From 250ffab4a28b0325aee48e8281fccca7fe03c06b Mon Sep 17 00:00:00 2001
From: Amara Emerson <aemerson@apple.com>
Date: Wed, 10 Apr 2019 23:06:08 +0000
Subject: [PATCH] [AArch64][GlobalISel] Scalarize vector SDIV.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@358142 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/CodeGen/GlobalISel/LegalizerHelper.cpp    |  1 +
 lib/Target/AArch64/AArch64LegalizerInfo.cpp   |  3 +-
 .../AArch64/GlobalISel/legalize-div.mir       | 37 +++++++++++++++++--
 3 files changed, 36 insertions(+), 5 deletions(-)

diff --git a/lib/CodeGen/GlobalISel/LegalizerHelper.cpp b/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
index f563b8c5db7..08759f28036 100644
--- a/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
+++ b/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
@@ -2190,6 +2190,7 @@ LegalizerHelper::fewerElementsVector(MachineInstr &MI, unsigned TypeIdx,
   case G_FSIN:
   case G_FSQRT:
   case G_BSWAP:
+  case G_SDIV:
     return fewerElementsVectorBasic(MI, TypeIdx, NarrowTy);
   case G_SHL:
   case G_LSHR:
diff --git a/lib/Target/AArch64/AArch64LegalizerInfo.cpp b/lib/Target/AArch64/AArch64LegalizerInfo.cpp
index b2d0a0ca0a1..3b4a1a90fb5 100644
--- a/lib/Target/AArch64/AArch64LegalizerInfo.cpp
+++ b/lib/Target/AArch64/AArch64LegalizerInfo.cpp
@@ -103,7 +103,8 @@ AArch64LegalizerInfo::AArch64LegalizerInfo(const AArch64Subtarget &ST) {
   getActionDefinitionsBuilder({G_SDIV, G_UDIV})
       .legalFor({s32, s64})
       .clampScalar(0, s32, s64)
-      .widenScalarToNextPow2(0);
+      .widenScalarToNextPow2(0)
+      .scalarize(0);
 
   getActionDefinitionsBuilder({G_LSHR, G_ASHR})
       .legalFor({{s32, s32}, {s64, s64}, {v2s32, v2s32}, {v4s32, v4s32}})
diff --git a/test/CodeGen/AArch64/GlobalISel/legalize-div.mir b/test/CodeGen/AArch64/GlobalISel/legalize-div.mir
index 4f4ff0a7fee..2b9bc347b49 100644
--- a/test/CodeGen/AArch64/GlobalISel/legalize-div.mir
+++ b/test/CodeGen/AArch64/GlobalISel/legalize-div.mir
@@ -9,12 +9,12 @@ body:             |
     ; CHECK: [[COPY1:%[0-9]+]]:_(s64) = COPY $x1
     ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
     ; CHECK: [[TRUNC:%[0-9]+]]:_(s32) = G_TRUNC [[COPY]](s64)
-    ; CHECK: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[TRUNC]], [[C]]
-    ; CHECK: [[ASHR:%[0-9]+]]:_(s32) = G_ASHR [[SHL]], [[C]]
+    ; CHECK: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[TRUNC]], [[C]](s32)
+    ; CHECK: [[ASHR:%[0-9]+]]:_(s32) = G_ASHR [[SHL]], [[C]](s32)
     ; CHECK: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
     ; CHECK: [[TRUNC1:%[0-9]+]]:_(s32) = G_TRUNC [[COPY1]](s64)
-    ; CHECK: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[TRUNC1]], [[C1]]
-    ; CHECK: [[ASHR1:%[0-9]+]]:_(s32) = G_ASHR [[SHL1]], [[C1]]
+    ; CHECK: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[TRUNC1]], [[C1]](s32)
+    ; CHECK: [[ASHR1:%[0-9]+]]:_(s32) = G_ASHR [[SHL1]], [[C1]](s32)
     ; CHECK: [[SDIV:%[0-9]+]]:_(s32) = G_SDIV [[ASHR]], [[ASHR1]]
     ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY [[SDIV]](s32)
     ; CHECK: $w0 = COPY [[COPY2]](s32)
@@ -39,3 +39,32 @@ body:             |
     $w0 = COPY %7(s32)
 
 ...
+---
+name:            sdiv_v4s32
+alignment:       2
+tracksRegLiveness: true
+machineFunctionInfo: {}
+body:             |
+  bb.1:
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: sdiv_v4s32
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $q1
+    ; CHECK: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
+    ; CHECK: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<4 x s32>)
+    ; CHECK: [[SDIV:%[0-9]+]]:_(s32) = G_SDIV [[UV]], [[UV4]]
+    ; CHECK: [[SDIV1:%[0-9]+]]:_(s32) = G_SDIV [[UV1]], [[UV5]]
+    ; CHECK: [[SDIV2:%[0-9]+]]:_(s32) = G_SDIV [[UV2]], [[UV6]]
+    ; CHECK: [[SDIV3:%[0-9]+]]:_(s32) = G_SDIV [[UV3]], [[UV7]]
+    ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[SDIV]](s32), [[SDIV1]](s32), [[SDIV2]](s32), [[SDIV3]](s32)
+    ; CHECK: $q0 = COPY [[BUILD_VECTOR]](<4 x s32>)
+    ; CHECK: RET_ReallyLR implicit $q0
+    %0:_(<4 x s32>) = COPY $q0
+    %1:_(<4 x s32>) = COPY $q1
+    %2:_(<4 x s32>) = G_SDIV %0, %1
+    $q0 = COPY %2(<4 x s32>)
+    RET_ReallyLR implicit $q0
+
+...
-- 
2.50.1