From a0d08405e7a64625982eb9c0c69a81fec657cfc6 Mon Sep 17 00:00:00 2001
From: Amara Emerson <aemerson@apple.com>
Date: Wed, 10 Apr 2019 23:06:11 +0000
Subject: [PATCH] [AArch64][GlobalISel] Add legalizer support for <8 x s16> and
 <16 x s8> G_ADD.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@358143 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/AArch64/AArch64LegalizerInfo.cpp   |  2 +-
 .../AArch64/GlobalISel/legalize-add.mir       | 46 ++++++++++++++
 .../AArch64/GlobalISel/select-binop.mir       | 61 +++++++++++++++++++
 3 files changed, 108 insertions(+), 1 deletion(-)

diff --git a/lib/Target/AArch64/AArch64LegalizerInfo.cpp b/lib/Target/AArch64/AArch64LegalizerInfo.cpp
index 3b4a1a90fb5..0c1e4b00fa0 100644
--- a/lib/Target/AArch64/AArch64LegalizerInfo.cpp
+++ b/lib/Target/AArch64/AArch64LegalizerInfo.cpp
@@ -76,7 +76,7 @@ AArch64LegalizerInfo::AArch64LegalizerInfo(const AArch64Subtarget &ST) {
       .widenScalarToNextPow2(0);
 
   getActionDefinitionsBuilder({G_ADD, G_SUB, G_MUL, G_AND, G_OR, G_XOR})
-      .legalFor({s32, s64, v2s32, v4s32, v2s64})
+      .legalFor({s32, s64, v2s32, v4s32, v2s64, v8s16, v16s8})
       .clampScalar(0, s32, s64)
       .widenScalarToNextPow2(0)
       .clampNumElements(0, v2s32, v4s32)
diff --git a/test/CodeGen/AArch64/GlobalISel/legalize-add.mir b/test/CodeGen/AArch64/GlobalISel/legalize-add.mir
index c3773f50cb0..3c2a52b7f0a 100644
--- a/test/CodeGen/AArch64/GlobalISel/legalize-add.mir
+++ b/test/CodeGen/AArch64/GlobalISel/legalize-add.mir
@@ -130,3 +130,49 @@ body:             |
     $q1 = COPY %8(<2 x s64>)
     $q2 = COPY %9(<2 x s64>)
 ...
+---
+name:            add_v8i16
+alignment:       2
+tracksRegLiveness: true
+machineFunctionInfo: {}
+body:             |
+  bb.1:
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: add_v8i16
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:_(<8 x s16>) = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $q1
+    ; CHECK: [[ADD:%[0-9]+]]:_(<8 x s16>) = G_ADD [[COPY]], [[COPY1]]
+    ; CHECK: $q0 = COPY [[ADD]](<8 x s16>)
+    ; CHECK: RET_ReallyLR implicit $q0
+    %0:_(<8 x s16>) = COPY $q0
+    %1:_(<8 x s16>) = COPY $q1
+    %2:_(<8 x s16>) = G_ADD %0, %1
+    $q0 = COPY %2(<8 x s16>)
+    RET_ReallyLR implicit $q0
+
+...
+---
+name:            add_v16i8
+alignment:       2
+tracksRegLiveness: true
+machineFunctionInfo: {}
+body:             |
+  bb.1:
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: add_v16i8
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:_(<16 x s8>) = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:_(<16 x s8>) = COPY $q1
+    ; CHECK: [[ADD:%[0-9]+]]:_(<16 x s8>) = G_ADD [[COPY]], [[COPY1]]
+    ; CHECK: $q0 = COPY [[ADD]](<16 x s8>)
+    ; CHECK: RET_ReallyLR implicit $q0
+    %0:_(<16 x s8>) = COPY $q0
+    %1:_(<16 x s8>) = COPY $q1
+    %2:_(<16 x s8>) = G_ADD %0, %1
+    $q0 = COPY %2(<16 x s8>)
+    RET_ReallyLR implicit $q0
+
+...
diff --git a/test/CodeGen/AArch64/GlobalISel/select-binop.mir b/test/CodeGen/AArch64/GlobalISel/select-binop.mir
index 8c858bcdb3e..9c4731e3b5d 100644
--- a/test/CodeGen/AArch64/GlobalISel/select-binop.mir
+++ b/test/CodeGen/AArch64/GlobalISel/select-binop.mir
@@ -54,6 +54,9 @@
   define void @fdiv_s32_fpr() { ret void }
   define void @fdiv_s64_fpr() { ret void }
 
+  define void @add_v8i16() { ret void }
+  define void @add_v16i8() { ret void }
+
 ...
 
 ---
@@ -912,3 +915,61 @@ body:             |
     %2(s64) = G_FDIV %0, %1
     $d0 = COPY %2(s64)
 ...
+---
+name:            add_v8i16
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1:
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: add_v8i16
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[ADDv8i16_:%[0-9]+]]:fpr128 = ADDv8i16 [[COPY]], [[COPY1]]
+    ; CHECK: $q0 = COPY [[ADDv8i16_]]
+    ; CHECK: RET_ReallyLR implicit $q0
+    %0:fpr(<8 x s16>) = COPY $q0
+    %1:fpr(<8 x s16>) = COPY $q1
+    %2:fpr(<8 x s16>) = G_ADD %0, %1
+    $q0 = COPY %2(<8 x s16>)
+    RET_ReallyLR implicit $q0
+
+...
+---
+name:            add_v16i8
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1:
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: add_v16i8
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[ADDv16i8_:%[0-9]+]]:fpr128 = ADDv16i8 [[COPY]], [[COPY1]]
+    ; CHECK: $q0 = COPY [[ADDv16i8_]]
+    ; CHECK: RET_ReallyLR implicit $q0
+    %0:fpr(<16 x s8>) = COPY $q0
+    %1:fpr(<16 x s8>) = COPY $q1
+    %2:fpr(<16 x s8>) = G_ADD %0, %1
+    $q0 = COPY %2(<16 x s8>)
+    RET_ReallyLR implicit $q0
+
+...
-- 
2.50.1