AArch64 & ARM: refactor crypto intrinsics to take scalars

Some of the SHA instructions take a scalar i32 as one argument (largely because
they work on 160-bit hash fragments). This wasn't reflected in the IR
previously, with ARM and AArch64 choosing different types (<4 x i32> and <1 x
i32> respectively) which was ugly.

This makes all the affected intrinsics take a uniform "i32", allowing them to
become non-polymorphic at the same time.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200706 91177308-0d34-0410-b5e6-96231b3b80d8

6 files changed, 145 insertions, 101 deletions

diff --git a/include/llvm/IR/IntrinsicsAArch64.td b/include/llvm/IR/IntrinsicsAArch64.td
index 5c289ce250..bb84b142ed 100644
--- a/include/llvm/IR/IntrinsicsAArch64.td
+++ b/include/llvm/IR/IntrinsicsAArch64.td
@@ -404,11 +404,4 @@ def int_aarch64_neon_vcvtfp2fxs_n :
 def int_aarch64_neon_vcvtfp2fxu_n :
   Intrinsic<[llvm_anyvector_ty], [llvm_anyfloat_ty, llvm_i32_ty], [IntrNoMem]>;
 
-class Neon_SHA_Intrinsic
-  : Intrinsic<[llvm_v4i32_ty], [llvm_v4i32_ty, llvm_v1i32_ty, llvm_v4i32_ty],
-              [IntrNoMem]>;
-
-def int_aarch64_neon_sha1c : Neon_SHA_Intrinsic;
-def int_aarch64_neon_sha1m : Neon_SHA_Intrinsic;
-def int_aarch64_neon_sha1p : Neon_SHA_Intrinsic;
 }
diff --git a/include/llvm/IR/IntrinsicsARM.td b/include/llvm/IR/IntrinsicsARM.td
index 0b50d64aeb..ff95421dfe 100644
--- a/include/llvm/IR/IntrinsicsARM.td
+++ b/include/llvm/IR/IntrinsicsARM.td
@@ -472,19 +472,37 @@ def int_arm_neon_vbsl : Intrinsic<[llvm_anyvector_ty],
 
 
 // Crypto instructions
-def int_arm_neon_aesd : Neon_2Arg_Intrinsic;
-def int_arm_neon_aese : Neon_2Arg_Intrinsic;
-def int_arm_neon_aesimc : Neon_1Arg_Intrinsic;
-def int_arm_neon_aesmc : Neon_1Arg_Intrinsic;
-def int_arm_neon_sha1h : Neon_1Arg_Intrinsic;
-def int_arm_neon_sha1su1 : Neon_2Arg_Intrinsic;
-def int_arm_neon_sha256su0 : Neon_2Arg_Intrinsic;
-def int_arm_neon_sha1c : Neon_3Arg_Intrinsic;
-def int_arm_neon_sha1m : Neon_3Arg_Intrinsic;
-def int_arm_neon_sha1p : Neon_3Arg_Intrinsic;
-def int_arm_neon_sha1su0: Neon_3Arg_Intrinsic;
-def int_arm_neon_sha256h: Neon_3Arg_Intrinsic;
-def int_arm_neon_sha256h2: Neon_3Arg_Intrinsic;
-def int_arm_neon_sha256su1: Neon_3Arg_Intrinsic;
+class AES_1Arg_Intrinsic : Intrinsic<[llvm_v16i8_ty],
+                                     [llvm_v16i8_ty], [IntrNoMem]>;
+class AES_2Arg_Intrinsic : Intrinsic<[llvm_v16i8_ty],
+                                     [llvm_v16i8_ty, llvm_v16i8_ty],
+                                     [IntrNoMem]>;
+
+class SHA_1Arg_Intrinsic : Intrinsic<[llvm_i32_ty], [llvm_i32_ty],
+                                     [IntrNoMem]>;
+class SHA_2Arg_Intrinsic : Intrinsic<[llvm_v4i32_ty],
+                                     [llvm_v4i32_ty, llvm_v4i32_ty],
+                                     [IntrNoMem]>;
+class SHA_3Arg_i32_Intrinsic : Intrinsic<[llvm_v4i32_ty],
+                                   [llvm_v4i32_ty, llvm_i32_ty, llvm_v4i32_ty],
+                                   [IntrNoMem]>;
+class SHA_3Arg_v4i32_Intrinsic : Intrinsic<[llvm_v4i32_ty],
+                                   [llvm_v4i32_ty, llvm_v4i32_ty,llvm_v4i32_ty],
+                                   [IntrNoMem]>;
+
+def int_arm_neon_aesd : AES_2Arg_Intrinsic;
+def int_arm_neon_aese : AES_2Arg_Intrinsic;
+def int_arm_neon_aesimc : AES_1Arg_Intrinsic;
+def int_arm_neon_aesmc : AES_1Arg_Intrinsic;
+def int_arm_neon_sha1h : SHA_1Arg_Intrinsic;
+def int_arm_neon_sha1su1 : SHA_2Arg_Intrinsic;
+def int_arm_neon_sha256su0 : SHA_2Arg_Intrinsic;
+def int_arm_neon_sha1c : SHA_3Arg_i32_Intrinsic;
+def int_arm_neon_sha1m : SHA_3Arg_i32_Intrinsic;
+def int_arm_neon_sha1p : SHA_3Arg_i32_Intrinsic;
+def int_arm_neon_sha1su0: SHA_3Arg_v4i32_Intrinsic;
+def int_arm_neon_sha256h: SHA_3Arg_v4i32_Intrinsic;
+def int_arm_neon_sha256h2: SHA_3Arg_v4i32_Intrinsic;
+def int_arm_neon_sha256su1: SHA_3Arg_v4i32_Intrinsic;
 
