[mips][msa] Added support for matching bsel and bseli from normal IR (i.e. not intrinsics)

This required correcting the definition of the bsel and bseli intrinsics.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191290 91177308-0d34-0410-b5e6-96231b3b80d8

4 files changed, 413 insertions, 7 deletions

diff --git a/test/CodeGen/Mips/msa/compare.ll b/test/CodeGen/Mips/msa/compare.ll
index 34e619b..fc83f44 100644
--- a/test/CodeGen/Mips/msa/compare.ll
+++ b/test/CodeGen/Mips/msa/compare.ll
@@ -639,3 +639,329 @@ define void @clti_u_v2i64(<2 x i64>* %c, <2 x i64>* %a) nounwind {
   ret void
   ; CHECK: .size clti_u_v2i64
 }
+
+define void @bsel_s_v16i8(<16 x i8>* %d, <16 x i8>* %a, <16 x i8>* %b,
+                        <16 x i8>* %c) nounwind {
+  ; CHECK: bsel_s_v16i8:
+
+  %1 = load <16 x i8>* %a
+  ; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
+  %2 = load <16 x i8>* %b
+  ; CHECK-DAG: ld.b [[R2:\$w[0-9]+]], 0($6)
+  %3 = load <16 x i8>* %c
+  ; CHECK-DAG: ld.b [[R3:\$w[0-9]+]], 0($7)
+  %4 = icmp sgt <16 x i8> %1, %2
+  ; CHECK-DAG: clt_s.b [[R4:\$w[0-9]+]], [[R2]], [[R1]]
+  %5 = select <16 x i1> %4, <16 x i8> %1, <16 x i8> %3
+  ; CHECK-DAG: bsel.v [[R4]], [[R1]], [[R3]]
+  store <16 x i8> %5, <16 x i8>* %d
+  ; CHECK-DAG: st.b [[R4]], 0($4)
+
+  ret void
+  ; CHECK: .size bsel_s_v16i8
+}
+
+define void @bsel_s_v8i16(<8 x i16>* %d, <8 x i16>* %a, <8 x i16>* %b,
+                        <8 x i16>* %c) nounwind {
+  ; CHECK: bsel_s_v8i16:
+
+  %1 = load <8 x i16>* %a
+  ; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
+  %2 = load <8 x i16>* %b
+  ; CHECK-DAG: ld.h [[R2:\$w[0-9]+]], 0($6)
+  %3 = load <8 x i16>* %c
+  ; CHECK-DAG: ld.h [[R3:\$w[0-9]+]], 0($7)
+  %4 = icmp sgt <8 x i16> %1, %2
+  ; CHECK-DAG: clt_s.h [[R4:\$w[0-9]+]], [[R2]], [[R1]]
+  %5 = select <8 x i1> %4, <8 x i16> %1, <8 x i16> %3
+  ; CHECK-DAG: bsel.v [[R4]], [[R1]], [[R3]]
+  store <8 x i16> %5, <8 x i16>* %d
+  ; CHECK-DAG: st.h [[R4]], 0($4)
+
+  ret void
+  ; CHECK: .size bsel_s_v8i16
+}
+
+define void @bsel_s_v4i32(<4 x i32>* %d, <4 x i32>* %a, <4 x i32>* %b,
+                        <4 x i32>* %c) nounwind {
+  ; CHECK: bsel_s_v4i32:
+
+  %1 = load <4 x i32>* %a
+  ; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
+  %2 = load <4 x i32>* %b
+  ; CHECK-DAG: ld.w [[R2:\$w[0-9]+]], 0($6)
+  %3 = load <4 x i32>* %c
+  ; CHECK-DAG: ld.w [[R3:\$w[0-9]+]], 0($7)
+  %4 = icmp sgt <4 x i32> %1, %2
+  ; CHECK-DAG: clt_s.w [[R4:\$w[0-9]+]], [[R2]], [[R1]]
+  %5 = select <4 x i1> %4, <4 x i32> %1, <4 x i32> %3
+  ; CHECK-DAG: bsel.v [[R4]], [[R1]], [[R3]]
