[InstCombine] foldICmpBinOp(): consider inverted check in 'unsigned sub overflow' check

A follow-up for r329011.
This may be changed to produce @llvm.sub.with.overflow in a later patch,
but for now just make things more consistent overall.

A few observations stem from this:
* There does not seem to be a similar one-instruction fold for uadd-overflow
* I'm not sure we'll want to canonicalize `B u> A` as `usub.with.overflow`,
  so since the `icmp` here no longer refers to `sub`,
  reconstructing `usub.with.overflow` will be problematic,
  and will likely require standalone pass (similar to DivRemPairs).

https://rise4fun.com/Alive/Zqs

Name: (A - B) u> A --> B u> A
  %t0 = sub i8 %A, %B
  %r = icmp ugt i8 %t0, %A
=>
  %r = icmp ugt i8 %B, %A

Name: (A - B) u<= A --> B u<= A
  %t0 = sub i8 %A, %B
  %r = icmp ule i8 %t0, %A
=>
  %r = icmp ule i8 %B, %A

Name: C u< (C - D) --> C u< D
  %t0 = sub i8 %C, %D
  %r = icmp ult i8 %C, %t0
=>
  %r = icmp ult i8 %C, %D

Name: C u>= (C - D) --> C u>= D
  %t0 = sub i8 %C, %D
  %r = icmp uge i8 %C, %t0
=>
  %r = icmp uge i8 %C, %D

llvm-svn: 371101

llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp

@@ -3791,12 +3791,13 @@ Instruction *InstCombiner::foldICmpBinOp(ICmpInst &I) {
   if (C == Op0 && NoOp1WrapProblem)
     return new ICmpInst(Pred, D, Constant::getNullValue(Op0->getType()));
 
-  // (A - B) >u A --> A <u B
-  if (A == Op1 && Pred == ICmpInst::ICMP_UGT)
-    return new ICmpInst(ICmpInst::ICMP_ULT, A, B);
-  // C <u (C - D) --> C <u D
-  if (C == Op0 && Pred == ICmpInst::ICMP_ULT)
-    return new ICmpInst(ICmpInst::ICMP_ULT, C, D);
+  // Convert sub-with-unsigned-overflow comparisons into a comparison of args.
+  // (A - B) u>/u<= A --> B u>/u<= A
+  if (A == Op1 && (Pred == ICmpInst::ICMP_UGT || Pred == ICmpInst::ICMP_ULE))
+    return new ICmpInst(Pred, B, A);
+  // C u</u>= (C - D) --> C u</u>= D
+  if (C == Op0 && (Pred == ICmpInst::ICMP_ULT || Pred == ICmpInst::ICMP_UGE))
+    return new ICmpInst(Pred, C, D);
 
   // icmp (Y-X), (Z-X) -> icmp Y, Z for equalities or if there is no overflow.
   if (B && D && B == D && NoOp0WrapProblem && NoOp1WrapProblem &&

llvm/test/Transforms/InstCombine/unsigned-sub-lack-of-overflow-check.ll

@@ -8,8 +8,7 @@
 
 define i1 @t0_basic(i8 %x, i8 %y) {
 ; CHECK-LABEL: @t0_basic(
-; CHECK-NEXT:    [[T0:%.*]] = sub i8 [[X:%.*]], [[Y:%.*]]
-; CHECK-NEXT:    [[R:%.*]] = icmp ule i8 [[T0]], [[X]]
+; CHECK-NEXT:    [[R:%.*]] = icmp ule i8 [[Y:%.*]], [[X:%.*]]
 ; CHECK-NEXT:    ret i1 [[R]]
 ;
   %t0 = sub i8 %x, %y
@@ -19,8 +18,7 @@ define i1 @t0_basic(i8 %x, i8 %y) {
 
 define <2 x i1> @t1_vec(<2 x i8> %x, <2 x i8> %y) {
 ; CHECK-LABEL: @t1_vec(
-; CHECK-NEXT:    [[T0:%.*]] = sub <2 x i8> [[X:%.*]], [[Y:%.*]]
-; CHECK-NEXT:    [[R:%.*]] = icmp ule <2 x i8> [[T0]], [[X]]
+; CHECK-NEXT:    [[R:%.*]] = icmp ule <2 x i8> [[Y:%.*]], [[X:%.*]]
 ; CHECK-NEXT:    ret <2 x i1> [[R]]
 ;
   %t0 = sub <2 x i8> %x, %y
@@ -32,8 +30,7 @@ define <2 x i1> @t1_vec(<2 x i8> %x, <2 x i8> %y) {
 
 define i1 @t2_commutative(i8 %x, i8 %y) {
 ; CHECK-LABEL: @t2_commutative(
-; CHECK-NEXT:    [[T0:%.*]] = sub i8 [[X:%.*]], [[Y:%.*]]
-; CHECK-NEXT:    [[R:%.*]] = icmp ule i8 [[T0]], [[X]]
+; CHECK-NEXT:    [[R:%.*]] = icmp ule i8 [[Y:%.*]], [[X:%.*]]
 ; CHECK-NEXT:    ret i1 [[R]]
 ;
   %t0 = sub i8 %x, %y
@@ -49,7 +46,7 @@ define i1 @t3_extrause0(i8 %x, i8 %y) {
 ; CHECK-LABEL: @t3_extrause0(
 ; CHECK-NEXT:    [[T0:%.*]] = sub i8 [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    call void @use8(i8 [[T0]])
-; CHECK-NEXT:    [[R:%.*]] = icmp ule i8 [[T0]], [[X]]
+; CHECK-NEXT:    [[R:%.*]] = icmp uge i8 [[X]], [[Y]]
 ; CHECK-NEXT:    ret i1 [[R]]
 ;
   %t0 = sub i8 %x, %y

llvm/test/Transforms/InstCombine/unsigned-sub-overflow-check.ll

@@ -8,7 +8,7 @@
 
 define i1 @t0_basic(i8 %x, i8 %y) {
 ; CHECK-LABEL: @t0_basic(
-; CHECK-NEXT:    [[R:%.*]] = icmp ult i8 [[X:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    [[R:%.*]] = icmp ugt i8 [[Y:%.*]], [[X:%.*]]
 ; CHECK-NEXT:    ret i1 [[R]]
 ;
   %t0 = sub i8 %x, %y
@@ -18,7 +18,7 @@ define i1 @t0_basic(i8 %x, i8 %y) {
 
 define <2 x i1> @t1_vec(<2 x i8> %x, <2 x i8> %y) {
 ; CHECK-LABEL: @t1_vec(
-; CHECK-NEXT:    [[R:%.*]] = icmp ult <2 x i8> [[X:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    [[R:%.*]] = icmp ugt <2 x i8> [[Y:%.*]], [[X:%.*]]
 ; CHECK-NEXT:    ret <2 x i1> [[R]]
 ;
   %t0 = sub <2 x i8> %x, %y
@@ -30,7 +30,7 @@ define <2 x i1> @t1_vec(<2 x i8> %x, <2 x i8> %y) {
 
 define i1 @t2_commutative(i8 %x, i8 %y) {
 ; CHECK-LABEL: @t2_commutative(
-; CHECK-NEXT:    [[R:%.*]] = icmp ult i8 [[X:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    [[R:%.*]] = icmp ugt i8 [[Y:%.*]], [[X:%.*]]
 ; CHECK-NEXT:    ret i1 [[R]]
 ;
   %t0 = sub i8 %x, %y