[mips][msa] Build all the tests in little and big endian modes and correct an incorrect test.

Summary:
This patch (correctly) breaks some MSA tests by exposing the cases when
SelectionDAG::getConstant() produces illegal types. These have been temporarily
marked XFAIL and the XFAIL flag will be removed when
SelectionDAG::getConstant() is fixed.

There are three categories of failure:
* Immediate instructions are not selected in one endian mode.
* Immediates used in ldi.[bhwd] must be different according to endianness.
  (this only affects cases where the 'wrong' ldi is used to load the correct
   bitpattern. E.g. (bitcast:v2i64 (build_vector:v4i32 ...)))
* Non-immediate instructions that rely on immediates affected by the
  previous two categories as part of their match pattern.
  For example, the bset match pattern is the vector equivalent of
  'ws | (1 << wt)'.

One test needed correcting to expect different output depending on whether big
or little endian was in use. This test was
test/CodeGen/Mips/msa/basic_operations.ll and experiences the second category
of failure shown above. The little endian version of this test is named
basic_operations_little.ll and will be merged back into basic_operations.ll in
a follow up commit now that FileCheck supports multiple check prefixes.

Reviewers: bkramer, jacksprat, dsanders

Reviewed By: dsanders

CC: llvm-commits

Differential Revision: http://llvm-reviews.chandlerc.com/D1972

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@194806 91177308-0d34-0410-b5e6-96231b3b80d8

59 files changed, 666 insertions, 3 deletions

diff --git a/test/CodeGen/Mips/msa/2r.ll b/test/CodeGen/Mips/msa/2r.ll
index 273a2a2..da35ad8 100644
--- a/test/CodeGen/Mips/msa/2r.ll
+++ b/test/CodeGen/Mips/msa/2r.ll
@@ -1,6 +1,7 @@
 ; Test the MSA intrinsics that are encoded with the 2R instruction format.
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_nloc_b_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_nloc_b_RES  = global <16 x i8> <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>, align 16
diff --git a/test/CodeGen/Mips/msa/2r_vector_scalar.ll b/test/CodeGen/Mips/msa/2r_vector_scalar.ll
index 83d99d7..6f6e1b9 100644
--- a/test/CodeGen/Mips/msa/2r_vector_scalar.ll
+++ b/test/CodeGen/Mips/msa/2r_vector_scalar.ll
@@ -2,6 +2,7 @@
 ; convert scalars to vectors.
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_fill_b_ARG1 = global i32 23, align 16
 @llvm_mips_fill_b_RES  = global <16 x i8> <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>, align 16
diff --git a/test/CodeGen/Mips/msa/2rf.ll b/test/CodeGen/Mips/msa/2rf.ll
index 4365c9e..b361ef5 100644
--- a/test/CodeGen/Mips/msa/2rf.ll
+++ b/test/CodeGen/Mips/msa/2rf.ll
@@ -1,6 +1,7 @@
 ; Test the MSA intrinsics that are encoded with the 2RF instruction format.
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_flog2_w_ARG1 = global <4 x float> <float 0.000000e+00, float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>, align 16
 @llvm_mips_flog2_w_RES  = global <4 x float> <float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00>, align 16
diff --git a/test/CodeGen/Mips/msa/2rf_exup.ll b/test/CodeGen/Mips/msa/2rf_exup.ll
index aad008b..8d7cc36 100644
--- a/test/CodeGen/Mips/msa/2rf_exup.ll
+++ b/test/CodeGen/Mips/msa/2rf_exup.ll
@@ -2,6 +2,7 @@
 ; are encoded with the 2RF instruction format.
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_fexupl_w_ARG1 = global <8 x half> <half 0.000000e+00, half 1.000000e+00, half 2.000000e+00, half 3.000000e+00, half 4.000000e+00, half 5.000000e+00, half 6.000000e+00, half 7.000000e+00>, align 16
 @llvm_mips_fexupl_w_RES  = global <4 x float> <float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00>, align 16
diff --git a/test/CodeGen/Mips/msa/2rf_float_int.ll b/test/CodeGen/Mips/msa/2rf_float_int.ll
index 1dbb228..3b5dfda 100644
--- a/test/CodeGen/Mips/msa/2rf_float_int.ll
+++ b/test/CodeGen/Mips/msa/2rf_float_int.ll
@@ -2,6 +2,7 @@
 ; with the 2RF instruction format.
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_ffint_s_w_ARG1 = global <4 x i32> <i32 0, i32 1, i32 2, i32 3>, align 16
 @llvm_mips_ffint_s_w_RES  = global <4 x float> <float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00>, align 16
diff --git a/test/CodeGen/Mips/msa/2rf_fq.ll b/test/CodeGen/Mips/msa/2rf_fq.ll
index da76fdc..021dd93 100644
--- a/test/CodeGen/Mips/msa/2rf_fq.ll
+++ b/test/CodeGen/Mips/msa/2rf_fq.ll
@@ -2,6 +2,7 @@
 ; encoded with the 2RF instruction format.
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_ffql_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_ffql_w_RES  = global <4 x float> <float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00>, align 16
diff --git a/test/CodeGen/Mips/msa/2rf_int_float.ll b/test/CodeGen/Mips/msa/2rf_int_float.ll
index 4a59a7c..4665ae0 100644
--- a/test/CodeGen/Mips/msa/2rf_int_float.ll
+++ b/test/CodeGen/Mips/msa/2rf_int_float.ll
@@ -3,6 +3,7 @@
 ; as fclass are also here.
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_fclass_w_ARG1 = global <4 x float> <float 0.000000e+00, float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>, align 16
 @llvm_mips_fclass_w_RES  = global <4 x i32> <i32 0, i32 0, i32 0, i32 0>, align 16
diff --git a/test/CodeGen/Mips/msa/2rf_tq.ll b/test/CodeGen/Mips/msa/2rf_tq.ll
index efd962c..6f3c508 100644
--- a/test/CodeGen/Mips/msa/2rf_tq.ll
+++ b/test/CodeGen/Mips/msa/2rf_tq.ll
@@ -2,6 +2,7 @@
 ; encoded with the 2RF instruction format.
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_ftq_h_ARG1 = global <4 x float> <float 0.000000e+00, float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>, align 16
 @llvm_mips_ftq_h_ARG2 = global <4 x float> <float 4.000000e+00, float 5.000000e+00, float 6.000000e+00, float 7.000000e+00>, align 16
diff --git a/test/CodeGen/Mips/msa/3r-a.ll b/test/CodeGen/Mips/msa/3r-a.ll
index 0ad02a0..dab15b6 100644
--- a/test/CodeGen/Mips/msa/3r-a.ll
+++ b/test/CodeGen/Mips/msa/3r-a.ll
@@ -2,6 +2,7 @@
 ; There are lots of these so this covers those beginning with 'a'
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 ; It should fail to compile without fp64.
 ; RUN: not llc -march=mips -mattr=+msa < %s 2>&1 | \
diff --git a/test/CodeGen/Mips/msa/3r-b.ll b/test/CodeGen/Mips/msa/3r-b.ll
index 5677d19..7c45b2b 100644
--- a/test/CodeGen/Mips/msa/3r-b.ll
+++ b/test/CodeGen/Mips/msa/3r-b.ll
@@ -2,6 +2,8 @@
 ; There are lots of these so this covers those beginning with 'b'
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
+; XFAIL: *
 
