Merging r195327:

------------------------------------------------------------------------
r195327 | apazos | 2013-11-20 23:37:04 -0800 (Wed, 20 Nov 2013) | 6 lines

Implemented Neon scalar by element intrinsics.

Intrinsics implemented: vqdmull_lane, vqdmulh_lane, vqrdmulh_lane,
vqdmlal_lane, vqdmlsl_lane scalar Neon intrinsics.


------------------------------------------------------------------------


git-svn-id: https://llvm.org/svn/llvm-project/llvm/branches/release_34@195611 91177308-0d34-0410-b5e6-96231b3b80d8

2 files changed, 37 insertions, 70 deletions

diff --git a/test/CodeGen/AArch64/neon-scalar-mul.ll b/test/CodeGen/AArch64/neon-scalar-mul.ll
index 4992a51fe1..991037f6cb 100644
--- a/test/CodeGen/AArch64/neon-scalar-mul.ll
+++ b/test/CodeGen/AArch64/neon-scalar-mul.ll
@@ -123,7 +123,7 @@ define i32 @test_vqdmullh_s16(i16 %a, i16 %b) {
 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)
+  %vqdmull2.i = call <1 x i32> @llvm.arm.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
 }
@@ -134,10 +134,10 @@ define i64 @test_vqdmulls_s32(i32 %a, i32 %b) {
 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)
+  %vqdmull2.i = call <1 x i64> @llvm.arm.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>)
+declare <1 x i32> @llvm.arm.neon.vqdmull.v1i32(<1 x i16>, <1 x i16>)
+declare <1 x i64> @llvm.arm.neon.vqdmull.v1i64(<1 x i32>, <1 x i32>)
diff --git a/test/CodeGen/AArch64/neon-scalar-saturating-add-sub.ll b/test/CodeGen/AArch64/neon-scalar-saturating-add-sub.ll
index 5f03565258..bd66f80ceb 100644
--- a/test/CodeGen/AArch64/neon-scalar-saturating-add-sub.ll
+++ b/test/CodeGen/AArch64/neon-scalar-saturating-add-sub.ll
@@ -1,171 +1,138 @@
 ; RUN: llc -mtriple=aarch64-none-linux-gnu -mattr=+neon < %s | FileCheck %s
 
-declare <1 x i64> @llvm.arm.neon.vqaddu.v1i64(<1 x i64>, <1 x i64>)
-declare <1 x i64> @llvm.arm.neon.vqadds.v1i64(<1 x i64>, <1 x i64>)
-
-define <1 x i64> @test_uqadd_v1i64(<1 x i64> %lhs, <1 x i64> %rhs) {
-; CHECK: test_uqadd_v1i64:
-  %tmp1 = call <1 x i64> @llvm.arm.neon.vqaddu.v1i64(<1 x i64> %lhs, <1 x i64> %rhs)
-; CHECK: uqadd d0, d0, d1
-  ret <1 x i64> %tmp1
-}
-
-define <1 x i64> @test_sqadd_v1i64(<1 x i64> %lhs, <1 x i64> %rhs) {
-; CHECK: test_sqadd_v1i64:
-  %tmp1 = call <1 x i64> @llvm.arm.neon.vqadds.v1i64(<1 x i64> %lhs, <1 x i64> %rhs)
-; CHECK: sqadd d0, d0, d1
-  ret <1 x i64> %tmp1
-}
-
-declare <1 x i64> @llvm.arm.neon.vqsubu.v1i64(<1 x i64>, <1 x i64>)
-declare <1 x i64> @llvm.arm.neon.vqsubs.v1i64(<1 x i64>, <1 x i64>)
-
-define <1 x i64> @test_uqsub_v1i64(<1 x i64> %lhs, <1 x i64> %rhs) {
-; CHECK: test_uqsub_v1i64:
-  %tmp1 = call <1 x i64> @llvm.arm.neon.vqsubu.v1i64(<1 x i64> %lhs, <1 x i64> %rhs)
-; CHECK: uqsub d0, d0, d1
-  ret <1 x i64> %tmp1
-}
-
-define <1 x i64> @test_sqsub_v1i64(<1 x i64> %lhs, <1 x i64> %rhs) {
-; CHECK: test_sqsub_v1i64:
-  %tmp1 = call <1 x i64> @llvm.arm.neon.vqsubs.v1i64(<1 x i64> %lhs, <1 x i64> %rhs)
-; CHECK: sqsub d0, d0, d1
-  ret <1 x i64> %tmp1
-}
-
-declare <1 x i8> @llvm.aarch64.neon.vqaddu.v1i8(<1 x i8>, <1 x i8>)
-declare <1 x i8> @llvm.aarch64.neon.vqadds.v1i8(<1 x i8>, <1 x i8>)
+declare <1 x i8> @llvm.arm.neon.vqaddu.v1i8(<1 x i8>, <1 x i8>)
+declare <1 x i8> @llvm.arm.neon.vqadds.v1i8(<1 x i8>, <1 x i8>)
 