+  store <4 x i32> %5, <4 x i32>* %d
+  ; CHECK-DAG: st.w [[R4]], 0($4)
+
+  ret void
+  ; CHECK: .size bsel_s_v4i32
+}
+
+define void @bsel_s_v2i64(<2 x i64>* %d, <2 x i64>* %a, <2 x i64>* %b,
+                        <2 x i64>* %c) nounwind {
+  ; CHECK: bsel_s_v2i64:
+
+  %1 = load <2 x i64>* %a
+  ; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
+  %2 = load <2 x i64>* %b
+  ; CHECK-DAG: ld.d [[R2:\$w[0-9]+]], 0($6)
+  %3 = load <2 x i64>* %c
+  ; CHECK-DAG: ld.d [[R3:\$w[0-9]+]], 0($7)
+  %4 = icmp sgt <2 x i64> %1, %2
+  ; CHECK-DAG: clt_s.d [[R4:\$w[0-9]+]], [[R2]], [[R1]]
+  %5 = select <2 x i1> %4, <2 x i64> %1, <2 x i64> %3
+  ; CHECK-DAG: bsel.v [[R4]], [[R1]], [[R3]]
+  store <2 x i64> %5, <2 x i64>* %d
+  ; CHECK-DAG: st.d [[R4]], 0($4)
+
+  ret void
+  ; CHECK: .size bsel_s_v2i64
+}
+
+define void @bsel_u_v16i8(<16 x i8>* %d, <16 x i8>* %a, <16 x i8>* %b,
+                        <16 x i8>* %c) nounwind {
+  ; CHECK: bsel_u_v16i8:
+
+  %1 = load <16 x i8>* %a
+  ; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
+  %2 = load <16 x i8>* %b
+  ; CHECK-DAG: ld.b [[R2:\$w[0-9]+]], 0($6)
+  %3 = load <16 x i8>* %c
+  ; CHECK-DAG: ld.b [[R3:\$w[0-9]+]], 0($7)
+  %4 = icmp ugt <16 x i8> %1, %2
+  ; CHECK-DAG: clt_u.b [[R4:\$w[0-9]+]], [[R2]], [[R1]]
+  %5 = select <16 x i1> %4, <16 x i8> %1, <16 x i8> %3
+  ; CHECK-DAG: bsel.v [[R4]], [[R1]], [[R3]]
+  store <16 x i8> %5, <16 x i8>* %d
+  ; CHECK-DAG: st.b [[R4]], 0($4)
+
+  ret void
+  ; CHECK: .size bsel_u_v16i8
+}
+
+define void @bsel_u_v8i16(<8 x i16>* %d, <8 x i16>* %a, <8 x i16>* %b,
+                        <8 x i16>* %c) nounwind {
+  ; CHECK: bsel_u_v8i16:
+
+  %1 = load <8 x i16>* %a
+  ; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
+  %2 = load <8 x i16>* %b
+  ; CHECK-DAG: ld.h [[R2:\$w[0-9]+]], 0($6)
+  %3 = load <8 x i16>* %c
+  ; CHECK-DAG: ld.h [[R3:\$w[0-9]+]], 0($7)
+  %4 = icmp ugt <8 x i16> %1, %2
+  ; CHECK-DAG: clt_u.h [[R4:\$w[0-9]+]], [[R2]], [[R1]]
+  %5 = select <8 x i1> %4, <8 x i16> %1, <8 x i16> %3
+  ; CHECK-DAG: bsel.v [[R4]], [[R1]], [[R3]]
+  store <8 x i16> %5, <8 x i16>* %d
+  ; CHECK-DAG: st.h [[R4]], 0($4)
+
+  ret void
+  ; CHECK: .size bsel_u_v8i16
+}
+
+define void @bsel_u_v4i32(<4 x i32>* %d, <4 x i32>* %a, <4 x i32>* %b,
+                        <4 x i32>* %c) nounwind {
+  ; CHECK: bsel_u_v4i32:
+
+  %1 = load <4 x i32>* %a
+  ; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
+  %2 = load <4 x i32>* %b
+  ; CHECK-DAG: ld.w [[R2:\$w[0-9]+]], 0($6)
+  %3 = load <4 x i32>* %c
+  ; CHECK-DAG: ld.w [[R3:\$w[0-9]+]], 0($7)
