[AArch64] Add support for NEON scalar integer compare instructions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@192596 91177308-0d34-0410-b5e6-96231b3b80d8

4 files changed, 395 insertions, 0 deletions

diff --git a/test/CodeGen/AArch64/neon-scalar-compare.ll b/test/CodeGen/AArch64/neon-scalar-compare.ll
new file mode 100644
index 0000000..831c10b
--- /dev/null
+++ b/test/CodeGen/AArch64/neon-scalar-compare.ll
@@ -0,0 +1,128 @@
+; RUN: llc -mtriple=aarch64-none-linux-gnu -mattr=+neon < %s | FileCheck %s
+
+;; Scalar Integer Compare
+
+define i64 @test_vceqd(i64 %a, i64 %b) {
+; CHECK: test_vceqd
+; CHECK: cmeq {{d[0-9]+}}, {{d[0-9]}}, {{d[0-9]}}
+entry:
+  %vceq.i = insertelement <1 x i64> undef, i64 %a, i32 0
+  %vceq1.i = insertelement <1 x i64> undef, i64 %b, i32 0
+  %vceq2.i = call <1 x i64> @llvm.aarch64.neon.vceq(<1 x i64> %vceq.i, <1 x i64> %vceq1.i)
+  %0 = extractelement <1 x i64> %vceq2.i, i32 0
+  ret i64 %0
+}
+
+define i64 @test_vceqzd(i64 %a) {
+; CHECK: test_vceqzd
+; CHECK: cmeq {{d[0-9]}}, {{d[0-9]}}, #0x0
+entry:
+  %vceqz.i = insertelement <1 x i64> undef, i64 %a, i32 0
+  %vceqz1.i = call <1 x i64> @llvm.aarch64.neon.vceq(<1 x i64> %vceqz.i, <1 x i64> zeroinitializer)
+  %0 = extractelement <1 x i64> %vceqz1.i, i32 0
+  ret i64 %0
+}
+
+define i64 @test_vcged(i64 %a, i64 %b) {
+; CHECK: test_vcged
+; CHECK: cmge {{d[0-9]}}, {{d[0-9]}}, {{d[0-9]}}
+entry:
+  %vcge.i = insertelement <1 x i64> undef, i64 %a, i32 0
+  %vcge1.i = insertelement <1 x i64> undef, i64 %b, i32 0
+  %vcge2.i = call <1 x i64> @llvm.aarch64.neon.vcge(<1 x i64> %vcge.i, <1 x i64> %vcge1.i)
+  %0 = extractelement <1 x i64> %vcge2.i, i32 0
+  ret i64 %0
+}
+
+define i64 @test_vcgezd(i64 %a) {
+; CHECK: test_vcgezd
+; CHECK: cmge {{d[0-9]}}, {{d[0-9]}}, #0x0
+entry:
+  %vcgez.i = insertelement <1 x i64> undef, i64 %a, i32 0
+  %vcgez1.i = call <1 x i64> @llvm.aarch64.neon.vcge(<1 x i64> %vcgez.i, <1 x i64> zeroinitializer)
+  %0 = extractelement <1 x i64> %vcgez1.i, i32 0
+  ret i64 %0
+}
+
+define i64 @test_vcgtd(i64 %a, i64 %b) {
+; CHECK: test_vcgtd
+; CHECK: cmgt {{d[0-9]}}, {{d[0-9]}}, {{d[0-9]}}
+entry:
+  %vcgt.i = insertelement <1 x i64> undef, i64 %a, i32 0
+  %vcgt1.i = insertelement <1 x i64> undef, i64 %b, i32 0
+  %vcgt2.i = call <1 x i64> @llvm.aarch64.neon.vcgt(<1 x i64> %vcgt.i, <1 x i64> %vcgt1.i)
+  %0 = extractelement <1 x i64> %vcgt2.i, i32 0
+  ret i64 %0
+}
+
+define i64 @test_vcgtzd(i64 %a) {
+; CHECK: test_vcgtzd
+; CHECK: cmgt {{d[0-9]}}, {{d[0-9]}}, #0x0
+entry:
+  %vcgtz.i = insertelement <1 x i64> undef, i64 %a, i32 0
+  %vcgtz1.i = call <1 x i64> @llvm.aarch64.neon.vcgt(<1 x i64> %vcgtz.i, <1 x i64> zeroinitializer)
+  %0 = extractelement <1 x i64> %vcgtz1.i, i32 0
+  ret i64 %0
+}
+
+define i64 @test_vcled(i64 %a, i64 %b) {
+; CHECK: test_vcled
+; CHECK: cmgt {{d[0-9]}}, {{d[0-9]}}, {{d[0-9]}}
+entry:
+  %vcgt.i = insertelement <1 x i64> undef, i64 %b, i32 0
+  %vcgt1.i = insertelement <1 x i64> undef, i64 %a, i32 0
+  %vcgt2.i = call <1 x i64> @llvm.aarch64.neon.vcgt(<1 x i64> %vcgt.i, <1 x i64> %vcgt1.i)
+  %0 = extractelement <1 x i64> %vcgt2.i, i32 0
+  ret i64 %0
+}
+
+define i64 @test_vclezd(i64 %a) {
+; CHECK: test_vclezd
+; CHECK: cmle {{d[0-9]}}, {{d[0-9]}}, #0x0
+entry:
+  %vclez.i = insertelement <1 x i64> undef, i64 %a, i32 0
+  %vclez1.i = call <1 x i64> @llvm.aarch64.neon.vclez(<1 x i64> %vclez.i, <1 x i64> zeroinitializer)
+  %0 = extractelement <1 x i64> %vclez1.i, i32 0
+  ret i64 %0
+}
+
+define i64 @test_vcltd(i64 %a, i64 %b) {
+; CHECK: test_vcltd
+; CHECK: cmge {{d[0-9]}}, {{d[0-9]}}, {{d[0-9]}}
+entry:
+  %vcge.i = insertelement <1 x i64> undef, i64 %b, i32 0
+  %vcge1.i = insertelement <1 x i64> undef, i64 %a, i32 0
+  %vcge2.i = call <1 x i64> @llvm.aarch64.neon.vcge(<1 x i64> %vcge.i, <1 x i64> %vcge1.i)
+  %0 = extractelement <1 x i64> %vcge2.i, i32 0
+  ret i64 %0
+}
+
+define i64 @test_vcltzd(i64 %a) {
+; CHECK: test_vcltzd
