Loop unroller: set thresholds for optsize and minsize functions to zero

Before r268509, Clang would disable the loop unroll pass when optimizing
for size. That commit enabled it to be able to support unroll pragmas
in -Os builds. However, this regressed binary size in one of Chromium's
DLLs with ~100 KB.

This restores the original behaviour of no unrolling at -Os, but doing it
in LLVM instead of Clang makes more sense, and also allows the pragmas to
keep working.

Differential revision: http://reviews.llvm.org/D20115

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269124 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Transforms/Scalar/LoopUnrollPass.cpp b/lib/Transforms/Scalar/LoopUnrollPass.cpp
index 2e5c55e73a4fcca4e0160af7254b8ba4916939a4..95ec12d5cb89c6d4536ad67ac28ac3c83262ddde 100644 (file)
--- a/lib/Transforms/Scalar/LoopUnrollPass.cpp
+++ b/lib/Transforms/Scalar/LoopUnrollPass.cpp
@@ -111,9 +111,9 @@ static TargetTransformInfo::UnrollingPreferences gatherUnrollingPreferences(
   UP.Threshold = 150;
   UP.PercentDynamicCostSavedThreshold = 20;
   UP.DynamicCostSavingsDiscount = 2000;
-  UP.OptSizeThreshold = 50;
+  UP.OptSizeThreshold = 0;
   UP.PartialThreshold = UP.Threshold;
-  UP.PartialOptSizeThreshold = UP.OptSizeThreshold;
+  UP.PartialOptSizeThreshold = 0;
   UP.Count = 0;
   UP.MaxCount = UINT_MAX;
   UP.FullUnrollMaxCount = UINT_MAX;

diff --git a/test/Transforms/LoopUnroll/PowerPC/a2-unrolling.ll b/test/Transforms/LoopUnroll/PowerPC/a2-unrolling.ll
index bf2b9cf40dd0e8bd3713e954b285b5c8a16ef947..77e92bd7506cf655eacd542987f070488ae48d0a 100644 (file)
--- a/test/Transforms/LoopUnroll/PowerPC/a2-unrolling.ll
+++ b/test/Transforms/LoopUnroll/PowerPC/a2-unrolling.ll
@@ -1,30 +1,5 @@
 ; RUN: opt < %s -S -mtriple=powerpc64-unknown-linux-gnu -mcpu=a2 -loop-unroll | FileCheck %s -check-prefix=EPILOG
 ; RUN: opt < %s -S -mtriple=powerpc64-unknown-linux-gnu -mcpu=a2 -loop-unroll -unroll-runtime-epilog=false | FileCheck %s -check-prefix=PROLOG
-define void @unroll_opt_for_size() nounwind optsize {
-entry:
-  br label %loop
-
-loop:
-  %iv = phi i32 [ 0, %entry ], [ %inc, %loop ]
-  %inc = add i32 %iv, 1
-  %exitcnd = icmp uge i32 %inc, 1024
-  br i1 %exitcnd, label %exit, label %loop
-
-exit:
-  ret void
-}
-
-; EPILOG-LABEL: @unroll_opt_for_size
-; EPILOG:      add
-; EPILOG-NEXT: add
-; EPILOG-NEXT: add
-; EPILOG: icmp
-
-; PROLOG-LABEL: @unroll_opt_for_size
-; PROLOG:      add
-; PROLOG-NEXT: add
-; PROLOG-NEXT: add
-; PROLOG: icmp
 
 define i32 @test(i32* nocapture %a, i32 %n) nounwind uwtable readonly {
 entry:

diff --git a/test/Transforms/LoopUnroll/PowerPC/p7-unrolling.ll b/test/Transforms/LoopUnroll/PowerPC/p7-unrolling.ll
index 7a50fc0a4f49e7dc8daa354fd49ac2e17ad3437e..c9677d83e377525e6f1fdd1b9f1658d3fa818c64 100644 (file)
--- a/test/Transforms/LoopUnroll/PowerPC/p7-unrolling.ll
+++ b/test/Transforms/LoopUnroll/PowerPC/p7-unrolling.ll
@@ -1,53 +1,4 @@
 ; RUN: opt < %s -S -mtriple=powerpc64-unknown-linux-gnu -mcpu=pwr7 -loop-unroll | FileCheck %s
-define void @unroll_opt_for_size() nounwind optsize {
-entry:
-  br label %loop
-
-loop:
-  %iv = phi i32 [ 0, %entry ], [ %inc, %loop ]
-  %inc = add i32 %iv, 1
-  %exitcnd = icmp uge i32 %inc, 1024
-  br i1 %exitcnd, label %exit, label %loop
-
-exit:
-  ret void
-}
-
-; CHECK-LABEL: @unroll_opt_for_size
-; CHECK:      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: 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:
   br label %loop

