[LLVM][MIPS] Regenerate ct.select test CHECK linesusers/wizardengineer/ct-select-mips

Update CHECK lines to match the new constant-time AND/OR/XOR expansion from the CT_SELECT legalization fix.
author: wizardengineer <juliuswoosebert@gmail.com> 2026-03-07 15:38:15 -0500
committer: wizardengineer <juliuswoosebert@gmail.com> 2026-03-07 16:39:57 -0500
commit: 9d45e799ab7c4dea36c4c8df5ed757a92886c814 (patch)
tree: 1d58dc1b5a08bf10e6732d72b8021280f759ff2a
parent: 89b18c0965ccec867ad8e234c77ac7f83c7edfc5 (diff)
download: llvm-users/wizardengineer/ct-select-mips.tar.gz
llvm-users/wizardengineer/ct-select-mips.tar.bz2
llvm-users/wizardengineer/ct-select-mips.zip
5 files changed, 617 insertions, 755 deletions
diff --git a/llvm/test/CodeGen/Mips/ctselect-fallback-edge-cases.ll b/llvm/test/CodeGen/Mips/ctselect-fallback-edge-cases.ll
index f1831a625d4a..401a742c27ea 100644
--- a/llvm/test/CodeGen/Mips/ctselect-fallback-edge-cases.ll
+++ b/llvm/test/CodeGen/Mips/ctselect-fallback-edge-cases.ll
@@ -8,22 +8,24 @@
 define i1 @test_ctselect_i1(i1 %cond, i1 %a, i1 %b) {
 ; M32-LABEL: test_ctselect_i1:
 ; M32:       # %bb.0:
-; M32-NEXT:    xori $2, $4, 1
-; M32-NEXT:    and $1, $4, $5
-; M32-NEXT:    and $2, $2, $6
+; M32-NEXT:    andi $2, $4, 1
+; M32-NEXT:    xor $1, $5, $6
+; M32-NEXT:    negu $2, $2
+; M32-NEXT:    and $1, $1, $2
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    or $2, $1, $2
+; M32-NEXT:    xor $2, $6, $1
 ;
 ; M64-LABEL: test_ctselect_i1:
 ; M64:       # %bb.0:
-; M64-NEXT:    sll $2, $4, 0
-; M64-NEXT:    sll $1, $6, 0
-; M64-NEXT:    xori $2, $2, 1
-; M64-NEXT:    and $1, $2, $1
-; M64-NEXT:    and $2, $4, $5
+; M64-NEXT:    sll $1, $4, 0
+; M64-NEXT:    xor $2, $5, $6
+; M64-NEXT:    andi $1, $1, 1
 ; M64-NEXT:    sll $2, $2, 0
+; M64-NEXT:    negu $1, $1
+; M64-NEXT:    and $1, $2, $1
+; M64-NEXT:    sll $2, $6, 0
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $2, $1
+; M64-NEXT:    xor $2, $2, $1
   %result = call i1 @llvm.ct.select.i1(i1 %cond, i1 %a, i1 %b)
   ret i1 %result
 }
@@ -32,30 +34,18 @@ define i1 @test_ctselect_i1(i1 %cond, i1 %a, i1 %b) {
 define i32 @test_ctselect_extremal_values(i1 %cond) {
 ; M32-LABEL: test_ctselect_extremal_values:
 ; M32:       # %bb.0:
-; M32-NEXT:    lui $3, 32767
 ; M32-NEXT:    andi $1, $4, 1
-; M32-NEXT:    negu $2, $1
-; M32-NEXT:    ori $3, $3, 65535
-; M32-NEXT:    addiu $1, $1, -1
-; M32-NEXT:    and $2, $2, $3
-; M32-NEXT:    lui $3, 32768
-; M32-NEXT:    and $1, $1, $3
+; M32-NEXT:    lui $2, 32768
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    or $2, $2, $1
+; M32-NEXT:    subu $2, $2, $1
 ;
 ; M64-LABEL: test_ctselect_extremal_values:
 ; M64:       # %bb.0:
 ; M64-NEXT:    sll $1, $4, 0
-; M64-NEXT:    lui $3, 32767
+; M64-NEXT:    lui $2, 32768
 ; M64-NEXT:    andi $1, $1, 1
-; M64-NEXT:    ori $3, $3, 65535
-; M64-NEXT:    negu $2, $1
-; M64-NEXT:    addiu $1, $1, -1
-; M64-NEXT:    and $2, $2, $3
-; M64-NEXT:    lui $3, 32768
-; M64-NEXT:    and $1, $1, $3
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $2, $1
+; M64-NEXT:    subu $2, $2, $1
   %result = call i32 @llvm.ct.select.i32(i1 %cond, i32 2147483647, i32 -2147483648)
   ret i32 %result
 }
@@ -67,14 +57,14 @@ define ptr @test_ctselect_null_ptr(i1 %cond, ptr %ptr) {
 ; M32-NEXT:    andi $1, $4, 1
 ; M32-NEXT:    negu $1, $1
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    and $2, $1, $5
+; M32-NEXT:    and $2, $5, $1
 ;
 ; M64-LABEL: test_ctselect_null_ptr:
 ; M64:       # %bb.0:
 ; M64-NEXT:    andi $1, $4, 1
 ; M64-NEXT:    dnegu $1, $1
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    and $2, $1, $5
+; M64-NEXT:    and $2, $5, $1
   %result = call ptr @llvm.ct.select.p0(i1 %cond, ptr %ptr, ptr null)
   ret ptr %result
 }
@@ -83,23 +73,21 @@ define ptr @test_ctselect_null_ptr(i1 %cond, ptr %ptr) {
 define ptr @test_ctselect_function_ptr(i1 %cond, ptr %func1, ptr %func2) {
 ; M32-LABEL: test_ctselect_function_ptr:
 ; M32:       # %bb.0:
-; M32-NEXT:    andi $1, $4, 1
-; M32-NEXT:    negu $2, $1
-; M32-NEXT:    addiu $1, $1, -1
-; M32-NEXT:    and $2, $2, $5
-; M32-NEXT:    and $1, $1, $6
+; M32-NEXT:    andi $2, $4, 1
+; M32-NEXT:    xor $1, $5, $6
+; M32-NEXT:    negu $2, $2
+; M32-NEXT:    and $1, $1, $2
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    or $2, $2, $1
+; M32-NEXT:    xor $2, $6, $1
 ;
 ; M64-LABEL: test_ctselect_function_ptr:
 ; M64:       # %bb.0:
-; M64-NEXT:    andi $1, $4, 1
-; M64-NEXT:    dnegu $2, $1
-; M64-NEXT:    daddiu $1, $1, -1
-; M64-NEXT:    and $2, $2, $5
-; M64-NEXT:    and $1, $1, $6
+; M64-NEXT:    andi $2, $4, 1
+; M64-NEXT:    xor $1, $5, $6
+; M64-NEXT:    dnegu $2, $2
+; M64-NEXT:    and $1, $1, $2
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $2, $1
+; M64-NEXT:    xor $2, $6, $1
   %result = call ptr @llvm.ct.select.p0(i1 %cond, ptr %func1, ptr %func2)
   ret ptr %result
 }
@@ -108,26 +96,25 @@ define ptr @test_ctselect_function_ptr(i1 %cond, ptr %func1, ptr %func2) {
 define ptr @test_ctselect_ptr_cmp(ptr %p1, ptr %p2, ptr %a, ptr %b) {
 ; M32-LABEL: test_ctselect_ptr_cmp:
 ; M32:       # %bb.0:
-; M32-NEXT:    xor $1, $4, $5
-; M32-NEXT:    sltu $1, $zero, $1
-; M32-NEXT:    addiu $1, $1, -1
-; M32-NEXT:    and $2, $1, $6
-; M32-NEXT:    not $1, $1
-; M32-NEXT:    and $1, $1, $7
+; M32-NEXT:    xor $2, $4, $5
+; M32-NEXT:    xor $1, $6, $7
+; M32-NEXT:    sltiu $2, $2, 1
+; M32-NEXT:    negu $2, $2
+; M32-NEXT:    and $1, $1, $2
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    or $2, $2, $1
+; M32-NEXT:    xor $2, $7, $1
 ;
 ; M64-LABEL: test_ctselect_ptr_cmp:
 ; M64:       # %bb.0:
-; M64-NEXT:    xor $1, $4, $5
-; M64-NEXT:    daddiu $3, $zero, -1
-; M64-NEXT:    daddiu $2, $zero, -1
-; M64-NEXT:    movn $3, $zero, $1
-; M64-NEXT:    xor $2, $3, $2
-; M64-NEXT:    and $1, $3, $6
-; M64-NEXT:    and $2, $2, $7
+; M64-NEXT:    xor $2, $4, $5
+; M64-NEXT:    xor $1, $6, $7
+; M64-NEXT:    sltiu $2, $2, 1
+; M64-NEXT:    dsll $2, $2, 32
+; M64-NEXT:    dsrl $2, $2, 32
+; M64-NEXT:    dnegu $2, $2
+; M64-NEXT:    and $1, $1, $2
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $1, $2
+; M64-NEXT:    xor $2, $7, $1
   %cmp = icmp eq ptr %p1, %p2
   %result = call ptr @llvm.ct.select.p0(i1 %cmp, ptr %a, ptr %b)
   ret ptr %result
@@ -139,23 +126,21 @@ define ptr @test_ctselect_ptr_cmp(ptr %p1, ptr %p2, ptr %a, ptr %b) {
 define ptr @test_ctselect_struct_ptr(i1 %cond, ptr %a, ptr %b) {
 ; M32-LABEL: test_ctselect_struct_ptr:
 ; M32:       # %bb.0:
-; M32-NEXT:    andi $1, $4, 1
-; M32-NEXT:    negu $2, $1
-; M32-NEXT:    addiu $1, $1, -1
-; M32-NEXT:    and $2, $2, $5
-; M32-NEXT:    and $1, $1, $6
+; M32-NEXT:    andi $2, $4, 1
+; M32-NEXT:    xor $1, $5, $6
+; M32-NEXT:    negu $2, $2
+; M32-NEXT:    and $1, $1, $2
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    or $2, $2, $1
+; M32-NEXT:    xor $2, $6, $1
 ;
 ; M64-LABEL: test_ctselect_struct_ptr:
 ; M64:       # %bb.0:
-; M64-NEXT:    andi $1, $4, 1
-; M64-NEXT:    dnegu $2, $1
-; M64-NEXT:    daddiu $1, $1, -1
-; M64-NEXT:    and $2, $2, $5
-; M64-NEXT:    and $1, $1, $6
+; M64-NEXT:    andi $2, $4, 1
+; M64-NEXT:    xor $1, $5, $6
+; M64-NEXT:    dnegu $2, $2
+; M64-NEXT:    and $1, $1, $2
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $2, $1
+; M64-NEXT:    xor $2, $6, $1
   %result = call ptr @llvm.ct.select.p0(i1 %cond, ptr %a, ptr %b)
   ret ptr %result
 }
@@ -164,73 +149,65 @@ define ptr @test_ctselect_struct_ptr(i1 %cond, ptr %a, ptr %b) {
 define i32 @test_ctselect_deeply_nested(i1 %c1, i1 %c2, i1 %c3, i1 %c4, i32 %a, i32 %b, i32 %c, i32 %d, i32 %e) {
 ; M32-LABEL: test_ctselect_deeply_nested:
 ; M32:       # %bb.0:
-; M32-NEXT:    andi $1, $4, 1
-; M32-NEXT:    lw $3, 16($sp)
-; M32-NEXT:    lw $9, 32($sp)
-; M32-NEXT:    lw $8, 28($sp)
