[mips][msa] Corrected the definition of the dotp_[su].[hwd] intrinsics

The elements of the operands should be half the width of the elements of
the result.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190505 91177308-0d34-0410-b5e6-96231b3b80d8

3 files changed, 105 insertions, 71 deletions

diff --git a/include/llvm/IR/IntrinsicsMips.td b/include/llvm/IR/IntrinsicsMips.td
index b4872d23d75..5bec80294c8 100644
--- a/include/llvm/IR/IntrinsicsMips.td
+++ b/include/llvm/IR/IntrinsicsMips.td
@@ -814,18 +814,18 @@ def int_mips_div_u_d : GCCBuiltin<"__builtin_msa_div_u_d">,
   Intrinsic<[llvm_v2i64_ty], [llvm_v2i64_ty, llvm_v2i64_ty], [IntrNoMem]>;
 
 def int_mips_dotp_s_h : GCCBuiltin<"__builtin_msa_dotp_s_h">,
-  Intrinsic<[llvm_v8i16_ty], [llvm_v8i16_ty, llvm_v8i16_ty], [IntrNoMem]>;
+  Intrinsic<[llvm_v8i16_ty], [llvm_v16i8_ty, llvm_v16i8_ty], [IntrNoMem]>;
 def int_mips_dotp_s_w : GCCBuiltin<"__builtin_msa_dotp_s_w">,
-  Intrinsic<[llvm_v4i32_ty], [llvm_v4i32_ty, llvm_v4i32_ty], [IntrNoMem]>;
+  Intrinsic<[llvm_v4i32_ty], [llvm_v8i16_ty, llvm_v8i16_ty], [IntrNoMem]>;
 def int_mips_dotp_s_d : GCCBuiltin<"__builtin_msa_dotp_s_d">,
-  Intrinsic<[llvm_v2i64_ty], [llvm_v2i64_ty, llvm_v2i64_ty], [IntrNoMem]>;
+  Intrinsic<[llvm_v2i64_ty], [llvm_v4i32_ty, llvm_v4i32_ty], [IntrNoMem]>;
 
 def int_mips_dotp_u_h : GCCBuiltin<"__builtin_msa_dotp_u_h">,
-  Intrinsic<[llvm_v8i16_ty], [llvm_v8i16_ty, llvm_v8i16_ty], [IntrNoMem]>;
+  Intrinsic<[llvm_v8i16_ty], [llvm_v16i8_ty, llvm_v16i8_ty], [IntrNoMem]>;
 def int_mips_dotp_u_w : GCCBuiltin<"__builtin_msa_dotp_u_w">,
-  Intrinsic<[llvm_v4i32_ty], [llvm_v4i32_ty, llvm_v4i32_ty], [IntrNoMem]>;
+  Intrinsic<[llvm_v4i32_ty], [llvm_v8i16_ty, llvm_v8i16_ty], [IntrNoMem]>;
 def int_mips_dotp_u_d : GCCBuiltin<"__builtin_msa_dotp_u_d">,
-  Intrinsic<[llvm_v2i64_ty], [llvm_v2i64_ty, llvm_v2i64_ty], [IntrNoMem]>;
+  Intrinsic<[llvm_v2i64_ty], [llvm_v4i32_ty, llvm_v4i32_ty], [IntrNoMem]>;
 
 def int_mips_dpadd_s_h : GCCBuiltin<"__builtin_msa_dpadd_s_h">,
   Intrinsic<[llvm_v8i16_ty], [llvm_v8i16_ty, llvm_v16i8_ty, llvm_v16i8_ty],
diff --git a/lib/Target/Mips/MipsMSAInstrInfo.td b/lib/Target/Mips/MipsMSAInstrInfo.td
index a7d47908bf2..66d7f37a71d 100644
--- a/lib/Target/Mips/MipsMSAInstrInfo.td
+++ b/lib/Target/Mips/MipsMSAInstrInfo.td
@@ -1238,19 +1238,19 @@ class DIV_U_H_DESC : MSA_3R_DESC_BASE<"div_u.h", int_mips_div_u_h, MSA128H>;
 class DIV_U_W_DESC : MSA_3R_DESC_BASE<"div_u.w", int_mips_div_u_w, MSA128W>;
 class DIV_U_D_DESC : MSA_3R_DESC_BASE<"div_u.d", int_mips_div_u_d, MSA128D>;
 