 } // end TargetPrefix
diff --git a/lib/Target/AArch64/AArch64InstrNEON.td b/lib/Target/AArch64/AArch64InstrNEON.td
index 2cf27b861b..edb619e1e7 100644
--- a/lib/Target/AArch64/AArch64InstrNEON.td
+++ b/lib/Target/AArch64/AArch64InstrNEON.td
@@ -8863,13 +8863,15 @@ class NeonI_Cryptosha_ss<bits<2> size, bits<5> opcode,
   : NeonI_Crypto_SHA<size, opcode,
                      (outs FPR32:$Rd), (ins FPR32:$Rn),
                      asmop # "\t$Rd, $Rn",
-                     [(set (v1i32 FPR32:$Rd),
-                        (v1i32 (opnode (v1i32 FPR32:$Rn))))],
-                     NoItinerary> {
+                     [], NoItinerary> {
   let Predicates = [HasNEON, HasCrypto];
+  let hasSideEffects = 0;
 }
 
 def SHA1H : NeonI_Cryptosha_ss<0b00, 0b00000, "sha1h", int_arm_neon_sha1h>;
+def : Pat<(i32 (int_arm_neon_sha1h i32:$Rn)),
+          (COPY_TO_REGCLASS (SHA1H (COPY_TO_REGCLASS i32:$Rn, FPR32)), GPR32)>;
+
 
 class NeonI_Cryptosha3_vvv<bits<2> size, bits<3> opcode, string asmop,
                            SDPatternOperator opnode>
@@ -8911,24 +8913,30 @@ def SHA256H : NeonI_Cryptosha3_qqv<0b00, 0b100, "sha256h",
 def SHA256H2 : NeonI_Cryptosha3_qqv<0b00, 0b101, "sha256h2",
                                     int_arm_neon_sha256h2>;
 
-class NeonI_Cryptosha3_qsv<bits<2> size, bits<3> opcode, string asmop,
-                           SDPatternOperator opnode>
+class NeonI_Cryptosha3_qsv<bits<2> size, bits<3> opcode, string asmop>
   : NeonI_Crypto_3VSHA<size, opcode,
                        (outs FPR128:$Rd),
                        (ins FPR128:$src, FPR32:$Rn, VPR128:$Rm),
                        asmop # "\t$Rd, $Rn, $Rm.4s",
-                       [(set (v4i32 FPR128:$Rd),
-                          (v4i32 (opnode (v4i32 FPR128:$src),
-                                         (v1i32 FPR32:$Rn),
-                                         (v4i32 VPR128:$Rm))))],
-                       NoItinerary> {
+                       [], NoItinerary> {
   let Constraints = "$src = $Rd";
+  let hasSideEffects = 0;
   let Predicates = [HasNEON, HasCrypto];
 }
 
-def SHA1C : NeonI_Cryptosha3_qsv<0b00, 0b000, "sha1c", int_aarch64_neon_sha1c>;
-def SHA1P : NeonI_Cryptosha3_qsv<0b00, 0b001, "sha1p", int_aarch64_neon_sha1p>;
-def SHA1M : NeonI_Cryptosha3_qsv<0b00, 0b010, "sha1m", int_aarch64_neon_sha1m>;
+def SHA1C : NeonI_Cryptosha3_qsv<0b00, 0b000, "sha1c">;
+def SHA1P : NeonI_Cryptosha3_qsv<0b00, 0b001, "sha1p">;
+def SHA1M : NeonI_Cryptosha3_qsv<0b00, 0b010, "sha1m">;
+
+def : Pat<(int_arm_neon_sha1c v4i32:$hash_abcd, i32:$hash_e, v4i32:$wk),
+          (SHA1C v4i32:$hash_abcd,
+                 (COPY_TO_REGCLASS i32:$hash_e, FPR32), v4i32:$wk)>;
+def : Pat<(int_arm_neon_sha1m v4i32:$hash_abcd, i32:$hash_e, v4i32:$wk),
+          (SHA1M v4i32:$hash_abcd,
+                 (COPY_TO_REGCLASS i32:$hash_e, FPR32), v4i32:$wk)>;
+def : Pat<(int_arm_neon_sha1p v4i32:$hash_abcd, i32:$hash_e, v4i32:$wk),
+          (SHA1P v4i32:$hash_abcd,
+                 (COPY_TO_REGCLASS i32:$hash_e, FPR32), v4i32:$wk)>;
 