 @llvm_mips_bclr_b_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_bclr_b_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
diff --git a/test/CodeGen/Mips/msa/3r-c.ll b/test/CodeGen/Mips/msa/3r-c.ll
index 84d8fa2..6ec92c2 100644
--- a/test/CodeGen/Mips/msa/3r-c.ll
+++ b/test/CodeGen/Mips/msa/3r-c.ll
@@ -2,6 +2,7 @@
 ; There are lots of these so this covers those beginning with 'c'
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_ceq_b_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_ceq_b_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
diff --git a/test/CodeGen/Mips/msa/3r-d.ll b/test/CodeGen/Mips/msa/3r-d.ll
index 48c83a6..0099554 100644
--- a/test/CodeGen/Mips/msa/3r-d.ll
+++ b/test/CodeGen/Mips/msa/3r-d.ll
@@ -2,6 +2,7 @@
 ; There are lots of these so this covers those beginning with 'd'
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_div_s_b_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_div_s_b_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
diff --git a/test/CodeGen/Mips/msa/3r-i.ll b/test/CodeGen/Mips/msa/3r-i.ll
index c04734d..2ef3047 100644
--- a/test/CodeGen/Mips/msa/3r-i.ll
+++ b/test/CodeGen/Mips/msa/3r-i.ll
@@ -2,6 +2,7 @@
 ; There are lots of these so this covers those beginning with 'i'
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_ilvev_b_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_ilvev_b_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
diff --git a/test/CodeGen/Mips/msa/3r-m.ll b/test/CodeGen/Mips/msa/3r-m.ll
index c8de6d2..ddfd720 100644
--- a/test/CodeGen/Mips/msa/3r-m.ll
+++ b/test/CodeGen/Mips/msa/3r-m.ll
@@ -2,6 +2,7 @@
 ; There are lots of these so this covers those beginning with 'm'
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_max_a_b_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_max_a_b_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
diff --git a/test/CodeGen/Mips/msa/3r-p.ll b/test/CodeGen/Mips/msa/3r-p.ll
index 4620f88..852023b 100644
--- a/test/CodeGen/Mips/msa/3r-p.ll
+++ b/test/CodeGen/Mips/msa/3r-p.ll
@@ -2,6 +2,7 @@
 ; There are lots of these so this covers those beginning with 'p'
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_pckev_b_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_pckev_b_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
diff --git a/test/CodeGen/Mips/msa/3r-s.ll b/test/CodeGen/Mips/msa/3r-s.ll
index 1b894cf..30cf265 100644
--- a/test/CodeGen/Mips/msa/3r-s.ll
+++ b/test/CodeGen/Mips/msa/3r-s.ll
@@ -2,6 +2,7 @@
 ; There are lots of these so this covers those beginning with 's'
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_sld_b_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_sld_b_ARG2 = global i32 10, align 16
diff --git a/test/CodeGen/Mips/msa/3r-v.ll b/test/CodeGen/Mips/msa/3r-v.ll
index fb547cf..c9693f9 100644
--- a/test/CodeGen/Mips/msa/3r-v.ll
+++ b/test/CodeGen/Mips/msa/3r-v.ll
@@ -2,6 +2,7 @@
 ; There are lots of these so this covers those beginning with 'v'
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_vshf_b_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_vshf_b_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
diff --git a/test/CodeGen/Mips/msa/3r_4r.ll b/test/CodeGen/Mips/msa/3r_4r.ll
index 2e341d0..b7fd728 100644
--- a/test/CodeGen/Mips/msa/3r_4r.ll
+++ b/test/CodeGen/Mips/msa/3r_4r.ll
@@ -2,6 +2,7 @@
 ; use the result as a third operand.
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_maddv_b_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_maddv_b_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
diff --git a/test/CodeGen/Mips/msa/3r_4r_widen.ll b/test/CodeGen/Mips/msa/3r_4r_widen.ll
index 6dd7fbc..7063e45 100644
--- a/test/CodeGen/Mips/msa/3r_4r_widen.ll
+++ b/test/CodeGen/Mips/msa/3r_4r_widen.ll
@@ -3,6 +3,7 @@
 ; operands had.
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_dpadd_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_dpadd_s_h_ARG2 = global <16 x i8> <i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 20, i8 21, i8 22, i8 23>, align 16
diff --git a/test/CodeGen/Mips/msa/3r_splat.ll b/test/CodeGen/Mips/msa/3r_splat.ll
index a10cf5c..6b0cb26 100644
--- a/test/CodeGen/Mips/msa/3r_splat.ll
+++ b/test/CodeGen/Mips/msa/3r_splat.ll
@@ -3,6 +3,8 @@
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | \
 ; RUN:     FileCheck -check-prefix=MIPS32 %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | \
+; RUN:     FileCheck -check-prefix=MIPS32 %s
 
 @llvm_mips_splat_b_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_splat_b_RES  = global <16 x i8> <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>, align 16
diff --git a/test/CodeGen/Mips/msa/3rf.ll b/test/CodeGen/Mips/msa/3rf.ll
index ea1291e..ae665af 100644
--- a/test/CodeGen/Mips/msa/3rf.ll
+++ b/test/CodeGen/Mips/msa/3rf.ll
@@ -1,6 +1,7 @@
 ; Test the MSA intrinsics that are encoded with the 3RF instruction format.
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_fadd_w_ARG1 = global <4 x float> <float 0.000000e+00, float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>, align 16
 @llvm_mips_fadd_w_ARG2 = global <4 x float> <float 4.000000e+00, float 5.000000e+00, float 6.000000e+00, float 7.000000e+00>, align 16
diff --git a/test/CodeGen/Mips/msa/3rf_4rf.ll b/test/CodeGen/Mips/msa/3rf_4rf.ll
index 9c446c9..67ef7fd 100644
--- a/test/CodeGen/Mips/msa/3rf_4rf.ll
+++ b/test/CodeGen/Mips/msa/3rf_4rf.ll
@@ -2,6 +2,7 @@
 ; use the result as a third operand.
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_fmadd_w_ARG1 = global <4 x float> <float 0.000000e+00, float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>, align 16
 @llvm_mips_fmadd_w_ARG2 = global <4 x float> <float 4.000000e+00, float 5.000000e+00, float 6.000000e+00, float 7.000000e+00>, align 16
diff --git a/test/CodeGen/Mips/msa/3rf_4rf_q.ll b/test/CodeGen/Mips/msa/3rf_4rf_q.ll
index 888a022..de28be0 100644
--- a/test/CodeGen/Mips/msa/3rf_4rf_q.ll
+++ b/test/CodeGen/Mips/msa/3rf_4rf_q.ll
@@ -2,6 +2,7 @@
 ; use the result as a third operand and perform fixed-point operations.
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_madd_q_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_madd_q_h_ARG2 = global <8 x i16> <i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15>, align 16
diff --git a/test/CodeGen/Mips/msa/3rf_exdo.ll b/test/CodeGen/Mips/msa/3rf_exdo.ll
index 16a3c81..8a7f268 100644
--- a/test/CodeGen/Mips/msa/3rf_exdo.ll
+++ b/test/CodeGen/Mips/msa/3rf_exdo.ll
@@ -2,6 +2,7 @@
 ; 3RF instruction format.
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_fexdo_h_ARG1 = global <4 x float> <float 0.000000e+00, float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>, align 16
 @llvm_mips_fexdo_h_ARG2 = global <4 x float> <float 4.000000e+00, float 5.000000e+00, float 6.000000e+00, float 7.000000e+00>, align 16
diff --git a/test/CodeGen/Mips/msa/3rf_float_int.ll b/test/CodeGen/Mips/msa/3rf_float_int.ll
index a446ebd..7b01e17 100644
--- a/test/CodeGen/Mips/msa/3rf_float_int.ll
+++ b/test/CodeGen/Mips/msa/3rf_float_int.ll
@@ -2,6 +2,7 @@
 ; take an integer as an operand.
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_fexp2_w_ARG1 = global <4 x float> <float 0.000000e+00, float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>, align 16
 @llvm_mips_fexp2_w_ARG2 = global <4 x i32> <i32 4, i32 5, i32 6, i32 7>, align 16
diff --git a/test/CodeGen/Mips/msa/3rf_int_float.ll b/test/CodeGen/Mips/msa/3rf_int_float.ll
index 3dbcc79..5624771 100644
--- a/test/CodeGen/Mips/msa/3rf_int_float.ll
+++ b/test/CodeGen/Mips/msa/3rf_int_float.ll
@@ -2,6 +2,7 @@
 ; produce an integer as a result.
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_fcaf_w_ARG1 = global <4 x float> <float 0.000000e+00, float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>, align 16
 @llvm_mips_fcaf_w_ARG2 = global <4 x float> <float 4.000000e+00, float 5.000000e+00, float 6.000000e+00, float 7.000000e+00>, align 16
diff --git a/test/CodeGen/Mips/msa/3rf_q.ll b/test/CodeGen/Mips/msa/3rf_q.ll
index 0504fe8..f7000ee 100644
--- a/test/CodeGen/Mips/msa/3rf_q.ll
+++ b/test/CodeGen/Mips/msa/3rf_q.ll
@@ -2,6 +2,7 @@
 ; format.
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_mul_q_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_mul_q_h_ARG2 = global <8 x i16> <i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15>, align 16
diff --git a/test/CodeGen/Mips/msa/arithmetic.ll b/test/CodeGen/Mips/msa/arithmetic.ll
index 512a1c1..09ee502 100644
--- a/test/CodeGen/Mips/msa/arithmetic.ll
+++ b/test/CodeGen/Mips/msa/arithmetic.ll
@@ -1,4 +1,5 @@
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 define void @add_v16i8(<16 x i8>* %c, <16 x i8>* %a, <16 x i8>* %b) nounwind {
   ; CHECK: add_v16i8:
diff --git a/test/CodeGen/Mips/msa/arithmetic_float.ll b/test/CodeGen/Mips/msa/arithmetic_float.ll
index 43a1f29..dc38721 100644
--- a/test/CodeGen/Mips/msa/arithmetic_float.ll
+++ b/test/CodeGen/Mips/msa/arithmetic_float.ll
@@ -1,4 +1,5 @@
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 define void @add_v4f32(<4 x float>* %c, <4 x float>* %a, <4 x float>* %b) nounwind {
   ; CHECK: add_v4f32:
diff --git a/test/CodeGen/Mips/msa/basic_operations.ll b/test/CodeGen/Mips/msa/basic_operations.ll
index 73c17dc..2f153bf 100644
--- a/test/CodeGen/Mips/msa/basic_operations.ll
+++ b/test/CodeGen/Mips/msa/basic_operations.ll
@@ -22,8 +22,8 @@ define void @const_v16i8() nounwind {
   store volatile <16 x i8> <i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6>, <16 x i8>*@v16i8
   ; MIPS32: ld.b  [[R1:\$w[0-9]+]], %lo(
 