-class DOTP_S_H_DESC : MSA_3R_DESC_BASE<"dotp_s.h", int_mips_dotp_s_h, MSA128H>,
-                      IsCommutable;
-class DOTP_S_W_DESC : MSA_3R_DESC_BASE<"dotp_s.w", int_mips_dotp_s_w, MSA128W>,
-                      IsCommutable;
-class DOTP_S_D_DESC : MSA_3R_DESC_BASE<"dotp_s.d", int_mips_dotp_s_d, MSA128D>,
-                      IsCommutable;
-
-class DOTP_U_H_DESC : MSA_3R_DESC_BASE<"dotp_u.h", int_mips_dotp_u_h, MSA128H>,
-                      IsCommutable;
-class DOTP_U_W_DESC : MSA_3R_DESC_BASE<"dotp_u.w", int_mips_dotp_u_w, MSA128W>,
-                      IsCommutable;
-class DOTP_U_D_DESC : MSA_3R_DESC_BASE<"dotp_u.d", int_mips_dotp_u_d, MSA128D>,
-                      IsCommutable;
+class DOTP_S_H_DESC : MSA_3R_DESC_BASE<"dotp_s.h", int_mips_dotp_s_h, MSA128H,
+                                       MSA128B, MSA128B>, IsCommutable;
+class DOTP_S_W_DESC : MSA_3R_DESC_BASE<"dotp_s.w", int_mips_dotp_s_w, MSA128W,
+                                       MSA128H, MSA128H>, IsCommutable;
+class DOTP_S_D_DESC : MSA_3R_DESC_BASE<"dotp_s.d", int_mips_dotp_s_d, MSA128D,
+                                       MSA128W, MSA128W>, IsCommutable;
+
+class DOTP_U_H_DESC : MSA_3R_DESC_BASE<"dotp_u.h", int_mips_dotp_u_h, MSA128H,
+                                       MSA128B, MSA128B>, IsCommutable;
+class DOTP_U_W_DESC : MSA_3R_DESC_BASE<"dotp_u.w", int_mips_dotp_u_w, MSA128W,
+                                       MSA128H, MSA128H>, IsCommutable;
+class DOTP_U_D_DESC : MSA_3R_DESC_BASE<"dotp_u.d", int_mips_dotp_u_d, MSA128D,
+                                       MSA128W, MSA128W>, IsCommutable;
 
 class DPADD_S_H_DESC : MSA_3R_4R_DESC_BASE<"dpadd_s.h", int_mips_dpadd_s_h,
                                            MSA128H, MSA128B, MSA128B>,
diff --git a/test/CodeGen/Mips/msa/3r-d.ll b/test/CodeGen/Mips/msa/3r-d.ll
index 5edf8ca983e..8d4d3d4e6e9 100644
--- a/test/CodeGen/Mips/msa/3r-d.ll
+++ b/test/CodeGen/Mips/msa/3r-d.ll
@@ -179,134 +179,168 @@ declare <2 x i64> @llvm.mips.div.u.d(<2 x i64>, <2 x i64>) nounwind
 ; CHECK: st.d
 ; CHECK: .size llvm_mips_div_u_d_test
 ;
-@llvm_mips_dotp_s_h_ARG1 = global <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>, align 16
-@llvm_mips_dotp_s_h_ARG2 = global <8 x i16> <i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15>, align 16
-@llvm_mips_dotp_s_h_RES  = global <8 x i16> <i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0>, align 16
+@llvm_mips_dotp_s_h_ARG1 = global <16 x i8> <i8  0, i8  1, i8  2, i8  3,
+                                             i8  4, i8  5, i8  6, i8  7,
+                                             i8  8, i8  9, i8 10, i8 11,
+                                             i8 12, i8 13, i8 14, i8 15>,
+                                            align 16
+@llvm_mips_dotp_s_h_ARG2 = global <16 x i8> <i8 16, i8 17, i8 18, i8 19,
+                                             i8 20, i8 21, i8 22, i8 23,
+                                             i8 24, i8 25, i8 26, i8 27,
+                                             i8 28, i8 29, i8 30, i8 31>,
+                                            align 16
+@llvm_mips_dotp_s_h_RES  = global <8 x i16> <i16 0, i16 0, i16 0, i16 0,
+                                             i16 0, i16 0, i16 0, i16 0>,
+                                            align 16
 