diff --git a/test/Transforms/LoopUnroll/partial-unroll-optsize.ll b/test/Transforms/LoopUnroll/partial-unroll-optsize.ll
deleted file mode 100644 (file)
index e5e0151..0000000
--- a/test/Transforms/LoopUnroll/partial-unroll-optsize.ll
+++ /dev/null
@@ -1,53 +0,0 @@
-; RUN: opt < %s -S -loop-unroll -unroll-allow-partial | FileCheck %s
-; RUN: sed -e 's/optsize/minsize/' %s | opt -S -loop-unroll -unroll-allow-partial | FileCheck %s
-
-; Loop size = 3, when the function has the optsize attribute, the
-; OptSizeUnrollThreshold, i.e. 50, is used, hence the loop should be unrolled
-; by 32 times because (1 * 32) + 2 < 50 (whereas (1 * 64 + 2) is not).
-define void @unroll_opt_for_size() nounwind optsize {
-entry:
-  br label %loop
-
-loop:
-  %iv = phi i32 [ 0, %entry ], [ %inc, %loop ]
-  %inc = add i32 %iv, 1
-  %exitcnd = icmp uge i32 %inc, 1024
-  br i1 %exitcnd, label %exit, label %loop
-
-exit:
-  ret void
-}
-
-; CHECK:      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: 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