-; M32-NEXT:    negu $2, $1
-; M32-NEXT:    addiu $1, $1, -1
+; M32-NEXT:    lw $1, 20($sp)
+; M32-NEXT:    lw $2, 16($sp)
+; M32-NEXT:    andi $3, $4, 1
+; M32-NEXT:    andi $4, $6, 1
+; M32-NEXT:    lw $6, 28($sp)
+; M32-NEXT:    negu $3, $3
+; M32-NEXT:    xor $2, $2, $1
 ; M32-NEXT:    and $2, $2, $3
-; M32-NEXT:    lw $3, 20($sp)
-; M32-NEXT:    and $1, $1, $3
 ; M32-NEXT:    andi $3, $5, 1
-; M32-NEXT:    or $1, $2, $1
-; M32-NEXT:    andi $2, $6, 1
-; M32-NEXT:    andi $6, $7, 1
-; M32-NEXT:    negu $4, $3
-; M32-NEXT:    addiu $3, $3, -1
-; M32-NEXT:    addiu $7, $6, -1
-; M32-NEXT:    and $1, $4, $1
-; M32-NEXT:    addiu $5, $2, -1
-; M32-NEXT:    negu $2, $2
-; M32-NEXT:    negu $6, $6
-; M32-NEXT:    and $4, $7, $9
-; M32-NEXT:    lw $7, 24($sp)
-; M32-NEXT:    and $5, $5, $8
-; M32-NEXT:    and $3, $3, $7
-; M32-NEXT:    or $1, $1, $3
-; M32-NEXT:    and $1, $2, $1
-; M32-NEXT:    or $1, $1, $5
-; M32-NEXT:    and $1, $6, $1
+; M32-NEXT:    lw $5, 32($sp)
+; M32-NEXT:    xor $1, $1, $2
+; M32-NEXT:    lw $2, 24($sp)
+; M32-NEXT:    negu $3, $3
+; M32-NEXT:    xor $1, $1, $2
+; M32-NEXT:    and $1, $1, $3
+; M32-NEXT:    andi $3, $7, 1
+; M32-NEXT:    xor $1, $2, $1
+; M32-NEXT:    negu $2, $4
+; M32-NEXT:    negu $3, $3
+; M32-NEXT:    xor $1, $1, $6
+; M32-NEXT:    and $1, $1, $2
+; M32-NEXT:    xor $1, $6, $1
+; M32-NEXT:    xor $1, $1, $5
+; M32-NEXT:    and $1, $1, $3
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    or $2, $1, $4
+; M32-NEXT:    xor $2, $5, $1
 ;
 ; M64-LABEL: test_ctselect_deeply_nested:
 ; M64:       # %bb.0:
 ; M64-NEXT:    sll $1, $4, 0
-; M64-NEXT:    sll $3, $8, 0
-; M64-NEXT:    sll $4, $5, 0
-; M64-NEXT:    lw $8, 0($sp)
+; M64-NEXT:    xor $2, $8, $9
+; M64-NEXT:    sll $5, $5, 0
+; M64-NEXT:    sll $3, $6, 0
+; M64-NEXT:    sll $6, $11, 0
+; M64-NEXT:    sll $4, $7, 0
+; M64-NEXT:    lw $7, 0($sp)
 ; M64-NEXT:    andi $1, $1, 1
+; M64-NEXT:    sll $2, $2, 0
+; M64-NEXT:    andi $5, $5, 1
+; M64-NEXT:    andi $3, $3, 1
 ; M64-NEXT:    andi $4, $4, 1
-; M64-NEXT:    negu $2, $1
-; M64-NEXT:    addiu $1, $1, -1
-; M64-NEXT:    negu $5, $4
-; M64-NEXT:    addiu $4, $4, -1
-; M64-NEXT:    and $2, $2, $3
-; M64-NEXT:    sll $3, $9, 0
-; M64-NEXT:    and $1, $1, $3
-; M64-NEXT:    sll $3, $11, 0
-; M64-NEXT:    or $1, $2, $1
-; M64-NEXT:    sll $2, $6, 0
-; M64-NEXT:    sll $6, $7, 0
-; M64-NEXT:    andi $2, $2, 1
-; M64-NEXT:    and $1, $5, $1
-; M64-NEXT:    andi $6, $6, 1
-; M64-NEXT:    addiu $5, $2, -1
-; M64-NEXT:    negu $2, $2
-; M64-NEXT:    addiu $7, $6, -1
-; M64-NEXT:    negu $6, $6
-; M64-NEXT:    and $3, $5, $3
-; M64-NEXT:    sll $5, $10, 0
-; M64-NEXT:    and $7, $7, $8
-; M64-NEXT:    and $4, $4, $5
-; M64-NEXT:    or $1, $1, $4
+; M64-NEXT:    negu $1, $1
+; M64-NEXT:    negu $5, $5
+; M64-NEXT:    negu $4, $4
 ; M64-NEXT:    and $1, $2, $1
-; M64-NEXT:    or $1, $1, $3
-; M64-NEXT:    and $1, $6, $1
+; M64-NEXT:    sll $2, $9, 0
+; M64-NEXT:    xor $1, $2, $1
+; M64-NEXT:    sll $2, $10, 0
+; M64-NEXT:    xor $1, $1, $2
+; M64-NEXT:    and $1, $1, $5
+; M64-NEXT:    xor $1, $2, $1
+; M64-NEXT:    negu $2, $3
+; M64-NEXT:    xor $1, $1, $6
+; M64-NEXT:    and $1, $1, $2
+; M64-NEXT:    xor $1, $6, $1
+; M64-NEXT:    xor $1, $1, $7
+; M64-NEXT:    and $1, $1, $4
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $1, $7
+; M64-NEXT:    xor $2, $7, $1
   %sel1 = call i32 @llvm.ct.select.i32(i1 %c1, i32 %a, i32 %b)
   %sel2 = call i32 @llvm.ct.select.i32(i1 %c2, i32 %sel1, i32 %c)
   %sel3 = call i32 @llvm.ct.select.i32(i1 %c3, i32 %sel2, i32 %d)
diff --git a/llvm/test/CodeGen/Mips/ctselect-fallback-patterns.ll b/llvm/test/CodeGen/Mips/ctselect-fallback-patterns.ll
index 2e65e586ce5f..a1c5d524c693 100644
--- a/llvm/test/CodeGen/Mips/ctselect-fallback-patterns.ll
+++ b/llvm/test/CodeGen/Mips/ctselect-fallback-patterns.ll
@@ -6,16 +6,18 @@
 define i32 @test_ctselect_smin_zero(i32 %x) {
 ; M32-LABEL: test_ctselect_smin_zero:
 ; M32:       # %bb.0:
-; M32-NEXT:    sra $1, $4, 31
+; M32-NEXT:    slti $1, $4, 0
+; M32-NEXT:    negu $1, $1
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    and $2, $1, $4
+; M32-NEXT:    and $2, $4, $1
 ;
 ; M64-LABEL: test_ctselect_smin_zero:
 ; M64:       # %bb.0:
 ; M64-NEXT:    sll $1, $4, 0
-; M64-NEXT:    sra $2, $1, 31
+; M64-NEXT:    slti $2, $1, 0
+; M64-NEXT:    negu $2, $2
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    and $2, $2, $1
+; M64-NEXT:    and $2, $1, $2
   %cmp = icmp slt i32 %x, 0
   %result = call i32 @llvm.ct.select.i32(i1 %cmp, i32 %x, i32 0)
   ret i32 %result
@@ -25,17 +27,18 @@ define i32 @test_ctselect_smin_zero(i32 %x) {
 define i32 @test_ctselect_smax_zero(i32 %x) {
 ; M32-LABEL: test_ctselect_smax_zero:
 ; M32:       # %bb.0:
-; M32-NEXT:    slti $1, $4, 1
-; M32-NEXT:    movn $4, $zero, $1
+; M32-NEXT:    slt $1, $zero, $4
+; M32-NEXT:    negu $1, $1
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    move $2, $4
+; M32-NEXT:    and $2, $4, $1
 ;
 ; M64-LABEL: test_ctselect_smax_zero:
 ; M64:       # %bb.0:
-; M64-NEXT:    sll $2, $4, 0
-; M64-NEXT:    slti $1, $2, 1
+; M64-NEXT:    sll $1, $4, 0
+; M64-NEXT:    slt $2, $zero, $1
+; M64-NEXT:    negu $2, $2
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    movn $2, $zero, $1
+; M64-NEXT:    and $2, $1, $2
   %cmp = icmp sgt i32 %x, 0
   %result = call i32 @llvm.ct.select.i32(i1 %cmp, i32 %x, i32 0)
   ret i32 %result
@@ -45,27 +48,23 @@ define i32 @test_ctselect_smax_zero(i32 %x) {
 define i32 @test_ctselect_smin_generic(i32 %x, i32 %y) {
 ; M32-LABEL: test_ctselect_smin_generic:
 ; M32:       # %bb.0:
-; M32-NEXT:    slt $1, $4, $5
-; M32-NEXT:    xori $1, $1, 1
-; M32-NEXT:    addiu $1, $1, -1
-; M32-NEXT:    and $2, $1, $4
-; M32-NEXT:    not $1, $1
-; M32-NEXT:    and $1, $1, $5
+; M32-NEXT:    slt $2, $4, $5
+; M32-NEXT:    xor $1, $4, $5
+; M32-NEXT:    negu $2, $2
+; M32-NEXT:    and $1, $1, $2
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    or $2, $2, $1
+; M32-NEXT:    xor $2, $5, $1
 ;
 ; M64-LABEL: test_ctselect_smin_generic:
 ; M64:       # %bb.0:
 ; M64-NEXT:    sll $1, $5, 0
 ; M64-NEXT:    sll $2, $4, 0
-; M64-NEXT:    slt $3, $2, $1
-; M64-NEXT:    xori $3, $3, 1
-; M64-NEXT:    addiu $3, $3, -1
+; M64-NEXT:    xor $3, $2, $1
+; M64-NEXT:    slt $2, $2, $1
+; M64-NEXT:    negu $2, $2
 ; M64-NEXT:    and $2, $3, $2
-; M64-NEXT:    not $3, $3
-; M64-NEXT:    and $1, $3, $1
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $2, $1
+; M64-NEXT:    xor $2, $1, $2
   %cmp = icmp slt i32 %x, %y
   %result = call i32 @llvm.ct.select.i32(i1 %cmp, i32 %x, i32 %y)
   ret i32 %result
@@ -75,27 +74,23 @@ define i32 @test_ctselect_smin_generic(i32 %x, i32 %y) {
 define i32 @test_ctselect_smax_generic(i32 %x, i32 %y) {
 ; M32-LABEL: test_ctselect_smax_generic:
 ; M32:       # %bb.0:
-; M32-NEXT:    slt $1, $5, $4
-; M32-NEXT:    xori $1, $1, 1
-; M32-NEXT:    addiu $1, $1, -1
-; M32-NEXT:    and $2, $1, $4
-; M32-NEXT:    not $1, $1
-; M32-NEXT:    and $1, $1, $5
+; M32-NEXT:    slt $2, $5, $4
+; M32-NEXT:    xor $1, $4, $5
+; M32-NEXT:    negu $2, $2
+; M32-NEXT:    and $1, $1, $2
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    or $2, $2, $1
+; M32-NEXT:    xor $2, $5, $1
 ;
 ; M64-LABEL: test_ctselect_smax_generic:
 ; M64:       # %bb.0:
-; M64-NEXT:    sll $1, $4, 0
-; M64-NEXT:    sll $2, $5, 0
-; M64-NEXT:    slt $3, $2, $1
-; M64-NEXT:    xori $3, $3, 1
-; M64-NEXT:    addiu $3, $3, -1
-; M64-NEXT:    and $1, $3, $1
-; M64-NEXT:    not $3, $3
+; M64-NEXT:    sll $1, $5, 0
+; M64-NEXT:    sll $2, $4, 0
+; M64-NEXT:    xor $3, $2, $1
+; M64-NEXT:    slt $2, $1, $2
+; M64-NEXT:    negu $2, $2
 ; M64-NEXT:    and $2, $3, $2
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $1, $2
+; M64-NEXT:    xor $2, $1, $2
   %cmp = icmp sgt i32 %x, %y
