[X86] Synchronize the loadiwkey builtin operand order with gcc version.

Author: topperc

clang/include/clang/Basic/BuiltinsX86.def

@@ -1901,7 +1901,7 @@ TARGET_BUILTIN(__builtin_ia32_enqcmd, "Ucv*vC*", "n", "enqcmd")
 TARGET_BUILTIN(__builtin_ia32_enqcmds, "Ucv*vC*", "n", "enqcmd")
 
 // KEY LOCKER
-TARGET_BUILTIN(__builtin_ia32_loadiwkey, "vUiV2OiV2OiV2Oi", "nV:128:", "kl")
+TARGET_BUILTIN(__builtin_ia32_loadiwkey, "vV2OiV2OiV2OiUi", "nV:128:", "kl")
 TARGET_BUILTIN(__builtin_ia32_encodekey128,
                "UiUiV2OiV2Oi*V2Oi*V2Oi*V2Oi*V2Oi*V2Oi*", "nV:128:", "kl")
 TARGET_BUILTIN(__builtin_ia32_encodekey256,

clang/lib/Headers/keylockerintrin.h

@@ -95,7 +95,7 @@
 static __inline__ void __DEFAULT_FN_ATTRS
 _mm_loadiwkey (unsigned int __ctl, __m128i __intkey,
                __m128i __enkey_lo, __m128i __enkey_hi) {
-  __builtin_ia32_loadiwkey (__ctl, __intkey, __enkey_lo, __enkey_hi);
+  __builtin_ia32_loadiwkey (__intkey, __enkey_lo, __enkey_hi, __ctl);
 }
 
 /// Wrap a 128-bit AES key from __key into a key handle and output in

llvm/include/llvm/IR/IntrinsicsX86.td

@@ -4953,7 +4953,7 @@ let TargetPrefix = "x86" in {
 // Key Locker
 let TargetPrefix = "x86" in {
   def int_x86_loadiwkey : GCCBuiltin<"__builtin_ia32_loadiwkey">,
-      Intrinsic<[], [llvm_i32_ty, llvm_v2i64_ty, llvm_v2i64_ty, llvm_v2i64_ty],
+      Intrinsic<[], [llvm_v2i64_ty, llvm_v2i64_ty, llvm_v2i64_ty, llvm_i32_ty],
                 []>;
   def int_x86_encodekey128 :
       Intrinsic<[llvm_i32_ty, llvm_v2i64_ty, llvm_v2i64_ty,

llvm/lib/Target/X86/X86InstrKL.td

@@ -20,7 +20,7 @@ let SchedRW = [WriteSystem], Predicates = [HasKL] in {
   let Uses = [XMM0, EAX], Defs = [EFLAGS] in {
     def LOADIWKEY : I<0xDC, MRMSrcReg, (outs), (ins VR128:$src1, VR128:$src2),
                       "loadiwkey\t{$src2, $src1|$src1, $src2}",
-                      [(int_x86_loadiwkey EAX, XMM0, VR128:$src1, VR128:$src2)]>, T8XS;
+                      [(int_x86_loadiwkey XMM0, VR128:$src1, VR128:$src2, EAX)]>, T8XS;
   }
 
   let Uses = [XMM0], Defs = [XMM0, XMM1, XMM2, XMM4, XMM5, XMM6, EFLAGS] in {

llvm/test/CodeGen/X86/keylocker-intrinsics.ll

@@ -4,7 +4,7 @@
 ; RUN: llc < %s -verify-machineinstrs -mtriple=x86_64-unkown-unknown -mattr=+widekl | FileCheck %s --check-prefix=X64
 ; RUN: llc < %s -verify-machineinstrs -mtriple=i386-unkown-unknown -mattr=+widekl -mattr=+avx2 | FileCheck %s --check-prefix=X32
 
-declare void @llvm.x86.loadiwkey(i32, <2 x i64>, <2 x i64>, <2 x i64>)
+declare void @llvm.x86.loadiwkey(<2 x i64>, <2 x i64>, <2 x i64>, i32)
 declare { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } @llvm.x86.encodekey128(i32, <2 x i64>)
 declare { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } @llvm.x86.encodekey256(i32, <2 x i64>, <2 x i64>)
 declare { i8, <2 x i64> } @llvm.x86.aesenc128kl(<2 x i64>, i8*)
@@ -29,7 +29,7 @@ define void @test_loadiwkey(i32 %ctl, <2 x i64> %intkey, <2 x i64> %enkey_lo, <2
 ; X32-NEXT:    loadiwkey %xmm2, %xmm1
 ; X32-NEXT:    retl
 entry:
-  tail call void @llvm.x86.loadiwkey(i32 %ctl, <2 x i64> %intkey, <2 x i64> %enkey_lo, <2 x i64> %enkey_hi)
+  tail call void @llvm.x86.loadiwkey(<2 x i64> %intkey, <2 x i64> %enkey_lo, <2 x i64> %enkey_hi, i32 %ctl)
   ret void
 }