 // Additional patterns to match shl to USHL.
 def : Pat<(v8i8 (shl (v8i8 VPR64:$Rn), (v8i8 VPR64:$Rm))),
diff --git a/lib/Target/ARM/ARMInstrNEON.td b/lib/Target/ARM/ARMInstrNEON.td
index 3d993eb964..bbf8590935 100644
--- a/lib/Target/ARM/ARMInstrNEON.td
+++ b/lib/Target/ARM/ARMInstrNEON.td
@@ -5876,7 +5876,7 @@ defm VRINTPN : VRINT_FPI<"p", 0b111, int_arm_neon_vrintp>;
 
 // Cryptography instructions
 let PostEncoderMethod = "NEONThumb2DataIPostEncoder",
-    DecoderNamespace = "v8Crypto" in {
+    DecoderNamespace = "v8Crypto", hasSideEffects = 0 in {
   class AES<string op, bit op7, bit op6, SDPatternOperator Int>
     : N2VQIntXnp<0b00, 0b00, 0b011, op6, op7, NoItinerary,
                  !strconcat("aes", op), "8", v16i8, v16i8, Int>,
@@ -5906,17 +5906,45 @@ def AESE : AES2Op<"e", 0, 0, int_arm_neon_aese>;
 def AESIMC : AES<"imc", 1, 1, int_arm_neon_aesimc>;
 def AESMC : AES<"mc", 1, 0, int_arm_neon_aesmc>;
 
-def SHA1H : N2SHA<"1h", 0b01, 0b010, 1, 1, int_arm_neon_sha1h>;
+def SHA1H : N2SHA<"1h", 0b01, 0b010, 1, 1, null_frag>;
 def SHA1SU1 : N2SHA2Op<"1su1", 0b10, 0b011, 1, 0, int_arm_neon_sha1su1>;
 def SHA256SU0 : N2SHA2Op<"256su0", 0b10, 0b011, 1, 1, int_arm_neon_sha256su0>;
-def SHA1C : N3SHA3Op<"1c", 0b00100, 0b00, int_arm_neon_sha1c>;
-def SHA1M : N3SHA3Op<"1m", 0b00100, 0b10, int_arm_neon_sha1m>;
-def SHA1P : N3SHA3Op<"1p", 0b00100, 0b01, int_arm_neon_sha1p>;
+def SHA1C : N3SHA3Op<"1c", 0b00100, 0b00, null_frag>;
+def SHA1M : N3SHA3Op<"1m", 0b00100, 0b10, null_frag>;
+def SHA1P : N3SHA3Op<"1p", 0b00100, 0b01, null_frag>;
 def SHA1SU0 : N3SHA3Op<"1su0", 0b00100, 0b11, int_arm_neon_sha1su0>;
 def SHA256H : N3SHA3Op<"256h", 0b00110, 0b00, int_arm_neon_sha256h>;
 def SHA256H2 : N3SHA3Op<"256h2", 0b00110, 0b01, int_arm_neon_sha256h2>;
 def SHA256SU1 : N3SHA3Op<"256su1", 0b00110, 0b10, int_arm_neon_sha256su1>;
 
+def : Pat<(i32 (int_arm_neon_sha1h i32:$Rn)),
+          (COPY_TO_REGCLASS (f32 (EXTRACT_SUBREG
+              (SHA1H (SUBREG_TO_REG (i64 0),
+                                    (f32 (COPY_TO_REGCLASS i32:$Rn, SPR)),
+                                    ssub_0)),
+              ssub_0)), GPR)>;
+
+def : Pat<(v4i32 (int_arm_neon_sha1c v4i32:$hash_abcd, i32:$hash_e, v4i32:$wk)),
+          (SHA1C v4i32:$hash_abcd,
+                 (SUBREG_TO_REG (i64 0),
+                                (f32 (COPY_TO_REGCLASS i32:$hash_e, SPR)),
+                                ssub_0),
+                 v4i32:$wk)>;
+
+def : Pat<(v4i32 (int_arm_neon_sha1m v4i32:$hash_abcd, i32:$hash_e, v4i32:$wk)),
+          (SHA1M v4i32:$hash_abcd,
+                 (SUBREG_TO_REG (i64 0),
+                                (f32 (COPY_TO_REGCLASS i32:$hash_e, SPR)),
+                                ssub_0),
+                 v4i32:$wk)>;
+
+def : Pat<(v4i32 (int_arm_neon_sha1p v4i32:$hash_abcd, i32:$hash_e, v4i32:$wk)),
+          (SHA1P v4i32:$hash_abcd,
+                 (SUBREG_TO_REG (i64 0),
+                                (f32 (COPY_TO_REGCLASS i32:$hash_e, SPR)),
+                                ssub_0),
+                 v4i32:$wk)>;
+
 //===----------------------------------------------------------------------===//
 // NEON instructions for single-precision FP math
 //===----------------------------------------------------------------------===//
diff --git a/test/CodeGen/AArch64/neon-crypto.ll b/test/CodeGen/AArch64/neon-crypto.ll
index 0283e0e7ca..c0014fa387 100644
--- a/test/CodeGen/AArch64/neon-crypto.ll
+++ b/test/CodeGen/AArch64/neon-crypto.ll
@@ -1,40 +1,40 @@
 ; RUN: llc < %s -verify-machineinstrs -mtriple=aarch64-none-linux-gnu -mattr=+neon -mattr=+crypto | FileCheck %s
 ; RUN: not llc < %s -verify-machineinstrs -mtriple=aarch64-none-linux-gnu -mattr=+neon 2>&1 | FileCheck --check-prefix=CHECK-NO-CRYPTO %s
 
