AArch64: Resolve some FIXMEs in CGBuiltin left over from backend merge

[kongyi]

Summary

Merge vrshr_n_v and vqshlu_n_v with ARM.

Remove FIXME comments for others as they can't actually be shared, at least not with current EmitCommonNeonBuiltinExpr implementation.

NFC.

[t.p.northover]

Bother, that's disappointing. Patch looks good though!

Thanks for doing this.

Tim.

[kongyi]

Closed by commit rL214173 (authored by @kongyi).

lib/CodeGen/CGBuiltin.cpp

static NeonIntrinsicInfo ARMSIMDIntrinsicMap [] = {
   NEONMAP1(vqrdmulh_v, arm_neon_vqrdmulh, Add1ArgType),
   NEONMAP1(vqrdmulhq_v, arm_neon_vqrdmulh, Add1ArgType),
   NEONMAP2(vqrshl_v, arm_neon_vqrshiftu, arm_neon_vqrshifts, Add1ArgType | UnsignedAlts),
   NEONMAP2(vqrshlq_v, arm_neon_vqrshiftu, arm_neon_vqrshifts, Add1ArgType | UnsignedAlts),
   NEONMAP2(vqshl_n_v, arm_neon_vqshiftu, arm_neon_vqshifts, UnsignedAlts),
   NEONMAP2(vqshl_v, arm_neon_vqshiftu, arm_neon_vqshifts, Add1ArgType | UnsignedAlts),
   NEONMAP2(vqshlq_n_v, arm_neon_vqshiftu, arm_neon_vqshifts, UnsignedAlts),
   NEONMAP2(vqshlq_v, arm_neon_vqshiftu, arm_neon_vqshifts, Add1ArgType | UnsignedAlts),
+  NEONMAP1(vqshlu_n_v, arm_neon_vqshiftsu, 0),
+  NEONMAP1(vqshluq_n_v, arm_neon_vqshiftsu, 0),
   NEONMAP2(vqsub_v, arm_neon_vqsubu, arm_neon_vqsubs, Add1ArgType | UnsignedAlts),
   NEONMAP2(vqsubq_v, arm_neon_vqsubu, arm_neon_vqsubs, Add1ArgType | UnsignedAlts),
   NEONMAP1(vraddhn_v, arm_neon_vraddhn, Add1ArgType),
   NEONMAP2(vrecpe_v, arm_neon_vrecpe, arm_neon_vrecpe, 0),
   NEONMAP2(vrecpeq_v, arm_neon_vrecpe, arm_neon_vrecpe, 0),
   NEONMAP1(vrecps_v, arm_neon_vrecps, Add1ArgType),
   NEONMAP1(vrecpsq_v, arm_neon_vrecps, Add1ArgType),
   NEONMAP2(vrhadd_v, arm_neon_vrhaddu, arm_neon_vrhadds, Add1ArgType | UnsignedAlts),
   NEONMAP2(vrhaddq_v, arm_neon_vrhaddu, arm_neon_vrhadds, Add1ArgType | UnsignedAlts),
   NEONMAP2(vrshl_v, arm_neon_vrshiftu, arm_neon_vrshifts, Add1ArgType | UnsignedAlts),
   NEONMAP2(vrshlq_v, arm_neon_vrshiftu, arm_neon_vrshifts, Add1ArgType | UnsignedAlts),
+  NEONMAP2(vrshr_n_v, arm_neon_vrshiftu, arm_neon_vrshifts, UnsignedAlts),
+  NEONMAP2(vrshrq_n_v, arm_neon_vrshiftu, arm_neon_vrshifts, UnsignedAlts),
   NEONMAP2(vrsqrte_v, arm_neon_vrsqrte, arm_neon_vrsqrte, 0),
   NEONMAP2(vrsqrteq_v, arm_neon_vrsqrte, arm_neon_vrsqrte, 0),
   NEONMAP1(vrsqrts_v, arm_neon_vrsqrts, Add1ArgType),
   NEONMAP1(vrsqrtsq_v, arm_neon_vrsqrts, Add1ArgType),
   NEONMAP1(vrsubhn_v, arm_neon_vrsubhn, Add1ArgType),
   NEONMAP1(vsha1su0q_v, arm_neon_sha1su0, 0),
   NEONMAP1(vsha1su1q_v, arm_neon_sha1su1, 0),
   NEONMAP1(vsha256h2q_v, arm_neon_sha256h2, 0),
static NeonIntrinsicInfo AArch64SIMDIntrinsicMap[] = {
   NEONMAP1(vqrdmulh_v, aarch64_neon_sqrdmulh, Add1ArgType),
   NEONMAP1(vqrdmulhq_v, aarch64_neon_sqrdmulh, Add1ArgType),
   NEONMAP2(vqrshl_v, aarch64_neon_uqrshl, aarch64_neon_sqrshl, Add1ArgType | UnsignedAlts),