-  store volatile <16 x i8> <i8 1, i8 2, i8 1, i8 2, i8 1, i8 2, i8 1, i8 2, i8 1, i8 2, i8 1, i8 2, i8 1, i8 2, i8 1, i8 2>, <16 x i8>*@v16i8
-  ; MIPS32: ldi.h [[R1:\$w[0-9]+]], 258
+  store volatile <16 x i8> <i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0>, <16 x i8>*@v16i8
+  ; MIPS32: ldi.h [[R1:\$w[0-9]+]], 256
 
   store volatile <16 x i8> <i8 1, i8 2, i8 3, i8 4, i8 1, i8 2, i8 3, i8 4, i8 1, i8 2, i8 3, i8 4, i8 1, i8 2, i8 3, i8 4>, <16 x i8>*@v16i8
   ; MIPS32-DAG: lui [[R2:\$[0-9]+]], 258
diff --git a/test/CodeGen/Mips/msa/basic_operations_float.ll b/test/CodeGen/Mips/msa/basic_operations_float.ll
index 7f6db85..1f53810 100644
--- a/test/CodeGen/Mips/msa/basic_operations_float.ll
+++ b/test/CodeGen/Mips/msa/basic_operations_float.ll
@@ -1,4 +1,5 @@
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck -check-prefix=MIPS32 %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck -check-prefix=MIPS32 %s
 
