From 6829815d96cdc4fc0eedb1944977509033ed9db3 Mon Sep 17 00:00:00 2001
From: Hal Finkel <hfinkel@anl.gov>
Date: Sat, 10 Jan 2015 00:31:10 +0000
Subject: [PATCH] [PowerPC] Readjust the loop unrolling threshold

Now that the way that the partial unrolling threshold for small loops is used
to compute the unrolling factor as been corrected, a slightly smaller threshold
is preferable. This is expected; other targets may need to re-tune as well.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@225566 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/PowerPC/PPCScheduleP7.td           |  4 +-
 lib/Target/PowerPC/PPCScheduleP8.td           |  4 +-
 .../LoopUnroll/PowerPC/p7-unrolling.ll        | 52 ++++++++++++++++++-
 3 files changed, 54 insertions(+), 6 deletions(-)

diff --git a/lib/Target/PowerPC/PPCScheduleP7.td b/lib/Target/PowerPC/PPCScheduleP7.td
index 00aa834ca4b..03693cbeada 100644
--- a/lib/Target/PowerPC/PPCScheduleP7.td
+++ b/lib/Target/PowerPC/PPCScheduleP7.td
@@ -380,8 +380,8 @@ def P7Model : SchedMachineModel {
                        // Itineraries are queried instead.
   let MispredictPenalty = 16;
 
-  // Try to make sure we have at least 10-11 dispatch groups in a loop.
-  let LoopMicroOpBufferSize = 44;
+  // Try to make sure we have at least 10 dispatch groups in a loop.
+  let LoopMicroOpBufferSize = 40;
 
   let Itineraries = P7Itineraries;
 }
diff --git a/lib/Target/PowerPC/PPCScheduleP8.td b/lib/Target/PowerPC/PPCScheduleP8.td
index 8cdc79e9d14..07971809c87 100644
--- a/lib/Target/PowerPC/PPCScheduleP8.td
+++ b/lib/Target/PowerPC/PPCScheduleP8.td
@@ -389,8 +389,8 @@ def P8Model : SchedMachineModel {
                        // Itineraries are queried instead.
   let MispredictPenalty = 16;
 
-  // Try to make sure we have at least 10-11 dispatch groups in a loop.
-  let LoopMicroOpBufferSize = 66;
+  // Try to make sure we have at least 10 dispatch groups in a loop.
+  let LoopMicroOpBufferSize = 60;
 
   let Itineraries = P8Itineraries;
 }
diff --git a/test/Transforms/LoopUnroll/PowerPC/p7-unrolling.ll b/test/Transforms/LoopUnroll/PowerPC/p7-unrolling.ll
index b2ea74a75da..7a50fc0a4f4 100644
--- a/test/Transforms/LoopUnroll/PowerPC/p7-unrolling.ll
+++ b/test/Transforms/LoopUnroll/PowerPC/p7-unrolling.ll
@@ -22,7 +22,31 @@ exit:
 ; CHECK-NEXT: add
 ; CHECK-NEXT: add
 ; CHECK-NEXT: add
-; CHECK: icmp
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: icmp
 
 define void @unroll_default() nounwind {
 entry:
@@ -47,5 +71,29 @@ exit:
 ; CHECK-NEXT: add
 ; CHECK-NEXT: add
 ; CHECK-NEXT: add
-; CHECK: icmp
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: add
+; CHECK-NEXT: icmp