   NEONMAP2(vqrshlq_v, aarch64_neon_uqrshl, aarch64_neon_sqrshl, Add1ArgType | UnsignedAlts),
   NEONMAP2(vqshl_n_v, aarch64_neon_uqshl, aarch64_neon_sqshl, UnsignedAlts),
   NEONMAP2(vqshl_v, aarch64_neon_uqshl, aarch64_neon_sqshl, Add1ArgType | UnsignedAlts),
   NEONMAP2(vqshlq_n_v, aarch64_neon_uqshl, aarch64_neon_sqshl,UnsignedAlts),
   NEONMAP2(vqshlq_v, aarch64_neon_uqshl, aarch64_neon_sqshl, Add1ArgType | UnsignedAlts),
+  NEONMAP1(vqshlu_n_v, aarch64_neon_sqshlu, 0),
+  NEONMAP1(vqshluq_n_v, aarch64_neon_sqshlu, 0),
   NEONMAP2(vqsub_v, aarch64_neon_uqsub, aarch64_neon_sqsub, Add1ArgType | UnsignedAlts),
   NEONMAP2(vqsubq_v, aarch64_neon_uqsub, aarch64_neon_sqsub, Add1ArgType | UnsignedAlts),
   NEONMAP1(vraddhn_v, aarch64_neon_raddhn, Add1ArgType),
   NEONMAP2(vrecpe_v, aarch64_neon_frecpe, aarch64_neon_urecpe, 0),
   NEONMAP2(vrecpeq_v, aarch64_neon_frecpe, aarch64_neon_urecpe, 0),
   NEONMAP1(vrecps_v, aarch64_neon_frecps, Add1ArgType),
   NEONMAP1(vrecpsq_v, aarch64_neon_frecps, Add1ArgType),
   NEONMAP2(vrhadd_v, aarch64_neon_urhadd, aarch64_neon_srhadd, Add1ArgType | UnsignedAlts),
   NEONMAP2(vrhaddq_v, aarch64_neon_urhadd, aarch64_neon_srhadd, Add1ArgType | UnsignedAlts),
   NEONMAP2(vrshl_v, aarch64_neon_urshl, aarch64_neon_srshl, Add1ArgType | UnsignedAlts),
   NEONMAP2(vrshlq_v, aarch64_neon_urshl, aarch64_neon_srshl, Add1ArgType | UnsignedAlts),
+  NEONMAP2(vrshr_n_v, aarch64_neon_urshl, aarch64_neon_srshl, UnsignedAlts),
+  NEONMAP2(vrshrq_n_v, aarch64_neon_urshl, aarch64_neon_srshl, UnsignedAlts),
   NEONMAP2(vrsqrte_v, aarch64_neon_frsqrte, aarch64_neon_ursqrte, 0),
   NEONMAP2(vrsqrteq_v, aarch64_neon_frsqrte, aarch64_neon_ursqrte, 0),
   NEONMAP1(vrsqrts_v, aarch64_neon_frsqrts, Add1ArgType),
   NEONMAP1(vrsqrtsq_v, aarch64_neon_frsqrts, Add1ArgType),
   NEONMAP1(vrsubhn_v, aarch64_neon_rsubhn, Add1ArgType),
   NEONMAP1(vsha1su0q_v, aarch64_crypto_sha1su0, 0),
   NEONMAP1(vsha1su1q_v, aarch64_crypto_sha1su1, 0),
   NEONMAP1(vsha256h2q_v, aarch64_crypto_sha256h2, 0),
case NEON::BI__builtin_neon_vqdmlsl_v: {
     AccumOps.push_back(Mul);
     return EmitNeonCall(CGM.getIntrinsic(AltLLVMIntrinsic, Ty),
                         AccumOps, NameHint);
   }
   case NEON::BI__builtin_neon_vqshl_n_v:
   case NEON::BI__builtin_neon_vqshlq_n_v:
     return EmitNeonCall(CGM.getIntrinsic(Int, Ty), Ops, "vqshl_n",
                         1, false);
+  case NEON::BI__builtin_neon_vqshlu_n_v:
+  case NEON::BI__builtin_neon_vqshluq_n_v:
+    return EmitNeonCall(CGM.getIntrinsic(Int, Ty), Ops, "vqshlu_n",
+                        1, false);
   case NEON::BI__builtin_neon_vrecpe_v:
   case NEON::BI__builtin_neon_vrecpeq_v:
   case NEON::BI__builtin_neon_vrsqrte_v:
   case NEON::BI__builtin_neon_vrsqrteq_v:
     Int = Ty->isFPOrFPVectorTy() ? LLVMIntrinsic : AltLLVMIntrinsic;
     return EmitNeonCall(CGM.getIntrinsic(Int, Ty), Ops, NameHint);
 