 @v4f32 = global <4 x float> <float 0.0, float 0.0, float 0.0, float 0.0>
 @v2f64 = global <2 x double> <double 0.0, double 0.0>
diff --git a/test/CodeGen/Mips/msa/basic_operations_little.ll b/test/CodeGen/Mips/msa/basic_operations_little.ll
new file mode 100644
index 0000000..e091ed8
--- /dev/null
+++ b/test/CodeGen/Mips/msa/basic_operations_little.ll
@@ -0,0 +1,469 @@
+; This test will be merged back into basic_operations.ll once FileCheck accepts multiple prefixes.
+
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck -check-prefix=MIPS32 %s
+
+@v4i8 = global <4 x i8> <i8 0, i8 0, i8 0, i8 0>
+@v16i8 = global <16 x i8> <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>
+@v8i16 = global <8 x i16> <i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0>
+@v4i32 = global <4 x i32> <i32 0, i32 0, i32 0, i32 0>
+@v2i64 = global <2 x i64> <i64 0, i64 0>
+@i64 = global i64 0
+
+define void @const_v16i8() nounwind {
+  ; MIPS32: const_v16i8:
+
+  store volatile <16 x i8> <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>, <16 x i8>*@v16i8
+  ; MIPS32: ldi.b [[R1:\$w[0-9]+]], 0
+
+  store volatile <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>, <16 x i8>*@v16i8
+  ; MIPS32: ldi.b [[R1:\$w[0-9]+]], 1
+
+  store volatile <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 31>, <16 x i8>*@v16i8
+  ; MIPS32: ld.b  [[R1:\$w[0-9]+]], %lo(
+
+  store volatile <16 x i8> <i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6>, <16 x i8>*@v16i8
+  ; MIPS32: ld.b  [[R1:\$w[0-9]+]], %lo(
+
+  store volatile <16 x i8> <i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0>, <16 x i8>*@v16i8
+  ; MIPS32: ldi.h [[R1:\$w[0-9]+]], 1
+
+  store volatile <16 x i8> <i8 1, i8 2, i8 3, i8 4, i8 1, i8 2, i8 3, i8 4, i8 1, i8 2, i8 3, i8 4, i8 1, i8 2, i8 3, i8 4>, <16 x i8>*@v16i8
+  ; MIPS32-DAG: lui [[R2:\$[0-9]+]], 1027
+  ; MIPS32-DAG: ori [[R2]], [[R2]], 513
+  ; MIPS32-DAG: fill.w [[R1:\$w[0-9]+]], [[R2]]
+
+  store volatile <16 x i8> <i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8>, <16 x i8>*@v16i8
+  ; MIPS32: ld.b  [[R1:\$w[0-9]+]], %lo(
+
+  ret void
+  ; MIPS32: .size const_v16i8
+}
+
+define void @const_v8i16() nounwind {
+  ; MIPS32: const_v8i16:
+
+  store volatile <8 x i16> <i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0>, <8 x i16>*@v8i16
+  ; MIPS32: ldi.b [[R1:\$w[0-9]+]], 0
+
+  store volatile <8 x i16> <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>, <8 x i16>*@v8i16
+  ; MIPS32: ldi.h [[R1:\$w[0-9]+]], 1
+
+  store volatile <8 x i16> <i16 1, i16 1, i16 1, i16 2, i16 1, i16 1, i16 1, i16 31>, <8 x i16>*@v8i16
+  ; MIPS32: ld.h  [[R1:\$w[0-9]+]], %lo(
+
+  store volatile <8 x i16> <i16 1028, i16 1028, i16 1028, i16 1028, i16 1028, i16 1028, i16 1028, i16 1028>, <8 x i16>*@v8i16
+  ; MIPS32: ldi.b [[R1:\$w[0-9]+]], 4
+
+  store volatile <8 x i16> <i16 1, i16 2, i16 1, i16 2, i16 1, i16 2, i16 1, i16 2>, <8 x i16>*@v8i16
+  ; MIPS32-DAG: lui [[R2:\$[0-9]+]], 2
+  ; MIPS32-DAG: ori [[R2]], [[R2]], 1
+  ; MIPS32-DAG: fill.w [[R1:\$w[0-9]+]], [[R2]]
+
+  store volatile <8 x i16> <i16 1, i16 2, i16 3, i16 4, i16 1, i16 2, i16 3, i16 4>, <8 x i16>*@v8i16
+  ; MIPS32: ld.h  [[R1:\$w[0-9]+]], %lo(
+
+  ret void
+  ; MIPS32: .size const_v8i16
+}
+
+define void @const_v4i32() nounwind {
+  ; MIPS32: const_v4i32:
+
+  store volatile <4 x i32> <i32 0, i32 0, i32 0, i32 0>, <4 x i32>*@v4i32
+  ; MIPS32: ldi.b [[R1:\$w[0-9]+]], 0
+
+  store volatile <4 x i32> <i32 1, i32 1, i32 1, i32 1>, <4 x i32>*@v4i32
+  ; MIPS32: ldi.w [[R1:\$w[0-9]+]], 1
+
+  store volatile <4 x i32> <i32 1, i32 1, i32 1, i32 31>, <4 x i32>*@v4i32
+  ; MIPS32: ld.w  [[R1:\$w[0-9]+]], %lo(
+
+  store volatile <4 x i32> <i32 16843009, i32 16843009, i32 16843009, i32 16843009>, <4 x i32>*@v4i32
+  ; MIPS32: ldi.b [[R1:\$w[0-9]+]], 1
+
+  store volatile <4 x i32> <i32 65537, i32 65537, i32 65537, i32 65537>, <4 x i32>*@v4i32
+  ; MIPS32: ldi.h [[R1:\$w[0-9]+]], 1
+
+  store volatile <4 x i32> <i32 1, i32 2, i32 1, i32 2>, <4 x i32>*@v4i32
+  ; MIPS32: ld.w  [[R1:\$w[0-9]+]], %lo(
+
+  store volatile <4 x i32> <i32 3, i32 4, i32 5, i32 6>, <4 x i32>*@v4i32
+  ; MIPS32: ld.w  [[R1:\$w[0-9]+]], %lo(
+
+  ret void
+  ; MIPS32: .size const_v4i32
+}
+
+define void @const_v2i64() nounwind {
+  ; MIPS32: const_v2i64:
+
+  store volatile <2 x i64> <i64 0, i64 0>, <2 x i64>*@v2i64
+  ; MIPS32: ldi.b [[R1:\$w[0-9]+]], 0
+
+  store volatile <2 x i64> <i64 72340172838076673, i64 72340172838076673>, <2 x i64>*@v2i64
+  ; MIPS32: ldi.b [[R1:\$w[0-9]+]], 1
+
+  store volatile <2 x i64> <i64 281479271743489, i64 281479271743489>, <2 x i64>*@v2i64
+  ; MIPS32: ldi.h [[R1:\$w[0-9]+]], 1
+
+  store volatile <2 x i64> <i64 4294967297, i64 4294967297>, <2 x i64>*@v2i64
+  ; MIPS32: ldi.w [[R1:\$w[0-9]+]], 1
+
+  store volatile <2 x i64> <i64 1, i64 1>, <2 x i64>*@v2i64
+  ; MIPS32: ldi.d [[R1:\$w[0-9]+]], 1
+
+  store volatile <2 x i64> <i64 1, i64 31>, <2 x i64>*@v2i64
+  ; MIPS32: ld.w  [[R1:\$w[0-9]+]], %lo(
+
+  store volatile <2 x i64> <i64 3, i64 4>, <2 x i64>*@v2i64
+  ; MIPS32: ld.w  [[R1:\$w[0-9]+]], %lo(
+
+  ret void
+  ; MIPS32: .size const_v2i64
+}
+
+define void @nonconst_v16i8(i8 %a, i8 %b, i8 %c, i8 %d, i8 %e, i8 %f, i8 %g, i8 %h) nounwind {
+  ; MIPS32: nonconst_v16i8:
+
+  %1 = insertelement <16 x i8> undef, i8 %a, i32 0
+  %2 = insertelement <16 x i8> %1, i8 %b, i32 1
+  %3 = insertelement <16 x i8> %2, i8 %c, i32 2
+  %4 = insertelement <16 x i8> %3, i8 %d, i32 3
+  %5 = insertelement <16 x i8> %4, i8 %e, i32 4
+  %6 = insertelement <16 x i8> %5, i8 %f, i32 5
+  %7 = insertelement <16 x i8> %6, i8 %g, i32 6
+  %8 = insertelement <16 x i8> %7, i8 %h, i32 7
+  %9 = insertelement <16 x i8> %8, i8 %h, i32 8
+  %10 = insertelement <16 x i8> %9, i8 %h, i32 9
+  %11 = insertelement <16 x i8> %10, i8 %h, i32 10
+  %12 = insertelement <16 x i8> %11, i8 %h, i32 11
+  %13 = insertelement <16 x i8> %12, i8 %h, i32 12
+  %14 = insertelement <16 x i8> %13, i8 %h, i32 13
+  %15 = insertelement <16 x i8> %14, i8 %h, i32 14
+  %16 = insertelement <16 x i8> %15, i8 %h, i32 15
+  ; MIPS32-DAG: insert.b [[R1:\$w[0-9]+]][0], $4
+  ; MIPS32-DAG: insert.b [[R1]][1], $5
+  ; MIPS32-DAG: insert.b [[R1]][2], $6
+  ; MIPS32-DAG: insert.b [[R1]][3], $7
+  ; MIPS32-DAG: lbu [[R2:\$[0-9]+]], 16($sp)
+  ; MIPS32-DAG: insert.b [[R1]][4], [[R2]]
+  ; MIPS32-DAG: lbu [[R3:\$[0-9]+]], 20($sp)
+  ; MIPS32-DAG: insert.b [[R1]][5], [[R3]]
+  ; MIPS32-DAG: lbu [[R4:\$[0-9]+]], 24($sp)
+  ; MIPS32-DAG: insert.b [[R1]][6], [[R4]]
+  ; MIPS32-DAG: lbu [[R5:\$[0-9]+]], 28($sp)
+  ; MIPS32-DAG: insert.b [[R1]][7], [[R5]]
+  ; MIPS32-DAG: insert.b [[R1]][8], [[R5]]
+  ; MIPS32-DAG: insert.b [[R1]][9], [[R5]]
+  ; MIPS32-DAG: insert.b [[R1]][10], [[R5]]
+  ; MIPS32-DAG: insert.b [[R1]][11], [[R5]]
+  ; MIPS32-DAG: insert.b [[R1]][12], [[R5]]
+  ; MIPS32-DAG: insert.b [[R1]][13], [[R5]]
+  ; MIPS32-DAG: insert.b [[R1]][14], [[R5]]
+  ; MIPS32-DAG: insert.b [[R1]][15], [[R5]]
+
+  store volatile <16 x i8> %16, <16 x i8>*@v16i8
+
+  ret void
+  ; MIPS32: .size nonconst_v16i8
+}
+
+define void @nonconst_v8i16(i16 %a, i16 %b, i16 %c, i16 %d, i16 %e, i16 %f, i16 %g, i16 %h) nounwind {
+  ; MIPS32: nonconst_v8i16:
+
+  %1 = insertelement <8 x i16> undef, i16 %a, i32 0
+  %2 = insertelement <8 x i16> %1, i16 %b, i32 1
+  %3 = insertelement <8 x i16> %2, i16 %c, i32 2
+  %4 = insertelement <8 x i16> %3, i16 %d, i32 3
+  %5 = insertelement <8 x i16> %4, i16 %e, i32 4
+  %6 = insertelement <8 x i16> %5, i16 %f, i32 5
+  %7 = insertelement <8 x i16> %6, i16 %g, i32 6
+  %8 = insertelement <8 x i16> %7, i16 %h, i32 7
+  ; MIPS32-DAG: insert.h [[R1:\$w[0-9]+]][0], $4
+  ; MIPS32-DAG: insert.h [[R1]][1], $5
+  ; MIPS32-DAG: insert.h [[R1]][2], $6
+  ; MIPS32-DAG: insert.h [[R1]][3], $7
+  ; MIPS32-DAG: lhu [[R2:\$[0-9]+]], 16($sp)
+  ; MIPS32-DAG: insert.h [[R1]][4], [[R2]]
+  ; MIPS32-DAG: lhu [[R2:\$[0-9]+]], 20($sp)
+  ; MIPS32-DAG: insert.h [[R1]][5], [[R2]]
+  ; MIPS32-DAG: lhu [[R2:\$[0-9]+]], 24($sp)
+  ; MIPS32-DAG: insert.h [[R1]][6], [[R2]]
+  ; MIPS32-DAG: lhu [[R2:\$[0-9]+]], 28($sp)
+  ; MIPS32-DAG: insert.h [[R1]][7], [[R2]]
+
+  store volatile <8 x i16> %8, <8 x i16>*@v8i16
+
+  ret void
+  ; MIPS32: .size nonconst_v8i16
+}
+
+define void @nonconst_v4i32(i32 %a, i32 %b, i32 %c, i32 %d) nounwind {
+  ; MIPS32: nonconst_v4i32:
+
+  %1 = insertelement <4 x i32> undef, i32 %a, i32 0
+  %2 = insertelement <4 x i32> %1, i32 %b, i32 1
+  %3 = insertelement <4 x i32> %2, i32 %c, i32 2
+  %4 = insertelement <4 x i32> %3, i32 %d, i32 3
+  ; MIPS32: insert.w [[R1:\$w[0-9]+]][0], $4
+  ; MIPS32: insert.w [[R1]][1], $5
+  ; MIPS32: insert.w [[R1]][2], $6
+  ; MIPS32: insert.w [[R1]][3], $7
+
+  store volatile <4 x i32> %4, <4 x i32>*@v4i32
+
+  ret void
+  ; MIPS32: .size nonconst_v4i32
+}
+
+define void @nonconst_v2i64(i64 %a, i64 %b) nounwind {
+  ; MIPS32: nonconst_v2i64:
+
+  %1 = insertelement <2 x i64> undef, i64 %a, i32 0
+  %2 = insertelement <2 x i64> %1, i64 %b, i32 1
+  ; MIPS32: insert.w [[R1:\$w[0-9]+]][0], $4
+  ; MIPS32: insert.w [[R1]][1], $5
+  ; MIPS32: insert.w [[R1]][2], $6
+  ; MIPS32: insert.w [[R1]][3], $7
+
+  store volatile <2 x i64> %2, <2 x i64>*@v2i64
+
+  ret void
+  ; MIPS32: .size nonconst_v2i64
+}
+
+define i32 @extract_sext_v16i8() nounwind {
+  ; MIPS32: extract_sext_v16i8:
+
+  %1 = load <16 x i8>* @v16i8
+  ; MIPS32-DAG: ld.b [[R1:\$w[0-9]+]],
+
+  %2 = add <16 x i8> %1, %1
+  ; MIPS32-DAG: addv.b [[R2:\$w[0-9]+]], [[R1]], [[R1]]
+
+  %3 = extractelement <16 x i8> %2, i32 1
+  %4 = sext i8 %3 to i32
+  ; MIPS32-DAG: copy_s.b [[R3:\$[0-9]+]], [[R1]][1]
+  ; MIPS32-NOT: sll
+  ; MIPS32-NOT: sra
+
+  ret i32 %4
+  ; MIPS32: .size extract_sext_v16i8
+}
+
+define i32 @extract_sext_v8i16() nounwind {
+  ; MIPS32: extract_sext_v8i16:
+
+  %1 = load <8 x i16>* @v8i16
+  ; MIPS32-DAG: ld.h [[R1:\$w[0-9]+]],
+
+  %2 = add <8 x i16> %1, %1
+  ; MIPS32-DAG: addv.h [[R2:\$w[0-9]+]], [[R1]], [[R1]]
+
+  %3 = extractelement <8 x i16> %2, i32 1
+  %4 = sext i16 %3 to i32
+  ; MIPS32-DAG: copy_s.h [[R3:\$[0-9]+]], [[R1]][1]
+  ; MIPS32-NOT: sll
+  ; MIPS32-NOT: sra
+
+  ret i32 %4
+  ; MIPS32: .size extract_sext_v8i16
+}
+
+define i32 @extract_sext_v4i32() nounwind {
+  ; MIPS32: extract_sext_v4i32:
+
+  %1 = load <4 x i32>* @v4i32
+  ; MIPS32-DAG: ld.w [[R1:\$w[0-9]+]],
+
+  %2 = add <4 x i32> %1, %1
+  ; MIPS32-DAG: addv.w [[R2:\$w[0-9]+]], [[R1]], [[R1]]
+
+  %3 = extractelement <4 x i32> %2, i32 1
+  ; MIPS32-DAG: copy_s.w [[R3:\$[0-9]+]], [[R1]][1]
