[AArch64] Add support for NEON scalar three register different instruction

class.  The instruction class includes the signed saturating doubling
multiply-add long, signed saturating doubling multiply-subtract long, and
the signed saturating doubling multiply long instructions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@192908 91177308-0d34-0410-b5e6-96231b3b80d8

4 files changed, 171 insertions, 0 deletions

diff --git a/test/CodeGen/AArch64/neon-scalar-mul.ll b/test/CodeGen/AArch64/neon-scalar-mul.ll
index 6914a08..3129df2 100644
--- a/test/CodeGen/AArch64/neon-scalar-mul.ll
+++ b/test/CodeGen/AArch64/neon-scalar-mul.ll
@@ -68,3 +68,78 @@ define double @test_vmulxd_f64(double %a, double %b) {
 
 declare <1 x float> @llvm.aarch64.neon.vmulx.v1f32(<1 x float>, <1 x float>)
 declare <1 x double> @llvm.aarch64.neon.vmulx.v1f64(<1 x double>, <1 x double>)
+
+define i32 @test_vqdmlalh_s16(i16 %a, i16 %b) {
+; CHECK: test_vqdmlalh_s16
+; CHECK: sqdmlal {{s[0-9]+}}, {{h[0-9]+}}, {{h[0-9]+}}
+entry:
+  %vqdmlal.i = insertelement <1 x i16> undef, i16 %a, i32 0
+  %vqdmlal1.i = insertelement <1 x i16> undef, i16 %b, i32 0
+  %vqdmlal2.i = call <1 x i32> @llvm.aarch64.neon.vqdmlal.v1i32(<1 x i16> %vqdmlal.i, <1 x i16> %vqdmlal1.i)
+  %0 = extractelement <1 x i32> %vqdmlal2.i, i32 0
+  ret i32 %0
+}
+
+define i64 @test_vqdmlals_s32(i32 %a, i32 %b) {
+; CHECK: test_vqdmlals_s32
+; CHECK: sqdmlal {{d[0-9]+}}, {{s[0-9]+}}, {{s[0-9]+}}
+entry:
+  %vqdmlal.i = insertelement <1 x i32> undef, i32 %a, i32 0
+  %vqdmlal1.i = insertelement <1 x i32> undef, i32 %b, i32 0
+  %vqdmlal2.i = call <1 x i64> @llvm.aarch64.neon.vqdmlal.v1i64(<1 x i32> %vqdmlal.i, <1 x i32> %vqdmlal1.i)
+  %0 = extractelement <1 x i64> %vqdmlal2.i, i32 0
+  ret i64 %0
+}
+
+declare <1 x i32> @llvm.aarch64.neon.vqdmlal.v1i32(<1 x i16>, <1 x i16>)
+declare <1 x i64> @llvm.aarch64.neon.vqdmlal.v1i64(<1 x i32>, <1 x i32>)
+
+define i32 @test_vqdmlslh_s16(i16 %a, i16 %b) {
+; CHECK: test_vqdmlslh_s16
+; CHECK: sqdmlsl {{s[0-9]+}}, {{h[0-9]+}}, {{h[0-9]+}}
+entry:
+  %vqdmlsl.i = insertelement <1 x i16> undef, i16 %a, i32 0
+  %vqdmlsl1.i = insertelement <1 x i16> undef, i16 %b, i32 0
+  %vqdmlsl2.i = call <1 x i32> @llvm.aarch64.neon.vqdmlsl.v1i32(<1 x i16> %vqdmlsl.i, <1 x i16> %vqdmlsl1.i)
+  %0 = extractelement <1 x i32> %vqdmlsl2.i, i32 0
+  ret i32 %0
+}
+
+define i64 @test_vqdmlsls_s32(i32 %a, i32 %b) {
+; CHECK: test_vqdmlsls_s32
+; CHECK: sqdmlsl {{d[0-9]+}}, {{s[0-9]+}}, {{s[0-9]+}}
+entry:
+  %vqdmlsl.i = insertelement <1 x i32> undef, i32 %a, i32 0
+  %vqdmlsl1.i = insertelement <1 x i32> undef, i32 %b, i32 0
+  %vqdmlsl2.i = call <1 x i64> @llvm.aarch64.neon.vqdmlsl.v1i64(<1 x i32> %vqdmlsl.i, <1 x i32> %vqdmlsl1.i)
+  %0 = extractelement <1 x i64> %vqdmlsl2.i, i32 0
+  ret i64 %0
+}
+
+declare <1 x i32> @llvm.aarch64.neon.vqdmlsl.v1i32(<1 x i16>, <1 x i16>)
+declare <1 x i64> @llvm.aarch64.neon.vqdmlsl.v1i64(<1 x i32>, <1 x i32>)
+
+define i32 @test_vqdmullh_s16(i16 %a, i16 %b) {
+; CHECK: test_vqdmullh_s16