 define <1 x i8> @test_uqadd_v1i8_aarch64(<1 x i8> %lhs, <1 x i8> %rhs) {
 ; CHECK: test_uqadd_v1i8_aarch64:
-  %tmp1 = call <1 x i8> @llvm.aarch64.neon.vqaddu.v1i8(<1 x i8> %lhs, <1 x i8> %rhs)
+  %tmp1 = call <1 x i8> @llvm.arm.neon.vqaddu.v1i8(<1 x i8> %lhs, <1 x i8> %rhs)
 ;CHECK: uqadd {{b[0-31]+}}, {{b[0-31]+}}, {{b[0-31]+}}
   ret <1 x i8> %tmp1
 }
 
 define <1 x i8> @test_sqadd_v1i8_aarch64(<1 x i8> %lhs, <1 x i8> %rhs) {
 ; CHECK: test_sqadd_v1i8_aarch64:
-  %tmp1 = call <1 x i8> @llvm.aarch64.neon.vqadds.v1i8(<1 x i8> %lhs, <1 x i8> %rhs)
+  %tmp1 = call <1 x i8> @llvm.arm.neon.vqadds.v1i8(<1 x i8> %lhs, <1 x i8> %rhs)
 ;CHECK: sqadd {{b[0-31]+}}, {{b[0-31]+}}, {{b[0-31]+}}
   ret <1 x i8> %tmp1
 }
 
-declare <1 x i8> @llvm.aarch64.neon.vqsubu.v1i8(<1 x i8>, <1 x i8>)
-declare <1 x i8> @llvm.aarch64.neon.vqsubs.v1i8(<1 x i8>, <1 x i8>)
+declare <1 x i8> @llvm.arm.neon.vqsubu.v1i8(<1 x i8>, <1 x i8>)
+declare <1 x i8> @llvm.arm.neon.vqsubs.v1i8(<1 x i8>, <1 x i8>)
 
 define <1 x i8> @test_uqsub_v1i8_aarch64(<1 x i8> %lhs, <1 x i8> %rhs) {
 ; CHECK: test_uqsub_v1i8_aarch64:
-  %tmp1 = call <1 x i8> @llvm.aarch64.neon.vqsubu.v1i8(<1 x i8> %lhs, <1 x i8> %rhs)
+  %tmp1 = call <1 x i8> @llvm.arm.neon.vqsubu.v1i8(<1 x i8> %lhs, <1 x i8> %rhs)
 ;CHECK: uqsub {{b[0-31]+}}, {{b[0-31]+}}, {{b[0-31]+}}
   ret <1 x i8> %tmp1
 }
 
 define <1 x i8> @test_sqsub_v1i8_aarch64(<1 x i8> %lhs, <1 x i8> %rhs) {
 ; CHECK: test_sqsub_v1i8_aarch64:
-  %tmp1 = call <1 x i8> @llvm.aarch64.neon.vqsubs.v1i8(<1 x i8> %lhs, <1 x i8> %rhs)
+  %tmp1 = call <1 x i8> @llvm.arm.neon.vqsubs.v1i8(<1 x i8> %lhs, <1 x i8> %rhs)
 ;CHECK: sqsub {{b[0-31]+}}, {{b[0-31]+}}, {{b[0-31]+}}
   ret <1 x i8> %tmp1
 }
 