   %result = call i32 @llvm.ct.select.i32(i1 %cmp, i32 %x, i32 %y)
   ret i32 %result
@@ -105,27 +100,23 @@ define i32 @test_ctselect_smax_generic(i32 %x, i32 %y) {
 define i32 @test_ctselect_umin_generic(i32 %x, i32 %y) {
 ; M32-LABEL: test_ctselect_umin_generic:
 ; M32:       # %bb.0:
-; M32-NEXT:    sltu $1, $4, $5
-; M32-NEXT:    xori $1, $1, 1
-; M32-NEXT:    addiu $1, $1, -1
-; M32-NEXT:    and $2, $1, $4
-; M32-NEXT:    not $1, $1
-; M32-NEXT:    and $1, $1, $5
+; M32-NEXT:    sltu $2, $4, $5
+; M32-NEXT:    xor $1, $4, $5
+; M32-NEXT:    negu $2, $2
+; M32-NEXT:    and $1, $1, $2
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    or $2, $2, $1
+; M32-NEXT:    xor $2, $5, $1
 ;
 ; M64-LABEL: test_ctselect_umin_generic:
 ; M64:       # %bb.0:
 ; M64-NEXT:    sll $1, $5, 0
 ; M64-NEXT:    sll $2, $4, 0
-; M64-NEXT:    sltu $3, $2, $1
-; M64-NEXT:    xori $3, $3, 1
-; M64-NEXT:    addiu $3, $3, -1
+; M64-NEXT:    xor $3, $2, $1
+; M64-NEXT:    sltu $2, $2, $1
+; M64-NEXT:    negu $2, $2
 ; M64-NEXT:    and $2, $3, $2
-; M64-NEXT:    not $3, $3
-; M64-NEXT:    and $1, $3, $1
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $2, $1
+; M64-NEXT:    xor $2, $1, $2
   %cmp = icmp ult i32 %x, %y
   %result = call i32 @llvm.ct.select.i32(i1 %cmp, i32 %x, i32 %y)
   ret i32 %result
@@ -135,27 +126,23 @@ define i32 @test_ctselect_umin_generic(i32 %x, i32 %y) {
 define i32 @test_ctselect_umax_generic(i32 %x, i32 %y) {
 ; M32-LABEL: test_ctselect_umax_generic:
 ; M32:       # %bb.0:
-; M32-NEXT:    sltu $1, $5, $4
-; M32-NEXT:    xori $1, $1, 1
-; M32-NEXT:    addiu $1, $1, -1
-; M32-NEXT:    and $2, $1, $4
-; M32-NEXT:    not $1, $1
-; M32-NEXT:    and $1, $1, $5
+; M32-NEXT:    sltu $2, $5, $4
+; M32-NEXT:    xor $1, $4, $5
+; M32-NEXT:    negu $2, $2
+; M32-NEXT:    and $1, $1, $2
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    or $2, $2, $1
+; M32-NEXT:    xor $2, $5, $1
 ;
 ; M64-LABEL: test_ctselect_umax_generic:
 ; M64:       # %bb.0:
-; M64-NEXT:    sll $1, $4, 0
-; M64-NEXT:    sll $2, $5, 0
-; M64-NEXT:    sltu $3, $2, $1
-; M64-NEXT:    xori $3, $3, 1
-; M64-NEXT:    addiu $3, $3, -1
-; M64-NEXT:    and $1, $3, $1
-; M64-NEXT:    not $3, $3
+; M64-NEXT:    sll $1, $5, 0
+; M64-NEXT:    sll $2, $4, 0
+; M64-NEXT:    xor $3, $2, $1
+; M64-NEXT:    sltu $2, $1, $2
+; M64-NEXT:    negu $2, $2
 ; M64-NEXT:    and $2, $3, $2
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $1, $2
+; M64-NEXT:    xor $2, $1, $2
   %cmp = icmp ugt i32 %x, %y
   %result = call i32 @llvm.ct.select.i32(i1 %cmp, i32 %x, i32 %y)
   ret i32 %result
@@ -165,24 +152,24 @@ define i32 @test_ctselect_umax_generic(i32 %x, i32 %y) {
 define i32 @test_ctselect_abs(i32 %x) {
 ; M32-LABEL: test_ctselect_abs:
 ; M32:       # %bb.0:
-; M32-NEXT:    negu $1, $4
-; M32-NEXT:    sra $2, $4, 31
+; M32-NEXT:    slti $1, $4, 0
+; M32-NEXT:    negu $2, $4
+; M32-NEXT:    negu $1, $1
+; M32-NEXT:    xor $2, $2, $4
 ; M32-NEXT:    and $1, $2, $1
-; M32-NEXT:    not $2, $2
-; M32-NEXT:    and $2, $2, $4
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    or $2, $1, $2
+; M32-NEXT:    xor $2, $4, $1
 ;
 ; M64-LABEL: test_ctselect_abs:
 ; M64:       # %bb.0:
 ; M64-NEXT:    sll $1, $4, 0
-; M64-NEXT:    negu $2, $1
-; M64-NEXT:    sra $3, $1, 31
+; M64-NEXT:    slti $2, $1, 0
+; M64-NEXT:    negu $3, $1
+; M64-NEXT:    negu $2, $2
+; M64-NEXT:    xor $3, $3, $1
 ; M64-NEXT:    and $2, $3, $2
-; M64-NEXT:    not $3, $3
-; M64-NEXT:    and $1, $3, $1
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $2, $1
+; M64-NEXT:    xor $2, $1, $2
   %neg = sub i32 0, %x
   %cmp = icmp slt i32 %x, 0
   %result = call i32 @llvm.ct.select.i32(i1 %cmp, i32 %neg, i32 %x)
@@ -193,24 +180,24 @@ define i32 @test_ctselect_abs(i32 %x) {
 define i32 @test_ctselect_nabs(i32 %x) {
 ; M32-LABEL: test_ctselect_nabs:
 ; M32:       # %bb.0:
-; M32-NEXT:    sra $1, $4, 31
-; M32-NEXT:    negu $3, $4
-; M32-NEXT:    and $2, $1, $4
-; M32-NEXT:    not $1, $1
-; M32-NEXT:    and $1, $1, $3
+; M32-NEXT:    slti $1, $4, 0
+; M32-NEXT:    negu $2, $4
+; M32-NEXT:    negu $1, $1
+; M32-NEXT:    xor $3, $4, $2
+; M32-NEXT:    and $1, $3, $1
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    or $2, $2, $1
+; M32-NEXT:    xor $2, $2, $1
 ;
 ; M64-LABEL: test_ctselect_nabs:
 ; M64:       # %bb.0:
 ; M64-NEXT:    sll $1, $4, 0
-; M64-NEXT:    sra $2, $1, 31
-; M64-NEXT:    and $3, $2, $1
-; M64-NEXT:    negu $1, $1
-; M64-NEXT:    not $2, $2
-; M64-NEXT:    and $1, $2, $1
+; M64-NEXT:    slti $2, $1, 0
+; M64-NEXT:    negu $3, $1
+; M64-NEXT:    negu $2, $2
+; M64-NEXT:    xor $1, $1, $3
+; M64-NEXT:    and $1, $1, $2
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $3, $1
+; M64-NEXT:    xor $2, $3, $1
   %neg = sub i32 0, %x
   %cmp = icmp slt i32 %x, 0
   %result = call i32 @llvm.ct.select.i32(i1 %cmp, i32 %x, i32 %neg)
@@ -221,14 +208,16 @@ define i32 @test_ctselect_nabs(i32 %x) {
 define i32 @test_ctselect_sign_extend(i32 %x) {
 ; M32-LABEL: test_ctselect_sign_extend:
 ; M32:       # %bb.0:
+; M32-NEXT:    slti $1, $4, 0
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    sra $2, $4, 31
+; M32-NEXT:    negu $2, $1
 ;
 ; M64-LABEL: test_ctselect_sign_extend:
 ; M64:       # %bb.0:
 ; M64-NEXT:    sll $1, $4, 0
+; M64-NEXT:    slti $1, $1, 0
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    sra $2, $1, 31
+; M64-NEXT:    negu $2, $1
   %cmp = icmp slt i32 %x, 0
   %result = call i32 @llvm.ct.select.i32(i1 %cmp, i32 -1, i32 0)
   ret i32 %result
@@ -270,13 +259,12 @@ define i32 @test_ctselect_constant_folding_false(i32 %a, i32 %b) {
 ; M32-LABEL: test_ctselect_constant_folding_false:
 ; M32:       # %bb.0:
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    or $2, $zero, $5
+; M32-NEXT:    move $2, $5
 ;
 ; M64-LABEL: test_ctselect_constant_folding_false:
 ; M64:       # %bb.0:
-; M64-NEXT:    sll $1, $5, 0
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $zero, $1
+; M64-NEXT:    sll $2, $5, 0
   %result = call i32 @llvm.ct.select.i32(i1 false, i32 %a, i32 %b)
   ret i32 %result
 }
@@ -285,25 +273,13 @@ define i32 @test_ctselect_constant_folding_false(i32 %a, i32 %b) {
 define i32 @test_ctselect_identical_operands(i1 %cond, i32 %x) {
 ; M32-LABEL: test_ctselect_identical_operands:
 ; M32:       # %bb.0:
-; M32-NEXT:    andi $1, $4, 1
-; M32-NEXT:    negu $2, $1
-; M32-NEXT:    addiu $1, $1, -1
-; M32-NEXT:    and $2, $2, $5
-; M32-NEXT:    and $1, $1, $5
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    or $2, $2, $1
+; M32-NEXT:    move $2, $5
 ;
 ; M64-LABEL: test_ctselect_identical_operands:
 ; M64:       # %bb.0:
-; M64-NEXT:    sll $1, $4, 0
-; M64-NEXT:    sll $3, $5, 0
-; M64-NEXT:    andi $1, $1, 1
-; M64-NEXT:    negu $2, $1
-; M64-NEXT:    addiu $1, $1, -1
-; M64-NEXT:    and $2, $2, $3
-; M64-NEXT:    and $1, $1, $3
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $2, $1
+; M64-NEXT:    sll $2, $5, 0
   %result = call i32 @llvm.ct.select.i32(i1 %cond, i32 %x, i32 %x)
   ret i32 %result
 }
@@ -312,29 +288,27 @@ define i32 @test_ctselect_identical_operands(i1 %cond, i32 %x) {
 define i32 @test_ctselect_inverted_condition(i32 %x, i32 %y, i32 %a, i32 %b) {
 ; M32-LABEL: test_ctselect_inverted_condition:
 ; M32:       # %bb.0:
-; M32-NEXT:    xor $1, $4, $5
-; M32-NEXT:    sltiu $1, $1, 1
-; M32-NEXT:    addiu $1, $1, -1
-; M32-NEXT:    and $2, $1, $6
-; M32-NEXT:    not $1, $1
-; M32-NEXT:    and $1, $1, $7
+; M32-NEXT:    xor $2, $4, $5
+; M32-NEXT:    xor $1, $7, $6
+; M32-NEXT:    sltiu $2, $2, 1
+; M32-NEXT:    negu $2, $2
+; M32-NEXT:    and $1, $1, $2
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    or $2, $2, $1
+; M32-NEXT:    xor $2, $6, $1
 ;
 ; M64-LABEL: test_ctselect_inverted_condition:
 ; M64:       # %bb.0:
 ; M64-NEXT:    sll $1, $5, 0
 ; M64-NEXT:    sll $2, $4, 0
-; M64-NEXT:    sll $3, $7, 0
 ; M64-NEXT:    xor $1, $2, $1
-; M64-NEXT:    sll $2, $6, 0
+; M64-NEXT:    xor $2, $7, $6
 ; M64-NEXT:    sltiu $1, $1, 1
-; M64-NEXT:    addiu $1, $1, -1
-; M64-NEXT:    and $2, $1, $2
-; M64-NEXT:    not $1, $1
-; M64-NEXT:    and $1, $1, $3
+; M64-NEXT:    sll $2, $2, 0
+; M64-NEXT:    negu $1, $1
+; M64-NEXT:    and $1, $2, $1
+; M64-NEXT:    sll $2, $6, 0
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $2, $1
+; M64-NEXT:    xor $2, $2, $1
   %cmp = icmp eq i32 %x, %y
   %not_cmp = xor i1 %cmp, true
   %result = call i32 @llvm.ct.select.i32(i1 %not_cmp, i32 %a, i32 %b)
@@ -345,57 +319,51 @@ define i32 @test_ctselect_inverted_condition(i32 %x, i32 %y, i32 %a, i32 %b) {
 define i32 @test_ctselect_chain(i1 %c1, i1 %c2, i1 %c3, i32 %a, i32 %b, i32 %c, i32 %d) {
 ; M32-LABEL: test_ctselect_chain:
 ; M32:       # %bb.0:
-; M32-NEXT:    andi $1, $4, 1
+; M32-NEXT:    lw $1, 16($sp)
+; M32-NEXT:    andi $3, $4, 1
+; M32-NEXT:    negu $3, $3
+; M32-NEXT:    xor $2, $7, $1
+; M32-NEXT:    and $2, $2, $3
 ; M32-NEXT:    andi $3, $5, 1
-; M32-NEXT:    lw $5, 16($sp)
-; M32-NEXT:    negu $2, $1
-; M32-NEXT:    addiu $1, $1, -1
-; M32-NEXT:    negu $4, $3
-; M32-NEXT:    addiu $3, $3, -1
-; M32-NEXT:    and $1, $1, $5
-; M32-NEXT:    and $2, $2, $7
-; M32-NEXT:    lw $5, 24($sp)
-; M32-NEXT:    or $1, $2, $1
+; M32-NEXT:    xor $1, $1, $2
+; M32-NEXT:    lw $2, 20($sp)
+; M32-NEXT:    negu $3, $3
+; M32-NEXT:    xor $1, $1, $2
+; M32-NEXT:    and $1, $1, $3
+; M32-NEXT:    lw $3, 24($sp)
+; M32-NEXT:    xor $1, $2, $1
 ; M32-NEXT:    andi $2, $6, 1
-; M32-NEXT:    and $1, $4, $1
-; M32-NEXT:    addiu $4, $2, -1
+; M32-NEXT:    xor $1, $1, $3
 ; M32-NEXT:    negu $2, $2
-; M32-NEXT:    and $4, $4, $5
-; M32-NEXT:    lw $5, 20($sp)
-; M32-NEXT:    and $3, $3, $5
-; M32-NEXT:    or $1, $1, $3
-; M32-NEXT:    and $1, $2, $1
+; M32-NEXT:    and $1, $1, $2
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    or $2, $1, $4
+; M32-NEXT:    xor $2, $3, $1
 ;
 ; M64-LABEL: test_ctselect_chain:
 ; M64:       # %bb.0:
 ; M64-NEXT:    sll $1, $4, 0
-; M64-NEXT:    sll $3, $7, 0
-; M64-NEXT:    sll $4, $5, 0
+; M64-NEXT:    xor $2, $7, $8
+; M64-NEXT:    sll $3, $5, 0
 ; M64-NEXT:    andi $1, $1, 1
-; M64-NEXT:    andi $4, $4, 1
-; M64-NEXT:    negu $2, $1
-; M64-NEXT:    addiu $1, $1, -1
-; M64-NEXT:    negu $5, $4
-; M64-NEXT:    addiu $4, $4, -1
-; M64-NEXT:    and $2, $2, $3
-; M64-NEXT:    sll $3, $8, 0
-; M64-NEXT:    and $1, $1, $3
-; M64-NEXT:    sll $3, $6, 0
-; M64-NEXT:    sll $6, $10, 0
-; M64-NEXT:    or $1, $2, $1
+; M64-NEXT:    sll $2, $2, 0
 ; M64-NEXT:    andi $3, $3, 1
-; M64-NEXT:    and $1, $5, $1
-; M64-NEXT:    sll $5, $9, 0
-; M64-NEXT:    addiu $2, $3, -1
+; M64-NEXT:    negu $1, $1
 ; M64-NEXT:    negu $3, $3
-; M64-NEXT:    and $4, $4, $5
-; M64-NEXT:    and $2, $2, $6
-; M64-NEXT:    or $1, $1, $4
-; M64-NEXT:    and $1, $3, $1
+; M64-NEXT:    and $1, $2, $1
+; M64-NEXT:    sll $2, $8, 0
+; M64-NEXT:    xor $1, $2, $1
+; M64-NEXT:    sll $2, $9, 0
+; M64-NEXT:    xor $1, $1, $2
+; M64-NEXT:    and $1, $1, $3
+; M64-NEXT:    sll $3, $6, 0
+; M64-NEXT:    xor $1, $2, $1
+; M64-NEXT:    andi $2, $3, 1
+; M64-NEXT:    sll $3, $10, 0
+; M64-NEXT:    xor $1, $1, $3
+; M64-NEXT:    negu $2, $2
+; M64-NEXT:    and $1, $1, $2
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $1, $2
+; M64-NEXT:    xor $2, $3, $1
   %sel1 = call i32 @llvm.ct.select.i32(i1 %c1, i32 %a, i32 %b)
   %sel2 = call i32 @llvm.ct.select.i32(i1 %c2, i32 %sel1, i32 %c)
   %sel3 = call i32 @llvm.ct.select.i32(i1 %c3, i32 %sel2, i32 %d)
@@ -406,16 +374,17 @@ define i32 @test_ctselect_chain(i1 %c1, i1 %c2, i1 %c3, i32 %a, i32 %b, i32 %c,
 define i64 @test_ctselect_i64_smin_zero(i64 %x) {
 ; M32-LABEL: test_ctselect_i64_smin_zero:
 ; M32:       # %bb.0:
-; M32-NEXT:    sra $1, $5, 31
-; M32-NEXT:    and $2, $1, $4
+; M32-NEXT:    slti $1, $5, 0
+; M32-NEXT:    negu $1, $1
+; M32-NEXT:    and $2, $4, $1
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    and $3, $1, $5
+; M32-NEXT:    and $3, $5, $1
 ;
 ; M64-LABEL: test_ctselect_i64_smin_zero:
 ; M64:       # %bb.0:
 ; M64-NEXT:    dsra $1, $4, 63
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    and $2, $1, $4
+; M64-NEXT:    and $2, $4, $1
   %cmp = icmp slt i64 %x, 0
   %result = call i64 @llvm.ct.select.i64(i1 %cmp, i64 %x, i64 0)
   ret i64 %result
diff --git a/llvm/test/CodeGen/Mips/ctselect-fallback-vector.ll b/llvm/test/CodeGen/Mips/ctselect-fallback-vector.ll
index 6222f6052e12..302e06b0a733 100644
--- a/llvm/test/CodeGen/Mips/ctselect-fallback-vector.ll
+++ b/llvm/test/CodeGen/Mips/ctselect-fallback-vector.ll
@@ -6,21 +6,19 @@
 define <4 x i32> @test_ctselect_v4i32(i1 %cond, <4 x i32> %a, <4 x i32> %b) {
 ; MIPS64-MSA-LABEL: test_ctselect_v4i32:
 ; MIPS64-MSA:       # %bb.0:
-; MIPS64-MSA-NEXT:    insert.d $w2[0], $7
+; MIPS64-MSA-NEXT:    insert.d $w0[0], $7
+; MIPS64-MSA-NEXT:    insert.d $w1[0], $5
 ; MIPS64-MSA-NEXT:    sll $1, $4, 0
-; MIPS64-MSA-NEXT:    ldi.b $w0, -1
-; MIPS64-MSA-NEXT:    fill.w $w1, $1
-; MIPS64-MSA-NEXT:    insert.d $w2[1], $8
-; MIPS64-MSA-NEXT:    slli.w $w1, $w1, 31
-; MIPS64-MSA-NEXT:    srai.w $w1, $w1, 31
-; MIPS64-MSA-NEXT:    shf.w $w2, $w2, 177
-; MIPS64-MSA-NEXT:    xor.v $w0, $w1, $w0
-; MIPS64-MSA-NEXT:    and.v $w0, $w0, $w2
-; MIPS64-MSA-NEXT:    insert.d $w2[0], $5
-; MIPS64-MSA-NEXT:    insert.d $w2[1], $6
-; MIPS64-MSA-NEXT:    shf.w $w2, $w2, 177
+; MIPS64-MSA-NEXT:    fill.w $w2, $1
+; MIPS64-MSA-NEXT:    insert.d $w0[1], $8
+; MIPS64-MSA-NEXT:    insert.d $w1[1], $6
+; MIPS64-MSA-NEXT:    slli.w $w2, $w2, 31
+; MIPS64-MSA-NEXT:    xor.v $w1, $w1, $w0
+; MIPS64-MSA-NEXT:    srai.w $w2, $w2, 31
+; MIPS64-MSA-NEXT:    shf.w $w0, $w0, 177
+; MIPS64-MSA-NEXT:    shf.w $w1, $w1, 177
 ; MIPS64-MSA-NEXT:    and.v $w1, $w1, $w2
-; MIPS64-MSA-NEXT:    or.v $w0, $w1, $w0
+; MIPS64-MSA-NEXT:    xor.v $w0, $w0, $w1
 ; MIPS64-MSA-NEXT:    shf.w $w0, $w0, 177
 ; MIPS64-MSA-NEXT:    copy_s.d $2, $w0[0]
 ; MIPS64-MSA-NEXT:    jr $ra
@@ -30,26 +28,24 @@ define <4 x i32> @test_ctselect_v4i32(i1 %cond, <4 x i32> %a, <4 x i32> %b) {
 ; MIPS32-MSA:       # %bb.0:
 ; MIPS32-MSA-NEXT:    lw $2, 24($sp)
 ; MIPS32-MSA-NEXT:    lw $1, 28($sp)
+; MIPS32-MSA-NEXT:    insert.w $w1[0], $6
 ; MIPS32-MSA-NEXT:    fill.w $w2, $4
-; MIPS32-MSA-NEXT:    ldi.b $w1, -1
 ; MIPS32-MSA-NEXT:    insert.w $w0[0], $2
+; MIPS32-MSA-NEXT:    insert.w $w1[1], $7
 ; MIPS32-MSA-NEXT:    slli.w $w2, $w2, 31
 ; MIPS32-MSA-NEXT:    srai.w $w2, $w2, 31
 ; MIPS32-MSA-NEXT:    insert.w $w0[1], $1
 ; MIPS32-MSA-NEXT:    lw $1, 32($sp)
-; MIPS32-MSA-NEXT:    xor.v $w1, $w2, $w1
 ; MIPS32-MSA-NEXT:    insert.w $w0[2], $1
 ; MIPS32-MSA-NEXT:    lw $1, 36($sp)
 ; MIPS32-MSA-NEXT:    insert.w $w0[3], $1
 ; MIPS32-MSA-NEXT:    lw $1, 16($sp)
-; MIPS32-MSA-NEXT:    and.v $w0, $w1, $w0
-; MIPS32-MSA-NEXT:    insert.w $w1[0], $6
-; MIPS32-MSA-NEXT:    insert.w $w1[1], $7
 ; MIPS32-MSA-NEXT:    insert.w $w1[2], $1
 ; MIPS32-MSA-NEXT:    lw $1, 20($sp)
 ; MIPS32-MSA-NEXT:    insert.w $w1[3], $1
-; MIPS32-MSA-NEXT:    and.v $w1, $w2, $w1
-; MIPS32-MSA-NEXT:    or.v $w0, $w1, $w0
+; MIPS32-MSA-NEXT:    xor.v $w1, $w1, $w0
+; MIPS32-MSA-NEXT:    and.v $w1, $w1, $w2
+; MIPS32-MSA-NEXT:    xor.v $w0, $w0, $w1
 ; MIPS32-MSA-NEXT:    copy_s.w $2, $w0[0]
 ; MIPS32-MSA-NEXT:    copy_s.w $3, $w0[1]
 ; MIPS32-MSA-NEXT:    copy_s.w $4, $w0[2]
@@ -63,21 +59,19 @@ define <4 x i32> @test_ctselect_v4i32(i1 %cond, <4 x i32> %a, <4 x i32> %b) {
 define <8 x i16> @test_ctselect_v8i16(i1 %cond, <8 x i16> %a, <8 x i16> %b) {
 ; MIPS64-MSA-LABEL: test_ctselect_v8i16:
 ; MIPS64-MSA:       # %bb.0:
-; MIPS64-MSA-NEXT:    insert.d $w2[0], $7
+; MIPS64-MSA-NEXT:    insert.d $w0[0], $7
