Use vAny type to get rid of Neon intrinsics that differed only in whether

the overloaded vector types allowed floating-point or integer vector elements.
Most of these operations actually depend on the element type, so bitcasting
was not an option.

If you include the vpadd intrinsics that I updated earlier, this gets rid
of 20 intrinsics.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@78646 91177308-0d34-0410-b5e6-96231b3b80d8

1 files changed, 44 insertions, 90 deletions

diff --git a/include/llvm/IntrinsicsARM.td b/include/llvm/IntrinsicsARM.td
index d86dd08710..7b72082763 100644
--- a/include/llvm/IntrinsicsARM.td
+++ b/include/llvm/IntrinsicsARM.td
@@ -27,53 +27,42 @@ let TargetPrefix = "arm" in {  // All intrinsics start with "llvm.arm.".
 
   // The following classes do not correspond directly to GCC builtins.
   class Neon_1Arg_Intrinsic
-    : Intrinsic<[llvm_anyint_ty], [LLVMMatchType<0>], [IntrNoMem]>;
-  class Neon_1Arg_Float_Intrinsic
-    : Intrinsic<[llvm_anyfloat_ty], [LLVMMatchType<0>], [IntrNoMem]>;
+    : Intrinsic<[llvm_anyvector_ty], [LLVMMatchType<0>], [IntrNoMem]>;
   class Neon_1Arg_Narrow_Intrinsic
-    : Intrinsic<[llvm_anyint_ty],
+    : Intrinsic<[llvm_anyvector_ty],
                 [LLVMExtendedElementVectorType<0>], [IntrNoMem]>;
   class Neon_1Arg_Long_Intrinsic
-    : Intrinsic<[llvm_anyint_ty],
+    : Intrinsic<[llvm_anyvector_ty],
                 [LLVMTruncatedElementVectorType<0>], [IntrNoMem]>;
   class Neon_2Arg_Intrinsic
-    : Intrinsic<[llvm_anyint_ty], [LLVMMatchType<0>, LLVMMatchType<0>],
-                [IntrNoMem]>;
-  class Neon_2Arg_Float_Intrinsic
-    : Intrinsic<[llvm_anyfloat_ty], [LLVMMatchType<0>, LLVMMatchType<0>],
-                [IntrNoMem]>;
-  class Neon_2Arg_Vector_Intrinsic
     : Intrinsic<[llvm_anyvector_ty], [LLVMMatchType<0>, LLVMMatchType<0>],
                 [IntrNoMem]>;
   class Neon_2Arg_Narrow_Intrinsic
-    : Intrinsic<[llvm_anyint_ty],
+    : Intrinsic<[llvm_anyvector_ty],
                 [LLVMExtendedElementVectorType<0>,
                  LLVMExtendedElementVectorType<0>],
                 [IntrNoMem]>;
   class Neon_2Arg_Long_Intrinsic
-    : Intrinsic<[llvm_anyint_ty],
+    : Intrinsic<[llvm_anyvector_ty],
                 [LLVMTruncatedElementVectorType<0>,
                  LLVMTruncatedElementVectorType<0>],
                 [IntrNoMem]>;
   class Neon_2Arg_Wide_Intrinsic
-    : Intrinsic<[llvm_anyint_ty],
+    : Intrinsic<[llvm_anyvector_ty],
                 [LLVMMatchType<0>, LLVMTruncatedElementVectorType<0>],
                 [IntrNoMem]>;
   class Neon_3Arg_Intrinsic
-    : Intrinsic<[llvm_anyint_ty],
+    : Intrinsic<[llvm_anyvector_ty],
                 [LLVMMatchType<0>, LLVMMatchType<0>, LLVMMatchType<0>],
                 [IntrNoMem]>;
   class Neon_3Arg_Long_Intrinsic
-    : Intrinsic<[llvm_anyint_ty],
+    : Intrinsic<[llvm_anyvector_ty],
                 [LLVMMatchType<0>,
                  LLVMTruncatedElementVectorType<0>,
                  LLVMTruncatedElementVectorType<0>],
                 [IntrNoMem]>;
   class Neon_2Result_Intrinsic
-    : Intrinsic<[llvm_anyint_ty, LLVMMatchType<0>],
-                [LLVMMatchType<0>, LLVMMatchType<0>], [IntrNoMem]>;
-  class Neon_2Result_Float_Intrinsic
-    : Intrinsic<[llvm_anyfloat_ty, LLVMMatchType<0>],
+    : Intrinsic<[llvm_anyvector_ty, LLVMMatchType<0>],
                 [LLVMMatchType<0>, LLVMMatchType<0>], [IntrNoMem]>;
   class Neon_CvtFxToFP_Intrinsic
     : Intrinsic<[llvm_anyfloat_ty], [llvm_anyint_ty, llvm_i32_ty], [IntrNoMem]>;
@@ -141,18 +130,16 @@ let Properties = [IntrNoMem, Commutative] in {
   // Vector Maximum.
   def int_arm_neon_vmaxs : Neon_2Arg_Intrinsic;
   def int_arm_neon_vmaxu : Neon_2Arg_Intrinsic;
-  def int_arm_neon_vmaxf : Neon_2Arg_Float_Intrinsic;
 