+  %4 = icmp ugt <4 x i32> %1, %2
+  ; CHECK-DAG: clt_u.w [[R4:\$w[0-9]+]], [[R2]], [[R1]]
+  %5 = select <4 x i1> %4, <4 x i32> %1, <4 x i32> %3
+  ; CHECK-DAG: bsel.v [[R4]], [[R1]], [[R3]]
+  store <4 x i32> %5, <4 x i32>* %d
+  ; CHECK-DAG: st.w [[R4]], 0($4)
+
+  ret void
+  ; CHECK: .size bsel_u_v4i32
+}
+
+define void @bsel_u_v2i64(<2 x i64>* %d, <2 x i64>* %a, <2 x i64>* %b,
+                        <2 x i64>* %c) nounwind {
+  ; CHECK: bsel_u_v2i64:
+
+  %1 = load <2 x i64>* %a
+  ; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
+  %2 = load <2 x i64>* %b
+  ; CHECK-DAG: ld.d [[R2:\$w[0-9]+]], 0($6)
+  %3 = load <2 x i64>* %c
+  ; CHECK-DAG: ld.d [[R3:\$w[0-9]+]], 0($7)
+  %4 = icmp ugt <2 x i64> %1, %2
+  ; CHECK-DAG: clt_u.d [[R4:\$w[0-9]+]], [[R2]], [[R1]]
+  %5 = select <2 x i1> %4, <2 x i64> %1, <2 x i64> %3
+  ; CHECK-DAG: bsel.v [[R4]], [[R1]], [[R3]]
+  store <2 x i64> %5, <2 x i64>* %d
+  ; CHECK-DAG: st.d [[R4]], 0($4)
+
+  ret void
+  ; CHECK: .size bsel_u_v2i64
+}
+
+define void @bseli_s_v16i8(<16 x i8>* %d, <16 x i8>* %a, <16 x i8>* %b,
+                        <16 x i8>* %c) nounwind {
+  ; CHECK: bseli_s_v16i8:
+
+  %1 = load <16 x i8>* %a
+  ; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
+  %2 = load <16 x i8>* %b
+  ; CHECK-DAG: ld.b [[R2:\$w[0-9]+]], 0($6)
+  %3 = icmp sgt <16 x i8> %1, %2
+  ; CHECK-DAG: clt_s.b [[R4:\$w[0-9]+]], [[R2]], [[R1]]
+  %4 = select <16 x i1> %3, <16 x i8> %1, <16 x i8> <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
+  ; CHECK-DAG: bseli.b [[R4]], [[R1]], 1
+  store <16 x i8> %4, <16 x i8>* %d
+  ; CHECK-DAG: st.b [[R4]], 0($4)
+
+  ret void
+  ; CHECK: .size bseli_s_v16i8
+}
+
+define void @bseli_s_v8i16(<8 x i16>* %d, <8 x i16>* %a, <8 x i16>* %b,
+                        <8 x i16>* %c) nounwind {
+  ; CHECK: bseli_s_v8i16:
+
+  %1 = load <8 x i16>* %a
+  ; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
+  %2 = load <8 x i16>* %b
+  ; CHECK-DAG: ld.h [[R2:\$w[0-9]+]], 0($6)
+  %3 = icmp sgt <8 x i16> %1, %2
+  ; CHECK-DAG: clt_s.h [[R4:\$w[0-9]+]], [[R2]], [[R1]]
+  %4 = select <8 x i1> %3, <8 x i16> %1, <8 x i16> <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
+  ; CHECK-DAG: ldi.h [[R3:\$w[0-9]+]], 1
+  ; CHECK-DAG: bsel.v [[R4]], [[R1]], [[R3]]
+  store <8 x i16> %4, <8 x i16>* %d
+  ; CHECK-DAG: st.h [[R4]], 0($4)
+
+  ret void
+  ; CHECK: .size bseli_s_v8i16
+}
+
+define void @bseli_s_v4i32(<4 x i32>* %d, <4 x i32>* %a, <4 x i32>* %b,
+                        <4 x i32>* %c) nounwind {
+  ; CHECK: bseli_s_v4i32:
+
+  %1 = load <4 x i32>* %a