+; MIPS64-MSA-NEXT:    insert.d $w1[0], $5
 ; MIPS64-MSA-NEXT:    sll $1, $4, 0
-; MIPS64-MSA-NEXT:    ldi.b $w0, -1
-; MIPS64-MSA-NEXT:    fill.h $w1, $1
-; MIPS64-MSA-NEXT:    insert.d $w2[1], $8
-; MIPS64-MSA-NEXT:    slli.h $w1, $w1, 15
-; MIPS64-MSA-NEXT:    srai.h $w1, $w1, 15
-; MIPS64-MSA-NEXT:    shf.h $w2, $w2, 27
-; MIPS64-MSA-NEXT:    xor.v $w0, $w1, $w0
-; MIPS64-MSA-NEXT:    and.v $w0, $w0, $w2
-; MIPS64-MSA-NEXT:    insert.d $w2[0], $5
-; MIPS64-MSA-NEXT:    insert.d $w2[1], $6
-; MIPS64-MSA-NEXT:    shf.h $w2, $w2, 27
+; MIPS64-MSA-NEXT:    fill.h $w2, $1
+; MIPS64-MSA-NEXT:    insert.d $w0[1], $8
+; MIPS64-MSA-NEXT:    insert.d $w1[1], $6
+; MIPS64-MSA-NEXT:    slli.h $w2, $w2, 15
+; MIPS64-MSA-NEXT:    xor.v $w1, $w1, $w0
+; MIPS64-MSA-NEXT:    srai.h $w2, $w2, 15
+; MIPS64-MSA-NEXT:    shf.h $w0, $w0, 27
+; MIPS64-MSA-NEXT:    shf.h $w1, $w1, 27
 ; MIPS64-MSA-NEXT:    and.v $w1, $w1, $w2
-; MIPS64-MSA-NEXT:    or.v $w0, $w1, $w0
+; MIPS64-MSA-NEXT:    xor.v $w0, $w0, $w1
 ; MIPS64-MSA-NEXT:    shf.h $w0, $w0, 27
 ; MIPS64-MSA-NEXT:    copy_s.d $2, $w0[0]
 ; MIPS64-MSA-NEXT:    jr $ra
@@ -87,28 +81,26 @@ define <8 x i16> @test_ctselect_v8i16(i1 %cond, <8 x i16> %a, <8 x i16> %b) {
 ; MIPS32-MSA:       # %bb.0:
 ; MIPS32-MSA-NEXT:    lw $2, 24($sp)
 ; MIPS32-MSA-NEXT:    lw $1, 28($sp)
-; MIPS32-MSA-NEXT:    fill.h $w1, $4
-; MIPS32-MSA-NEXT:    ldi.b $w0, -1
-; MIPS32-MSA-NEXT:    insert.w $w2[0], $2
-; MIPS32-MSA-NEXT:    slli.h $w1, $w1, 15
-; MIPS32-MSA-NEXT:    srai.h $w1, $w1, 15
-; MIPS32-MSA-NEXT:    insert.w $w2[1], $1
+; MIPS32-MSA-NEXT:    insert.w $w1[0], $6
+; MIPS32-MSA-NEXT:    fill.h $w2, $4
+; MIPS32-MSA-NEXT:    insert.w $w0[0], $2
+; MIPS32-MSA-NEXT:    insert.w $w1[1], $7
+; MIPS32-MSA-NEXT:    slli.h $w2, $w2, 15
+; MIPS32-MSA-NEXT:    srai.h $w2, $w2, 15
+; MIPS32-MSA-NEXT:    insert.w $w0[1], $1
 ; MIPS32-MSA-NEXT:    lw $1, 32($sp)
-; MIPS32-MSA-NEXT:    xor.v $w0, $w1, $w0
-; MIPS32-MSA-NEXT:    insert.w $w2[2], $1
+; MIPS32-MSA-NEXT:    insert.w $w0[2], $1
 ; MIPS32-MSA-NEXT:    lw $1, 36($sp)
-; MIPS32-MSA-NEXT:    insert.w $w2[3], $1
+; MIPS32-MSA-NEXT:    insert.w $w0[3], $1
 ; MIPS32-MSA-NEXT:    lw $1, 16($sp)
-; MIPS32-MSA-NEXT:    shf.h $w2, $w2, 177
-; MIPS32-MSA-NEXT:    and.v $w0, $w0, $w2
-; MIPS32-MSA-NEXT:    insert.w $w2[0], $6
-; MIPS32-MSA-NEXT:    insert.w $w2[1], $7
-; MIPS32-MSA-NEXT:    insert.w $w2[2], $1
+; MIPS32-MSA-NEXT:    insert.w $w1[2], $1
 ; MIPS32-MSA-NEXT:    lw $1, 20($sp)
-; MIPS32-MSA-NEXT:    insert.w $w2[3], $1
-; MIPS32-MSA-NEXT:    shf.h $w2, $w2, 177
+; MIPS32-MSA-NEXT:    insert.w $w1[3], $1
+; MIPS32-MSA-NEXT:    xor.v $w1, $w1, $w0
+; MIPS32-MSA-NEXT:    shf.h $w0, $w0, 177
+; MIPS32-MSA-NEXT:    shf.h $w1, $w1, 177
 ; MIPS32-MSA-NEXT:    and.v $w1, $w1, $w2
-; MIPS32-MSA-NEXT:    or.v $w0, $w1, $w0
+; MIPS32-MSA-NEXT:    xor.v $w0, $w0, $w1
 ; MIPS32-MSA-NEXT:    shf.h $w0, $w0, 177
 ; MIPS32-MSA-NEXT:    copy_s.w $2, $w0[0]
 ; MIPS32-MSA-NEXT:    copy_s.w $3, $w0[1]
@@ -123,22 +115,21 @@ define <8 x i16> @test_ctselect_v8i16(i1 %cond, <8 x i16> %a, <8 x i16> %b) {
 define <16 x i8> @test_ctselect_v16i8(i1 %cond, <16 x i8> %a, <16 x i8> %b) {
 ; MIPS64-MSA-LABEL: test_ctselect_v16i8:
 ; MIPS64-MSA:       # %bb.0:
-; MIPS64-MSA-NEXT:    insert.d $w0[0], $5
-; MIPS64-MSA-NEXT:    insert.d $w1[0], $7
+; MIPS64-MSA-NEXT:    insert.d $w0[0], $7
+; MIPS64-MSA-NEXT:    insert.d $w1[0], $5
 ; MIPS64-MSA-NEXT:    sll $1, $4, 0
 ; MIPS64-MSA-NEXT:    fill.b $w2, $1
-; MIPS64-MSA-NEXT:    insert.d $w0[1], $6
-; MIPS64-MSA-NEXT:    insert.d $w1[1], $8
+; MIPS64-MSA-NEXT:    insert.d $w0[1], $8
+; MIPS64-MSA-NEXT:    insert.d $w1[1], $6
 ; MIPS64-MSA-NEXT:    slli.b $w2, $w2, 7
+; MIPS64-MSA-NEXT:    xor.v $w1, $w1, $w0
 ; MIPS64-MSA-NEXT:    shf.b $w0, $w0, 27
-; MIPS64-MSA-NEXT:    shf.b $w1, $w1, 27
 ; MIPS64-MSA-NEXT:    srai.b $w2, $w2, 7
+; MIPS64-MSA-NEXT:    shf.b $w1, $w1, 27
 ; MIPS64-MSA-NEXT:    shf.w $w0, $w0, 177
 ; MIPS64-MSA-NEXT:    shf.w $w1, $w1, 177
-; MIPS64-MSA-NEXT:    and.v $w0, $w2, $w0
-; MIPS64-MSA-NEXT:    xori.b $w2, $w2, 255
-; MIPS64-MSA-NEXT:    and.v $w1, $w2, $w1
-; MIPS64-MSA-NEXT:    or.v $w0, $w0, $w1
+; MIPS64-MSA-NEXT:    and.v $w1, $w1, $w2
+; MIPS64-MSA-NEXT:    xor.v $w0, $w0, $w1
 ; MIPS64-MSA-NEXT:    shf.b $w0, $w0, 27
 ; MIPS64-MSA-NEXT:    shf.w $w0, $w0, 177
 ; MIPS64-MSA-NEXT:    copy_s.d $2, $w0[0]
@@ -147,29 +138,28 @@ define <16 x i8> @test_ctselect_v16i8(i1 %cond, <16 x i8> %a, <16 x i8> %b) {
 ;
 ; MIPS32-MSA-LABEL: test_ctselect_v16i8:
 ; MIPS32-MSA:       # %bb.0:
-; MIPS32-MSA-NEXT:    insert.w $w0[0], $6
-; MIPS32-MSA-NEXT:    lw $1, 16($sp)
 ; MIPS32-MSA-NEXT:    lw $2, 24($sp)
+; MIPS32-MSA-NEXT:    lw $1, 28($sp)
+; MIPS32-MSA-NEXT:    insert.w $w1[0], $6
 ; MIPS32-MSA-NEXT:    fill.b $w2, $4
-; MIPS32-MSA-NEXT:    insert.w $w0[1], $7
-; MIPS32-MSA-NEXT:    insert.w $w1[0], $2
+; MIPS32-MSA-NEXT:    insert.w $w0[0], $2
+; MIPS32-MSA-NEXT:    insert.w $w1[1], $7
 ; MIPS32-MSA-NEXT:    slli.b $w2, $w2, 7
 ; MIPS32-MSA-NEXT:    srai.b $w2, $w2, 7
+; MIPS32-MSA-NEXT:    insert.w $w0[1], $1
+; MIPS32-MSA-NEXT:    lw $1, 32($sp)
 ; MIPS32-MSA-NEXT:    insert.w $w0[2], $1
-; MIPS32-MSA-NEXT:    lw $1, 20($sp)
+; MIPS32-MSA-NEXT:    lw $1, 36($sp)
 ; MIPS32-MSA-NEXT:    insert.w $w0[3], $1
-; MIPS32-MSA-NEXT:    lw $1, 28($sp)
-; MIPS32-MSA-NEXT:    insert.w $w1[1], $1
-; MIPS32-MSA-NEXT:    lw $1, 32($sp)
-; MIPS32-MSA-NEXT:    shf.b $w0, $w0, 27
+; MIPS32-MSA-NEXT:    lw $1, 16($sp)
 ; MIPS32-MSA-NEXT:    insert.w $w1[2], $1
-; MIPS32-MSA-NEXT:    lw $1, 36($sp)
-; MIPS32-MSA-NEXT:    and.v $w0, $w2, $w0
-; MIPS32-MSA-NEXT:    xori.b $w2, $w2, 255
+; MIPS32-MSA-NEXT:    lw $1, 20($sp)
 ; MIPS32-MSA-NEXT:    insert.w $w1[3], $1
+; MIPS32-MSA-NEXT:    xor.v $w1, $w1, $w0
+; MIPS32-MSA-NEXT:    shf.b $w0, $w0, 27
 ; MIPS32-MSA-NEXT:    shf.b $w1, $w1, 27
-; MIPS32-MSA-NEXT:    and.v $w1, $w2, $w1
-; MIPS32-MSA-NEXT:    or.v $w0, $w0, $w1
+; MIPS32-MSA-NEXT:    and.v $w1, $w1, $w2
+; MIPS32-MSA-NEXT:    xor.v $w0, $w0, $w1
 ; MIPS32-MSA-NEXT:    shf.b $w0, $w0, 27
 ; MIPS32-MSA-NEXT:    copy_s.w $2, $w0[0]
 ; MIPS32-MSA-NEXT:    copy_s.w $3, $w0[1]
@@ -184,18 +174,16 @@ define <16 x i8> @test_ctselect_v16i8(i1 %cond, <16 x i8> %a, <16 x i8> %b) {
 define <2 x i64> @test_ctselect_v2i64(i1 %cond, <2 x i64> %a, <2 x i64> %b) {
 ; MIPS64-MSA-LABEL: test_ctselect_v2i64:
 ; MIPS64-MSA:       # %bb.0:
-; MIPS64-MSA-NEXT:    fill.d $w2, $4
 ; MIPS64-MSA-NEXT:    insert.d $w0[0], $7
-; MIPS64-MSA-NEXT:    ldi.b $w1, -1
-; MIPS64-MSA-NEXT:    slli.d $w2, $w2, 63
-; MIPS64-MSA-NEXT:    insert.d $w0[1], $8
-; MIPS64-MSA-NEXT:    srai.d $w2, $w2, 63
-; MIPS64-MSA-NEXT:    xor.v $w1, $w2, $w1
-; MIPS64-MSA-NEXT:    and.v $w0, $w1, $w0
 ; MIPS64-MSA-NEXT:    insert.d $w1[0], $5
+; MIPS64-MSA-NEXT:    fill.d $w2, $4
+; MIPS64-MSA-NEXT:    insert.d $w0[1], $8
 ; MIPS64-MSA-NEXT:    insert.d $w1[1], $6
-; MIPS64-MSA-NEXT:    and.v $w1, $w2, $w1
-; MIPS64-MSA-NEXT:    or.v $w0, $w1, $w0
+; MIPS64-MSA-NEXT:    slli.d $w2, $w2, 63
+; MIPS64-MSA-NEXT:    srai.d $w2, $w2, 63
+; MIPS64-MSA-NEXT:    xor.v $w1, $w1, $w0
+; MIPS64-MSA-NEXT:    and.v $w1, $w1, $w2
+; MIPS64-MSA-NEXT:    xor.v $w0, $w0, $w1
 ; MIPS64-MSA-NEXT:    copy_s.d $2, $w0[0]
 ; MIPS64-MSA-NEXT:    jr $ra
 ; MIPS64-MSA-NEXT:    copy_s.d $3, $w0[1]
@@ -214,31 +202,28 @@ define <2 x i64> @test_ctselect_v2i64(i1 %cond, <2 x i64> %a, <2 x i64> %b) {
 ; MIPS32-MSA-NEXT:    and $sp, $sp, $1
 ; MIPS32-MSA-NEXT:    lw $2, 56($fp)
 ; MIPS32-MSA-NEXT:    lw $1, 60($fp)
+; MIPS32-MSA-NEXT:    insert.w $w1[0], $6
 ; MIPS32-MSA-NEXT:    sw $4, 12($sp)
 ; MIPS32-MSA-NEXT:    sw $4, 4($sp)
-; MIPS32-MSA-NEXT:    ldi.b $w0, -1
-; MIPS32-MSA-NEXT:    ld.d $w1, 0($sp)
-; MIPS32-MSA-NEXT:    shf.w $w0, $w0, 177
-; MIPS32-MSA-NEXT:    insert.w $w2[0], $2
-; MIPS32-MSA-NEXT:    slli.d $w1, $w1, 63
-; MIPS32-MSA-NEXT:    insert.w $w2[1], $1
+; MIPS32-MSA-NEXT:    ld.d $w2, 0($sp)
+; MIPS32-MSA-NEXT:    insert.w $w0[0], $2
+; MIPS32-MSA-NEXT:    insert.w $w1[1], $7
+; MIPS32-MSA-NEXT:    slli.d $w2, $w2, 63
+; MIPS32-MSA-NEXT:    insert.w $w0[1], $1
 ; MIPS32-MSA-NEXT:    lw $1, 64($fp)
-; MIPS32-MSA-NEXT:    srai.d $w1, $w1, 63
-; MIPS32-MSA-NEXT:    xor.v $w0, $w1, $w0
-; MIPS32-MSA-NEXT:    insert.w $w2[2], $1
+; MIPS32-MSA-NEXT:    srai.d $w2, $w2, 63
+; MIPS32-MSA-NEXT:    insert.w $w0[2], $1
 ; MIPS32-MSA-NEXT:    lw $1, 68($fp)
-; MIPS32-MSA-NEXT:    insert.w $w2[3], $1
+; MIPS32-MSA-NEXT:    insert.w $w0[3], $1
 ; MIPS32-MSA-NEXT:    lw $1, 48($fp)
-; MIPS32-MSA-NEXT:    shf.w $w2, $w2, 177
-; MIPS32-MSA-NEXT:    and.v $w0, $w0, $w2
-; MIPS32-MSA-NEXT:    insert.w $w2[0], $6
-; MIPS32-MSA-NEXT:    insert.w $w2[1], $7
-; MIPS32-MSA-NEXT:    insert.w $w2[2], $1
+; MIPS32-MSA-NEXT:    insert.w $w1[2], $1
 ; MIPS32-MSA-NEXT:    lw $1, 52($fp)
-; MIPS32-MSA-NEXT:    insert.w $w2[3], $1
-; MIPS32-MSA-NEXT:    shf.w $w2, $w2, 177
+; MIPS32-MSA-NEXT:    insert.w $w1[3], $1
+; MIPS32-MSA-NEXT:    xor.v $w1, $w1, $w0
+; MIPS32-MSA-NEXT:    shf.w $w0, $w0, 177
+; MIPS32-MSA-NEXT:    shf.w $w1, $w1, 177
 ; MIPS32-MSA-NEXT:    and.v $w1, $w1, $w2
-; MIPS32-MSA-NEXT:    or.v $w0, $w1, $w0
+; MIPS32-MSA-NEXT:    xor.v $w0, $w0, $w1
 ; MIPS32-MSA-NEXT:    shf.w $w0, $w0, 177
 ; MIPS32-MSA-NEXT:    copy_s.w $2, $w0[0]
 ; MIPS32-MSA-NEXT:    copy_s.w $3, $w0[1]
@@ -257,21 +242,19 @@ define <2 x i64> @test_ctselect_v2i64(i1 %cond, <2 x i64> %a, <2 x i64> %b) {
 define <4 x float> @test_ctselect_v4f32(i1 %cond, <4 x float> %a, <4 x float> %b) {
 ; MIPS64-MSA-LABEL: test_ctselect_v4f32:
 ; MIPS64-MSA:       # %bb.0:
-; MIPS64-MSA-NEXT:    insert.d $w2[0], $7
+; MIPS64-MSA-NEXT:    insert.d $w0[0], $7
+; MIPS64-MSA-NEXT:    insert.d $w1[0], $5
 ; MIPS64-MSA-NEXT:    sll $1, $4, 0
-; MIPS64-MSA-NEXT:    ldi.b $w0, -1
-; MIPS64-MSA-NEXT:    fill.w $w1, $1
-; MIPS64-MSA-NEXT:    insert.d $w2[1], $8
-; MIPS64-MSA-NEXT:    slli.w $w1, $w1, 31
-; MIPS64-MSA-NEXT:    srai.w $w1, $w1, 31
-; MIPS64-MSA-NEXT:    shf.w $w2, $w2, 177
-; MIPS64-MSA-NEXT:    xor.v $w0, $w1, $w0
-; MIPS64-MSA-NEXT:    and.v $w0, $w0, $w2
-; MIPS64-MSA-NEXT:    insert.d $w2[0], $5
-; MIPS64-MSA-NEXT:    insert.d $w2[1], $6
-; MIPS64-MSA-NEXT:    shf.w $w2, $w2, 177
+; MIPS64-MSA-NEXT:    fill.w $w2, $1
+; MIPS64-MSA-NEXT:    insert.d $w0[1], $8
+; MIPS64-MSA-NEXT:    insert.d $w1[1], $6
+; MIPS64-MSA-NEXT:    slli.w $w2, $w2, 31
+; MIPS64-MSA-NEXT:    xor.v $w1, $w1, $w0
+; MIPS64-MSA-NEXT:    srai.w $w2, $w2, 31
+; MIPS64-MSA-NEXT:    shf.w $w0, $w0, 177
+; MIPS64-MSA-NEXT:    shf.w $w1, $w1, 177
 ; MIPS64-MSA-NEXT:    and.v $w1, $w1, $w2
-; MIPS64-MSA-NEXT:    or.v $w0, $w1, $w0
+; MIPS64-MSA-NEXT:    xor.v $w0, $w0, $w1
 ; MIPS64-MSA-NEXT:    shf.w $w0, $w0, 177
 ; MIPS64-MSA-NEXT:    copy_s.d $2, $w0[0]
 ; MIPS64-MSA-NEXT:    jr $ra
@@ -281,26 +264,24 @@ define <4 x float> @test_ctselect_v4f32(i1 %cond, <4 x float> %a, <4 x float> %b
 ; MIPS32-MSA:       # %bb.0:
 ; MIPS32-MSA-NEXT:    lw $2, 24($sp)
 ; MIPS32-MSA-NEXT:    lw $1, 28($sp)
+; MIPS32-MSA-NEXT:    insert.w $w1[0], $6
 ; MIPS32-MSA-NEXT:    fill.w $w2, $5
-; MIPS32-MSA-NEXT:    ldi.b $w1, -1
 ; MIPS32-MSA-NEXT:    insert.w $w0[0], $2
+; MIPS32-MSA-NEXT:    insert.w $w1[1], $7
 ; MIPS32-MSA-NEXT:    slli.w $w2, $w2, 31
 ; MIPS32-MSA-NEXT:    srai.w $w2, $w2, 31
 ; MIPS32-MSA-NEXT:    insert.w $w0[1], $1
 ; MIPS32-MSA-NEXT:    lw $1, 32($sp)
-; MIPS32-MSA-NEXT:    xor.v $w1, $w2, $w1
 ; MIPS32-MSA-NEXT:    insert.w $w0[2], $1
 ; MIPS32-MSA-NEXT:    lw $1, 36($sp)
 ; MIPS32-MSA-NEXT:    insert.w $w0[3], $1
 ; MIPS32-MSA-NEXT:    lw $1, 16($sp)
-; MIPS32-MSA-NEXT:    and.v $w0, $w1, $w0
-; MIPS32-MSA-NEXT:    insert.w $w1[0], $6
-; MIPS32-MSA-NEXT:    insert.w $w1[1], $7
 ; MIPS32-MSA-NEXT:    insert.w $w1[2], $1
 ; MIPS32-MSA-NEXT:    lw $1, 20($sp)
 ; MIPS32-MSA-NEXT:    insert.w $w1[3], $1
-; MIPS32-MSA-NEXT:    and.v $w1, $w2, $w1
-; MIPS32-MSA-NEXT:    or.v $w0, $w1, $w0
+; MIPS32-MSA-NEXT:    xor.v $w1, $w1, $w0
+; MIPS32-MSA-NEXT:    and.v $w1, $w1, $w2
+; MIPS32-MSA-NEXT:    xor.v $w0, $w0, $w1
 ; MIPS32-MSA-NEXT:    jr $ra
 ; MIPS32-MSA-NEXT:    st.w $w0, 0($4)
   %result = call <4 x float> @llvm.ct.select.v4f32(i1 %cond, <4 x float> %a, <4 x float> %b)
@@ -311,18 +292,16 @@ define <4 x float> @test_ctselect_v4f32(i1 %cond, <4 x float> %a, <4 x float> %b
 define <2 x double> @test_ctselect_v2f64(i1 %cond, <2 x double> %a, <2 x double> %b) {
 ; MIPS64-MSA-LABEL: test_ctselect_v2f64:
 ; MIPS64-MSA:       # %bb.0:
-; MIPS64-MSA-NEXT:    fill.d $w2, $4
 ; MIPS64-MSA-NEXT:    insert.d $w0[0], $7
-; MIPS64-MSA-NEXT:    ldi.b $w1, -1
-; MIPS64-MSA-NEXT:    slli.d $w2, $w2, 63
-; MIPS64-MSA-NEXT:    insert.d $w0[1], $8
-; MIPS64-MSA-NEXT:    srai.d $w2, $w2, 63
-; MIPS64-MSA-NEXT:    xor.v $w1, $w2, $w1
-; MIPS64-MSA-NEXT:    and.v $w0, $w1, $w0
 ; MIPS64-MSA-NEXT:    insert.d $w1[0], $5
+; MIPS64-MSA-NEXT:    fill.d $w2, $4
+; MIPS64-MSA-NEXT:    insert.d $w0[1], $8
 ; MIPS64-MSA-NEXT:    insert.d $w1[1], $6
-; MIPS64-MSA-NEXT:    and.v $w1, $w2, $w1
-; MIPS64-MSA-NEXT:    or.v $w0, $w1, $w0
+; MIPS64-MSA-NEXT:    slli.d $w2, $w2, 63
+; MIPS64-MSA-NEXT:    srai.d $w2, $w2, 63
+; MIPS64-MSA-NEXT:    xor.v $w1, $w1, $w0
+; MIPS64-MSA-NEXT:    and.v $w1, $w1, $w2
+; MIPS64-MSA-NEXT:    xor.v $w0, $w0, $w1
 ; MIPS64-MSA-NEXT:    copy_s.d $2, $w0[0]
 ; MIPS64-MSA-NEXT:    jr $ra
 ; MIPS64-MSA-NEXT:    copy_s.d $3, $w0[1]
@@ -341,31 +320,28 @@ define <2 x double> @test_ctselect_v2f64(i1 %cond, <2 x double> %a, <2 x double>
 ; MIPS32-MSA-NEXT:    and $sp, $sp, $1
 ; MIPS32-MSA-NEXT:    lw $2, 56($fp)
 ; MIPS32-MSA-NEXT:    lw $1, 60($fp)
+; MIPS32-MSA-NEXT:    insert.w $w1[0], $6
 ; MIPS32-MSA-NEXT:    sw $5, 12($sp)
 ; MIPS32-MSA-NEXT:    sw $5, 4($sp)
-; MIPS32-MSA-NEXT:    ldi.b $w0, -1
-; MIPS32-MSA-NEXT:    ld.d $w1, 0($sp)
-; MIPS32-MSA-NEXT:    shf.w $w0, $w0, 177
-; MIPS32-MSA-NEXT:    insert.w $w2[0], $2
-; MIPS32-MSA-NEXT:    slli.d $w1, $w1, 63
-; MIPS32-MSA-NEXT:    insert.w $w2[1], $1
+; MIPS32-MSA-NEXT:    ld.d $w2, 0($sp)
+; MIPS32-MSA-NEXT:    insert.w $w0[0], $2
+; MIPS32-MSA-NEXT:    insert.w $w1[1], $7
+; MIPS32-MSA-NEXT:    slli.d $w2, $w2, 63
+; MIPS32-MSA-NEXT:    insert.w $w0[1], $1
 ; MIPS32-MSA-NEXT:    lw $1, 64($fp)
-; MIPS32-MSA-NEXT:    srai.d $w1, $w1, 63
-; MIPS32-MSA-NEXT:    xor.v $w0, $w1, $w0
-; MIPS32-MSA-NEXT:    insert.w $w2[2], $1
+; MIPS32-MSA-NEXT:    srai.d $w2, $w2, 63
+; MIPS32-MSA-NEXT:    insert.w $w0[2], $1
 ; MIPS32-MSA-NEXT:    lw $1, 68($fp)
-; MIPS32-MSA-NEXT:    insert.w $w2[3], $1
+; MIPS32-MSA-NEXT:    insert.w $w0[3], $1
 ; MIPS32-MSA-NEXT:    lw $1, 48($fp)
-; MIPS32-MSA-NEXT:    shf.w $w2, $w2, 177
-; MIPS32-MSA-NEXT:    and.v $w0, $w0, $w2
-; MIPS32-MSA-NEXT:    insert.w $w2[0], $6
-; MIPS32-MSA-NEXT:    insert.w $w2[1], $7
-; MIPS32-MSA-NEXT:    insert.w $w2[2], $1
+; MIPS32-MSA-NEXT:    insert.w $w1[2], $1
 ; MIPS32-MSA-NEXT:    lw $1, 52($fp)
-; MIPS32-MSA-NEXT:    insert.w $w2[3], $1
-; MIPS32-MSA-NEXT:    shf.w $w2, $w2, 177
+; MIPS32-MSA-NEXT:    insert.w $w1[3], $1
+; MIPS32-MSA-NEXT:    xor.v $w1, $w1, $w0
+; MIPS32-MSA-NEXT:    shf.w $w0, $w0, 177
+; MIPS32-MSA-NEXT:    shf.w $w1, $w1, 177
 ; MIPS32-MSA-NEXT:    and.v $w1, $w1, $w2
-; MIPS32-MSA-NEXT:    or.v $w0, $w1, $w0
+; MIPS32-MSA-NEXT:    xor.v $w0, $w0, $w1
 ; MIPS32-MSA-NEXT:    st.d $w0, 0($4)
 ; MIPS32-MSA-NEXT:    move $sp, $fp
 ; MIPS32-MSA-NEXT:    lw $fp, 24($sp) # 4-byte Folded Reload
@@ -381,16 +357,14 @@ define <4 x i32> @test_ctselect_v4i32_aligned_load(i1 %cond, ptr %p1, ptr %p2) {
 ; MIPS64-MSA-LABEL: test_ctselect_v4i32_aligned_load:
 ; MIPS64-MSA:       # %bb.0:
 ; MIPS64-MSA-NEXT:    sll $1, $4, 0
+; MIPS64-MSA-NEXT:    ld.w $w0, 0($6)
 ; MIPS64-MSA-NEXT:    ld.w $w1, 0($5)
-; MIPS64-MSA-NEXT:    ldi.b $w2, -1
-; MIPS64-MSA-NEXT:    fill.w $w0, $1
-; MIPS64-MSA-NEXT:    slli.w $w0, $w0, 31
-; MIPS64-MSA-NEXT:    srai.w $w0, $w0, 31
-; MIPS64-MSA-NEXT:    and.v $w1, $w0, $w1
-; MIPS64-MSA-NEXT:    xor.v $w0, $w0, $w2
-; MIPS64-MSA-NEXT:    ld.w $w2, 0($6)
-; MIPS64-MSA-NEXT:    and.v $w0, $w0, $w2
-; MIPS64-MSA-NEXT:    or.v $w0, $w1, $w0
+; MIPS64-MSA-NEXT:    fill.w $w2, $1
+; MIPS64-MSA-NEXT:    xor.v $w1, $w1, $w0
+; MIPS64-MSA-NEXT:    slli.w $w2, $w2, 31
+; MIPS64-MSA-NEXT:    srai.w $w2, $w2, 31
+; MIPS64-MSA-NEXT:    and.v $w1, $w1, $w2
+; MIPS64-MSA-NEXT:    xor.v $w0, $w0, $w1
 ; MIPS64-MSA-NEXT:    shf.w $w0, $w0, 177
 ; MIPS64-MSA-NEXT:    copy_s.d $2, $w0[0]
 ; MIPS64-MSA-NEXT:    jr $ra
@@ -398,16 +372,14 @@ define <4 x i32> @test_ctselect_v4i32_aligned_load(i1 %cond, ptr %p1, ptr %p2) {
 ;
 ; MIPS32-MSA-LABEL: test_ctselect_v4i32_aligned_load:
 ; MIPS32-MSA:       # %bb.0:
-; MIPS32-MSA-NEXT:    fill.w $w0, $4
+; MIPS32-MSA-NEXT:    fill.w $w2, $4
+; MIPS32-MSA-NEXT:    ld.w $w0, 0($6)
 ; MIPS32-MSA-NEXT:    ld.w $w1, 0($5)
-; MIPS32-MSA-NEXT:    ldi.b $w2, -1
-; MIPS32-MSA-NEXT:    slli.w $w0, $w0, 31
-; MIPS32-MSA-NEXT:    srai.w $w0, $w0, 31
-; MIPS32-MSA-NEXT:    and.v $w1, $w0, $w1
-; MIPS32-MSA-NEXT:    xor.v $w0, $w0, $w2
-; MIPS32-MSA-NEXT:    ld.w $w2, 0($6)
-; MIPS32-MSA-NEXT:    and.v $w0, $w0, $w2
-; MIPS32-MSA-NEXT:    or.v $w0, $w1, $w0
+; MIPS32-MSA-NEXT:    slli.w $w2, $w2, 31
+; MIPS32-MSA-NEXT:    xor.v $w1, $w1, $w0
+; MIPS32-MSA-NEXT:    srai.w $w2, $w2, 31
+; MIPS32-MSA-NEXT:    and.v $w1, $w1, $w2
+; MIPS32-MSA-NEXT:    xor.v $w0, $w0, $w1
 ; MIPS32-MSA-NEXT:    copy_s.w $2, $w0[0]
 ; MIPS32-MSA-NEXT:    copy_s.w $3, $w0[1]
 ; MIPS32-MSA-NEXT:    copy_s.w $4, $w0[2]
@@ -424,16 +396,14 @@ define <4 x i32> @test_ctselect_v4i32_unaligned_load(i1 %cond, ptr %p1, ptr %p2)
 ; MIPS64-MSA-LABEL: test_ctselect_v4i32_unaligned_load:
 ; MIPS64-MSA:       # %bb.0:
 ; MIPS64-MSA-NEXT:    sll $1, $4, 0
+; MIPS64-MSA-NEXT:    ld.w $w0, 0($6)
 ; MIPS64-MSA-NEXT:    ld.w $w1, 0($5)
-; MIPS64-MSA-NEXT:    ldi.b $w2, -1
-; MIPS64-MSA-NEXT:    fill.w $w0, $1
-; MIPS64-MSA-NEXT:    slli.w $w0, $w0, 31
-; MIPS64-MSA-NEXT:    srai.w $w0, $w0, 31
-; MIPS64-MSA-NEXT:    and.v $w1, $w0, $w1
-; MIPS64-MSA-NEXT:    xor.v $w0, $w0, $w2
-; MIPS64-MSA-NEXT:    ld.w $w2, 0($6)
-; MIPS64-MSA-NEXT:    and.v $w0, $w0, $w2
-; MIPS64-MSA-NEXT:    or.v $w0, $w1, $w0
+; MIPS64-MSA-NEXT:    fill.w $w2, $1
+; MIPS64-MSA-NEXT:    xor.v $w1, $w1, $w0
+; MIPS64-MSA-NEXT:    slli.w $w2, $w2, 31
+; MIPS64-MSA-NEXT:    srai.w $w2, $w2, 31
+; MIPS64-MSA-NEXT:    and.v $w1, $w1, $w2
+; MIPS64-MSA-NEXT:    xor.v $w0, $w0, $w1
 ; MIPS64-MSA-NEXT:    shf.w $w0, $w0, 177
 ; MIPS64-MSA-NEXT:    copy_s.d $2, $w0[0]
 ; MIPS64-MSA-NEXT:    jr $ra
@@ -441,16 +411,14 @@ define <4 x i32> @test_ctselect_v4i32_unaligned_load(i1 %cond, ptr %p1, ptr %p2)
 ;
 ; MIPS32-MSA-LABEL: test_ctselect_v4i32_unaligned_load:
 ; MIPS32-MSA:       # %bb.0:
-; MIPS32-MSA-NEXT:    fill.w $w0, $4
+; MIPS32-MSA-NEXT:    fill.w $w2, $4
+; MIPS32-MSA-NEXT:    ld.w $w0, 0($6)
 ; MIPS32-MSA-NEXT:    ld.w $w1, 0($5)
-; MIPS32-MSA-NEXT:    ldi.b $w2, -1
-; MIPS32-MSA-NEXT:    slli.w $w0, $w0, 31
-; MIPS32-MSA-NEXT:    srai.w $w0, $w0, 31
-; MIPS32-MSA-NEXT:    and.v $w1, $w0, $w1
-; MIPS32-MSA-NEXT:    xor.v $w0, $w0, $w2
-; MIPS32-MSA-NEXT:    ld.w $w2, 0($6)
-; MIPS32-MSA-NEXT:    and.v $w0, $w0, $w2
-; MIPS32-MSA-NEXT:    or.v $w0, $w1, $w0
+; MIPS32-MSA-NEXT:    slli.w $w2, $w2, 31
+; MIPS32-MSA-NEXT:    xor.v $w1, $w1, $w0
+; MIPS32-MSA-NEXT:    srai.w $w2, $w2, 31
+; MIPS32-MSA-NEXT:    and.v $w1, $w1, $w2
+; MIPS32-MSA-NEXT:    xor.v $w0, $w0, $w1
 ; MIPS32-MSA-NEXT:    copy_s.w $2, $w0[0]
 ; MIPS32-MSA-NEXT:    copy_s.w $3, $w0[1]
 ; MIPS32-MSA-NEXT:    copy_s.w $4, $w0[2]
@@ -466,21 +434,19 @@ define <4 x i32> @test_ctselect_v4i32_unaligned_load(i1 %cond, ptr %p1, ptr %p2)
 define void @test_ctselect_v4i32_store(i1 %cond, <4 x i32> %a, <4 x i32> %b, ptr %out) {
 ; MIPS64-MSA-LABEL: test_ctselect_v4i32_store:
 ; MIPS64-MSA:       # %bb.0:
-; MIPS64-MSA-NEXT:    insert.d $w2[0], $7
+; MIPS64-MSA-NEXT:    insert.d $w0[0], $7
+; MIPS64-MSA-NEXT:    insert.d $w1[0], $5
 ; MIPS64-MSA-NEXT:    sll $1, $4, 0
-; MIPS64-MSA-NEXT:    ldi.b $w0, -1
-; MIPS64-MSA-NEXT:    fill.w $w1, $1
-; MIPS64-MSA-NEXT:    insert.d $w2[1], $8
-; MIPS64-MSA-NEXT:    slli.w $w1, $w1, 31
-; MIPS64-MSA-NEXT:    srai.w $w1, $w1, 31
-; MIPS64-MSA-NEXT:    shf.w $w2, $w2, 177
-; MIPS64-MSA-NEXT:    xor.v $w0, $w1, $w0
-; MIPS64-MSA-NEXT:    and.v $w0, $w0, $w2
-; MIPS64-MSA-NEXT:    insert.d $w2[0], $5
-; MIPS64-MSA-NEXT:    insert.d $w2[1], $6
-; MIPS64-MSA-NEXT:    shf.w $w2, $w2, 177
+; MIPS64-MSA-NEXT:    fill.w $w2, $1
+; MIPS64-MSA-NEXT:    insert.d $w0[1], $8
+; MIPS64-MSA-NEXT:    insert.d $w1[1], $6
+; MIPS64-MSA-NEXT:    slli.w $w2, $w2, 31
+; MIPS64-MSA-NEXT:    xor.v $w1, $w1, $w0
+; MIPS64-MSA-NEXT:    srai.w $w2, $w2, 31
+; MIPS64-MSA-NEXT:    shf.w $w0, $w0, 177
+; MIPS64-MSA-NEXT:    shf.w $w1, $w1, 177
 ; MIPS64-MSA-NEXT:    and.v $w1, $w1, $w2
-; MIPS64-MSA-NEXT:    or.v $w0, $w1, $w0
+; MIPS64-MSA-NEXT:    xor.v $w0, $w0, $w1
 ; MIPS64-MSA-NEXT:    jr $ra
 ; MIPS64-MSA-NEXT:    st.w $w0, 0($9)
 ;
@@ -488,27 +454,25 @@ define void @test_ctselect_v4i32_store(i1 %cond, <4 x i32> %a, <4 x i32> %b, ptr
 ; MIPS32-MSA:       # %bb.0:
 ; MIPS32-MSA-NEXT:    lw $2, 24($sp)
 ; MIPS32-MSA-NEXT:    lw $1, 28($sp)
+; MIPS32-MSA-NEXT:    insert.w $w1[0], $6
 ; MIPS32-MSA-NEXT:    fill.w $w2, $4
-; MIPS32-MSA-NEXT:    ldi.b $w1, -1
 ; MIPS32-MSA-NEXT:    insert.w $w0[0], $2
+; MIPS32-MSA-NEXT:    insert.w $w1[1], $7
 ; MIPS32-MSA-NEXT:    slli.w $w2, $w2, 31
 ; MIPS32-MSA-NEXT:    srai.w $w2, $w2, 31
 ; MIPS32-MSA-NEXT:    insert.w $w0[1], $1
 ; MIPS32-MSA-NEXT:    lw $1, 32($sp)
-; MIPS32-MSA-NEXT:    xor.v $w1, $w2, $w1
 ; MIPS32-MSA-NEXT:    insert.w $w0[2], $1
 ; MIPS32-MSA-NEXT:    lw $1, 36($sp)
 ; MIPS32-MSA-NEXT:    insert.w $w0[3], $1
 ; MIPS32-MSA-NEXT:    lw $1, 16($sp)
-; MIPS32-MSA-NEXT:    and.v $w0, $w1, $w0
-; MIPS32-MSA-NEXT:    insert.w $w1[0], $6
-; MIPS32-MSA-NEXT:    insert.w $w1[1], $7
 ; MIPS32-MSA-NEXT:    insert.w $w1[2], $1
 ; MIPS32-MSA-NEXT:    lw $1, 20($sp)
 ; MIPS32-MSA-NEXT:    insert.w $w1[3], $1
 ; MIPS32-MSA-NEXT:    lw $1, 40($sp)
-; MIPS32-MSA-NEXT:    and.v $w1, $w2, $w1
-; MIPS32-MSA-NEXT:    or.v $w0, $w1, $w0
+; MIPS32-MSA-NEXT:    xor.v $w1, $w1, $w0
+; MIPS32-MSA-NEXT:    and.v $w1, $w1, $w2
+; MIPS32-MSA-NEXT:    xor.v $w0, $w0, $w1
 ; MIPS32-MSA-NEXT:    jr $ra
 ; MIPS32-MSA-NEXT:    st.w $w0, 0($1)
   %result = call <4 x i32> @llvm.ct.select.v4i32(i1 %cond, <4 x i32> %a, <4 x i32> %b)
@@ -521,31 +485,28 @@ define <4 x i32> @test_ctselect_v4i32_chain(i1 %cond1, i1 %cond2, <4 x i32> %a,
 ; MIPS64-MSA-LABEL: test_ctselect_v4i32_chain:
 ; MIPS64-MSA:       # %bb.0:
 ; MIPS64-MSA-NEXT:    insert.d $w0[0], $8
+; MIPS64-MSA-NEXT:    insert.d $w1[0], $6
 ; MIPS64-MSA-NEXT:    sll $1, $4, 0
-; MIPS64-MSA-NEXT:    ldi.b $w1, -1
 ; MIPS64-MSA-NEXT:    fill.w $w2, $1
 ; MIPS64-MSA-NEXT:    sll $1, $5, 0
 ; MIPS64-MSA-NEXT:    insert.d $w0[1], $9
+; MIPS64-MSA-NEXT:    insert.d $w1[1], $7
 ; MIPS64-MSA-NEXT:    slli.w $w2, $w2, 31
+; MIPS64-MSA-NEXT:    xor.v $w1, $w1, $w0
 ; MIPS64-MSA-NEXT:    srai.w $w2, $w2, 31
 ; MIPS64-MSA-NEXT:    shf.w $w0, $w0, 177
-; MIPS64-MSA-NEXT:    xor.v $w3, $w2, $w1
-; MIPS64-MSA-NEXT:    and.v $w0, $w3, $w0
-; MIPS64-MSA-NEXT:    insert.d $w3[0], $6
-; MIPS64-MSA-NEXT:    insert.d $w3[1], $7
-; MIPS64-MSA-NEXT:    shf.w $w3, $w3, 177