   // Vector Minimum.
   def int_arm_neon_vmins : Neon_2Arg_Intrinsic;
   def int_arm_neon_vminu : Neon_2Arg_Intrinsic;
-  def int_arm_neon_vminf : Neon_2Arg_Float_Intrinsic;
 
   // Vector Reciprocal Step.
-  def int_arm_neon_vrecps : Neon_2Arg_Float_Intrinsic;
+  def int_arm_neon_vrecps : Neon_2Arg_Intrinsic;
 
   // Vector Reciprocal Square Root Step.
-  def int_arm_neon_vrsqrts : Neon_2Arg_Float_Intrinsic;
+  def int_arm_neon_vrsqrts : Neon_2Arg_Intrinsic;
 }
 
 // Vector Subtract.
@@ -186,7 +173,6 @@ let TargetPrefix = "arm" in {
 // Vector Absolute Differences.
 def int_arm_neon_vabds : Neon_2Arg_Intrinsic;
 def int_arm_neon_vabdu : Neon_2Arg_Intrinsic;
-def int_arm_neon_vabdf : Neon_2Arg_Float_Intrinsic;
 def int_arm_neon_vabdls : Neon_2Arg_Long_Intrinsic;
 def int_arm_neon_vabdlu : Neon_2Arg_Long_Intrinsic;
 
@@ -197,16 +183,16 @@ def int_arm_neon_vabals : Neon_3Arg_Long_Intrinsic;
 def int_arm_neon_vabalu : Neon_3Arg_Long_Intrinsic;
 
 // Vector Pairwise Add.
-def int_arm_neon_vpadd : Neon_2Arg_Vector_Intrinsic;
+def int_arm_neon_vpadd : Neon_2Arg_Intrinsic;
 
 // Vector Pairwise Add Long.
 // Note: This is different than the other "long" NEON intrinsics because
 // the result vector has half as many elements as the source vector.
 // The source and destination vector types must be specified separately.
 let TargetPrefix = "arm" in {
-  def int_arm_neon_vpaddls : Intrinsic<[llvm_anyint_ty], [llvm_anyint_ty],
+  def int_arm_neon_vpaddls : Intrinsic<[llvm_anyvector_ty], [llvm_anyvector_ty],
                                        [IntrNoMem]>;
-  def int_arm_neon_vpaddlu : Intrinsic<[llvm_anyint_ty], [llvm_anyint_ty],
+  def int_arm_neon_vpaddlu : Intrinsic<[llvm_anyvector_ty], [llvm_anyvector_ty],
                                        [IntrNoMem]>;
 }
 
@@ -214,21 +200,19 @@ let TargetPrefix = "arm" in {
 // Note: This is similar to vpaddl but the destination vector also appears
 // as the first argument.
 let TargetPrefix = "arm" in {
-  def int_arm_neon_vpadals : Intrinsic<[llvm_anyint_ty],
-                                       [LLVMMatchType<0>, llvm_anyint_ty],
+  def int_arm_neon_vpadals : Intrinsic<[llvm_anyvector_ty],
+                                       [LLVMMatchType<0>, llvm_anyvector_ty],
                                        [IntrNoMem]>;
-  def int_arm_neon_vpadalu : Intrinsic<[llvm_anyint_ty],
-                                       [LLVMMatchType<0>, llvm_anyint_ty],
+  def int_arm_neon_vpadalu : Intrinsic<[llvm_anyvector_ty],
+                                       [LLVMMatchType<0>, llvm_anyvector_ty],
                                        [IntrNoMem]>;
 }
 