-declare <1 x i16> @llvm.aarch64.neon.vqaddu.v1i16(<1 x i16>, <1 x i16>)
-declare <1 x i16> @llvm.aarch64.neon.vqadds.v1i16(<1 x i16>, <1 x i16>)
+declare <1 x i16> @llvm.arm.neon.vqaddu.v1i16(<1 x i16>, <1 x i16>)
+declare <1 x i16> @llvm.arm.neon.vqadds.v1i16(<1 x i16>, <1 x i16>)
 
 define <1 x i16> @test_uqadd_v1i16_aarch64(<1 x i16> %lhs, <1 x i16> %rhs) {
 ; CHECK: test_uqadd_v1i16_aarch64:
-  %tmp1 = call <1 x i16> @llvm.aarch64.neon.vqaddu.v1i16(<1 x i16> %lhs, <1 x i16> %rhs)
+  %tmp1 = call <1 x i16> @llvm.arm.neon.vqaddu.v1i16(<1 x i16> %lhs, <1 x i16> %rhs)
 ;CHECK: uqadd {{h[0-31]+}}, {{h[0-31]+}}, {{h[0-31]+}}
   ret <1 x i16> %tmp1
 }
 
 define <1 x i16> @test_sqadd_v1i16_aarch64(<1 x i16> %lhs, <1 x i16> %rhs) {
 ; CHECK: test_sqadd_v1i16_aarch64:
-  %tmp1 = call <1 x i16> @llvm.aarch64.neon.vqadds.v1i16(<1 x i16> %lhs, <1 x i16> %rhs)
+  %tmp1 = call <1 x i16> @llvm.arm.neon.vqadds.v1i16(<1 x i16> %lhs, <1 x i16> %rhs)
 ;CHECK: sqadd {{h[0-31]+}}, {{h[0-31]+}}, {{h[0-31]+}}
   ret <1 x i16> %tmp1
 }
 
-declare <1 x i16> @llvm.aarch64.neon.vqsubu.v1i16(<1 x i16>, <1 x i16>)
-declare <1 x i16> @llvm.aarch64.neon.vqsubs.v1i16(<1 x i16>, <1 x i16>)
+declare <1 x i16> @llvm.arm.neon.vqsubu.v1i16(<1 x i16>, <1 x i16>)
+declare <1 x i16> @llvm.arm.neon.vqsubs.v1i16(<1 x i16>, <1 x i16>)
 
 define <1 x i16> @test_uqsub_v1i16_aarch64(<1 x i16> %lhs, <1 x i16> %rhs) {
 ; CHECK: test_uqsub_v1i16_aarch64:
-  %tmp1 = call <1 x i16> @llvm.aarch64.neon.vqsubu.v1i16(<1 x i16> %lhs, <1 x i16> %rhs)
+  %tmp1 = call <1 x i16> @llvm.arm.neon.vqsubu.v1i16(<1 x i16> %lhs, <1 x i16> %rhs)
 ;CHECK: uqsub {{h[0-31]+}}, {{h[0-31]+}}, {{h[0-31]+}}
   ret <1 x i16> %tmp1
 }
 
 define <1 x i16> @test_sqsub_v1i16_aarch64(<1 x i16> %lhs, <1 x i16> %rhs) {
 ; CHECK: test_sqsub_v1i16_aarch64:
-  %tmp1 = call <1 x i16> @llvm.aarch64.neon.vqsubs.v1i16(<1 x i16> %lhs, <1 x i16> %rhs)
+  %tmp1 = call <1 x i16> @llvm.arm.neon.vqsubs.v1i16(<1 x i16> %lhs, <1 x i16> %rhs)
 ;CHECK: sqsub {{h[0-31]+}}, {{h[0-31]+}}, {{h[0-31]+}}
   ret <1 x i16> %tmp1
 }
 