+  case NEON::BI__builtin_neon_vrshr_n_v:
+  case NEON::BI__builtin_neon_vrshrq_n_v:
+    return EmitNeonCall(CGM.getIntrinsic(Int, Ty), Ops, "vrshr_n",
+                        1, true);
   case NEON::BI__builtin_neon_vshl_n_v:
   case NEON::BI__builtin_neon_vshlq_n_v:
     Ops[1] = EmitNeonShiftVector(Ops[1], Ty, false);
     return Builder.CreateShl(Builder.CreateBitCast(Ops[0],Ty), Ops[1],
                              "vshl_n");
   case NEON::BI__builtin_neon_vshll_n_v: {
     llvm::Type *SrcTy = llvm::VectorType::getTruncatedElementVectorType(VTy);
     Ops[0] = Builder.CreateBitCast(Ops[0], SrcTy);
Value *CodeGenFunction::EmitARMBuiltinExpr(unsigned BuiltinID,
   case NEON::BI__builtin_neon_vqrshrn_n_v:
     Int =
       usgn ? Intrinsic::arm_neon_vqrshiftnu : Intrinsic::arm_neon_vqrshiftns;
     return EmitNeonCall(CGM.getIntrinsic(Int, Ty), Ops, "vqrshrn_n",
                         1, true);
   case NEON::BI__builtin_neon_vqrshrun_n_v:
     return EmitNeonCall(CGM.getIntrinsic(Intrinsic::arm_neon_vqrshiftnsu, Ty),
                         Ops, "vqrshrun_n", 1, true);
-  case NEON::BI__builtin_neon_vqshlu_n_v:
-  case NEON::BI__builtin_neon_vqshluq_n_v:
-    return EmitNeonCall(CGM.getIntrinsic(Intrinsic::arm_neon_vqshiftsu, Ty),
-                        Ops, "vqshlu", 1, false);
   case NEON::BI__builtin_neon_vqshrn_n_v:
     Int = usgn ? Intrinsic::arm_neon_vqshiftnu : Intrinsic::arm_neon_vqshiftns;
     return EmitNeonCall(CGM.getIntrinsic(Int, Ty), Ops, "vqshrn_n",
                         1, true);
   case NEON::BI__builtin_neon_vqshrun_n_v:
     return EmitNeonCall(CGM.getIntrinsic(Intrinsic::arm_neon_vqshiftnsu, Ty),
                         Ops, "vqshrun_n", 1, true);
   case NEON::BI__builtin_neon_vrecpe_v:
   case NEON::BI__builtin_neon_vrecpeq_v:
     return EmitNeonCall(CGM.getIntrinsic(Intrinsic::arm_neon_vrecpe, Ty),
                         Ops, "vrecpe");
   case NEON::BI__builtin_neon_vrshrn_n_v:
     return EmitNeonCall(CGM.getIntrinsic(Intrinsic::arm_neon_vrshiftn, Ty),
                         Ops, "vrshrn_n", 1, true);
-  case NEON::BI__builtin_neon_vrshr_n_v:
-  case NEON::BI__builtin_neon_vrshrq_n_v:
-    Int = usgn ? Intrinsic::arm_neon_vrshiftu : Intrinsic::arm_neon_vrshifts;
-    return EmitNeonCall(CGM.getIntrinsic(Int, Ty), Ops, "vrshr_n", 1, true);
   case NEON::BI__builtin_neon_vrsra_n_v:
   case NEON::BI__builtin_neon_vrsraq_n_v:
     Ops[0] = Builder.CreateBitCast(Ops[0], Ty);
     Ops[1] = Builder.CreateBitCast(Ops[1], Ty);
     Ops[2] = EmitNeonShiftVector(Ops[2], Ty, true);
     Int = usgn ? Intrinsic::arm_neon_vrshiftu : Intrinsic::arm_neon_vrshifts;