 define void @llvm_mips_dotp_s_h_test() nounwind {
 entry:
-  %0 = load <8 x i16>* @llvm_mips_dotp_s_h_ARG1
-  %1 = load <8 x i16>* @llvm_mips_dotp_s_h_ARG2
-  %2 = tail call <8 x i16> @llvm.mips.dotp.s.h(<8 x i16> %0, <8 x i16> %1)
+  %0 = load <16 x i8>* @llvm_mips_dotp_s_h_ARG1
+  %1 = load <16 x i8>* @llvm_mips_dotp_s_h_ARG2
+  %2 = tail call <8 x i16> @llvm.mips.dotp.s.h(<16 x i8> %0, <16 x i8> %1)
   store <8 x i16> %2, <8 x i16>* @llvm_mips_dotp_s_h_RES
   ret void
 }
 
-declare <8 x i16> @llvm.mips.dotp.s.h(<8 x i16>, <8 x i16>) nounwind
+declare <8 x i16> @llvm.mips.dotp.s.h(<16 x i8>, <16 x i8>) nounwind
 
 ; CHECK: llvm_mips_dotp_s_h_test:
-; CHECK: ld.h
-; CHECK: ld.h
+; CHECK: ld.b
+; CHECK: ld.b
 ; CHECK: dotp_s.h
 ; CHECK: st.h
 ; CHECK: .size llvm_mips_dotp_s_h_test
 ;
-@llvm_mips_dotp_s_w_ARG1 = global <4 x i32> <i32 0, i32 1, i32 2, i32 3>, align 16
-@llvm_mips_dotp_s_w_ARG2 = global <4 x i32> <i32 4, i32 5, i32 6, i32 7>, align 16
-@llvm_mips_dotp_s_w_RES  = global <4 x i32> <i32 0, i32 0, i32 0, i32 0>, align 16
+@llvm_mips_dotp_s_w_ARG1 = global <8 x i16> <i16 0, i16 1, i16 2, i16 3,
+                                             i16 4, i16 5, i16 6, i16 7>,
+                                            align 16
+@llvm_mips_dotp_s_w_ARG2 = global <8 x i16> <i16  4, i16  5, i16  6, i16  7,
+                                             i16  8, i16  9, i16 10, i16 11>,
+                                            align 16
+@llvm_mips_dotp_s_w_RES  = global <4 x i32> <i32 0, i32 0, i32 0, i32 0>,
+                                            align 16
 
 define void @llvm_mips_dotp_s_w_test() nounwind {
 entry:
-  %0 = load <4 x i32>* @llvm_mips_dotp_s_w_ARG1
-  %1 = load <4 x i32>* @llvm_mips_dotp_s_w_ARG2
-  %2 = tail call <4 x i32> @llvm.mips.dotp.s.w(<4 x i32> %0, <4 x i32> %1)
+  %0 = load <8 x i16>* @llvm_mips_dotp_s_w_ARG1
+  %1 = load <8 x i16>* @llvm_mips_dotp_s_w_ARG2
+  %2 = tail call <4 x i32> @llvm.mips.dotp.s.w(<8 x i16> %0, <8 x i16> %1)
   store <4 x i32> %2, <4 x i32>* @llvm_mips_dotp_s_w_RES
   ret void
 }
 
-declare <4 x i32> @llvm.mips.dotp.s.w(<4 x i32>, <4 x i32>) nounwind
+declare <4 x i32> @llvm.mips.dotp.s.w(<8 x i16>, <8 x i16>) nounwind
 
 ; CHECK: llvm_mips_dotp_s_w_test:
-; CHECK: ld.w
-; CHECK: ld.w
+; CHECK: ld.h
+; CHECK: ld.h
 ; CHECK: dotp_s.w
 ; CHECK: st.w
 ; CHECK: .size llvm_mips_dotp_s_w_test
 ;
-@llvm_mips_dotp_s_d_ARG1 = global <2 x i64> <i64 0, i64 1>, align 16
-@llvm_mips_dotp_s_d_ARG2 = global <2 x i64> <i64 2, i64 3>, align 16
+@llvm_mips_dotp_s_d_ARG1 = global <4 x i32> <i32 0, i32 1, i32 0, i32 1>,
+                                            align 16
+@llvm_mips_dotp_s_d_ARG2 = global <4 x i32> <i32 2, i32 3, i32 2, i32 3>,
+                                            align 16
 @llvm_mips_dotp_s_d_RES  = global <2 x i64> <i64 0, i64 0>, align 16
 