 // Vector Pairwise Maximum and Minimum.
 def int_arm_neon_vpmaxs : Neon_2Arg_Intrinsic;
 def int_arm_neon_vpmaxu : Neon_2Arg_Intrinsic;
-def int_arm_neon_vpmaxf : Neon_2Arg_Float_Intrinsic;
 def int_arm_neon_vpmins : Neon_2Arg_Intrinsic;
 def int_arm_neon_vpminu : Neon_2Arg_Intrinsic;
-def int_arm_neon_vpminf : Neon_2Arg_Float_Intrinsic;
 
 // Vector Shifts:
 //
@@ -283,7 +267,6 @@ def int_arm_neon_vshiftins : Neon_3Arg_Intrinsic;
 
 // Vector Absolute Value and Saturating Absolute Value.
 def int_arm_neon_vabs : Neon_1Arg_Intrinsic;
-def int_arm_neon_vabsf : Neon_1Arg_Float_Intrinsic;
 def int_arm_neon_vqabs : Neon_1Arg_Intrinsic;
 
 // Vector Saturating Negate.
@@ -298,11 +281,9 @@ def int_arm_neon_vcnt : Neon_1Arg_Intrinsic;
 
 // Vector Reciprocal Estimate.
 def int_arm_neon_vrecpe : Neon_1Arg_Intrinsic;
-def int_arm_neon_vrecpef : Neon_1Arg_Float_Intrinsic;
 
 // Vector Reciprocal Square Root Estimate.
 def int_arm_neon_vrsqrte : Neon_1Arg_Intrinsic;
-def int_arm_neon_vrsqrtef : Neon_1Arg_Float_Intrinsic;
 
 // Vector Conversions Between Floating-point and Fixed-point.
 def int_arm_neon_vcvtfp2fxs : Neon_CvtFPToFx_Intrinsic;
@@ -331,68 +312,41 @@ def int_arm_neon_vtbx3 : Neon_Tbl5Arg_Intrinsic;
 def int_arm_neon_vtbx4 : Neon_Tbl6Arg_Intrinsic;
 
 // Vector Transpose.
-def int_arm_neon_vtrni : Neon_2Result_Intrinsic;
-def int_arm_neon_vtrnf : Neon_2Result_Float_Intrinsic;
+def int_arm_neon_vtrn : Neon_2Result_Intrinsic;
 
 // Vector Interleave (vzip).
-def int_arm_neon_vzipi : Neon_2Result_Intrinsic;
-def int_arm_neon_vzipf : Neon_2Result_Float_Intrinsic;
+def int_arm_neon_vzip : Neon_2Result_Intrinsic;
 
 // Vector Deinterleave (vuzp).
-def int_arm_neon_vuzpi : Neon_2Result_Intrinsic;
-def int_arm_neon_vuzpf : Neon_2Result_Float_Intrinsic;
+def int_arm_neon_vuzp : Neon_2Result_Intrinsic;
 
 let TargetPrefix = "arm" in {
 
   // De-interleaving vector loads from N-element structures.
-  def int_arm_neon_vld1i : Intrinsic<[llvm_anyint_ty],
-                                     [llvm_ptr_ty], [IntrReadArgMem]>;
-  def int_arm_neon_vld1f : Intrinsic<[llvm_anyfloat_ty],
-                                     [llvm_ptr_ty], [IntrReadArgMem]>;
-  def int_arm_neon_vld2i : Intrinsic<[llvm_anyint_ty, LLVMMatchType<0>],
-                                     [llvm_ptr_ty], [IntrReadArgMem]>;
-  def int_arm_neon_vld2f : Intrinsic<[llvm_anyfloat_ty, LLVMMatchType<0>],
-                                     [llvm_ptr_ty], [IntrReadArgMem]>;
-  def int_arm_neon_vld3i : Intrinsic<[llvm_anyint_ty, LLVMMatchType<0>,
-                                      LLVMMatchType<0>],
-                                     [llvm_ptr_ty], [IntrReadArgMem]>;
-  def int_arm_neon_vld3f : Intrinsic<[llvm_anyfloat_ty, LLVMMatchType<0>,
-                                      LLVMMatchType<0>],
-                                     [llvm_ptr_ty], [IntrReadArgMem]>;
-  def int_arm_neon_vld4i : Intrinsic<[llvm_anyint_ty, LLVMMatchType<0>,
-                                      LLVMMatchType<0>, LLVMMatchType<0>],
-                                     [llvm_ptr_ty], [IntrReadArgMem]>;
-  def int_arm_neon_vld4f : Intrinsic<[llvm_anyfloat_ty, LLVMMatchType<0>,
-                                      LLVMMatchType<0>, LLVMMatchType<0>],
-                                     [llvm_ptr_ty], [IntrReadArgMem]>;
+  def int_arm_neon_vld1 : Intrinsic<[llvm_anyvector_ty],
+                                    [llvm_ptr_ty], [IntrReadArgMem]>;
+  def int_arm_neon_vld2 : Intrinsic<[llvm_anyvector_ty, LLVMMatchType<0>],
+                                    [llvm_ptr_ty], [IntrReadArgMem]>;
+  def int_arm_neon_vld3 : Intrinsic<[llvm_anyvector_ty, LLVMMatchType<0>,
+                                     LLVMMatchType<0>],
+                                    [llvm_ptr_ty], [IntrReadArgMem]>;
+  def int_arm_neon_vld4 : Intrinsic<[llvm_anyvector_ty, LLVMMatchType<0>,
+                                     LLVMMatchType<0>, LLVMMatchType<0>],
+                                    [llvm_ptr_ty], [IntrReadArgMem]>;
 