+  ; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
+  %2 = load <4 x i32>* %b
+  ; CHECK-DAG: ld.w [[R2:\$w[0-9]+]], 0($6)
+  %3 = icmp sgt <4 x i32> %1, %2
+  ; CHECK-DAG: clt_s.w [[R4:\$w[0-9]+]], [[R2]], [[R1]]
+  %4 = select <4 x i1> %3, <4 x i32> %1, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
+  ; CHECK-DAG: ldi.w [[R3:\$w[0-9]+]], 1
+  ; CHECK-DAG: bsel.v [[R4]], [[R1]], [[R3]]
+  store <4 x i32> %4, <4 x i32>* %d
+  ; CHECK-DAG: st.w [[R4]], 0($4)
+
+  ret void
+  ; CHECK: .size bseli_s_v4i32
+}
+
+define void @bseli_s_v2i64(<2 x i64>* %d, <2 x i64>* %a, <2 x i64>* %b,
+                        <2 x i64>* %c) nounwind {
+  ; CHECK: bseli_s_v2i64:
+
+  %1 = load <2 x i64>* %a
+  ; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
+  %2 = load <2 x i64>* %b
+  ; CHECK-DAG: ld.d [[R2:\$w[0-9]+]], 0($6)
+  %3 = icmp sgt <2 x i64> %1, %2
+  ; CHECK-DAG: clt_s.d [[R4:\$w[0-9]+]], [[R2]], [[R1]]
+  %4 = select <2 x i1> %3, <2 x i64> %1, <2 x i64> <i64 1, i64 1>
+  ; CHECK-DAG: ldi.d [[R3:\$w[0-9]+]], 1
+  ; CHECK-DAG: bsel.v [[R4]], [[R1]], [[R3]]
+  store <2 x i64> %4, <2 x i64>* %d
+  ; CHECK-DAG: st.d [[R4]], 0($4)
+
+  ret void
+  ; CHECK: .size bseli_s_v2i64
+}
+
+define void @bseli_u_v16i8(<16 x i8>* %d, <16 x i8>* %a, <16 x i8>* %b,
+                        <16 x i8>* %c) nounwind {
+  ; CHECK: bseli_u_v16i8:
+
+  %1 = load <16 x i8>* %a
+  ; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
+  %2 = load <16 x i8>* %b
+  ; CHECK-DAG: ld.b [[R2:\$w[0-9]+]], 0($6)
+  %3 = icmp ugt <16 x i8> %1, %2
+  ; CHECK-DAG: clt_u.b [[R4:\$w[0-9]+]], [[R2]], [[R1]]
+  %4 = select <16 x i1> %3, <16 x i8> %1, <16 x i8> <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
+  ; CHECK-DAG: bseli.b [[R4]], [[R1]], 1
+  store <16 x i8> %4, <16 x i8>* %d
+  ; CHECK-DAG: st.b [[R4]], 0($4)
+
+  ret void
+  ; CHECK: .size bseli_u_v16i8
+}
+
+define void @bseli_u_v8i16(<8 x i16>* %d, <8 x i16>* %a, <8 x i16>* %b,
+                        <8 x i16>* %c) nounwind {
+  ; CHECK: bseli_u_v8i16:
+
+  %1 = load <8 x i16>* %a
+  ; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
+  %2 = load <8 x i16>* %b
+  ; CHECK-DAG: ld.h [[R2:\$w[0-9]+]], 0($6)
+  %3 = icmp ugt <8 x i16> %1, %2
+  ; CHECK-DAG: clt_u.h [[R4:\$w[0-9]+]], [[R2]], [[R1]]
+  %4 = select <8 x i1> %3, <8 x i16> %1, <8 x i16> <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
+  ; CHECK-DAG: ldi.h [[R3:\$w[0-9]+]], 1
+  ; CHECK-DAG: bsel.v [[R4]], [[R1]], [[R3]]
+  store <8 x i16> %4, <8 x i16>* %d
+  ; CHECK-DAG: st.h [[R4]], 0($4)
+
+  ret void
+  ; CHECK: .size bseli_u_v8i16
+}
+