-declare <4 x i32> @llvm.arm.neon.sha256su1.v4i32(<4 x i32>, <4 x i32>, <4 x i32>) #1
+declare <4 x i32> @llvm.arm.neon.sha256su1(<4 x i32>, <4 x i32>, <4 x i32>) #1
 
-declare <4 x i32> @llvm.arm.neon.sha256h2.v4i32(<4 x i32>, <4 x i32>, <4 x i32>) #1
+declare <4 x i32> @llvm.arm.neon.sha256h2(<4 x i32>, <4 x i32>, <4 x i32>) #1
 
-declare <4 x i32> @llvm.arm.neon.sha256h.v4i32(<4 x i32>, <4 x i32>, <4 x i32>) #1
+declare <4 x i32> @llvm.arm.neon.sha256h(<4 x i32>, <4 x i32>, <4 x i32>) #1
 
-declare <4 x i32> @llvm.arm.neon.sha1su0.v4i32(<4 x i32>, <4 x i32>, <4 x i32>) #1
+declare <4 x i32> @llvm.arm.neon.sha1su0(<4 x i32>, <4 x i32>, <4 x i32>) #1
 
-declare <4 x i32> @llvm.aarch64.neon.sha1m(<4 x i32>, <1 x i32>, <4 x i32>) #1
+declare <4 x i32> @llvm.arm.neon.sha1m(<4 x i32>, i32, <4 x i32>) #1
 
-declare <4 x i32> @llvm.aarch64.neon.sha1p(<4 x i32>, <1 x i32>, <4 x i32>) #1
+declare <4 x i32> @llvm.arm.neon.sha1p(<4 x i32>, i32, <4 x i32>) #1
 
-declare <4 x i32> @llvm.aarch64.neon.sha1c(<4 x i32>, <1 x i32>, <4 x i32>) #1
+declare <4 x i32> @llvm.arm.neon.sha1c(<4 x i32>, i32, <4 x i32>) #1
 
-declare <4 x i32> @llvm.arm.neon.sha256su0.v4i32(<4 x i32>, <4 x i32>) #1
+declare <4 x i32> @llvm.arm.neon.sha256su0(<4 x i32>, <4 x i32>) #1
 
-declare <4 x i32> @llvm.arm.neon.sha1su1.v4i32(<4 x i32>, <4 x i32>) #1
+declare <4 x i32> @llvm.arm.neon.sha1su1(<4 x i32>, <4 x i32>) #1
 
-declare <1 x i32> @llvm.arm.neon.sha1h.v1i32(<1 x i32>) #1
+declare i32 @llvm.arm.neon.sha1h(i32) #1
 
-declare <16 x i8> @llvm.arm.neon.aesimc.v16i8(<16 x i8>) #1
+declare <16 x i8> @llvm.arm.neon.aesimc(<16 x i8>) #1
 
-declare <16 x i8> @llvm.arm.neon.aesmc.v16i8(<16 x i8>) #1
+declare <16 x i8> @llvm.arm.neon.aesmc(<16 x i8>) #1
 
-declare <16 x i8> @llvm.arm.neon.aesd.v16i8(<16 x i8>, <16 x i8>) #1
+declare <16 x i8> @llvm.arm.neon.aesd(<16 x i8>, <16 x i8>) #1
 
-declare <16 x i8> @llvm.arm.neon.aese.v16i8(<16 x i8>, <16 x i8>) #1
+declare <16 x i8> @llvm.arm.neon.aese(<16 x i8>, <16 x i8>) #1
 
 define <16 x i8> @test_vaeseq_u8(<16 x i8> %data, <16 x i8> %key) {
 ; CHECK: test_vaeseq_u8:
 ; CHECK: aese {{v[0-9]+}}.16b, {{v[0-9]+}}.16b
 ; CHECK-NO-CRYPTO: Cannot select: intrinsic %llvm.arm.neon.aese
 entry:
-  %aese.i = tail call <16 x i8> @llvm.arm.neon.aese.v16i8(<16 x i8> %data, <16 x i8> %key)
+  %aese.i = tail call <16 x i8> @llvm.arm.neon.aese(<16 x i8> %data, <16 x i8> %key)
   ret <16 x i8> %aese.i
 }
 
