From 3fb13b8484dcbec085da047879bf89ccb1b65b12 Mon Sep 17 00:00:00 2001
From: Sanjay Patel <spatel@rotateright.com>
Date: Tue, 28 Jul 2020 15:13:19 -0400
Subject: [PATCH] [InstSimplify] allow undefs in icmp with vector constant
 folds

This is the main icmp simplification shortcoming seen in D84655.

Alive2 agrees that the basic examples are correct at least:

define <2 x i1> @src(<2 x i8> %x) {
%0:
  %r = icmp sle <2 x i8> { undef, 128 }, %x
  ret <2 x i1> %r
}
=>
define <2 x i1> @tgt(<2 x i8> %x) {
%0:
  ret <2 x i1> { 1, 1 }
}
Transformation seems to be correct!

define <2 x i1> @src(<2 x i32> %X) {
%0:
  %A = or <2 x i32> %X, { 63, 63 }
  %B = icmp ult <2 x i32> %A, { undef, 50 }
  ret <2 x i1> %B
}
=>
define <2 x i1> @tgt(<2 x i32> %X) {
%0:
  ret <2 x i1> { 0, 0 }
}
Transformation seems to be correct!

https://alive2.llvm.org/ce/z/omt2ee
https://alive2.llvm.org/ce/z/GW4nP_

Differential Revision: https://reviews.llvm.org/D84762
---
 llvm/lib/Analysis/InstructionSimplify.cpp          |  2 +-
 llvm/test/Transforms/InstSimplify/icmp-constant.ll | 22 ++++++----------------
 2 files changed, 7 insertions(+), 17 deletions(-)

diff --git a/llvm/lib/Analysis/InstructionSimplify.cpp b/llvm/lib/Analysis/InstructionSimplify.cpp
index c920fb3..01b2eb1 100644
--- a/llvm/lib/Analysis/InstructionSimplify.cpp
+++ b/llvm/lib/Analysis/InstructionSimplify.cpp
@@ -2732,7 +2732,7 @@ static Value *simplifyICmpWithConstant(CmpInst::Predicate Pred, Value *LHS,
   }
 
   const APInt *C;
-  if (!match(RHS, m_APInt(C)))
+  if (!match(RHS, m_APIntAllowUndef(C)))
     return nullptr;
 
   // Rule out tautological comparisons (eg., ult 0 or uge 0).
diff --git a/llvm/test/Transforms/InstSimplify/icmp-constant.ll b/llvm/test/Transforms/InstSimplify/icmp-constant.ll
index 8f51ba3..94126bf 100644
--- a/llvm/test/Transforms/InstSimplify/icmp-constant.ll
+++ b/llvm/test/Transforms/InstSimplify/icmp-constant.ll
@@ -21,8 +21,7 @@ define <2 x i1> @tautological_ule_vec(<2 x i8> %x) {
 
 define <2 x i1> @tautological_ule_vec_partial_undef(<2 x i8> %x) {
 ; CHECK-LABEL: @tautological_ule_vec_partial_undef(
-; CHECK-NEXT:    [[CMP:%.*]] = icmp ule <2 x i8> [[X:%.*]], <i8 -1, i8 undef>
-; CHECK-NEXT:    ret <2 x i1> [[CMP]]
+; CHECK-NEXT:    ret <2 x i1> <i1 true, i1 true>
 ;
   %cmp = icmp ule <2 x i8> %x, <i8 255, i8 undef>
   ret <2 x i1> %cmp
@@ -46,8 +45,7 @@ define <2 x i1> @tautological_ugt_vec(<2 x i8> %x) {
 
 define <2 x i1> @tautological_ugt_vec_partial_undef(<2 x i8> %x) {
 ; CHECK-LABEL: @tautological_ugt_vec_partial_undef(
-; CHECK-NEXT:    [[CMP:%.*]] = icmp ugt <2 x i8> [[X:%.*]], <i8 undef, i8 -1>
-; CHECK-NEXT:    ret <2 x i1> [[CMP]]
+; CHECK-NEXT:    ret <2 x i1> zeroinitializer
 ;
   %cmp = icmp ugt <2 x i8> %x, <i8 undef, i8 255>
   ret <2 x i1> %cmp
@@ -74,9 +72,7 @@ define <2 x i1> @urem3_vec(<2 x i32> %X) {
 
 define <2 x i1> @urem3_vec_partial_undef(<2 x i32> %X) {
 ; CHECK-LABEL: @urem3_vec_partial_undef(
-; CHECK-NEXT:    [[A:%.*]] = urem <2 x i32> [[X:%.*]], <i32 10, i32 10>
-; CHECK-NEXT:    [[B:%.*]] = icmp ult <2 x i32> [[A]], <i32 undef, i32 15>
-; CHECK-NEXT:    ret <2 x i1> [[B]]
+; CHECK-NEXT:    ret <2 x i1> <i1 true, i1 true>
 ;
   %A = urem <2 x i32> %X, <i32 10, i32 10>
   %B = icmp ult <2 x i32> %A, <i32 undef, i32 15>
@@ -104,9 +100,7 @@ define <2 x i1> @srem1_vec(<2 x i32> %X) {
 
 define <2 x i1> @srem1_vec_partial_undef(<2 x i32> %X) {
 ; CHECK-LABEL: @srem1_vec_partial_undef(
-; CHECK-NEXT:    [[A:%.*]] = srem <2 x i32> [[X:%.*]], <i32 -5, i32 -5>
-; CHECK-NEXT:    [[B:%.*]] = icmp sgt <2 x i32> [[A]], <i32 5, i32 undef>
-; CHECK-NEXT:    ret <2 x i1> [[B]]
+; CHECK-NEXT:    ret <2 x i1> zeroinitializer
 ;
   %A = srem <2 x i32> %X, <i32 -5, i32 -5>
   %B = icmp sgt <2 x i32> %A, <i32 5, i32 undef>
@@ -211,9 +205,7 @@ define <2 x i1> @shl5_vec(<2 x i32> %X) {
 
 define <2 x i1> @shl5_vec_partial_undef(<2 x i32> %X) {
 ; CHECK-LABEL: @shl5_vec_partial_undef(
-; CHECK-NEXT:    [[SUB:%.*]] = shl nuw <2 x i32> <i32 4, i32 4>, [[X:%.*]]
-; CHECK-NEXT:    [[CMP:%.*]] = icmp ugt <2 x i32> [[SUB]], <i32 undef, i32 3>
-; CHECK-NEXT:    ret <2 x i1> [[CMP]]
+; CHECK-NEXT:    ret <2 x i1> <i1 true, i1 true>
 ;
   %sub = shl nuw <2 x i32> <i32 4, i32 4>, %X
   %cmp = icmp ugt <2 x i32> %sub, <i32 undef, i32 3>
@@ -431,9 +423,7 @@ define <2 x i1> @or1_vec(<2 x i32> %X) {
 
 define <2 x i1> @or1_vec_partial_undef(<2 x i32> %X) {
 ; CHECK-LABEL: @or1_vec_partial_undef(
-; CHECK-NEXT:    [[A:%.*]] = or <2 x i32> [[X:%.*]], <i32 62, i32 62>
-; CHECK-NEXT:    [[B:%.*]] = icmp ult <2 x i32> [[A]], <i32 undef, i32 50>
-; CHECK-NEXT:    ret <2 x i1> [[B]]
+; CHECK-NEXT:    ret <2 x i1> zeroinitializer
 ;
   %A = or <2 x i32> %X, <i32 62, i32 62>
   %B = icmp ult <2 x i32> %A, <i32 undef, i32 50>
-- 
2.7.4