+define void @bseli_u_v4i32(<4 x i32>* %d, <4 x i32>* %a, <4 x i32>* %b,
+                        <4 x i32>* %c) nounwind {
+  ; CHECK: bseli_u_v4i32:
+
+  %1 = load <4 x i32>* %a
+  ; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
+  %2 = load <4 x i32>* %b
+  ; CHECK-DAG: ld.w [[R2:\$w[0-9]+]], 0($6)
+  %3 = icmp ugt <4 x i32> %1, %2
+  ; CHECK-DAG: clt_u.w [[R4:\$w[0-9]+]], [[R2]], [[R1]]
+  %4 = select <4 x i1> %3, <4 x i32> %1, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
+  ; CHECK-DAG: ldi.w [[R3:\$w[0-9]+]], 1
+  ; CHECK-DAG: bsel.v [[R4]], [[R1]], [[R3]]
+  store <4 x i32> %4, <4 x i32>* %d
+  ; CHECK-DAG: st.w [[R4]], 0($4)
+
+  ret void
+  ; CHECK: .size bseli_u_v4i32
+}
+
+define void @bseli_u_v2i64(<2 x i64>* %d, <2 x i64>* %a, <2 x i64>* %b,
+                        <2 x i64>* %c) nounwind {
+  ; CHECK: bseli_u_v2i64:
+
+  %1 = load <2 x i64>* %a
+  ; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
+  %2 = load <2 x i64>* %b
+  ; CHECK-DAG: ld.d [[R2:\$w[0-9]+]], 0($6)
+  %3 = icmp ugt <2 x i64> %1, %2
+  ; CHECK-DAG: clt_u.d [[R4:\$w[0-9]+]], [[R2]], [[R1]]
+  %4 = select <2 x i1> %3, <2 x i64> %1, <2 x i64> <i64 1, i64 1>
+  ; CHECK-DAG: ldi.d [[R3:\$w[0-9]+]], 1
+  ; CHECK-DAG: bsel.v [[R4]], [[R1]], [[R3]]
+  store <2 x i64> %4, <2 x i64>* %d
+  ; CHECK-DAG: st.d [[R4]], 0($4)
+
+  ret void
+  ; CHECK: .size bseli_u_v2i64
+}
diff --git a/test/CodeGen/Mips/msa/compare_float.ll b/test/CodeGen/Mips/msa/compare_float.ll
index b75f839..106653f 100644
--- a/test/CodeGen/Mips/msa/compare_float.ll
+++ b/test/CodeGen/Mips/msa/compare_float.ll
@@ -516,3 +516,83 @@ define void @true_v2f64(<2 x i64>* %c, <2 x double>* %a, <2 x double>* %b) nounw
   ; CHECK-DAG: st.d [[R4]], 0($4)
   ; CHECK: .size true_v2f64
 }
+
+define void @bsel_v4f32(<4 x float>* %d, <4 x float>* %a, <4 x float>* %b,
+                          <4 x float>* %c) nounwind {
+  ; CHECK: bsel_v4f32:
+
+  %1 = load <4 x float>* %a
+  ; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
+  %2 = load <4 x float>* %b
+  ; CHECK-DAG: ld.w [[R2:\$w[0-9]+]], 0($6)
+  %3 = load <4 x float>* %c
+  ; CHECK-DAG: ld.w [[R3:\$w[0-9]+]], 0($7)
+  %4 = fcmp ogt <4 x float> %1, %2
+  ; CHECK-DAG: fclt.w [[R4:\$w[0-9]+]], [[R2]], [[R1]]
+  %5 = select <4 x i1> %4, <4 x float> %1, <4 x float> %3
+  ; CHECK-DAG: bsel.v [[R4]], [[R1]], [[R3]]
+  store <4 x float> %5, <4 x float>* %d
+  ; CHECK-DAG: st.w [[R4]], 0($4)
+
+  ret void
+  ; CHECK: .size bsel_v4f32
+}
+
+define void @bsel_v2f64(<2 x double>* %d, <2 x double>* %a, <2 x double>* %b,
+                          <2 x double>* %c) nounwind {