@@ -42,7 +42,7 @@ define <16 x i8> @test_vaesdq_u8(<16 x i8> %data, <16 x i8> %key) {
 ; CHECK: test_vaesdq_u8:
 ; CHECK: aesd {{v[0-9]+}}.16b, {{v[0-9]+}}.16b
 entry:
-  %aesd.i = tail call <16 x i8> @llvm.arm.neon.aesd.v16i8(<16 x i8> %data, <16 x i8> %key)
+  %aesd.i = tail call <16 x i8> @llvm.arm.neon.aesd(<16 x i8> %data, <16 x i8> %key)
   ret <16 x i8> %aesd.i
 }
 
@@ -50,7 +50,7 @@ define <16 x i8> @test_vaesmcq_u8(<16 x i8> %data) {
 ; CHECK: test_vaesmcq_u8:
 ; CHECK: aesmc {{v[0-9]+}}.16b, {{v[0-9]+}}.16b
 entry:
-  %aesmc.i = tail call <16 x i8> @llvm.arm.neon.aesmc.v16i8(<16 x i8> %data)
+  %aesmc.i = tail call <16 x i8> @llvm.arm.neon.aesmc(<16 x i8> %data)
   ret <16 x i8> %aesmc.i
 }
 
@@ -58,7 +58,7 @@ define <16 x i8> @test_vaesimcq_u8(<16 x i8> %data) {
 ; CHECK: test_vaesimcq_u8:
 ; CHECK: aesimc {{v[0-9]+}}.16b, {{v[0-9]+}}.16b
 entry:
-  %aesimc.i = tail call <16 x i8> @llvm.arm.neon.aesimc.v16i8(<16 x i8> %data)
+  %aesimc.i = tail call <16 x i8> @llvm.arm.neon.aesimc(<16 x i8> %data)
   ret <16 x i8> %aesimc.i
 }
 
@@ -66,17 +66,15 @@ define i32 @test_vsha1h_u32(i32 %hash_e) {
 ; CHECK: test_vsha1h_u32:
 ; CHECK: sha1h {{s[0-9]+}}, {{s[0-9]+}}
 entry:
-  %sha1h.i = insertelement <1 x i32> undef, i32 %hash_e, i32 0
-  %sha1h1.i = tail call <1 x i32> @llvm.arm.neon.sha1h.v1i32(<1 x i32> %sha1h.i)
-  %0 = extractelement <1 x i32> %sha1h1.i, i32 0
-  ret i32 %0
+  %sha1h1.i = tail call i32 @llvm.arm.neon.sha1h(i32 %hash_e)
+  ret i32 %sha1h1.i
 }
 
 define <4 x i32> @test_vsha1su1q_u32(<4 x i32> %tw0_3, <4 x i32> %w12_15) {
 ; CHECK: test_vsha1su1q_u32:
 ; CHECK: sha1su1 {{v[0-9]+}}.4s, {{v[0-9]+}}.4s
 entry:
-  %sha1su12.i = tail call <4 x i32> @llvm.arm.neon.sha1su1.v4i32(<4 x i32> %tw0_3, <4 x i32> %w12_15)
+  %sha1su12.i = tail call <4 x i32> @llvm.arm.neon.sha1su1(<4 x i32> %tw0_3, <4 x i32> %w12_15)
   ret <4 x i32> %sha1su12.i
 }
 
@@ -84,7 +82,7 @@ define <4 x i32> @test_vsha256su0q_u32(<4 x i32> %w0_3, <4 x i32> %w4_7) {
 ; CHECK: test_vsha256su0q_u32:
 ; CHECK: sha256su0 {{v[0-9]+}}.4s, {{v[0-9]+}}.4s
 entry:
-  %sha256su02.i = tail call <4 x i32> @llvm.arm.neon.sha256su0.v4i32(<4 x i32> %w0_3, <4 x i32> %w4_7)
+  %sha256su02.i = tail call <4 x i32> @llvm.arm.neon.sha256su0(<4 x i32> %w0_3, <4 x i32> %w4_7)
   ret <4 x i32> %sha256su02.i
 }
 
@@ -92,8 +90,7 @@ define <4 x i32> @test_vsha1cq_u32(<4 x i32> %hash_abcd, i32 %hash_e, <4 x i32>
 ; CHECK: test_vsha1cq_u32:
 ; CHECK: sha1c {{q[0-9]+}}, {{s[0-9]+}}, {{v[0-9]+}}.4s
 entry:
-  %sha1c.i = insertelement <1 x i32> undef, i32 %hash_e, i32 0
-  %sha1c1.i = tail call <4 x i32> @llvm.aarch64.neon.sha1c(<4 x i32> %hash_abcd, <1 x i32> %sha1c.i, <4 x i32> %wk)
+  %sha1c1.i = tail call <4 x i32> @llvm.arm.neon.sha1c(<4 x i32> %hash_abcd, i32 %hash_e, <4 x i32> %wk)
   ret <4 x i32> %sha1c1.i
 }
 