+; CHECK: sqdmull {{s[0-9]+}}, {{h[0-9]+}}, {{h[0-9]+}}
+entry:
+  %vqdmull.i = insertelement <1 x i16> undef, i16 %a, i32 0
+  %vqdmull1.i = insertelement <1 x i16> undef, i16 %b, i32 0
+  %vqdmull2.i = call <1 x i32> @llvm.aarch64.neon.vqdmull.v1i32(<1 x i16> %vqdmull.i, <1 x i16> %vqdmull1.i)
+  %0 = extractelement <1 x i32> %vqdmull2.i, i32 0
+  ret i32 %0
+}
+
+define i64 @test_vqdmulls_s32(i32 %a, i32 %b) {
+; CHECK: test_vqdmulls_s32
+; CHECK: sqdmull {{d[0-9]+}}, {{s[0-9]+}}, {{s[0-9]+}}
+entry:
+  %vqdmull.i = insertelement <1 x i32> undef, i32 %a, i32 0
+  %vqdmull1.i = insertelement <1 x i32> undef, i32 %b, i32 0
+  %vqdmull2.i = call <1 x i64> @llvm.aarch64.neon.vqdmull.v1i64(<1 x i32> %vqdmull.i, <1 x i32> %vqdmull1.i)
+  %0 = extractelement <1 x i64> %vqdmull2.i, i32 0
+  ret i64 %0
+}
+
+declare <1 x i32> @llvm.aarch64.neon.vqdmull.v1i32(<1 x i16>, <1 x i16>)
+declare <1 x i64> @llvm.aarch64.neon.vqdmull.v1i64(<1 x i32>, <1 x i32>)
diff --git a/test/MC/AArch64/neon-diagnostics.s b/test/MC/AArch64/neon-diagnostics.s
index 411ea9f..44e4561 100644
--- a/test/MC/AArch64/neon-diagnostics.s
+++ b/test/MC/AArch64/neon-diagnostics.s
@@ -4460,3 +4460,45 @@
 // CHECK-ERROR: error: invalid operand for instruction
 // CHECK-ERROR:        neg d29, s24
 // CHECK-ERROR:                 ^
+
+//----------------------------------------------------------------------
+// Signed Saturating Doubling Multiply-Add Long
+//----------------------------------------------------------------------
+
+    sqdmlal s17, h27, s12
+    sqdmlal d19, s24, d12
+
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:        sqdmlal s17, h27, s12
+// CHECK-ERROR:                          ^
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:        sqdmlal d19, s24, d12
+// CHECK-ERROR:                          ^
+
+//----------------------------------------------------------------------
+// Signed Saturating Doubling Multiply-Subtract Long
+//----------------------------------------------------------------------
+
+    sqdmlsl s14, h12, s25
+    sqdmlsl d12, s23, d13
+
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:        sqdmlsl s14, h12, s25
+// CHECK-ERROR:                          ^
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:        sqdmlsl d12, s23, d13
+// CHECK-ERROR:                          ^
+
+//----------------------------------------------------------------------
+// Signed Saturating Doubling Multiply Long
+//----------------------------------------------------------------------
+
+    sqdmull s12, h22, s12
+    sqdmull d15, s22, d12
+
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:        sqdmull s12, h22, s12
+// CHECK-ERROR:                          ^
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:        sqdmull d15, s22, d12
+// CHECK-ERROR:                          ^
diff --git a/test/MC/AArch64/neon-scalar-mul.s b/test/MC/AArch64/neon-scalar-mul.s
index 8caddb4..e33bdad 100644
--- a/test/MC/AArch64/neon-scalar-mul.s
+++ b/test/MC/AArch64/neon-scalar-mul.s
@@ -31,3 +31,33 @@
 