+
+  ret i32 %3
+  ; MIPS32: .size extract_sext_v4i32
+}
+
+define i64 @extract_sext_v2i64() nounwind {
+  ; MIPS32: extract_sext_v2i64:
+
+  %1 = load <2 x i64>* @v2i64
+  ; MIPS32-DAG: ld.d [[R1:\$w[0-9]+]],
+
+  %2 = add <2 x i64> %1, %1
+  ; MIPS32-DAG: addv.d [[R2:\$w[0-9]+]], [[R1]], [[R1]]
+
+  %3 = extractelement <2 x i64> %2, i32 1
+  ; MIPS32-DAG: copy_s.w [[R3:\$[0-9]+]], [[R1]][2]
+  ; MIPS32-DAG: copy_s.w [[R4:\$[0-9]+]], [[R1]][3]
+  ; MIPS32-NOT: sll
+  ; MIPS32-NOT: sra
+
+  ret i64 %3
+  ; MIPS32: .size extract_sext_v2i64
+}
+
+define i32 @extract_zext_v16i8() nounwind {
+  ; MIPS32: extract_zext_v16i8:
+
+  %1 = load <16 x i8>* @v16i8
+  ; MIPS32-DAG: ld.b [[R1:\$w[0-9]+]],
+
+  %2 = add <16 x i8> %1, %1
+  ; MIPS32-DAG: addv.b [[R2:\$w[0-9]+]], [[R1]], [[R1]]
+
+  %3 = extractelement <16 x i8> %2, i32 1
+  %4 = zext i8 %3 to i32
+  ; MIPS32-DAG: copy_u.b [[R3:\$[0-9]+]], [[R1]][1]
+  ; MIPS32-NOT: andi
+
+  ret i32 %4
+  ; MIPS32: .size extract_zext_v16i8
+}
+
+define i32 @extract_zext_v8i16() nounwind {
+  ; MIPS32: extract_zext_v8i16:
+
+  %1 = load <8 x i16>* @v8i16
+  ; MIPS32-DAG: ld.h [[R1:\$w[0-9]+]],
+
+  %2 = add <8 x i16> %1, %1
+  ; MIPS32-DAG: addv.h [[R2:\$w[0-9]+]], [[R1]], [[R1]]
+
+  %3 = extractelement <8 x i16> %2, i32 1
+  %4 = zext i16 %3 to i32
+  ; MIPS32-DAG: copy_u.h [[R3:\$[0-9]+]], [[R1]][1]
+  ; MIPS32-NOT: andi
+
+  ret i32 %4
+  ; MIPS32: .size extract_zext_v8i16
+}
+
+define i32 @extract_zext_v4i32() nounwind {
+  ; MIPS32: extract_zext_v4i32:
+
+  %1 = load <4 x i32>* @v4i32
+  ; MIPS32-DAG: ld.w [[R1:\$w[0-9]+]],
+
+  %2 = add <4 x i32> %1, %1
+  ; MIPS32-DAG: addv.w [[R2:\$w[0-9]+]], [[R1]], [[R1]]
+
+  %3 = extractelement <4 x i32> %2, i32 1
+  ; MIPS32-DAG: copy_{{[su]}}.w [[R3:\$[0-9]+]], [[R1]][1]
+
+  ret i32 %3
+  ; MIPS32: .size extract_zext_v4i32
+}
+
+define i64 @extract_zext_v2i64() nounwind {
+  ; MIPS32: extract_zext_v2i64:
+
+  %1 = load <2 x i64>* @v2i64
+  ; MIPS32-DAG: ld.d [[R1:\$w[0-9]+]],
+
+  %2 = add <2 x i64> %1, %1
+  ; MIPS32-DAG: addv.d [[R2:\$w[0-9]+]], [[R1]], [[R1]]
+
+  %3 = extractelement <2 x i64> %2, i32 1
+  ; MIPS32-DAG: copy_{{[su]}}.w [[R3:\$[0-9]+]], [[R1]][2]
+  ; MIPS32-DAG: copy_{{[su]}}.w [[R4:\$[0-9]+]], [[R1]][3]
+  ; MIPS32-NOT: andi
+
+  ret i64 %3
+  ; MIPS32: .size extract_zext_v2i64
+}
+
+define void @insert_v16i8(i32 %a) nounwind {
+  ; MIPS32: insert_v16i8:
+
+  %1 = load <16 x i8>* @v16i8
+  ; MIPS32-DAG: ld.b [[R1:\$w[0-9]+]],
+
+  %a2 = trunc i32 %a to i8
+  %a3 = sext i8 %a2 to i32
+  %a4 = trunc i32 %a3 to i8
+  ; MIPS32-NOT: andi
+  ; MIPS32-NOT: sra
+
+  %2 = insertelement <16 x i8> %1, i8 %a4, i32 1
+  ; MIPS32-DAG: insert.b [[R1]][1], $4
+
+  store <16 x i8> %2, <16 x i8>* @v16i8
+  ; MIPS32-DAG: st.b [[R1]]
+
+  ret void
+  ; MIPS32: .size insert_v16i8
+}
+
+define void @insert_v8i16(i32 %a) nounwind {
+  ; MIPS32: insert_v8i16:
+
+  %1 = load <8 x i16>* @v8i16
+  ; MIPS32-DAG: ld.h [[R1:\$w[0-9]+]],
+
+  %a2 = trunc i32 %a to i16
+  %a3 = sext i16 %a2 to i32
+  %a4 = trunc i32 %a3 to i16
+  ; MIPS32-NOT: andi
+  ; MIPS32-NOT: sra
+
+  %2 = insertelement <8 x i16> %1, i16 %a4, i32 1
+  ; MIPS32-DAG: insert.h [[R1]][1], $4
+
+  store <8 x i16> %2, <8 x i16>* @v8i16
+  ; MIPS32-DAG: st.h [[R1]]
+
+  ret void
+  ; MIPS32: .size insert_v8i16
+}
+
+define void @insert_v4i32(i32 %a) nounwind {
+  ; MIPS32: insert_v4i32:
+
+  %1 = load <4 x i32>* @v4i32
+  ; MIPS32-DAG: ld.w [[R1:\$w[0-9]+]],
+
+  ; MIPS32-NOT: andi
+  ; MIPS32-NOT: sra
+
+  %2 = insertelement <4 x i32> %1, i32 %a, i32 1
+  ; MIPS32-DAG: insert.w [[R1]][1], $4
+
+  store <4 x i32> %2, <4 x i32>* @v4i32
+  ; MIPS32-DAG: st.w [[R1]]
+
+  ret void
+  ; MIPS32: .size insert_v4i32
+}
+
+define void @insert_v2i64(i64 %a) nounwind {
+  ; MIPS32: insert_v2i64:
+
+  %1 = load <2 x i64>* @v2i64
+  ; MIPS32-DAG: ld.w [[R1:\$w[0-9]+]],
+
+  ; MIPS32-NOT: andi
+  ; MIPS32-NOT: sra
+
+  %2 = insertelement <2 x i64> %1, i64 %a, i32 1
+  ; MIPS32-DAG: insert.w [[R1]][2], $4
+  ; MIPS32-DAG: insert.w [[R1]][3], $5
+
+  store <2 x i64> %2, <2 x i64>* @v2i64
+  ; MIPS32-DAG: st.w [[R1]]
+
+  ret void
+  ; MIPS32: .size insert_v2i64
+}
+
+define void @truncstore() nounwind {
+  ; MIPS32: truncstore:
+
+  store volatile <4 x i8> <i8 -1, i8 -1, i8 -1, i8 -1>, <4 x i8>*@v4i8
+  ; TODO: What code should be emitted?
+
+  ret void
+  ; MIPS32: .size truncstore
+}
diff --git a/test/CodeGen/Mips/msa/bit.ll b/test/CodeGen/Mips/msa/bit.ll
index 1fdf370..dc8bb8f 100644
--- a/test/CodeGen/Mips/msa/bit.ll
+++ b/test/CodeGen/Mips/msa/bit.ll
@@ -1,6 +1,11 @@
+; Both endians should emit the same output for immediate instructions.
+; This is not currently true.
+; XFAIL: *
+
 ; Test the MSA intrinsics that are encoded with the BIT instruction format.
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_sat_s_b_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_sat_s_b_RES  = global <16 x i8> <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>, align 16
diff --git a/test/CodeGen/Mips/msa/bitcast.ll b/test/CodeGen/Mips/msa/bitcast.ll
index 12edf83..8e880ec 100644
--- a/test/CodeGen/Mips/msa/bitcast.ll
+++ b/test/CodeGen/Mips/msa/bitcast.ll
@@ -1,7 +1,7 @@
 ; Test the bitcast operation for big-endian and little-endian.
 
-; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck -check-prefix=LITENDIAN %s
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck -check-prefix=BIGENDIAN %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck -check-prefix=LITENDIAN %s
 
 define void @v16i8_to_v16i8(<16 x i8>* %src, <16 x i8>* %dst) nounwind {
 entry:
diff --git a/test/CodeGen/Mips/msa/bitwise.ll b/test/CodeGen/Mips/msa/bitwise.ll
index 5831a08..9a88c47 100644
--- a/test/CodeGen/Mips/msa/bitwise.ll
+++ b/test/CodeGen/Mips/msa/bitwise.ll
@@ -1,4 +1,5 @@
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 define void @and_v16i8(<16 x i8>* %c, <16 x i8>* %a, <16 x i8>* %b) nounwind {
   ; CHECK: and_v16i8:
diff --git a/test/CodeGen/Mips/msa/compare.ll b/test/CodeGen/Mips/msa/compare.ll
index a7c704e..36569a9 100644
--- a/test/CodeGen/Mips/msa/compare.ll
+++ b/test/CodeGen/Mips/msa/compare.ll
@@ -1,4 +1,5 @@
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 define void @ceq_v16i8(<16 x i8>* %c, <16 x i8>* %a, <16 x i8>* %b) nounwind {
   ; CHECK: ceq_v16i8:
diff --git a/test/CodeGen/Mips/msa/compare_float.ll b/test/CodeGen/Mips/msa/compare_float.ll
index 4849928..2fc61f8 100644
--- a/test/CodeGen/Mips/msa/compare_float.ll
+++ b/test/CodeGen/Mips/msa/compare_float.ll
@@ -1,4 +1,5 @@
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 declare <4 x float> @llvm.mips.fmax.w(<4 x float>, <4 x float>) nounwind
 declare <2 x double> @llvm.mips.fmax.d(<2 x double>, <2 x double>) nounwind
diff --git a/test/CodeGen/Mips/msa/elm_copy.ll b/test/CodeGen/Mips/msa/elm_copy.ll
index 4bf041e..ed3e52c 100644
--- a/test/CodeGen/Mips/msa/elm_copy.ll
+++ b/test/CodeGen/Mips/msa/elm_copy.ll
@@ -2,6 +2,7 @@
 ; are element extraction operations.
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_copy_s_b_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_copy_s_b_RES  = global i32 0, align 16
diff --git a/test/CodeGen/Mips/msa/elm_cxcmsa.ll b/test/CodeGen/Mips/msa/elm_cxcmsa.ll
index dc1fcf7..8d6b0ee 100644
--- a/test/CodeGen/Mips/msa/elm_cxcmsa.ll
+++ b/test/CodeGen/Mips/msa/elm_cxcmsa.ll
@@ -2,6 +2,7 @@
 ; instruction format).
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 define i32 @msa_ir_cfcmsa_test() nounwind {
 entry:
diff --git a/test/CodeGen/Mips/msa/elm_insv.ll b/test/CodeGen/Mips/msa/elm_insv.ll
index a34002a..fa7ceaf 100644
--- a/test/CodeGen/Mips/msa/elm_insv.ll
+++ b/test/CodeGen/Mips/msa/elm_insv.ll
@@ -2,6 +2,7 @@
 ; instruction format.
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_insert_b_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_insert_b_ARG3 = global i32 27, align 16
diff --git a/test/CodeGen/Mips/msa/elm_move.ll b/test/CodeGen/Mips/msa/elm_move.ll
index 37fde15..98c06c7 100644
--- a/test/CodeGen/Mips/msa/elm_move.ll
+++ b/test/CodeGen/Mips/msa/elm_move.ll
@@ -2,6 +2,7 @@
 ; format).
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_move_vb_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_move_vb_RES  = global <16 x i8> <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>, align 16
diff --git a/test/CodeGen/Mips/msa/elm_shift_slide.ll b/test/CodeGen/Mips/msa/elm_shift_slide.ll
index d3863c8..3b3504b 100644
--- a/test/CodeGen/Mips/msa/elm_shift_slide.ll
+++ b/test/CodeGen/Mips/msa/elm_shift_slide.ll
@@ -1,7 +1,12 @@
+; Both endians should emit the same output for immediate instructions.
+; This is not currently true.
+; XFAIL: *
+
 ; Test the MSA intrinsics that are encoded with the ELM instruction format and
 ; are either shifts or slides.
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_sldi_b_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_sldi_b_RES  = global <16 x i8> <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>, align 16
diff --git a/test/CodeGen/Mips/msa/endian.ll b/test/CodeGen/Mips/msa/endian.ll
new file mode 100644
index 0000000..44d1925
--- /dev/null
+++ b/test/CodeGen/Mips/msa/endian.ll
@@ -0,0 +1,107 @@
+; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck -check-prefix=BIGENDIAN %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck -check-prefix=LITENDIAN %s
+
+@v16i8 = global <16 x i8> <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>
+@v8i16 = global <8 x i16> <i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0>
+@v4i32 = global <4 x i32> <i32 0, i32 0, i32 0, i32 0>
+@v2i64 = global <2 x i64> <i64 0, i64 0>
+
+define void @const_v16i8() nounwind {
+  ; LITENDIAN: .byte 0
+  ; LITENDIAN: .byte 1
+  ; LITENDIAN: .byte 2
+  ; LITENDIAN: .byte 3
+  ; LITENDIAN: .byte 4
+  ; LITENDIAN: .byte 5
+  ; LITENDIAN: .byte 6
+  ; LITENDIAN: .byte 7
+  ; LITENDIAN: .byte 8
+  ; LITENDIAN: .byte 9
+  ; LITENDIAN: .byte 10
+  ; LITENDIAN: .byte 11
+  ; LITENDIAN: .byte 12
+  ; LITENDIAN: .byte 13
+  ; LITENDIAN: .byte 14
+  ; LITENDIAN: .byte 15
+  ; LITENDIAN: const_v16i8:
+  ; BIGENDIAN: .byte 0
+  ; BIGENDIAN: .byte 1
+  ; BIGENDIAN: .byte 2
+  ; BIGENDIAN: .byte 3
+  ; BIGENDIAN: .byte 4
+  ; BIGENDIAN: .byte 5
+  ; BIGENDIAN: .byte 6
+  ; BIGENDIAN: .byte 7
+  ; BIGENDIAN: .byte 8
+  ; BIGENDIAN: .byte 9
+  ; BIGENDIAN: .byte 10
+  ; BIGENDIAN: .byte 11
+  ; BIGENDIAN: .byte 12
+  ; BIGENDIAN: .byte 13
+  ; BIGENDIAN: .byte 14
+  ; BIGENDIAN: .byte 15
+  ; BIGENDIAN: const_v16i8:
+
+  store volatile <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>, <16 x i8>*@v16i8
+
+  ret void
+}
+
+define void @const_v8i16() nounwind {
+  ; LITENDIAN: .2byte 0
+  ; LITENDIAN: .2byte 1
+  ; LITENDIAN: .2byte 2
+  ; LITENDIAN: .2byte 3
+  ; LITENDIAN: .2byte 4
+  ; LITENDIAN: .2byte 5
+  ; LITENDIAN: .2byte 6
+  ; LITENDIAN: .2byte 7
+  ; LITENDIAN: const_v8i16:
+  ; BIGENDIAN: .2byte 0
+  ; BIGENDIAN: .2byte 1
+  ; BIGENDIAN: .2byte 2
+  ; BIGENDIAN: .2byte 3
+  ; BIGENDIAN: .2byte 4
+  ; BIGENDIAN: .2byte 5
+  ; BIGENDIAN: .2byte 6
+  ; BIGENDIAN: .2byte 7
+  ; BIGENDIAN: const_v8i16:
+
+  store volatile <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>, <8 x i16>*@v8i16
+
+  ret void
+}
+
+define void @const_v4i32() nounwind {
+  ; LITENDIAN: .4byte 0
+  ; LITENDIAN: .4byte 1
+  ; LITENDIAN: .4byte 2
+  ; LITENDIAN: .4byte 3
+  ; LITENDIAN: const_v4i32:
+  ; BIGENDIAN: .4byte 0
+  ; BIGENDIAN: .4byte 1
+  ; BIGENDIAN: .4byte 2
+  ; BIGENDIAN: .4byte 3
+  ; BIGENDIAN: const_v4i32:
+
+  store volatile <4 x i32> <i32 0, i32 1, i32 2, i32 3>, <4 x i32>*@v4i32
+
+  ret void
+}
+
+define void @const_v2i64() nounwind {
+  ; LITENDIAN: .4byte 1
+  ; LITENDIAN: .4byte 0
+  ; LITENDIAN: .4byte 2
+  ; LITENDIAN: .4byte 0
+  ; LITENDIAN: const_v2i64:
+  ; BIGENDIAN: .4byte 0
+  ; BIGENDIAN: .4byte 1
+  ; BIGENDIAN: .4byte 0
+  ; BIGENDIAN: .4byte 2
+  ; BIGENDIAN: const_v2i64:
+
+  store volatile <2 x i64> <i64 1, i64 2>, <2 x i64>*@v2i64
+
+  ret void
+}
diff --git a/test/CodeGen/Mips/msa/i10.ll b/test/CodeGen/Mips/msa/i10.ll
index 5e8d259..c5a9617 100644
--- a/test/CodeGen/Mips/msa/i10.ll
+++ b/test/CodeGen/Mips/msa/i10.ll
@@ -1,6 +1,7 @@
 ; Test the MSA intrinsics that are encoded with the I10 instruction format.
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_bnz_b_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
 
