From a971f6f967db2edc027bb5c1e59e5c9c3ab0b13b Mon Sep 17 00:00:00 2001
From: Evan Cheng <evan.cheng@apple.com>
Date: Thu, 23 Mar 2006 01:57:24 +0000
Subject: [PATCH] Add 128-bit integer vector load and add (for testing).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26967 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/X86/X86ISelLowering.cpp |  8 ++++-
 lib/Target/X86/X86InstrSSE.td      | 48 +++++++++++++++++++++++++++---
 2 files changed, 51 insertions(+), 5 deletions(-)

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index ea3832f70c1..e7e8c4f1987 100644
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -290,14 +290,20 @@ X86TargetLowering::X86TargetLowering(TargetMachine &TM)
 
 
     setOperationAction(ISD::ADD,              MVT::v2f64, Legal);
+    setOperationAction(ISD::ADD,              MVT::v16i8, Legal);
+    setOperationAction(ISD::ADD,              MVT::v8i16, Legal);
+    setOperationAction(ISD::ADD,              MVT::v4i32, Legal);
     setOperationAction(ISD::SUB,              MVT::v2f64, Legal);
     setOperationAction(ISD::MUL,              MVT::v2f64, Legal);
     setOperationAction(ISD::LOAD,             MVT::v2f64, Legal);
+    setOperationAction(ISD::LOAD,             MVT::v16i8, Legal);
+    setOperationAction(ISD::LOAD,             MVT::v8i16, Legal);
+    setOperationAction(ISD::LOAD,             MVT::v4i32, Legal);
+    setOperationAction(ISD::LOAD,             MVT::v2i64, Legal);
     setOperationAction(ISD::BUILD_VECTOR,     MVT::v2f64, Expand);
     setOperationAction(ISD::BUILD_VECTOR,     MVT::v16i8, Expand);
     setOperationAction(ISD::BUILD_VECTOR,     MVT::v8i16, Expand);
     setOperationAction(ISD::BUILD_VECTOR,     MVT::v4i32, Expand);
-    setOperationAction(ISD::BUILD_VECTOR,     MVT::v2i64, Expand);
     setOperationAction(ISD::SCALAR_TO_VECTOR, MVT::v16i8, Custom);
     setOperationAction(ISD::SCALAR_TO_VECTOR, MVT::v8i16, Custom);
     setOperationAction(ISD::VECTOR_SHUFFLE,   MVT::v2f64, Custom);
diff --git a/lib/Target/X86/X86InstrSSE.td b/lib/Target/X86/X86InstrSSE.td
index f254112ec4b..518b6659ed8 100644
--- a/lib/Target/X86/X86InstrSSE.td
+++ b/lib/Target/X86/X86InstrSSE.td
@@ -770,6 +770,32 @@ def MOVQ128rm : I<0x7E, MRMSrcMem, (ops VR128:$dst, i64mem:$src),
 def MOVQ128mr : PDI<0xD6, MRMSrcMem, (ops i64mem:$dst, VR128:$src),
                   "movq {$src, $dst|$dst, $src}", []>;
 
+// 128-bit Integer Arithmetic
+let isTwoAddress = 1 in {
+let isCommutable = 1 in {
+def PADDBrr : PDI<0xFC, MRMSrcReg, (ops VR128:$dst, VR128:$src1, VR128:$src2),
+                  "paddb {$src2, $dst|$dst, $src2}",
+              [(set VR128:$dst, (v16i8 (add VR128:$src1, VR128:$src2)))]>;
+def PADDWrr : PDI<0xFD, MRMSrcReg, (ops VR128:$dst, VR128:$src1, VR128:$src2),
+                  "paddw {$src2, $dst|$dst, $src2}",
+              [(set VR128:$dst, (v8i16 (add VR128:$src1, VR128:$src2)))]>;
+def PADDDrr : PDI<0xFE, MRMSrcReg, (ops VR128:$dst, VR128:$src1, VR128:$src2),
+                  "paddd {$src2, $dst|$dst, $src2}",
+              [(set VR128:$dst, (v4i32 (add VR128:$src1, VR128:$src2)))]>;
+}
+def PADDBrm : PDI<0xFC, MRMSrcMem, (ops VR128:$dst, VR128:$src1, f128mem:$src2),
+                  "paddb {$src2, $dst|$dst, $src2}",
+              [(set VR128:$dst, (v16i8 (add VR128:$src1,
+                                        (load addr:$src2))))]>;
+def PADDWrm : PDI<0xFD, MRMSrcMem, (ops VR128:$dst, VR128:$src1, f128mem:$src2),
+                  "paddw {$src2, $dst|$dst, $src2}",
+              [(set VR128:$dst, (v8i16 (add VR128:$src1,
+                                        (load addr:$src2))))]>;
+def PADDDrm : PDI<0xFE, MRMSrcMem, (ops VR128:$dst, VR128:$src1, f128mem:$src2),
+                  "paddd {$src2, $dst|$dst, $src2}",
+              [(set VR128:$dst, (v4i32 (add VR128:$src1,
+                                        (load addr:$src2))))]>;
+}
 
 //===----------------------------------------------------------------------===//
 // Alias Instructions
@@ -796,11 +822,25 @@ def : Pat<(v8i16 (undef)), (IMPLICIT_DEF_VR128)>, Requires<[HasSSE2]>;
 def : Pat<(v4i32 (undef)), (IMPLICIT_DEF_VR128)>, Requires<[HasSSE2]>;
 def : Pat<(v2i64 (undef)), (IMPLICIT_DEF_VR128)>, Requires<[HasSSE2]>;
 
+// Load 128-bit integer vector values.
+def : Pat<(v16i8 (load addr:$src)), (MOVAPSrm addr:$src)>,
+  Requires<[HasSSE1]>;
+def : Pat<(v8i16 (load addr:$src)), (MOVAPSrm addr:$src)>,
+  Requires<[HasSSE1]>;
+def : Pat<(v4i32 (load addr:$src)), (MOVAPSrm addr:$src)>,
+  Requires<[HasSSE1]>;
+def : Pat<(v2i64 (load addr:$src)), (MOVAPDrm addr:$src)>,
+  Requires<[HasSSE2]>;
+
 // Store 128-bit integer vector values.
-def : Pat<(store (v16i8 VR128:$src), addr:$dst), (MOVAPSmr addr:$dst, VR128:$src)>;
-def : Pat<(store (v8i16 VR128:$src), addr:$dst), (MOVAPSmr addr:$dst, VR128:$src)>;
-def : Pat<(store (v4i32 VR128:$src), addr:$dst), (MOVAPSmr addr:$dst, VR128:$src)>;
-def : Pat<(store (v2i64 VR128:$src), addr:$dst), (MOVAPSmr addr:$dst, VR128:$src)>;
+def : Pat<(store (v16i8 VR128:$src), addr:$dst), (MOVAPSmr addr:$dst, VR128:$src)>,
+  Requires<[HasSSE1]>;
+def : Pat<(store (v8i16 VR128:$src), addr:$dst), (MOVAPSmr addr:$dst, VR128:$src)>,
+  Requires<[HasSSE1]>;
+def : Pat<(store (v4i32 VR128:$src), addr:$dst), (MOVAPSmr addr:$dst, VR128:$src)>,
+  Requires<[HasSSE1]>;
+def : Pat<(store (v2i64 VR128:$src), addr:$dst), (MOVAPDmr addr:$dst, VR128:$src)>,
+  Requires<[HasSSE2]>;
 
 // Scalar to v8i16 / v16i8. The source may be a R32, but only the lower 8 or
 // 16-bits matter.