-declare <1 x i32> @llvm.aarch64.neon.vqaddu.v1i32(<1 x i32>, <1 x i32>)
-declare <1 x i32> @llvm.aarch64.neon.vqadds.v1i32(<1 x i32>, <1 x i32>)
+declare <1 x i32> @llvm.arm.neon.vqaddu.v1i32(<1 x i32>, <1 x i32>)
+declare <1 x i32> @llvm.arm.neon.vqadds.v1i32(<1 x i32>, <1 x i32>)
 
 define <1 x i32> @test_uqadd_v1i32_aarch64(<1 x i32> %lhs, <1 x i32> %rhs) {
 ; CHECK: test_uqadd_v1i32_aarch64:
-  %tmp1 = call <1 x i32> @llvm.aarch64.neon.vqaddu.v1i32(<1 x i32> %lhs, <1 x i32> %rhs)
+  %tmp1 = call <1 x i32> @llvm.arm.neon.vqaddu.v1i32(<1 x i32> %lhs, <1 x i32> %rhs)
 ;CHECK: uqadd {{s[0-31]+}}, {{s[0-31]+}}, {{s[0-31]+}}
   ret <1 x i32> %tmp1
 }
 
 define <1 x i32> @test_sqadd_v1i32_aarch64(<1 x i32> %lhs, <1 x i32> %rhs) {
 ; CHECK: test_sqadd_v1i32_aarch64:
-  %tmp1 = call <1 x i32> @llvm.aarch64.neon.vqadds.v1i32(<1 x i32> %lhs, <1 x i32> %rhs)
+  %tmp1 = call <1 x i32> @llvm.arm.neon.vqadds.v1i32(<1 x i32> %lhs, <1 x i32> %rhs)
 ;CHECK: sqadd {{s[0-31]+}}, {{s[0-31]+}}, {{s[0-31]+}}
   ret <1 x i32> %tmp1
 }
 
-declare <1 x i32> @llvm.aarch64.neon.vqsubu.v1i32(<1 x i32>, <1 x i32>)
-declare <1 x i32> @llvm.aarch64.neon.vqsubs.v1i32(<1 x i32>, <1 x i32>)
+declare <1 x i32> @llvm.arm.neon.vqsubu.v1i32(<1 x i32>, <1 x i32>)
+declare <1 x i32> @llvm.arm.neon.vqsubs.v1i32(<1 x i32>, <1 x i32>)
 
 define <1 x i32> @test_uqsub_v1i32_aarch64(<1 x i32> %lhs, <1 x i32> %rhs) {
 ; CHECK: test_uqsub_v1i32_aarch64:
-  %tmp1 = call <1 x i32> @llvm.aarch64.neon.vqsubu.v1i32(<1 x i32> %lhs, <1 x i32> %rhs)
+  %tmp1 = call <1 x i32> @llvm.arm.neon.vqsubu.v1i32(<1 x i32> %lhs, <1 x i32> %rhs)
 ;CHECK: uqsub {{s[0-31]+}}, {{s[0-31]+}}, {{s[0-31]+}}
   ret <1 x i32> %tmp1
 }
 
+
 define <1 x i32> @test_sqsub_v1i32_aarch64(<1 x i32> %lhs, <1 x i32> %rhs) {
 ; CHECK: test_sqsub_v1i32_aarch64:
-  %tmp1 = call <1 x i32> @llvm.aarch64.neon.vqsubs.v1i32(<1 x i32> %lhs, <1 x i32> %rhs)
+  %tmp1 = call <1 x i32> @llvm.arm.neon.vqsubs.v1i32(<1 x i32> %lhs, <1 x i32> %rhs)
 ;CHECK: sqsub {{s[0-31]+}}, {{s[0-31]+}}, {{s[0-31]+}}
   ret <1 x i32> %tmp1
 }
 