+; CHECK: cmlt {{d[0-9]}}, {{d[0-9]}}, #0x0
+entry:
+  %vcltz.i = insertelement <1 x i64> undef, i64 %a, i32 0
+  %vcltz1.i = call <1 x i64> @llvm.aarch64.neon.vcltz(<1 x i64> %vcltz.i, <1 x i64> zeroinitializer)
+  %0 = extractelement <1 x i64> %vcltz1.i, i32 0
+  ret i64 %0
+}
+
+define i64 @test_vtstd(i64 %a, i64 %b) {
+; CHECK: test_vtstd
+; CHECK: cmtst {{d[0-9]}}, {{d[0-9]}}, {{d[0-9]}}
+entry:
+  %vtst.i = insertelement <1 x i64> undef, i64 %a, i32 0
+  %vtst1.i = insertelement <1 x i64> undef, i64 %b, i32 0
+  %vtst2.i = call <1 x i64> @llvm.aarch64.neon.vtstd(<1 x i64> %vtst.i, <1 x i64> %vtst1.i)
+  %0 = extractelement <1 x i64> %vtst2.i, i32 0
+  ret i64 %0
+}
+
+declare <1 x i64> @llvm.aarch64.neon.vtstd(<1 x i64>, <1 x i64>)
+declare <1 x i64> @llvm.aarch64.neon.vcltz(<1 x i64>, <1 x i64>)
+declare <1 x i64> @llvm.aarch64.neon.vchs(<1 x i64>, <1 x i64>)
+declare <1 x i64> @llvm.aarch64.neon.vcge(<1 x i64>, <1 x i64>)
+declare <1 x i64> @llvm.aarch64.neon.vclez(<1 x i64>, <1 x i64>)
+declare <1 x i64> @llvm.aarch64.neon.vchi(<1 x i64>, <1 x i64>)
+declare <1 x i64> @llvm.aarch64.neon.vcgt(<1 x i64>, <1 x i64>)
+declare <1 x i64> @llvm.aarch64.neon.vceq(<1 x i64>, <1 x i64>)
diff --git a/test/MC/AArch64/neon-diagnostics.s b/test/MC/AArch64/neon-diagnostics.s
index 2af7798..e08c5cd 100644
--- a/test/MC/AArch64/neon-diagnostics.s
+++ b/test/MC/AArch64/neon-diagnostics.s
@@ -4286,3 +4286,113 @@
 // CHECK-ERROR         dup v5.2d, w0
 // CHECK-ERROR                    ^
 
+//----------------------------------------------------------------------
+// Scalar Compare Bitwise Equal
+//----------------------------------------------------------------------
+
+         cmeq b20, d21, d22
+
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:          cmeq b20, d21, d22
+// CHECK-ERROR:               ^
+
+//----------------------------------------------------------------------
+// Scalar Compare Bitwise Equal To Zero
+//----------------------------------------------------------------------
+
+         cmeq d20, b21, #0
+
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:          cmeq d20, b21, #0
+// CHECK-ERROR:                    ^
+
+//----------------------------------------------------------------------
+// Scalar Compare Unsigned Higher Or Same
+//----------------------------------------------------------------------
+
+         cmhs b20, d21, d22
+
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:          cmhs b20, d21, d22
+// CHECK-ERROR:               ^
+
+        
+//----------------------------------------------------------------------
+// Scalar Compare Signed Greather Than Or Equal
+//----------------------------------------------------------------------
+
+         cmge b20, d21, d22
+
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:          cmge b20, d21, d22
+// CHECK-ERROR:               ^
+
+//----------------------------------------------------------------------
+// Scalar Compare Signed Greather Than Or Equal To Zero
+//----------------------------------------------------------------------
+
+         cmge d20, b21, #0
+
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:          cmge d20, b21, #0
+// CHECK-ERROR:                    ^
+
+//----------------------------------------------------------------------
+// Scalar Compare Unsigned Higher
+//----------------------------------------------------------------------
+
+         cmhi b20, d21, d22
+
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:          cmhi b20, d21, d22
+// CHECK-ERROR:               ^
+
+//----------------------------------------------------------------------
+// Scalar Compare Signed Greater Than
+//----------------------------------------------------------------------
+
+         cmgt b20, d21, d22
+
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:          cmgt b20, d21, d22
+// CHECK-ERROR:               ^
+
+//----------------------------------------------------------------------