     Ops[1] = Builder.CreateCall2(CGM.getIntrinsic(Int, Ty), Ops[1], Ops[2]);
     return Builder.CreateAdd(Ops[0], Ops[1], "vrsra_n");
return EmitNeonCall(CGM.getIntrinsic(Intrinsic::aarch64_neon_frecps, f32Type),
                         Ops, "vrecps");
   }
   case NEON::BI__builtin_neon_vrecpsd_f64: {
     llvm::Type *f64Type = llvm::Type::getDoubleTy(getLLVMContext());
     Ops.push_back(EmitScalarExpr(E->getArg(1)));
     return EmitNeonCall(CGM.getIntrinsic(Intrinsic::aarch64_neon_frecps, f64Type),
                         Ops, "vrecps");
   }
-  case NEON::BI__builtin_neon_vrshr_n_v:
-  case NEON::BI__builtin_neon_vrshrq_n_v:
-    // FIXME: this can be shared with 32-bit ARM, but not AArch64 at the
-    // moment. After the final merge it should be added to
-    // EmitCommonNeonBuiltinExpr.
-    Int = usgn ? Intrinsic::aarch64_neon_urshl : Intrinsic::aarch64_neon_srshl;
-    return EmitNeonCall(CGM.getIntrinsic(Int, Ty), Ops, "vrshr_n", 1, true);
-  case NEON::BI__builtin_neon_vqshlu_n_v:
-  case NEON::BI__builtin_neon_vqshluq_n_v:
-    // FIXME: AArch64 and ARM use different intrinsics for this, but are
-    // essentially compatible. It should be in EmitCommonNeonBuiltinExpr after
-    // the final merge.
-    Int = Intrinsic::aarch64_neon_sqshlu;
-    return EmitNeonCall(CGM.getIntrinsic(Int, Ty), Ops, "vqshlu_n", 1, false);
   case NEON::BI__builtin_neon_vqshrun_n_v:
-    // FIXME: as above
     Int = Intrinsic::aarch64_neon_sqshrun;
     return EmitNeonCall(CGM.getIntrinsic(Int, Ty), Ops, "vqshrun_n");
   case NEON::BI__builtin_neon_vqrshrun_n_v:
-    // FIXME: and again.
     Int = Intrinsic::aarch64_neon_sqrshrun;
     return EmitNeonCall(CGM.getIntrinsic(Int, Ty), Ops, "vqrshrun_n");
   case NEON::BI__builtin_neon_vqshrn_n_v:
-    // FIXME: guess
     Int = usgn ? Intrinsic::aarch64_neon_uqshrn : Intrinsic::aarch64_neon_sqshrn;
     return EmitNeonCall(CGM.getIntrinsic(Int, Ty), Ops, "vqshrn_n");
   case NEON::BI__builtin_neon_vrshrn_n_v:
-    // FIXME: there might be a pattern here.
     Int = Intrinsic::aarch64_neon_rshrn;
     return EmitNeonCall(CGM.getIntrinsic(Int, Ty), Ops, "vrshrn_n");
   case NEON::BI__builtin_neon_vqrshrn_n_v:
-    // FIXME: another one
     Int = usgn ? Intrinsic::aarch64_neon_uqrshrn : Intrinsic::aarch64_neon_sqrshrn;
     return EmitNeonCall(CGM.getIntrinsic(Int, Ty), Ops, "vqrshrn_n");
   case NEON::BI__builtin_neon_vrnda_v:
   case NEON::BI__builtin_neon_vrndaq_v: {
     Int = Intrinsic::round;
     return EmitNeonCall(CGM.getIntrinsic(Int, Ty), Ops, "vrnda");
   }
   case NEON::BI__builtin_neon_vrndi_v: