From 1e61e358a4ed048c3900c32dc84a2665c592cc4c Mon Sep 17 00:00:00 2001
From: Roman Lebedev <lebedev.ri@gmail.com>
Date: Sat, 14 Jul 2018 20:08:26 +0000
Subject: [PATCH] [InstCombine] Fold  x s<= x & (-1 >> y)  to  x s<= (-1 >> y)

https://bugs.llvm.org/show_bug.cgi?id=38123
https://rise4fun.com/Alive/I3O

This pattern is not commutative!
We must make sure not to fold the commuted version!

llvm-svn: 337107
---
 .../Transforms/InstCombine/InstCombineCompares.cpp |  6 ++++
 ...nstant-low-bit-mask-and-icmp-sle-to-icmp-sle.ll | 34 +++++++++-------------
 2 files changed, 20 insertions(+), 20 deletions(-)

diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp
index 4e42b9e..9bd5d53 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp
@@ -2918,6 +2918,12 @@ static Value *foldICmpWithLowBitMaskedVal(ICmpInst &I,
       return nullptr;         // Ignore the other case.
     DstPred = ICmpInst::Predicate::ICMP_SGT;
     break;
+  case ICmpInst::Predicate::ICMP_SLE:
+    //  x s<= x & (-1 >> y)    ->    x s<= (-1 >> y)
+    if (X != I.getOperand(0)) // X must be on LHS of comparison!
+      return nullptr;         // Ignore the other case.
+    DstPred = ICmpInst::Predicate::ICMP_SLE;
+    break;
   // TODO: more folds are possible, https://bugs.llvm.org/show_bug.cgi?id=38123
   default:
     return nullptr;
diff --git a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-sle-to-icmp-sle.ll b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-sle-to-icmp-sle.ll
index 6c9b9ed..11596fc 100644
--- a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-sle-to-icmp-sle.ll
+++ b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-sle-to-icmp-sle.ll
@@ -22,9 +22,8 @@ declare <3 x i8> @gen3x8()
 define i1 @p0() {
 ; CHECK-LABEL: @p0(
 ; CHECK-NEXT:    [[X:%.*]] = call i8 @gen8()
-; CHECK-NEXT:    [[TMP0:%.*]] = and i8 [[X]], 3
-; CHECK-NEXT:    [[RET:%.*]] = icmp sle i8 [[X]], [[TMP0]]
-; CHECK-NEXT:    ret i1 [[RET]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp slt i8 [[X]], 4
+; CHECK-NEXT:    ret i1 [[TMP1]]
 ;
   %x = call i8 @gen8()
   %tmp0 = and i8 %x, 3
@@ -36,9 +35,8 @@ define i1 @pv(i8 %y) {
 ; CHECK-LABEL: @pv(
 ; CHECK-NEXT:    [[X:%.*]] = call i8 @gen8()
 ; CHECK-NEXT:    [[TMP0:%.*]] = lshr i8 -1, [[Y:%.*]]
-; CHECK-NEXT:    [[TMP1:%.*]] = and i8 [[TMP0]], [[X]]
-; CHECK-NEXT:    [[RET:%.*]] = icmp sle i8 [[X]], [[TMP1]]
-; CHECK-NEXT:    ret i1 [[RET]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp sle i8 [[X]], [[TMP0]]
+; CHECK-NEXT:    ret i1 [[TMP1]]
 ;
   %x = call i8 @gen8()
   %tmp0 = lshr i8 -1, %y
@@ -54,9 +52,8 @@ define i1 @pv(i8 %y) {