   // Interleaving vector stores from N-element structures.
-  def int_arm_neon_vst1i : Intrinsic<[llvm_void_ty],
-                                     [llvm_ptr_ty, llvm_anyint_ty],
-                                     [IntrWriteArgMem]>;
-  def int_arm_neon_vst1f : Intrinsic<[llvm_void_ty],
-                                     [llvm_ptr_ty, llvm_anyfloat_ty],
+  def int_arm_neon_vst1 : Intrinsic<[llvm_void_ty],
+                                    [llvm_ptr_ty, llvm_anyvector_ty],
+                                    [IntrWriteArgMem]>;
+  def int_arm_neon_vst2 : Intrinsic<[llvm_void_ty],
+                                    [llvm_ptr_ty, llvm_anyvector_ty,
+                                     LLVMMatchType<0>], [IntrWriteArgMem]>;
+  def int_arm_neon_vst3 : Intrinsic<[llvm_void_ty],
+                                    [llvm_ptr_ty, llvm_anyvector_ty,
+                                     LLVMMatchType<0>, LLVMMatchType<0>],
                                      [IntrWriteArgMem]>;
-  def int_arm_neon_vst2i : Intrinsic<[llvm_void_ty],
-                                     [llvm_ptr_ty, llvm_anyint_ty,
-                                      LLVMMatchType<0>], [IntrWriteArgMem]>;
-  def int_arm_neon_vst2f : Intrinsic<[llvm_void_ty],
-                                     [llvm_ptr_ty, llvm_anyfloat_ty,
-                                      LLVMMatchType<0>], [IntrWriteArgMem]>;
-  def int_arm_neon_vst3i : Intrinsic<[llvm_void_ty],
-                                     [llvm_ptr_ty, llvm_anyint_ty,
-                                      LLVMMatchType<0>, LLVMMatchType<0>],
-                                      [IntrWriteArgMem]>;
-  def int_arm_neon_vst3f : Intrinsic<[llvm_void_ty],
-                                     [llvm_ptr_ty, llvm_anyfloat_ty,
-                                      LLVMMatchType<0>, LLVMMatchType<0>],
-                                      [IntrWriteArgMem]>;
-  def int_arm_neon_vst4i : Intrinsic<[llvm_void_ty],
-                                     [llvm_ptr_ty, llvm_anyint_ty,
-                                      LLVMMatchType<0>, LLVMMatchType<0>,
-                                      LLVMMatchType<0>], [IntrWriteArgMem]>;
-  def int_arm_neon_vst4f : Intrinsic<[llvm_void_ty],
-                                     [llvm_ptr_ty, llvm_anyfloat_ty,
-                                      LLVMMatchType<0>, LLVMMatchType<0>,
-                                      LLVMMatchType<0>], [IntrWriteArgMem]>;
+  def int_arm_neon_vst4 : Intrinsic<[llvm_void_ty],
+                                    [llvm_ptr_ty, llvm_anyvector_ty,
+                                     LLVMMatchType<0>, LLVMMatchType<0>,
+                                     LLVMMatchType<0>], [IntrWriteArgMem]>;
 }