From cfcc42bdc2b880a3036337ac259f329ae56d8c22 Mon Sep 17 00:00:00 2001
From: Sanjay Patel <spatel@rotateright.com>
Date: Sun, 13 Nov 2016 20:04:52 +0000
Subject: [PATCH] [ValueTracking] recognize even more variants of smin/smax

Similar to:
https://reviews.llvm.org/rL285499
https://reviews.llvm.org/rL286318

We can't minimally expose this in IR tests because we don't have min/max intrinsics,
but the difference is visible in codegen because SelectionDAGBuilder::visitSelect()
uses matchSelectPattern().

We're not canonicalizing these patterns in IR (yet), so I don't expect there to be any
regressions as noted here:
http://lists.llvm.org/pipermail/llvm-dev/2016-November/106868.html

llvm-svn: 286776
---
 llvm/lib/Analysis/ValueTracking.cpp       | 20 ++++++++++++++++++++
 llvm/test/CodeGen/X86/vec_minmax_match.ll | 24 ++++++++++++------------
 2 files changed, 32 insertions(+), 12 deletions(-)

diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index 5a6ca94..18117f6 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -3861,6 +3861,26 @@ static SelectPatternResult matchMinMax(CmpInst::Predicate Pred,
   if (Pred != CmpInst::ICMP_SGT && Pred != CmpInst::ICMP_SLT)
     return {SPF_UNKNOWN, SPNB_NA, false};
 
+  // Z = X -nsw Y
+  // (X >s Y) ? 0 : Z ==> (Z >s 0) ? 0 : Z ==> SMIN(Z, 0)
+  // (X <s Y) ? 0 : Z ==> (Z <s 0) ? 0 : Z ==> SMAX(Z, 0)
+  if (match(TrueVal, m_Zero()) &&
+      match(FalseVal, m_NSWSub(m_Specific(CmpLHS), m_Specific(CmpRHS)))) {
+    LHS = TrueVal;
+    RHS = FalseVal;
+    return {Pred == CmpInst::ICMP_SGT ? SPF_SMIN : SPF_SMAX, SPNB_NA, false};
+  }
+
+  // Z = X -nsw Y
+  // (X >s Y) ? Z : 0 ==> (Z >s 0) ? Z : 0 ==> SMAX(Z, 0)
+  // (X <s Y) ? Z : 0 ==> (Z <s 0) ? Z : 0 ==> SMIN(Z, 0)
+  if (match(FalseVal, m_Zero()) &&
+      match(TrueVal, m_NSWSub(m_Specific(CmpLHS), m_Specific(CmpRHS)))) {
+    LHS = TrueVal;
+    RHS = FalseVal;
+    return {Pred == CmpInst::ICMP_SGT ? SPF_SMAX : SPF_SMIN, SPNB_NA, false};
+  }
+
   const APInt *C1;
   if (!match(CmpRHS, m_APInt(C1)))
     return {SPF_UNKNOWN, SPNB_NA, false};
diff --git a/llvm/test/CodeGen/X86/vec_minmax_match.ll b/llvm/test/CodeGen/X86/vec_minmax_match.ll
index 6293a61..af4410a 100644
--- a/llvm/test/CodeGen/X86/vec_minmax_match.ll
+++ b/llvm/test/CodeGen/X86/vec_minmax_match.ll
@@ -37,9 +37,9 @@ define <4 x i32> @smin_vec2(<4 x i32> %x) {
 define <4 x i32> @smin_vec3(<4 x i32> %x, <4 x i32> %y) {
 ; CHECK-LABEL: smin_vec3:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    vpsubd %xmm1, %xmm0, %xmm2
-; CHECK-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
-; CHECK-NEXT:    vpandn %xmm2, %xmm0, %xmm0
+; CHECK-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
+; CHECK-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-NEXT:    vpminsd %xmm1, %xmm0, %xmm0
 ; CHECK-NEXT:    retq
 ;
   %sub = sub nsw <4 x i32> %x, %y
@@ -53,9 +53,9 @@ define <4 x i32> @smin_vec3(<4 x i32> %x, <4 x i32> %y) {
 define <4 x i32> @smin_vec4(<4 x i32> %x, <4 x i32> %y) {
 ; CHECK-LABEL: smin_vec4:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    vpsubd %xmm1, %xmm0, %xmm2
-; CHECK-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
-; CHECK-NEXT:    vpand %xmm2, %xmm0, %xmm0
+; CHECK-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
+; CHECK-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-NEXT:    vpminsd %xmm1, %xmm0, %xmm0
 ; CHECK-NEXT:    retq
 ;
   %sub = sub nsw <4 x i32> %x, %y
@@ -97,9 +97,9 @@ define <4 x i32> @smax_vec2(<4 x i32> %x) {
 define <4 x i32> @smax_vec3(<4 x i32> %x, <4 x i32> %y) {
 ; CHECK-LABEL: smax_vec3:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    vpsubd %xmm1, %xmm0, %xmm2
-; CHECK-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
-; CHECK-NEXT:    vpandn %xmm2, %xmm0, %xmm0
+; CHECK-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
+; CHECK-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-NEXT:    vpmaxsd %xmm1, %xmm0, %xmm0
 ; CHECK-NEXT:    retq
 ;
   %sub = sub nsw <4 x i32> %x, %y
@@ -113,9 +113,9 @@ define <4 x i32> @smax_vec3(<4 x i32> %x, <4 x i32> %y) {
 define <4 x i32> @smax_vec4(<4 x i32> %x, <4 x i32> %y) {
 ; CHECK-LABEL: smax_vec4:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    vpsubd %xmm1, %xmm0, %xmm2
-; CHECK-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
-; CHECK-NEXT:    vpand %xmm2, %xmm0, %xmm0
+; CHECK-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
+; CHECK-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-NEXT:    vpmaxsd %xmm1, %xmm0, %xmm0
 ; CHECK-NEXT:    retq
 ;
   %sub = sub nsw <4 x i32> %x, %y
-- 
2.7.4