 define <2 x i1> @p1_vec_splat() {
 ; CHECK-LABEL: @p1_vec_splat(
 ; CHECK-NEXT:    [[X:%.*]] = call <2 x i8> @gen2x8()
-; CHECK-NEXT:    [[TMP0:%.*]] = and <2 x i8> [[X]], <i8 3, i8 3>
-; CHECK-NEXT:    [[RET:%.*]] = icmp sle <2 x i8> [[X]], [[TMP0]]
-; CHECK-NEXT:    ret <2 x i1> [[RET]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp slt <2 x i8> [[X]], <i8 4, i8 4>
+; CHECK-NEXT:    ret <2 x i1> [[TMP1]]
 ;
   %x = call <2 x i8> @gen2x8()
   %tmp0 = and <2 x i8> %x, <i8 3, i8 3>
@@ -67,9 +64,8 @@ define <2 x i1> @p1_vec_splat() {
 define <2 x i1> @p2_vec_nonsplat() {
 ; CHECK-LABEL: @p2_vec_nonsplat(
 ; CHECK-NEXT:    [[X:%.*]] = call <2 x i8> @gen2x8()
-; CHECK-NEXT:    [[TMP0:%.*]] = and <2 x i8> [[X]], <i8 3, i8 15>
-; CHECK-NEXT:    [[RET:%.*]] = icmp sle <2 x i8> [[X]], [[TMP0]]
-; CHECK-NEXT:    ret <2 x i1> [[RET]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp slt <2 x i8> [[X]], <i8 4, i8 16>
+; CHECK-NEXT:    ret <2 x i1> [[TMP1]]
 ;
   %x = call <2 x i8> @gen2x8()
   %tmp0 = and <2 x i8> %x, <i8 3, i8 15> ; doesn't have to be splat.
@@ -80,9 +76,8 @@ define <2 x i1> @p2_vec_nonsplat() {
 define <3 x i1> @p3_vec_splat_undef() {
 ; CHECK-LABEL: @p3_vec_splat_undef(
 ; CHECK-NEXT:    [[X:%.*]] = call <3 x i8> @gen3x8()
-; CHECK-NEXT:    [[TMP0:%.*]] = and <3 x i8> [[X]], <i8 3, i8 undef, i8 3>
-; CHECK-NEXT:    [[RET:%.*]] = icmp sle <3 x i8> [[X]], [[TMP0]]
-; CHECK-NEXT:    ret <3 x i1> [[RET]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp slt <3 x i8> [[X]], <i8 4, i8 undef, i8 4>
+; CHECK-NEXT:    ret <3 x i1> [[TMP1]]
 ;
   %x = call <3 x i8> @gen3x8()
   %tmp0 = and <3 x i8> %x, <i8 3, i8 undef, i8 3>
@@ -101,8 +96,8 @@ define i1 @oneuse0() {
 ; CHECK-NEXT:    [[X:%.*]] = call i8 @gen8()
 ; CHECK-NEXT:    [[TMP0:%.*]] = and i8 [[X]], 3
 ; CHECK-NEXT:    call void @use8(i8 [[TMP0]])
-; CHECK-NEXT:    [[RET:%.*]] = icmp sle i8 [[X]], [[TMP0]]
-; CHECK-NEXT:    ret i1 [[RET]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp slt i8 [[X]], 4
+; CHECK-NEXT:    ret i1 [[TMP1]]
 ;
   %x = call i8 @gen8()
   %tmp0 = and i8 %x, 3
@@ -139,9 +134,8 @@ define i1 @cv0_GOOD(i8 %y) {
 ; CHECK-LABEL: @cv0_GOOD(
 ; CHECK-NEXT:    [[X:%.*]] = call i8 @gen8()
 ; CHECK-NEXT:    [[TMP0:%.*]] = lshr i8 -1, [[Y:%.*]]
-; CHECK-NEXT:    [[TMP1:%.*]] = and i8 [[TMP0]], [[X]]
-; CHECK-NEXT:    [[RET:%.*]] = icmp sle i8 [[X]], [[TMP1]]
-; CHECK-NEXT:    ret i1 [[RET]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp sle i8 [[X]], [[TMP0]]
+; CHECK-NEXT:    ret i1 [[TMP1]]
 ;
   %x = call i8 @gen8()
   %tmp0 = lshr i8 -1, %y
-- 
2.7.4