 // CHECK: fmulx s20, s22, s15   // encoding: [0xd4,0xde,0x2f,0x5e]
 // CHECK: fmulx d23, d11, d1    // encoding: [0x77,0xdd,0x61,0x5e]
+
+//----------------------------------------------------------------------
+// Signed Saturating Doubling Multiply-Add Long
+//----------------------------------------------------------------------
+
+    sqdmlal s17, h27, h12
+    sqdmlal d19, s24, s12
+
+// CHECK: sqdmlal s17, h27, h12  // encoding: [0x71,0x93,0x6c,0x5e]
+// CHECK: sqdmlal d19, s24, s12  // encoding: [0x13,0x93,0xac,0x5e]
+
+//----------------------------------------------------------------------
+// Signed Saturating Doubling Multiply-Subtract Long
+//----------------------------------------------------------------------
+
+    sqdmlsl s14, h12, h25
+    sqdmlsl d12, s23, s13
+
+// CHECK: sqdmlsl s14, h12, h25  // encoding: [0x8e,0xb1,0x79,0x5e]
+// CHECK: sqdmlsl d12, s23, s13  // encoding: [0xec,0xb2,0xad,0x5e]
+
+//----------------------------------------------------------------------
+// Signed Saturating Doubling Multiply Long
+//----------------------------------------------------------------------
+
+    sqdmull s12, h22, h12
+    sqdmull d15, s22, s12
+
+// CHECK: sqdmull s12, h22, h12  // encoding: [0xcc,0xd2,0x6c,0x5e]
+// CHECK: sqdmull d15, s22, s12  // encoding: [0xcf,0xd2,0xac,0x5e]
diff --git a/test/MC/Disassembler/AArch64/neon-instructions.txt b/test/MC/Disassembler/AArch64/neon-instructions.txt
index c70a2f6..364259f 100644
--- a/test/MC/Disassembler/AArch64/neon-instructions.txt
+++ b/test/MC/Disassembler/AArch64/neon-instructions.txt
@@ -1659,3 +1659,27 @@
 0xf4,0x39,0x60,0x7e
 0x95,0x39,0xa0,0x7e
 0xd2,0x3a,0xe0,0x7e
+
+#----------------------------------------------------------------------
+# Signed Saturating Doubling Multiply-Add Long
+#----------------------------------------------------------------------
+# CHECK: sqdmlal s17, h27, h12
+# CHECK: sqdmlal d19, s24, s12
+0x71,0x93,0x6c,0x5e
+0x13,0x93,0xac,0x5e
+        
+#----------------------------------------------------------------------
+# Signed Saturating Doubling Multiply-Subtract Long
+#----------------------------------------------------------------------
+# CHECK: sqdmlsl s14, h12, h25
+# CHECK: sqdmlsl d12, s23, s13
+0x8e,0xb1,0x79,0x5e
+0xec,0xb2,0xad,0x5e
+        
+#----------------------------------------------------------------------
+# Signed Saturating Doubling Multiply Long
+#----------------------------------------------------------------------
+# CHECK: sqdmull s12, h22, h12
+# CHECK: sqdmull d15, s22, s12
+0xcc,0xd2,0x6c,0x5e
+0xcf,0xd2,0xac,0x5e