@@ -101,8 +98,7 @@ define <4 x i32> @test_vsha1pq_u32(<4 x i32> %hash_abcd, i32 %hash_e, <4 x i32>
 ; CHECK: test_vsha1pq_u32:
 ; CHECK: sha1p {{q[0-9]+}}, {{s[0-9]+}}, {{v[0-9]+}}.4s
 entry:
-  %sha1p.i = insertelement <1 x i32> undef, i32 %hash_e, i32 0
-  %sha1p1.i = tail call <4 x i32> @llvm.aarch64.neon.sha1p(<4 x i32> %hash_abcd, <1 x i32> %sha1p.i, <4 x i32> %wk)
+  %sha1p1.i = tail call <4 x i32> @llvm.arm.neon.sha1p(<4 x i32> %hash_abcd, i32 %hash_e, <4 x i32> %wk)
   ret <4 x i32> %sha1p1.i
 }
 
@@ -110,8 +106,7 @@ define <4 x i32> @test_vsha1mq_u32(<4 x i32> %hash_abcd, i32 %hash_e, <4 x i32>
 ; CHECK: test_vsha1mq_u32:
 ; CHECK: sha1m {{q[0-9]+}}, {{s[0-9]+}}, {{v[0-9]+}}.4s
 entry:
-  %sha1m.i = insertelement <1 x i32> undef, i32 %hash_e, i32 0
-  %sha1m1.i = tail call <4 x i32> @llvm.aarch64.neon.sha1m(<4 x i32> %hash_abcd, <1 x i32> %sha1m.i, <4 x i32> %wk)
+  %sha1m1.i = tail call <4 x i32> @llvm.arm.neon.sha1m(<4 x i32> %hash_abcd, i32 %hash_e, <4 x i32> %wk)
   ret <4 x i32> %sha1m1.i
 }
 
@@ -119,7 +114,7 @@ define <4 x i32> @test_vsha1su0q_u32(<4 x i32> %w0_3, <4 x i32> %w4_7, <4 x i32>
 ; CHECK: test_vsha1su0q_u32:
 ; CHECK: sha1su0 {{v[0-9]+}}.4s, {{v[0-9]+}}.4s, {{v[0-9]+}}.4s
 entry:
-  %sha1su03.i = tail call <4 x i32> @llvm.arm.neon.sha1su0.v4i32(<4 x i32> %w0_3, <4 x i32> %w4_7, <4 x i32> %w8_11)
+  %sha1su03.i = tail call <4 x i32> @llvm.arm.neon.sha1su0(<4 x i32> %w0_3, <4 x i32> %w4_7, <4 x i32> %w8_11)
   ret <4 x i32> %sha1su03.i
 }
 
@@ -127,7 +122,7 @@ define <4 x i32> @test_vsha256hq_u32(<4 x i32> %hash_abcd, <4 x i32> %hash_efgh,
 ; CHECK: test_vsha256hq_u32:
 ; CHECK: sha256h {{q[0-9]+}}, {{q[0-9]+}}, {{v[0-9]+}}.4s
 entry:
-  %sha256h3.i = tail call <4 x i32> @llvm.arm.neon.sha256h.v4i32(<4 x i32> %hash_abcd, <4 x i32> %hash_efgh, <4 x i32> %wk)
+  %sha256h3.i = tail call <4 x i32> @llvm.arm.neon.sha256h(<4 x i32> %hash_abcd, <4 x i32> %hash_efgh, <4 x i32> %wk)
   ret <4 x i32> %sha256h3.i
 }
 
@@ -135,7 +130,7 @@ define <4 x i32> @test_vsha256h2q_u32(<4 x i32> %hash_efgh, <4 x i32> %hash_abcd
 ; CHECK: test_vsha256h2q_u32:
 ; CHECK: sha256h2 {{q[0-9]+}}, {{q[0-9]+}}, {{v[0-9]+}}.4s
 entry:
-  %sha256h23.i = tail call <4 x i32> @llvm.arm.neon.sha256h2.v4i32(<4 x i32> %hash_efgh, <4 x i32> %hash_abcd, <4 x i32> %wk)
+  %sha256h23.i = tail call <4 x i32> @llvm.arm.neon.sha256h2(<4 x i32> %hash_efgh, <4 x i32> %hash_abcd, <4 x i32> %wk)
   ret <4 x i32> %sha256h23.i
 }
 
@@ -143,7 +138,7 @@ define <4 x i32> @test_vsha256su1q_u32(<4 x i32> %tw0_3, <4 x i32> %w8_11, <4 x
 ; CHECK: test_vsha256su1q_u32:
 ; CHECK: sha256su1 {{v[0-9]+}}.4s, {{v[0-9]+}}.4s, {{v[0-9]+}}.4s
 entry:
-  %sha256su13.i = tail call <4 x i32> @llvm.arm.neon.sha256su1.v4i32(<4 x i32> %tw0_3, <4 x i32> %w8_11, <4 x i32> %w12_15)
+  %sha256su13.i = tail call <4 x i32> @llvm.arm.neon.sha256su1(<4 x i32> %tw0_3, <4 x i32> %w8_11, <4 x i32> %w12_15)
   ret <4 x i32> %sha256su13.i
 }
 