diff --git a/test/Transforms/LoopUnroll/unroll-opt-attribute.ll b/test/Transforms/LoopUnroll/unroll-opt-attribute.ll
index 334162f5605525afa588a0b43506db78046ef940..f4a76c71368d1c31cedda8ac2f3b186914b63e33 100644 (file)
--- a/test/Transforms/LoopUnroll/unroll-opt-attribute.ll
+++ b/test/Transforms/LoopUnroll/unroll-opt-attribute.ll
@@ -4,10 +4,10 @@
 \r
 ;///////////////////// TEST 1 //////////////////////////////\r
 \r
-; This test shows that with optsize attribute, the loop is unrolled\r
-; according to the specified unroll factor.\r
+; This test shows that the loop is unrolled according to the specified\r
+; unroll factor.\r
 \r
-define void @Test1() nounwind optsize {\r
+define void @Test1() nounwind {\r
 entry:\r
   br label %loop\r
 \r
@@ -32,38 +32,10 @@ exit:
 \r
 ;///////////////////// TEST 2 //////////////////////////////\r
 \r
-; This test shows that with minsize attribute, the loop is unrolled\r
-; according to the specified unroll factor.\r
-\r
-define void @Test2() nounwind minsize {\r
-entry:\r
-  br label %loop\r
-\r
-loop:\r
-  %iv = phi i32 [ 0, %entry ], [ %inc, %loop ]\r
-  %inc = add i32 %iv, 1\r
-  %exitcnd = icmp uge i32 %inc, 1024\r
-  br i1 %exitcnd, label %exit, label %loop\r
-\r
-exit:\r
-  ret void\r
-}\r
-\r
-; CHECK_COUNT4-LABEL: @Test2\r
-; CHECK_COUNT4:      phi\r
-; CHECK_COUNT4-NEXT: add\r
-; CHECK_COUNT4-NEXT: add\r
-; CHECK_COUNT4-NEXT: add\r
-; CHECK_COUNT4-NEXT: add\r
-; CHECK_COUNT4-NEXT: icmp\r
-\r
-\r
-;///////////////////// TEST 3 //////////////////////////////\r
-\r
 ; This test shows that with optnone attribute, the loop is not unrolled\r
 ; even if an unroll factor was specified.\r
 \r
-define void @Test3() nounwind optnone noinline {\r
+define void @Test2() nounwind optnone noinline {\r
 entry:\r
   br label %loop\r
 \r
@@ -77,20 +49,19 @@ exit:
   ret void\r
 }\r
 \r
-; CHECK_COUNT4-LABEL: @Test3\r
+; CHECK_COUNT4-LABEL: @Test2\r
 ; CHECK_COUNT4:      phi\r
 ; CHECK_COUNT4-NEXT: add\r
 ; CHECK_COUNT4-NEXT: icmp\r
 \r
 \r
-;///////////////////// TEST 4 //////////////////////////////\r
+;///////////////////// TEST 3 //////////////////////////////\r
 \r
-; This test shows that without any attribute, this loop is fully unrolled \r
-; by default.\r
+; This test shows that this loop is fully unrolled by default.\r
 \r
 @tab = common global [24 x i32] zeroinitializer, align 4\r
 \r
-define i32 @Test4() {\r
+define i32 @Test3() {\r
 entry:\r
   br label %for.body\r
 \r
@@ -106,7 +77,7 @@ for.end:                                          ; preds = %for.body
   ret i32 42\r
 }\r
 \r
-; CHECK_NOCOUNT-LABEL: @Test4\r
+; CHECK_NOCOUNT-LABEL: @Test3\r
 ; CHECK_NOCOUNT:      store\r
 ; CHECK_NOCOUNT-NEXT: store\r
 ; CHECK_NOCOUNT-NEXT: store\r
@@ -134,12 +105,11 @@ for.end:                                          ; preds = %for.body
 ; CHECK_NOCOUNT-NEXT: ret\r
 \r
 \r
-;///////////////////// TEST 5 //////////////////////////////\r
+;///////////////////// TEST 4 //////////////////////////////\r
 \r
-; This test shows that with optsize attribute, this loop is not unrolled \r
-; by default.\r
+; This test shows that with optsize attribute, this loop is not unrolled.\r
 \r
-define i32 @Test5() optsize {\r
+define i32 @Test4() optsize {\r
 entry:\r
   br label %for.body\r
 \r
@@ -155,6 +125,6 @@ for.end:                                          ; preds = %for.body
   ret i32 42\r
 }\r
 \r
-; CHECK_NOCOUNT-LABEL: @Test5\r
+; CHECK_NOCOUNT-LABEL: @Test4\r
 ; CHECK_NOCOUNT:      phi\r
 ; CHECK_NOCOUNT:      icmp\r

diff --git a/test/Transforms/LoopUnroll/unroll-pragmas.ll b/test/Transforms/LoopUnroll/unroll-pragmas.ll
index 88f32c92d694f353ef5fdbbaacb5574425603918..5b405a030a163456906844e2f047bed6d3dd7e09 100644 (file)
--- a/test/Transforms/LoopUnroll/unroll-pragmas.ll
+++ b/test/Transforms/LoopUnroll/unroll-pragmas.ll
@@ -108,6 +108,29 @@ for.end:                                          ; preds = %for.body
 !3 = !{!3, !4}
 !4 = !{!"llvm.loop.unroll.full"}
 
+; #pragma clang loop unroll(full)
+; Loop should be fully unrolled, even for optsize.
+;
+; CHECK-LABEL: @loop64_with_full_optsize(
+; CHECK-NOT: br i1
+define void @loop64_with_full_optsize(i32* nocapture %a) optsize {
+entry:
+  br label %for.body
+
+for.body:                                         ; preds = %for.body, %entry
+  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
+  %arrayidx = getelementptr inbounds i32, i32* %a, i64 %indvars.iv
+  %0 = load i32, i32* %arrayidx, align 4
+  %inc = add nsw i32 %0, 1
+  store i32 %inc, i32* %arrayidx, align 4
+  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+  %exitcond = icmp eq i64 %indvars.iv.next, 64
+  br i1 %exitcond, label %for.end, label %for.body, !llvm.loop !3
+
+for.end:                                          ; preds = %for.body
+  ret void
+}
+
 ; #pragma clang loop unroll_count(4)
 ; Loop should be unrolled 4 times.
 ;