+  ; CHECK: bsel_v2f64:
+
+  %1 = load <2 x double>* %a
+  ; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
+  %2 = load <2 x double>* %b
+  ; CHECK-DAG: ld.d [[R2:\$w[0-9]+]], 0($6)
+  %3 = load <2 x double>* %c
+  ; CHECK-DAG: ld.d [[R3:\$w[0-9]+]], 0($7)
+  %4 = fcmp ogt <2 x double> %1, %2
+  ; CHECK-DAG: fclt.d [[R4:\$w[0-9]+]], [[R2]], [[R1]]
+  %5 = select <2 x i1> %4, <2 x double> %1, <2 x double> %3
+  ; CHECK-DAG: bsel.v [[R4]], [[R1]], [[R3]]
+  store <2 x double> %5, <2 x double>* %d
+  ; CHECK-DAG: st.d [[R4]], 0($4)
+
+  ret void
+  ; CHECK: .size bsel_v2f64
+}
+
+define void @bseli_v4f32(<4 x float>* %d, <4 x float>* %a, <4 x float>* %b,
+                          <4 x float>* %c) nounwind {
+  ; CHECK: bseli_v4f32:
+
+  %1 = load <4 x float>* %a
+  ; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
+  %2 = load <4 x float>* %b
+  ; CHECK-DAG: ld.w [[R2:\$w[0-9]+]], 0($6)
+  %3 = fcmp ogt <4 x float> %1, %2
+  ; CHECK-DAG: fclt.w [[R4:\$w[0-9]+]], [[R2]], [[R1]]
+  %4 = select <4 x i1> %3, <4 x float> %1, <4 x float> zeroinitializer
+  ; CHECK-DAG: bsel.v [[R4]], [[R1]], [[R3:\$w[0-9]+]]
+  store <4 x float> %4, <4 x float>* %d
+  ; CHECK-DAG: st.w [[R4]], 0($4)
+
+  ret void
+  ; CHECK: .size bseli_v4f32
+}
+
+define void @bseli_v2f64(<2 x double>* %d, <2 x double>* %a, <2 x double>* %b,
+                          <2 x double>* %c) nounwind {
+  ; CHECK: bseli_v2f64:
+
+  %1 = load <2 x double>* %a
+  ; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
+  %2 = load <2 x double>* %b
+  ; CHECK-DAG: ld.d [[R2:\$w[0-9]+]], 0($6)
+  %3 = fcmp ogt <2 x double> %1, %2
+  ; CHECK-DAG: fclt.d [[R4:\$w[0-9]+]], [[R2]], [[R1]]
+  %4 = select <2 x i1> %3, <2 x double> %1, <2 x double> zeroinitializer
+  ; CHECK-DAG: bsel.v [[R4]], [[R1]], [[R3:\$w[0-9]+]]
+  store <2 x double> %4, <2 x double>* %d
+  ; CHECK-DAG: st.d [[R4]], 0($4)
+
+  ret void
+  ; CHECK: .size bseli_v2f64
+}
diff --git a/test/CodeGen/Mips/msa/i8.ll b/test/CodeGen/Mips/msa/i8.ll
index e629b04..ad8d0a4 100644
--- a/test/CodeGen/Mips/msa/i8.ll
+++ b/test/CodeGen/Mips/msa/i8.ll
@@ -65,12 +65,12 @@ declare <16 x i8> @llvm.mips.bmzi.b(<16 x i8>, i32) nounwind
 define void @llvm_mips_bseli_b_test() nounwind {
 entry:
   %0 = load <16 x i8>* @llvm_mips_bseli_b_ARG1
-  %1 = tail call <16 x i8> @llvm.mips.bseli.b(<16 x i8> %0, i32 25)
+  %1 = tail call <16 x i8> @llvm.mips.bseli.b(<16 x i8> %0, <16 x i8> %0, i32 25)
   store <16 x i8> %1, <16 x i8>* @llvm_mips_bseli_b_RES
   ret void
 }
 
-declare <16 x i8> @llvm.mips.bseli.b(<16 x i8>, i32) nounwind