diff --git a/test/CodeGen/ARM/intrinsics-crypto.ll b/test/CodeGen/ARM/intrinsics-crypto.ll
index c038fe6da8..96413d341e 100644
--- a/test/CodeGen/ARM/intrinsics-crypto.ll
+++ b/test/CodeGen/ARM/intrinsics-crypto.ll
@@ -3,13 +3,13 @@
 define arm_aapcs_vfpcc <16 x i8> @test_aesde(<16 x i8>* %a, <16 x i8> *%b) {
   %tmp = load <16 x i8>* %a
   %tmp2 = load <16 x i8>* %b
-  %tmp3 = call <16 x i8> @llvm.arm.neon.aesd.v16i8(<16 x i8> %tmp, <16 x i8> %tmp2)
+  %tmp3 = call <16 x i8> @llvm.arm.neon.aesd(<16 x i8> %tmp, <16 x i8> %tmp2)
   ; CHECK: aesd.8 q{{[0-9]+}}, q{{[0-9]+}}
-  %tmp4 = call <16 x i8> @llvm.arm.neon.aese.v16i8(<16 x i8> %tmp3, <16 x i8> %tmp2)
+  %tmp4 = call <16 x i8> @llvm.arm.neon.aese(<16 x i8> %tmp3, <16 x i8> %tmp2)
   ; CHECK: aese.8 q{{[0-9]+}}, q{{[0-9]+}}
-  %tmp5 = call <16 x i8> @llvm.arm.neon.aesimc.v16i8(<16 x i8> %tmp4)
+  %tmp5 = call <16 x i8> @llvm.arm.neon.aesimc(<16 x i8> %tmp4)
   ; CHECK: aesimc.8 q{{[0-9]+}}, q{{[0-9]+}}
-  %tmp6 = call <16 x i8> @llvm.arm.neon.aesmc.v16i8(<16 x i8> %tmp5)
+  %tmp6 = call <16 x i8> @llvm.arm.neon.aesmc(<16 x i8> %tmp5)
   ; CHECK: aesmc.8 q{{[0-9]+}}, q{{[0-9]+}}
   ret <16 x i8> %tmp6
 }
@@ -18,40 +18,42 @@ define arm_aapcs_vfpcc <4 x i32> @test_sha(<4 x i32> *%a, <4 x i32> *%b, <4 x i3
   %tmp = load <4 x i32>* %a
   %tmp2 = load <4 x i32>* %b
   %tmp3 = load <4 x i32>* %c
-  %res1 = call <4 x i32> @llvm.arm.neon.sha1h.v4i32(<4 x i32> %tmp)
+  %scalar = extractelement <4 x i32> %tmp, i32 0
+  %resscalar = call i32 @llvm.arm.neon.sha1h(i32 %scalar)
+  %res1 = insertelement <4 x i32> undef, i32 %resscalar, i32 0
   ; CHECK: sha1h.32 q{{[0-9]+}}, q{{[0-9]+}}
-  %res2 = call <4 x i32> @llvm.arm.neon.sha1c.v4i32(<4 x i32> %tmp2, <4 x i32> %tmp3, <4 x i32> %res1)
+  %res2 = call <4 x i32> @llvm.arm.neon.sha1c(<4 x i32> %tmp2, i32 %scalar, <4 x i32> %res1)
   ; CHECK: sha1c.32 q{{[0-9]+}}, q{{[0-9]+}}, q{{[0-9]+}}
-  %res3 = call <4 x i32> @llvm.arm.neon.sha1m.v4i32(<4 x i32> %res2, <4 x i32> %tmp3, <4 x i32> %res1)
+  %res3 = call <4 x i32> @llvm.arm.neon.sha1m(<4 x i32> %res2, i32 %scalar, <4 x i32> %res1)
   ; CHECK: sha1m.32 q{{[0-9]+}}, q{{[0-9]+}}, q{{[0-9]+}}
-  %res4 = call <4 x i32> @llvm.arm.neon.sha1p.v4i32(<4 x i32> %res3, <4 x i32> %tmp3, <4 x i32> %res1)
+  %res4 = call <4 x i32> @llvm.arm.neon.sha1p(<4 x i32> %res3, i32 %scalar, <4 x i32> %res1)
   ; CHECK: sha1p.32 q{{[0-9]+}}, q{{[0-9]+}}, q{{[0-9]+}}
-  %res5 = call <4 x i32> @llvm.arm.neon.sha1su0.v4i32(<4 x i32> %res4, <4 x i32> %tmp3, <4 x i32> %res1)
+  %res5 = call <4 x i32> @llvm.arm.neon.sha1su0(<4 x i32> %res4, <4 x i32> %tmp3, <4 x i32> %res1)
   ; CHECK: sha1su0.32 q{{[0-9]+}}, q{{[0-9]+}}, q{{[0-9]+}}
-  %res6 = call <4 x i32> @llvm.arm.neon.sha1su1.v4i32(<4 x i32> %res5, <4 x i32> %res1)
+  %res6 = call <4 x i32> @llvm.arm.neon.sha1su1(<4 x i32> %res5, <4 x i32> %res1)
   ; CHECK: sha1su1.32 q{{[0-9]+}}, q{{[0-9]+}}
-  %res7 = call <4 x i32> @llvm.arm.neon.sha256h.v4i32(<4 x i32> %res6, <4 x i32> %tmp3, <4 x i32> %res1)
+  %res7 = call <4 x i32> @llvm.arm.neon.sha256h(<4 x i32> %res6, <4 x i32> %tmp3, <4 x i32> %res1)
   ; CHECK: sha256h.32 q{{[0-9]+}}, q{{[0-9]+}}, q{{[0-9]+}}
-  %res8 = call <4 x i32> @llvm.arm.neon.sha256h2.v4i32(<4 x i32> %res7, <4 x i32> %tmp3, <4 x i32> %res1)
+  %res8 = call <4 x i32> @llvm.arm.neon.sha256h2(<4 x i32> %res7, <4 x i32> %tmp3, <4 x i32> %res1)
   ; CHECK: sha256h2.32 q{{[0-9]+}}, q{{[0-9]+}}, q{{[0-9]+}}
-  %res9 = call <4 x i32> @llvm.arm.neon.sha256su1.v4i32(<4 x i32> %res8, <4 x i32> %tmp3, <4 x i32> %res1)
+  %res9 = call <4 x i32> @llvm.arm.neon.sha256su1(<4 x i32> %res8, <4 x i32> %tmp3, <4 x i32> %res1)
   ; CHECK: sha256su1.32 q{{[0-9]+}}, q{{[0-9]+}}, q{{[0-9]+}}
-  %res10 = call <4 x i32> @llvm.arm.neon.sha256su0.v4i32(<4 x i32> %res9, <4 x i32> %tmp3)
+  %res10 = call <4 x i32> @llvm.arm.neon.sha256su0(<4 x i32> %res9, <4 x i32> %tmp3)
   ; CHECK: sha256su0.32 q{{[0-9]+}}, q{{[0-9]+}}
   ret <4 x i32> %res10
 }
 