+// Scalar Compare Signed Greater Than Zero
+//----------------------------------------------------------------------
+
+         cmgt d20, b21, #0
+
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:          cmgt d20, b21, #0
+// CHECK-ERROR:                    ^
+
+//----------------------------------------------------------------------
+// Scalar Compare Signed Less Than Or Equal To Zero
+//----------------------------------------------------------------------
+
+         cmle d20, b21, #0
+
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:          cmle d20, b21, #0
+// CHECK-ERROR:                    ^
+
+//----------------------------------------------------------------------
+// Scalar Compare Less Than Zero
+//----------------------------------------------------------------------
+
+         cmlt d20, b21, #0
+
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:          cmlt d20, b21, #0
+// CHECK-ERROR:                    ^
+
+//----------------------------------------------------------------------
+// Scalar Compare Bitwise Test Bits
+//----------------------------------------------------------------------
+
+         cmtst b20, d21, d22
+
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:          cmtst b20, d21, d22
+// CHECK-ERROR:                ^
diff --git a/test/MC/AArch64/neon-scalar-compare.s b/test/MC/AArch64/neon-scalar-compare.s
new file mode 100644
index 0000000..55ade0e
--- /dev/null
+++ b/test/MC/AArch64/neon-scalar-compare.s
@@ -0,0 +1,90 @@
+// RUN: llvm-mc -triple aarch64-none-linux-gnu -mattr=+neon -show-encoding < %s | FileCheck %s
+
+// Check that the assembler can handle the documented syntax for AArch64
+
+//----------------------------------------------------------------------
+// Scalar Compare Bitwise Equal
+//----------------------------------------------------------------------
+
+         cmeq d20, d21, d22
+
+// CHECK: cmeq d20, d21, d22   // encoding: [0xb4,0x8e,0xf6,0x7e]
+
+//----------------------------------------------------------------------
+// Scalar Compare Bitwise Equal To Zero
+//----------------------------------------------------------------------
+
+         cmeq d20, d21, #0x0
+
+// CHECK: cmeq d20, d21, #0x0   // encoding: [0xb4,0x9a,0xe0,0x5e]
+
+//----------------------------------------------------------------------
+// Scalar Compare Unsigned Higher Or Same
+//----------------------------------------------------------------------
+
+         cmhs d20, d21, d22
+
+// CHECK: cmhs d20, d21, d22   // encoding: [0xb4,0x3e,0xf6,0x7e]
+
+//----------------------------------------------------------------------
+// Scalar Compare Signed Greather Than Or Equal
+//----------------------------------------------------------------------
+
+         cmge d20, d21, d22
+
+// CHECK: cmge d20, d21, d22    // encoding: [0xb4,0x3e,0xf6,0x5e]
+
+//----------------------------------------------------------------------
+// Scalar Compare Signed Greather Than Or Equal To Zero
+//----------------------------------------------------------------------
+
+         cmge d20, d21, #0x0
+
+// CHECK: cmge d20, d21, #0x0   // encoding: [0xb4,0x8a,0xe0,0x7e]
+
+//----------------------------------------------------------------------
+// Scalar Compare Unsigned Higher
+//----------------------------------------------------------------------
+
+         cmhi d20, d21, d22
+
+// CHECK: cmhi d20, d21, d22   // encoding: [0xb4,0x36,0xf6,0x7e]
+//----------------------------------------------------------------------
+// Scalar Compare Signed Greater Than
+//----------------------------------------------------------------------
+
+         cmgt d20, d21, d22
+
+// CHECK: cmgt d20, d21, d22   // encoding: [0xb4,0x36,0xf6,0x5e]
+
+//----------------------------------------------------------------------
+// Scalar Compare Signed Greater Than Zero
+//----------------------------------------------------------------------
+
+         cmgt d20, d21, #0x0
