)
 create mode 100644 test/Transforms/LoopUnroll/ARM/upperbound.ll

diff --git a/lib/Target/ARM/ARMTargetTransformInfo.cpp b/lib/Target/ARM/ARMTargetTransformInfo.cpp
index 6a8647ee0fd..c626c4149c7 100644
--- a/lib/Target/ARM/ARMTargetTransformInfo.cpp
+++ b/lib/Target/ARM/ARMTargetTransformInfo.cpp
@@ -683,6 +683,7 @@ void ARMTTIImpl::getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
 
   UP.Partial = true;
   UP.Runtime = true;
+  UP.UpperBound = true;
   UP.UnrollRemainder = true;
   UP.DefaultUnrollRuntimeCount = 4;
   UP.UnrollAndJam = true;
diff --git a/test/Transforms/LoopUnroll/ARM/upperbound.ll b/test/Transforms/LoopUnroll/ARM/upperbound.ll
new file mode 100644
index 00000000000..b47bdc92cdc
--- /dev/null
+++ b/test/Transforms/LoopUnroll/ARM/upperbound.ll
@@ -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
+}
+
-- 
2.50.1