diff --git a/test/CodeGen/Mips/msa/i5-a.ll b/test/CodeGen/Mips/msa/i5-a.ll
index 1fe83b3..f45df07 100644
--- a/test/CodeGen/Mips/msa/i5-a.ll
+++ b/test/CodeGen/Mips/msa/i5-a.ll
@@ -1,7 +1,12 @@
+; Both endians should emit the same output for immediate instructions.
+; This is not currently true.
+; XFAIL: *
+
 ; Test the MSA intrinsics that are encoded with the I5 instruction format.
 ; There are lots of these so this covers those beginning with 'a'
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_addvi_b_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_addvi_b_RES  = global <16 x i8> <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>, align 16
diff --git a/test/CodeGen/Mips/msa/i5-b.ll b/test/CodeGen/Mips/msa/i5-b.ll
index d1df6e9..f4477a0 100644
--- a/test/CodeGen/Mips/msa/i5-b.ll
+++ b/test/CodeGen/Mips/msa/i5-b.ll
@@ -2,6 +2,8 @@
 ; There are lots of these so this covers those beginning with 'b'
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
+; XFAIL: *
 
 @llvm_mips_bclri_b_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_bclri_b_RES  = global <16 x i8> <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>, align 16
diff --git a/test/CodeGen/Mips/msa/i5-c.ll b/test/CodeGen/Mips/msa/i5-c.ll
index d23a2df..d63b3cc 100644
--- a/test/CodeGen/Mips/msa/i5-c.ll
+++ b/test/CodeGen/Mips/msa/i5-c.ll
@@ -1,7 +1,12 @@
+; Both endians should emit the same output for immediate instructions.
+; This is not currently true.
+; XFAIL: *
+
 ; Test the MSA intrinsics that are encoded with the I5 instruction format.
 ; There are lots of these so this covers those beginning with 'c'
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_ceqi_b_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_ceqi_b_RES  = global <16 x i8> <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>, align 16
diff --git a/test/CodeGen/Mips/msa/i5-m.ll b/test/CodeGen/Mips/msa/i5-m.ll
index 3017cd4..74e698b 100644
--- a/test/CodeGen/Mips/msa/i5-m.ll
+++ b/test/CodeGen/Mips/msa/i5-m.ll
@@ -1,7 +1,12 @@
+; Both endians should emit the same output for immediate instructions.
+; This is not currently true.
+; XFAIL: *
+
 ; Test the MSA intrinsics that are encoded with the I5 instruction format.
 ; There are lots of these so this covers those beginning with 'm'
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_maxi_s_b_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_maxi_s_b_RES  = global <16 x i8> <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>, align 16
diff --git a/test/CodeGen/Mips/msa/i5-s.ll b/test/CodeGen/Mips/msa/i5-s.ll
index 77f52bf..60ba8e1 100644
--- a/test/CodeGen/Mips/msa/i5-s.ll
+++ b/test/CodeGen/Mips/msa/i5-s.ll
@@ -1,7 +1,12 @@
+; Both endians should emit the same output for immediate instructions.
+; This is not currently true.
+; XFAIL: *
+
 ; Test the MSA intrinsics that are encoded with the I5 instruction format.
 ; There are lots of these so this covers those beginning with 's'
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_subvi_b_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_subvi_b_RES  = global <16 x i8> <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>, align 16
diff --git a/test/CodeGen/Mips/msa/i5_ld_st.ll b/test/CodeGen/Mips/msa/i5_ld_st.ll
index 7ec9326..7cc55f2 100644
--- a/test/CodeGen/Mips/msa/i5_ld_st.ll
+++ b/test/CodeGen/Mips/msa/i5_ld_st.ll
@@ -2,6 +2,7 @@
 ; are loads or stores.
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_ld_b_ARG = 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_ld_b_RES = global <16 x i8> <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>, align 16
diff --git a/test/CodeGen/Mips/msa/i8.ll b/test/CodeGen/Mips/msa/i8.ll
index f3e8dfc..d2931a7 100644
--- a/test/CodeGen/Mips/msa/i8.ll
+++ b/test/CodeGen/Mips/msa/i8.ll
@@ -1,6 +1,7 @@
 ; Test the MSA intrinsics that are encoded with the I8 instruction format.
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_andi_b_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_andi_b_RES  = global <16 x i8> <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>, align 16
diff --git a/test/CodeGen/Mips/msa/llvm-stress-s1935737938.ll b/test/CodeGen/Mips/msa/llvm-stress-s1935737938.ll
index 23675f5..f9cab03 100644
--- a/test/CodeGen/Mips/msa/llvm-stress-s1935737938.ll
+++ b/test/CodeGen/Mips/msa/llvm-stress-s1935737938.ll
@@ -1,5 +1,7 @@
 ; RUN: llc -march=mips < %s
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s
+; RUN: llc -march=mipsel < %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s
 
 ; This test originally failed for MSA with a
 ; `Opc && "Cannot copy registers"' assertion.
diff --git a/test/CodeGen/Mips/msa/llvm-stress-s3997499501.ll b/test/CodeGen/Mips/msa/llvm-stress-s3997499501.ll
index 6528113..bef75f3 100644
--- a/test/CodeGen/Mips/msa/llvm-stress-s3997499501.ll
+++ b/test/CodeGen/Mips/msa/llvm-stress-s3997499501.ll
@@ -1,5 +1,7 @@
 ; RUN: llc -march=mips < %s
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s
+; RUN: llc -march=mipsel < %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s
 
 ; This test originally failed to select instructions for extract_vector_elt for
 ; v4f32 on MSA.
diff --git a/test/CodeGen/Mips/msa/llvm-stress-s525530439.ll b/test/CodeGen/Mips/msa/llvm-stress-s525530439.ll
index 685c407..697871d 100644
--- a/test/CodeGen/Mips/msa/llvm-stress-s525530439.ll
+++ b/test/CodeGen/Mips/msa/llvm-stress-s525530439.ll
@@ -1,5 +1,7 @@
 ; RUN: llc -march=mips < %s
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s
+; RUN: llc -march=mipsel < %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s
 
 ; This test originally failed for MSA with a
 ; `Num < NumOperands && "Invalid child # of SDNode!"' assertion.
diff --git a/test/CodeGen/Mips/msa/llvm-stress-s997348632.ll b/test/CodeGen/Mips/msa/llvm-stress-s997348632.ll
index 399d3a0..dc4200a 100644
--- a/test/CodeGen/Mips/msa/llvm-stress-s997348632.ll
+++ b/test/CodeGen/Mips/msa/llvm-stress-s997348632.ll
@@ -1,5 +1,7 @@
 ; RUN: llc -march=mips < %s
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s
+; RUN: llc -march=mipsel < %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s
 
 ; This test originally failed to select instructions for extract_vector_elt for
 ; v2f64 on MSA.
diff --git a/test/CodeGen/Mips/msa/llvm-stress-sz1-s742806235.ll b/test/CodeGen/Mips/msa/llvm-stress-sz1-s742806235.ll
index 36afffe..8c4fcba 100644
--- a/test/CodeGen/Mips/msa/llvm-stress-sz1-s742806235.ll
+++ b/test/CodeGen/Mips/msa/llvm-stress-sz1-s742806235.ll
@@ -1,5 +1,7 @@
 ; RUN: llc -march=mips < %s
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s
+; RUN: llc -march=mipsel < %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s
 
 ; This test originally failed to select code for a truncstore of a
 ; build_vector.
diff --git a/test/CodeGen/Mips/msa/shuffle.ll b/test/CodeGen/Mips/msa/shuffle.ll
index b5df0e0..316c669 100644
--- a/test/CodeGen/Mips/msa/shuffle.ll
+++ b/test/CodeGen/Mips/msa/shuffle.ll
@@ -1,4 +1,5 @@
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 define void @vshf_v16i8_0(<16 x i8>* %c, <16 x i8>* %a, <16 x i8>* %b) nounwind {
   ; CHECK: vshf_v16i8_0:
diff --git a/test/CodeGen/Mips/msa/spill.ll b/test/CodeGen/Mips/msa/spill.ll
index d73a8b5..66f896a 100644
--- a/test/CodeGen/Mips/msa/spill.ll
+++ b/test/CodeGen/Mips/msa/spill.ll
@@ -2,6 +2,7 @@
 ; to have 33 live MSA registers simultaneously
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 define i32 @test_i8(<16 x i8>* %p0, <16 x i8>* %q1) nounwind {
 entry:
diff --git a/test/CodeGen/Mips/msa/vecs10.ll b/test/CodeGen/Mips/msa/vecs10.ll
index a00a488..e22e075 100644
--- a/test/CodeGen/Mips/msa/vecs10.ll
+++ b/test/CodeGen/Mips/msa/vecs10.ll
@@ -1,6 +1,7 @@
 ; Test the MSA intrinsics that are encoded with the VECS10 instruction format.
 
 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
+; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
 
 @llvm_mips_bnz_v_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