+
+// CHECK: cmgt d20, d21, #0x0   // encoding: [0xb4,0x8a,0xe0,0x5e]
+
+//----------------------------------------------------------------------
+// Scalar Compare Signed Less Than Or Equal To Zero
+//----------------------------------------------------------------------
+
+         cmle d20, d21, #0x0
+
+// CHECK: cmle d20, d21, #0x0   // encoding: [0xb4,0x9a,0xe0,0x7e]
+
+//----------------------------------------------------------------------
+// Scalar Compare Less Than Zero
+//----------------------------------------------------------------------
+
+         cmlt d20, d21, #0x0
+
+// CHECK: cmlt d20, d21, #0x0   // encoding: [0xb4,0xaa,0xe0,0x5e]
+
+//----------------------------------------------------------------------
+// Scalar Compare Bitwise Test Bits
+//----------------------------------------------------------------------
+
+         cmtst d20, d21, d22
+
+// CHECK: cmtst d20, d21, d22   // encoding: [0xb4,0x8e,0xf6,0x5e]
diff --git a/test/MC/Disassembler/AArch64/neon-instructions.txt b/test/MC/Disassembler/AArch64/neon-instructions.txt
index b05d3cb..e6c45c0 100644
--- a/test/MC/Disassembler/AArch64/neon-instructions.txt
+++ b/test/MC/Disassembler/AArch64/neon-instructions.txt
@@ -1532,3 +1532,70 @@
 # CHECK: frsqrte d21, d12
 0xb6,0xd9,0xa1,0x7e
 0x95,0xd9,0xe1,0x7e
+
+#----------------------------------------------------------------------
+# Scalar Compare Bitwise Equal
+#----------------------------------------------------------------------
+# CHECK: cmeq d20, d21, d22
+0xb4,0x8e,0xf6,0x7e
+
+#----------------------------------------------------------------------
+# Scalar Compare Bitwise Equal To Zero
+#----------------------------------------------------------------------
+# CHECK: cmeq d20, d21, #0x0
+0xb4,0x9a,0xe0,0x5e
+
+#----------------------------------------------------------------------
+# Scalar Compare Unsigned Higher Or Same
+#----------------------------------------------------------------------
+# CHECK: cmhs d20, d21, d22
+0xb4,0x3e,0xf6,0x7e
+
+        
+#----------------------------------------------------------------------
+# Scalar Compare Signed Greather Than Or Equal
+#----------------------------------------------------------------------
+# CHECK: cmge d20, d21, d22
+0xb4,0x3e,0xf6,0x5e
+
+#----------------------------------------------------------------------
+# Scalar Compare Signed Greather Than Or Equal To Zero
+#----------------------------------------------------------------------
+# CHECK: cmge d20, d21, #0x0
+0xb4,0x8a,0xe0,0x7e
+
+#----------------------------------------------------------------------
+# Scalar Compare Unsigned Higher
+#----------------------------------------------------------------------
+# CHECK: cmhi d20, d21, d22
+0xb4,0x36,0xf6,0x7e
+
+#----------------------------------------------------------------------
+# Scalar Compare Signed Greater Than
+#----------------------------------------------------------------------
+# CHECK: cmgt d20, d21, d22
+0xb4,0x36,0xf6,0x5e
+
+#----------------------------------------------------------------------
+# Scalar Compare Signed Greater Than Zero
+#----------------------------------------------------------------------
+# CHECK: cmgt d20, d21, #0x0
+0xb4,0x8a,0xe0,0x5e
+
+#----------------------------------------------------------------------
+# Scalar Compare Signed Less Than Or Equal To Zero
+#----------------------------------------------------------------------
+# CHECK: cmle d20, d21, #0x0
+0xb4,0x9a,0xe0,0x7e
+
+#----------------------------------------------------------------------
+# Scalar Compare Less Than Zero
+#----------------------------------------------------------------------
+# CHECK: cmlt d20, d21, #0x0
+0xb4,0xaa,0xe0,0x5e
+
+#----------------------------------------------------------------------
+# Scalar Compare Bitwise Test Bits
+#----------------------------------------------------------------------
+# CHECK: cmtst d20, d21, d22
+0xb4,0x8e,0xf6,0x5e