[NewPM] Run non-trivial loop unswitching under -O2/3/s/z

[aeubanks]

Summary

Fixes https://bugs.llvm.org/show_bug.cgi?id=48715.

[aeubanks]

parameter comment

[asbirlea]

lgtm.

[nikic]

FYI this change had some pretty significant impact on compile-time (https://llvm-compile-time-tracker.com/compare.php?from=4718ec01669b01373180f4cd1256c6e2dd6f3999&to=f748e92295515ea7b39cd687a718915b559de6ec&stat=instructions) and code size (https://llvm-compile-time-tracker.com/compare.php?from=4718ec01669b01373180f4cd1256c6e2dd6f3999&to=f748e92295515ea7b39cd687a718915b559de6ec&stat=size-text). Is that expected?

[aeubanks]

In D94448#2494328, @nikic wrote:

FYI this change had some pretty significant impact on compile-time (https://llvm-compile-time-tracker.com/compare.php?from=4718ec01669b01373180f4cd1256c6e2dd6f3999&to=f748e92295515ea7b39cd687a718915b559de6ec&stat=instructions) and code size (https://llvm-compile-time-tracker.com/compare.php?from=4718ec01669b01373180f4cd1256c6e2dd6f3999&to=f748e92295515ea7b39cd687a718915b559de6ec&stat=size-text). Is that expected?

I took a closer look at what the legacy unswitch pass does in regards to -O2/3/s/z, and it skips non-trivial unswitching for anything other than -O3. So in the -O3 case I think the increased compile time and code size is expected. But the other cases should be fixed, I'll do that.
(weird that sqlite3 is an outlier and compiles 11% faster with this change though)

(also it would be nice to have -O1/2/s/z runs on llvm-compile-time-tracker, but of course you can only test so many configurations)

[asbirlea]

I'm curious if the compile time impact is complemented by impact in run time. Looking into that for O3.

llvm/lib/Passes/PassBuilder.cpp

PassBuilder::buildFunctionSimplificationPipeline(OptimizationLevel Level,
   // implications on the outer loop.
   LPM1.addPass(LoopInstSimplifyPass());
   LPM1.addPass(LoopSimplifyCFGPass());
 
   // Disable header duplication in loop rotation at -Oz.
   LPM1.addPass(LoopRotatePass(Level != OptimizationLevel::Oz));
   // TODO: Investigate promotion cap for O1.
   LPM1.addPass(LICMPass(PTO.LicmMssaOptCap, PTO.LicmMssaNoAccForPromotionCap));
-  LPM1.addPass(SimpleLoopUnswitchPass());
+  LPM1.addPass(SimpleLoopUnswitchPass(/* NonTrivial */ true));
   LPM2.addPass(LoopIdiomRecognizePass());
   LPM2.addPass(IndVarSimplifyPass());
 
   for (auto &C : LateLoopOptimizationsEPCallbacks)
     C(LPM2, Level);
 
   LPM2.addPass(LoopDeletionPass());
   // Do not enable unrolling in PreLinkThinLTO phase during sample PGO

llvm/test/Transforms/LoopUnroll/opt-levels.ll

 ; RUN: opt < %s -S -passes="default<O2>" -unroll-runtime=true -unroll-threshold-default=0 -unroll-threshold-aggressive=300 | FileCheck %s -check-prefix=O2
 ; RUN: opt < %s -S -passes="default<O3>" -unroll-runtime=true -unroll-threshold-default=0 -unroll-threshold-aggressive=300 | FileCheck %s -check-prefix=O3
 ; RUN: opt < %s -S -passes="default<Os>" -unroll-runtime=true -unroll-threshold-default=0 -unroll-threshold-aggressive=300 | FileCheck %s -check-prefix=Os
 ; RUN: opt < %s -S -passes="default<Oz>" -unroll-runtime=true -unroll-threshold-default=0 -unroll-threshold-aggressive=300 | FileCheck %s -check-prefix=Oz
 
 ; Check that Os and Oz are optimized like O2, not like O3. To easily highlight
 ; the behavior, we artificially disable unrolling for anything but O3 by setting
 ; the default threshold to 0.
 
-; O3:     loop2.preheader
-; O2-NOT: loop2.preheader
-; Os-NOT: loop2.preheader
-; Oz-NOT: loop2.preheader
+; O3:     loop1.preheader
+; O2-NOT: loop1.preheader
+; Os-NOT: loop1.preheader
+; Oz-NOT: loop1.preheader
 
 define void @unroll(i32 %iter, i32* %addr1, i32* %addr2) nounwind {
 entry:
   br label %loop1
 
 loop1:
   %iv1 = phi i32 [ 0, %entry ], [ %inc1, %loop1.latch ]
   %offset1 = getelementptr i32, i32* %addr1, i32 %iv1

llvm/test/Transforms/SimpleLoopUnswitch/pipeline.ll

+; RUN: opt < %s -S -passes="default<O1>" | FileCheck %s -check-prefix=O1
+; RUN: opt < %s -S -passes="default<O2>" | FileCheck %s -check-prefix=O2
+
+declare i32 @a()
+declare i32 @b()
+declare i32 @c()
+
+; O1-NOT: loop_begin.us:
+; O2: loop_begin.us:
+
+define i32 @test1(i1* %ptr, i1 %cond1, i1 %cond2) {
+entry:
+  br label %loop_begin
+
+loop_begin:
+  br i1 %cond1, label %loop_a, label %loop_b
+
+loop_a:
+  call i32 @a()
+  br label %latch
+
+loop_b:
+  br i1 %cond2, label %loop_b_a, label %loop_b_b
+
+loop_b_a:
+  call i32 @b()
+  br label %latch
+
+loop_b_b:
+  call i32 @c()
+  br label %latch
+
+latch:
+  %v = load i1, i1* %ptr
+  br i1 %v, label %loop_begin, label %loop_exit
+
+loop_exit:
+  ret i32 0
+}