+declare <16 x i8> @llvm.mips.bseli.b(<16 x i8>, <16 x i8>, i32) nounwind
 
 ; CHECK: llvm_mips_bseli_b_test:
 ; CHECK: ld.b
diff --git a/test/CodeGen/Mips/msa/vec.ll b/test/CodeGen/Mips/msa/vec.ll
index 7ad640b..ee13493 100644
--- a/test/CodeGen/Mips/msa/vec.ll
+++ b/test/CodeGen/Mips/msa/vec.ll
@@ -355,7 +355,7 @@ entry:
   %1 = load <16 x i8>* @llvm_mips_bsel_v_b_ARG2
   %2 = bitcast <16 x i8> %0 to <16 x i8>
   %3 = bitcast <16 x i8> %1 to <16 x i8>
-  %4 = tail call <16 x i8> @llvm.mips.bsel.v(<16 x i8> %2, <16 x i8> %3)
+  %4 = tail call <16 x i8> @llvm.mips.bsel.v(<16 x i8> %2, <16 x i8> %2, <16 x i8> %3)
   %5 = bitcast <16 x i8> %4 to <16 x i8>
   store <16 x i8> %5, <16 x i8>* @llvm_mips_bsel_v_b_RES
   ret void
@@ -378,7 +378,7 @@ entry:
   %1 = load <8 x i16>* @llvm_mips_bsel_v_h_ARG2
   %2 = bitcast <8 x i16> %0 to <16 x i8>
   %3 = bitcast <8 x i16> %1 to <16 x i8>
-  %4 = tail call <16 x i8> @llvm.mips.bsel.v(<16 x i8> %2, <16 x i8> %3)
+  %4 = tail call <16 x i8> @llvm.mips.bsel.v(<16 x i8> %2, <16 x i8> %2, <16 x i8> %3)
   %5 = bitcast <16 x i8> %4 to <8 x i16>
   store <8 x i16> %5, <8 x i16>* @llvm_mips_bsel_v_h_RES
   ret void
@@ -401,7 +401,7 @@ entry:
   %1 = load <4 x i32>* @llvm_mips_bsel_v_w_ARG2
   %2 = bitcast <4 x i32> %0 to <16 x i8>
   %3 = bitcast <4 x i32> %1 to <16 x i8>
-  %4 = tail call <16 x i8> @llvm.mips.bsel.v(<16 x i8> %2, <16 x i8> %3)
+  %4 = tail call <16 x i8> @llvm.mips.bsel.v(<16 x i8> %2, <16 x i8> %2, <16 x i8> %3)
   %5 = bitcast <16 x i8> %4 to <4 x i32>
   store <4 x i32> %5, <4 x i32>* @llvm_mips_bsel_v_w_RES
   ret void
@@ -424,7 +424,7 @@ entry:
   %1 = load <2 x i64>* @llvm_mips_bsel_v_d_ARG2
   %2 = bitcast <2 x i64> %0 to <16 x i8>
   %3 = bitcast <2 x i64> %1 to <16 x i8>
-  %4 = tail call <16 x i8> @llvm.mips.bsel.v(<16 x i8> %2, <16 x i8> %3)
+  %4 = tail call <16 x i8> @llvm.mips.bsel.v(<16 x i8> %2, <16 x i8> %2, <16 x i8> %3)
   %5 = bitcast <16 x i8> %4 to <2 x i64>
   store <2 x i64> %5, <2 x i64>* @llvm_mips_bsel_v_d_RES
   ret void
@@ -848,7 +848,7 @@ entry:
 declare <16 x i8> @llvm.mips.and.v(<16 x i8>, <16 x i8>) nounwind
 declare <16 x i8> @llvm.mips.bmnz.v(<16 x i8>, <16 x i8>) nounwind
 declare <16 x i8> @llvm.mips.bmz.v(<16 x i8>, <16 x i8>) nounwind
-declare <16 x i8> @llvm.mips.bsel.v(<16 x i8>, <16 x i8>) nounwind
+declare <16 x i8> @llvm.mips.bsel.v(<16 x i8>, <16 x i8>, <16 x i8>) nounwind
 declare <16 x i8> @llvm.mips.nor.v(<16 x i8>, <16 x i8>) nounwind
 declare <16 x i8> @llvm.mips.or.v(<16 x i8>, <16 x i8>) nounwind
 declare <16 x i8> @llvm.mips.xor.v(<16 x i8>, <16 x i8>) nounwind