-declare <16 x i8> @llvm.arm.neon.aesd.v16i8(<16 x i8>, <16 x i8>)
-declare <16 x i8> @llvm.arm.neon.aese.v16i8(<16 x i8>, <16 x i8>)
-declare <16 x i8> @llvm.arm.neon.aesimc.v16i8(<16 x i8>)
-declare <16 x i8> @llvm.arm.neon.aesmc.v16i8(<16 x i8>)
-declare <4 x i32> @llvm.arm.neon.sha1h.v4i32(<4 x i32>)
-declare <4 x i32> @llvm.arm.neon.sha1c.v4i32(<4 x i32>, <4 x i32>, <4 x i32>)
-declare <4 x i32> @llvm.arm.neon.sha1m.v4i32(<4 x i32>, <4 x i32>, <4 x i32>)
-declare <4 x i32> @llvm.arm.neon.sha1p.v4i32(<4 x i32>, <4 x i32>, <4 x i32>)
-declare <4 x i32> @llvm.arm.neon.sha1su0.v4i32(<4 x i32>, <4 x i32>, <4 x i32>)
-declare <4 x i32> @llvm.arm.neon.sha256h.v4i32(<4 x i32>, <4 x i32>, <4 x i32>)
-declare <4 x i32> @llvm.arm.neon.sha256h2.v4i32(<4 x i32>, <4 x i32>, <4 x i32>)
-declare <4 x i32> @llvm.arm.neon.sha256su1.v4i32(<4 x i32>, <4 x i32>, <4 x i32>)
-declare <4 x i32> @llvm.arm.neon.sha256su0.v4i32(<4 x i32>, <4 x i32>)
-declare <4 x i32> @llvm.arm.neon.sha1su1.v4i32(<4 x i32>, <4 x i32>)
+declare <16 x i8> @llvm.arm.neon.aesd(<16 x i8>, <16 x i8>)
+declare <16 x i8> @llvm.arm.neon.aese(<16 x i8>, <16 x i8>)
+declare <16 x i8> @llvm.arm.neon.aesimc(<16 x i8>)
+declare <16 x i8> @llvm.arm.neon.aesmc(<16 x i8>)
+declare i32 @llvm.arm.neon.sha1h(i32)
+declare <4 x i32> @llvm.arm.neon.sha1c(<4 x i32>, i32, <4 x i32>)
+declare <4 x i32> @llvm.arm.neon.sha1m(<4 x i32>, i32, <4 x i32>)
+declare <4 x i32> @llvm.arm.neon.sha1p(<4 x i32>, i32, <4 x i32>)
+declare <4 x i32> @llvm.arm.neon.sha1su0(<4 x i32>, <4 x i32>, <4 x i32>)
+declare <4 x i32> @llvm.arm.neon.sha256h(<4 x i32>, <4 x i32>, <4 x i32>)
+declare <4 x i32> @llvm.arm.neon.sha256h2(<4 x i32>, <4 x i32>, <4 x i32>)
+declare <4 x i32> @llvm.arm.neon.sha256su1(<4 x i32>, <4 x i32>, <4 x i32>)
+declare <4 x i32> @llvm.arm.neon.sha256su0(<4 x i32>, <4 x i32>)
+declare <4 x i32> @llvm.arm.neon.sha1su1(<4 x i32>, <4 x i32>)