-declare <1 x i64> @llvm.aarch64.neon.vqaddu.v1i64(<1 x i64>, <1 x i64>)
-declare <1 x i64> @llvm.aarch64.neon.vqadds.v1i64(<1 x i64>, <1 x i64>)
+declare <1 x i64> @llvm.arm.neon.vqaddu.v1i64(<1 x i64>, <1 x i64>)
+declare <1 x i64> @llvm.arm.neon.vqadds.v1i64(<1 x i64>, <1 x i64>)
 
 define <1 x i64> @test_uqadd_v1i64_aarch64(<1 x i64> %lhs, <1 x i64> %rhs) {
 ; CHECK: test_uqadd_v1i64_aarch64:
-  %tmp1 = call <1 x i64> @llvm.aarch64.neon.vqaddu.v1i64(<1 x i64> %lhs, <1 x i64> %rhs)
+  %tmp1 = call <1 x i64> @llvm.arm.neon.vqaddu.v1i64(<1 x i64> %lhs, <1 x i64> %rhs)
 ;CHECK: uqadd {{d[0-31]+}}, {{d[0-31]+}}, {{d[0-31]+}}
   ret <1 x i64> %tmp1
 }
 
 define <1 x i64> @test_sqadd_v1i64_aarch64(<1 x i64> %lhs, <1 x i64> %rhs) {
 ; CHECK: test_sqadd_v1i64_aarch64:
-  %tmp1 = call <1 x i64> @llvm.aarch64.neon.vqadds.v1i64(<1 x i64> %lhs, <1 x i64> %rhs)
+  %tmp1 = call <1 x i64> @llvm.arm.neon.vqadds.v1i64(<1 x i64> %lhs, <1 x i64> %rhs)
 ;CHECK: sqadd {{d[0-31]+}}, {{d[0-31]+}}, {{d[0-31]+}}
   ret <1 x i64> %tmp1
 }
 
-declare <1 x i64> @llvm.aarch64.neon.vqsubu.v1i64(<1 x i64>, <1 x i64>)
-declare <1 x i64> @llvm.aarch64.neon.vqsubs.v1i64(<1 x i64>, <1 x i64>)
+declare <1 x i64> @llvm.arm.neon.vqsubu.v1i64(<1 x i64>, <1 x i64>)
+declare <1 x i64> @llvm.arm.neon.vqsubs.v1i64(<1 x i64>, <1 x i64>)
 
 define <1 x i64> @test_uqsub_v1i64_aarch64(<1 x i64> %lhs, <1 x i64> %rhs) {
 ; CHECK: test_uqsub_v1i64_aarch64:
-  %tmp1 = call <1 x i64> @llvm.aarch64.neon.vqsubu.v1i64(<1 x i64> %lhs, <1 x i64> %rhs)
+  %tmp1 = call <1 x i64> @llvm.arm.neon.vqsubu.v1i64(<1 x i64> %lhs, <1 x i64> %rhs)
 ;CHECK: uqsub {{d[0-31]+}}, {{d[0-31]+}}, {{d[0-31]+}}
   ret <1 x i64> %tmp1
 }
 
 define <1 x i64> @test_sqsub_v1i64_aarch64(<1 x i64> %lhs, <1 x i64> %rhs) {
 ; CHECK: test_sqsub_v1i64_aarch64:
-  %tmp1 = call <1 x i64> @llvm.aarch64.neon.vqsubs.v1i64(<1 x i64> %lhs, <1 x i64> %rhs)
+  %tmp1 = call <1 x i64> @llvm.arm.neon.vqsubs.v1i64(<1 x i64> %lhs, <1 x i64> %rhs)
 ;CHECK: sqsub {{d[0-31]+}}, {{d[0-31]+}}, {{d[0-31]+}}
   ret <1 x i64> %tmp1
 }