 define void @llvm_mips_dotp_s_d_test() nounwind {
 entry:
-  %0 = load <2 x i64>* @llvm_mips_dotp_s_d_ARG1
-  %1 = load <2 x i64>* @llvm_mips_dotp_s_d_ARG2
-  %2 = tail call <2 x i64> @llvm.mips.dotp.s.d(<2 x i64> %0, <2 x i64> %1)
+  %0 = load <4 x i32>* @llvm_mips_dotp_s_d_ARG1
+  %1 = load <4 x i32>* @llvm_mips_dotp_s_d_ARG2
+  %2 = tail call <2 x i64> @llvm.mips.dotp.s.d(<4 x i32> %0, <4 x i32> %1)
   store <2 x i64> %2, <2 x i64>* @llvm_mips_dotp_s_d_RES
   ret void
 }
 
-declare <2 x i64> @llvm.mips.dotp.s.d(<2 x i64>, <2 x i64>) nounwind
+declare <2 x i64> @llvm.mips.dotp.s.d(<4 x i32>, <4 x i32>) nounwind
 
 ; CHECK: llvm_mips_dotp_s_d_test:
-; CHECK: ld.d
-; CHECK: ld.d
+; CHECK: ld.w
+; CHECK: ld.w
 ; CHECK: dotp_s.d
 ; CHECK: st.d
 ; CHECK: .size llvm_mips_dotp_s_d_test
 ;
-@llvm_mips_dotp_u_h_ARG1 = global <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>, align 16
-@llvm_mips_dotp_u_h_ARG2 = global <8 x i16> <i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15>, align 16
-@llvm_mips_dotp_u_h_RES  = global <8 x i16> <i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0>, align 16
+@llvm_mips_dotp_u_h_ARG1 = global <16 x i8> <i8  0, i8  1, i8  2, i8  3,
+                                             i8  4, i8  5, i8  6, i8  7,
+                                             i8  8, i8  9, i8 10, i8 11,
+                                             i8 12, i8 13, i8 14, i8 15>,
+                                            align 16
+@llvm_mips_dotp_u_h_ARG2 = global <16 x i8> <i8 16, i8 17, i8 18, i8 19,
+                                             i8 20, i8 21, i8 22, i8 23,
+                                             i8 24, i8 25, i8 26, i8 27,
+                                             i8 28, i8 29, i8 30, i8 31>,
+                                            align 16
+@llvm_mips_dotp_u_h_RES  = global <8 x i16> <i16 0, i16 0, i16 0, i16 0,
+                                             i16 0, i16 0, i16 0, i16 0>,
+                                            align 16
 
 define void @llvm_mips_dotp_u_h_test() nounwind {
 entry:
-  %0 = load <8 x i16>* @llvm_mips_dotp_u_h_ARG1
-  %1 = load <8 x i16>* @llvm_mips_dotp_u_h_ARG2
-  %2 = tail call <8 x i16> @llvm.mips.dotp.u.h(<8 x i16> %0, <8 x i16> %1)
+  %0 = load <16 x i8>* @llvm_mips_dotp_u_h_ARG1
+  %1 = load <16 x i8>* @llvm_mips_dotp_u_h_ARG2
+  %2 = tail call <8 x i16> @llvm.mips.dotp.u.h(<16 x i8> %0, <16 x i8> %1)
   store <8 x i16> %2, <8 x i16>* @llvm_mips_dotp_u_h_RES
   ret void
 }
 
-declare <8 x i16> @llvm.mips.dotp.u.h(<8 x i16>, <8 x i16>) nounwind
+declare <8 x i16> @llvm.mips.dotp.u.h(<16 x i8>, <16 x i8>) nounwind
 
 ; CHECK: llvm_mips_dotp_u_h_test:
-; CHECK: ld.h
-; CHECK: ld.h
+; CHECK: ld.b
+; CHECK: ld.b
 ; CHECK: dotp_u.h
 ; CHECK: st.h
 ; CHECK: .size llvm_mips_dotp_u_h_test
 ;
-@llvm_mips_dotp_u_w_ARG1 = global <4 x i32> <i32 0, i32 1, i32 2, i32 3>, align 16
-@llvm_mips_dotp_u_w_ARG2 = global <4 x i32> <i32 4, i32 5, i32 6, i32 7>, align 16
-@llvm_mips_dotp_u_w_RES  = global <4 x i32> <i32 0, i32 0, i32 0, i32 0>, align 16
+@llvm_mips_dotp_u_w_ARG1 = global <8 x i16> <i16 0, i16 1, i16 2, i16 3,
+                                             i16 4, i16 5, i16 6, i16 7>,
+                                            align 16
+@llvm_mips_dotp_u_w_ARG2 = global <8 x i16> <i16  4, i16  5, i16  6, i16  7,
+                                             i16  8, i16  9, i16 10, i16 11>,
+                                            align 16
+@llvm_mips_dotp_u_w_RES  = global <4 x i32> <i32 0, i32 0, i32 0, i32 0>,
+                                            align 16
 
 define void @llvm_mips_dotp_u_w_test() nounwind {
 entry:
-  %0 = load <4 x i32>* @llvm_mips_dotp_u_w_ARG1
-  %1 = load <4 x i32>* @llvm_mips_dotp_u_w_ARG2
-  %2 = tail call <4 x i32> @llvm.mips.dotp.u.w(<4 x i32> %0, <4 x i32> %1)
+  %0 = load <8 x i16>* @llvm_mips_dotp_u_w_ARG1
+  %1 = load <8 x i16>* @llvm_mips_dotp_u_w_ARG2
+  %2 = tail call <4 x i32> @llvm.mips.dotp.u.w(<8 x i16> %0, <8 x i16> %1)
   store <4 x i32> %2, <4 x i32>* @llvm_mips_dotp_u_w_RES
   ret void
 }
 
-declare <4 x i32> @llvm.mips.dotp.u.w(<4 x i32>, <4 x i32>) nounwind
+declare <4 x i32> @llvm.mips.dotp.u.w(<8 x i16>, <8 x i16>) nounwind
 
 ; CHECK: llvm_mips_dotp_u_w_test:
-; CHECK: ld.w
-; CHECK: ld.w
+; CHECK: ld.h
+; CHECK: ld.h
 ; CHECK: dotp_u.w
 ; CHECK: st.w
 ; CHECK: .size llvm_mips_dotp_u_w_test
 ;
-@llvm_mips_dotp_u_d_ARG1 = global <2 x i64> <i64 0, i64 1>, align 16
-@llvm_mips_dotp_u_d_ARG2 = global <2 x i64> <i64 2, i64 3>, align 16
+@llvm_mips_dotp_u_d_ARG1 = global <4 x i32> <i32 0, i32 1, i32 0, i32 1>,
+                                            align 16
+@llvm_mips_dotp_u_d_ARG2 = global <4 x i32> <i32 2, i32 3, i32 2, i32 3>,
+                                            align 16
 @llvm_mips_dotp_u_d_RES  = global <2 x i64> <i64 0, i64 0>, align 16
 
 define void @llvm_mips_dotp_u_d_test() nounwind {
 entry:
-  %0 = load <2 x i64>* @llvm_mips_dotp_u_d_ARG1
-  %1 = load <2 x i64>* @llvm_mips_dotp_u_d_ARG2
-  %2 = tail call <2 x i64> @llvm.mips.dotp.u.d(<2 x i64> %0, <2 x i64> %1)
+  %0 = load <4 x i32>* @llvm_mips_dotp_u_d_ARG1
+  %1 = load <4 x i32>* @llvm_mips_dotp_u_d_ARG2
+  %2 = tail call <2 x i64> @llvm.mips.dotp.u.d(<4 x i32> %0, <4 x i32> %1)
   store <2 x i64> %2, <2 x i64>* @llvm_mips_dotp_u_d_RES
   ret void
 }
 
-declare <2 x i64> @llvm.mips.dotp.u.d(<2 x i64>, <2 x i64>) nounwind
+declare <2 x i64> @llvm.mips.dotp.u.d(<4 x i32>, <4 x i32>) nounwind
 
 ; CHECK: llvm_mips_dotp_u_d_test:
-; CHECK: ld.d
-; CHECK: ld.d
+; CHECK: ld.w
+; CHECK: ld.w
 ; CHECK: dotp_u.d
 ; CHECK: st.d
 ; CHECK: .size llvm_mips_dotp_u_d_test