-; MIPS64-MSA-NEXT:    and.v $w2, $w2, $w3
-; MIPS64-MSA-NEXT:    or.v $w0, $w2, $w0
+; MIPS64-MSA-NEXT:    shf.w $w1, $w1, 177
+; MIPS64-MSA-NEXT:    and.v $w1, $w1, $w2
 ; MIPS64-MSA-NEXT:    fill.w $w2, $1
+; MIPS64-MSA-NEXT:    xor.v $w0, $w0, $w1
+; MIPS64-MSA-NEXT:    insert.d $w1[0], $10
 ; MIPS64-MSA-NEXT:    slli.w $w2, $w2, 31
+; MIPS64-MSA-NEXT:    insert.d $w1[1], $11
 ; MIPS64-MSA-NEXT:    srai.w $w2, $w2, 31
-; MIPS64-MSA-NEXT:    and.v $w0, $w2, $w0
-; MIPS64-MSA-NEXT:    xor.v $w1, $w2, $w1
-; MIPS64-MSA-NEXT:    insert.d $w2[0], $10
-; MIPS64-MSA-NEXT:    insert.d $w2[1], $11
-; MIPS64-MSA-NEXT:    shf.w $w2, $w2, 177
-; MIPS64-MSA-NEXT:    and.v $w1, $w1, $w2
-; MIPS64-MSA-NEXT:    or.v $w0, $w0, $w1
+; MIPS64-MSA-NEXT:    shf.w $w1, $w1, 177
+; MIPS64-MSA-NEXT:    xor.v $w0, $w0, $w1
+; MIPS64-MSA-NEXT:    and.v $w0, $w0, $w2
+; MIPS64-MSA-NEXT:    xor.v $w0, $w1, $w0
 ; MIPS64-MSA-NEXT:    shf.w $w0, $w0, 177
 ; MIPS64-MSA-NEXT:    copy_s.d $2, $w0[0]
 ; MIPS64-MSA-NEXT:    jr $ra
@@ -555,41 +516,38 @@ define <4 x i32> @test_ctselect_v4i32_chain(i1 %cond1, i1 %cond2, <4 x i32> %a,
 ; MIPS32-MSA:       # %bb.0:
 ; MIPS32-MSA-NEXT:    lw $2, 24($sp)
 ; MIPS32-MSA-NEXT:    lw $1, 28($sp)
+; MIPS32-MSA-NEXT:    insert.w $w1[0], $6
 ; MIPS32-MSA-NEXT:    fill.w $w2, $4
-; MIPS32-MSA-NEXT:    ldi.b $w1, -1
 ; MIPS32-MSA-NEXT:    insert.w $w0[0], $2
+; MIPS32-MSA-NEXT:    insert.w $w1[1], $7
 ; MIPS32-MSA-NEXT:    slli.w $w2, $w2, 31
 ; MIPS32-MSA-NEXT:    lw $2, 40($sp)
 ; MIPS32-MSA-NEXT:    srai.w $w2, $w2, 31
 ; MIPS32-MSA-NEXT:    insert.w $w0[1], $1
 ; MIPS32-MSA-NEXT:    lw $1, 32($sp)
-; MIPS32-MSA-NEXT:    xor.v $w3, $w2, $w1
 ; MIPS32-MSA-NEXT:    insert.w $w0[2], $1
 ; MIPS32-MSA-NEXT:    lw $1, 36($sp)
 ; MIPS32-MSA-NEXT:    insert.w $w0[3], $1
 ; MIPS32-MSA-NEXT:    lw $1, 16($sp)
-; MIPS32-MSA-NEXT:    and.v $w0, $w3, $w0
-; MIPS32-MSA-NEXT:    insert.w $w3[0], $6
-; MIPS32-MSA-NEXT:    insert.w $w3[1], $7
-; MIPS32-MSA-NEXT:    insert.w $w3[2], $1
+; MIPS32-MSA-NEXT:    insert.w $w1[2], $1
 ; MIPS32-MSA-NEXT:    lw $1, 20($sp)
-; MIPS32-MSA-NEXT:    insert.w $w3[3], $1
+; MIPS32-MSA-NEXT:    insert.w $w1[3], $1
 ; MIPS32-MSA-NEXT:    lw $1, 44($sp)
-; MIPS32-MSA-NEXT:    and.v $w2, $w2, $w3
-; MIPS32-MSA-NEXT:    or.v $w0, $w2, $w0
+; MIPS32-MSA-NEXT:    xor.v $w1, $w1, $w0
+; MIPS32-MSA-NEXT:    and.v $w1, $w1, $w2
 ; MIPS32-MSA-NEXT:    fill.w $w2, $5
+; MIPS32-MSA-NEXT:    xor.v $w0, $w0, $w1
+; MIPS32-MSA-NEXT:    insert.w $w1[0], $2
 ; MIPS32-MSA-NEXT:    slli.w $w2, $w2, 31
-; MIPS32-MSA-NEXT:    srai.w $w2, $w2, 31
-; MIPS32-MSA-NEXT:    and.v $w0, $w2, $w0
-; MIPS32-MSA-NEXT:    xor.v $w1, $w2, $w1
-; MIPS32-MSA-NEXT:    insert.w $w2[0], $2
-; MIPS32-MSA-NEXT:    insert.w $w2[1], $1
+; MIPS32-MSA-NEXT:    insert.w $w1[1], $1
 ; MIPS32-MSA-NEXT:    lw $1, 48($sp)
-; MIPS32-MSA-NEXT:    insert.w $w2[2], $1
+; MIPS32-MSA-NEXT:    srai.w $w2, $w2, 31
+; MIPS32-MSA-NEXT:    insert.w $w1[2], $1
 ; MIPS32-MSA-NEXT:    lw $1, 52($sp)
-; MIPS32-MSA-NEXT:    insert.w $w2[3], $1
-; MIPS32-MSA-NEXT:    and.v $w1, $w1, $w2
-; MIPS32-MSA-NEXT:    or.v $w0, $w0, $w1
+; MIPS32-MSA-NEXT:    insert.w $w1[3], $1
+; MIPS32-MSA-NEXT:    xor.v $w0, $w0, $w1
+; MIPS32-MSA-NEXT:    and.v $w0, $w0, $w2
+; MIPS32-MSA-NEXT:    xor.v $w0, $w1, $w0
 ; MIPS32-MSA-NEXT:    copy_s.w $2, $w0[0]
 ; MIPS32-MSA-NEXT:    copy_s.w $3, $w0[1]
 ; MIPS32-MSA-NEXT:    copy_s.w $4, $w0[2]
@@ -607,20 +565,18 @@ define <4 x float> @test_ctselect_v4f32_arithmetic(i1 %cond, <4 x float> %x, <4
 ; MIPS64-MSA-NEXT:    insert.d $w0[0], $7
 ; MIPS64-MSA-NEXT:    insert.d $w1[0], $5
 ; MIPS64-MSA-NEXT:    sll $1, $4, 0
-; MIPS64-MSA-NEXT:    fill.w $w3, $1
 ; MIPS64-MSA-NEXT:    insert.d $w0[1], $8
 ; MIPS64-MSA-NEXT:    insert.d $w1[1], $6
-; MIPS64-MSA-NEXT:    slli.w $w3, $w3, 31
 ; MIPS64-MSA-NEXT:    shf.w $w0, $w0, 177
 ; MIPS64-MSA-NEXT:    shf.w $w1, $w1, 177
-; MIPS64-MSA-NEXT:    srai.w $w3, $w3, 31
 ; MIPS64-MSA-NEXT:    fadd.w $w2, $w1, $w0
 ; MIPS64-MSA-NEXT:    fsub.w $w0, $w1, $w0
-; MIPS64-MSA-NEXT:    ldi.b $w1, -1
-; MIPS64-MSA-NEXT:    xor.v $w1, $w3, $w1
-; MIPS64-MSA-NEXT:    and.v $w2, $w3, $w2
-; MIPS64-MSA-NEXT:    and.v $w0, $w1, $w0
-; MIPS64-MSA-NEXT:    or.v $w0, $w2, $w0
+; MIPS64-MSA-NEXT:    xor.v $w1, $w2, $w0
+; MIPS64-MSA-NEXT:    fill.w $w2, $1
+; MIPS64-MSA-NEXT:    slli.w $w2, $w2, 31
+; MIPS64-MSA-NEXT:    srai.w $w2, $w2, 31
+; MIPS64-MSA-NEXT:    and.v $w1, $w1, $w2
+; MIPS64-MSA-NEXT:    xor.v $w0, $w0, $w1
 ; MIPS64-MSA-NEXT:    shf.w $w0, $w0, 177
 ; MIPS64-MSA-NEXT:    copy_s.d $2, $w0[0]
 ; MIPS64-MSA-NEXT:    jr $ra
@@ -631,11 +587,8 @@ define <4 x float> @test_ctselect_v4f32_arithmetic(i1 %cond, <4 x float> %x, <4
 ; MIPS32-MSA-NEXT:    lw $2, 24($sp)
 ; MIPS32-MSA-NEXT:    lw $1, 28($sp)
 ; MIPS32-MSA-NEXT:    insert.w $w1[0], $6
-; MIPS32-MSA-NEXT:    fill.w $w3, $5
 ; MIPS32-MSA-NEXT:    insert.w $w0[0], $2
 ; MIPS32-MSA-NEXT:    insert.w $w1[1], $7
-; MIPS32-MSA-NEXT:    slli.w $w3, $w3, 31
-; MIPS32-MSA-NEXT:    srai.w $w3, $w3, 31
 ; MIPS32-MSA-NEXT:    insert.w $w0[1], $1
 ; MIPS32-MSA-NEXT:    lw $1, 32($sp)
 ; MIPS32-MSA-NEXT:    insert.w $w0[2], $1
@@ -647,11 +600,12 @@ define <4 x float> @test_ctselect_v4f32_arithmetic(i1 %cond, <4 x float> %x, <4
 ; MIPS32-MSA-NEXT:    insert.w $w1[3], $1
 ; MIPS32-MSA-NEXT:    fadd.w $w2, $w1, $w0
 ; MIPS32-MSA-NEXT:    fsub.w $w0, $w1, $w0
-; MIPS32-MSA-NEXT:    ldi.b $w1, -1
-; MIPS32-MSA-NEXT:    xor.v $w1, $w3, $w1
-; MIPS32-MSA-NEXT:    and.v $w2, $w3, $w2
-; MIPS32-MSA-NEXT:    and.v $w0, $w1, $w0
-; MIPS32-MSA-NEXT:    or.v $w0, $w2, $w0
+; MIPS32-MSA-NEXT:    xor.v $w1, $w2, $w0
+; MIPS32-MSA-NEXT:    fill.w $w2, $5
+; MIPS32-MSA-NEXT:    slli.w $w2, $w2, 31
+; MIPS32-MSA-NEXT:    srai.w $w2, $w2, 31
+; MIPS32-MSA-NEXT:    and.v $w1, $w1, $w2
+; MIPS32-MSA-NEXT:    xor.v $w0, $w0, $w1
 ; MIPS32-MSA-NEXT:    jr $ra
 ; MIPS32-MSA-NEXT:    st.w $w0, 0($4)
   %sum = fadd <4 x float> %x, %y
@@ -664,36 +618,32 @@ define <4 x float> @test_ctselect_v4f32_arithmetic(i1 %cond, <4 x float> %x, <4
 define void @test_ctselect_v4i32_mixed(i1 %cond, ptr %p1, ptr %p2, ptr %out) {
 ; MIPS64-MSA-LABEL: test_ctselect_v4i32_mixed:
 ; MIPS64-MSA:       # %bb.0:
+; MIPS64-MSA-NEXT:    ld.w $w0, 0($6)
+; MIPS64-MSA-NEXT:    ld.w $w1, 0($5)
 ; MIPS64-MSA-NEXT:    sll $1, $4, 0
-; MIPS64-MSA-NEXT:    ld.w $w0, 0($5)
-; MIPS64-MSA-NEXT:    ldi.b $w2, -1
-; MIPS64-MSA-NEXT:    fill.w $w1, $1
-; MIPS64-MSA-NEXT:    addvi.w $w0, $w0, 1
-; MIPS64-MSA-NEXT:    slli.w $w1, $w1, 31
-; MIPS64-MSA-NEXT:    srai.w $w1, $w1, 31
-; MIPS64-MSA-NEXT:    and.v $w0, $w1, $w0
-; MIPS64-MSA-NEXT:    xor.v $w1, $w1, $w2
-; MIPS64-MSA-NEXT:    ld.w $w2, 0($6)
-; MIPS64-MSA-NEXT:    addvi.w $w2, $w2, 2
+; MIPS64-MSA-NEXT:    fill.w $w2, $1
+; MIPS64-MSA-NEXT:    addvi.w $w0, $w0, 2
+; MIPS64-MSA-NEXT:    addvi.w $w1, $w1, 1
+; MIPS64-MSA-NEXT:    slli.w $w2, $w2, 31
+; MIPS64-MSA-NEXT:    xor.v $w1, $w1, $w0
+; MIPS64-MSA-NEXT:    srai.w $w2, $w2, 31
 ; MIPS64-MSA-NEXT:    and.v $w1, $w1, $w2
-; MIPS64-MSA-NEXT:    or.v $w0, $w0, $w1
+; MIPS64-MSA-NEXT:    xor.v $w0, $w0, $w1
 ; MIPS64-MSA-NEXT:    jr $ra
 ; MIPS64-MSA-NEXT:    st.w $w0, 0($7)
 ;
 ; MIPS32-MSA-LABEL: test_ctselect_v4i32_mixed:
 ; MIPS32-MSA:       # %bb.0:
-; MIPS32-MSA-NEXT:    ld.w $w0, 0($5)
-; MIPS32-MSA-NEXT:    fill.w $w1, $4
-; MIPS32-MSA-NEXT:    ldi.b $w2, -1
-; MIPS32-MSA-NEXT:    slli.w $w1, $w1, 31
-; MIPS32-MSA-NEXT:    addvi.w $w0, $w0, 1
-; MIPS32-MSA-NEXT:    srai.w $w1, $w1, 31
-; MIPS32-MSA-NEXT:    and.v $w0, $w1, $w0
-; MIPS32-MSA-NEXT:    xor.v $w1, $w1, $w2
-; MIPS32-MSA-NEXT:    ld.w $w2, 0($6)
-; MIPS32-MSA-NEXT:    addvi.w $w2, $w2, 2
+; MIPS32-MSA-NEXT:    ld.w $w0, 0($6)
+; MIPS32-MSA-NEXT:    ld.w $w1, 0($5)
+; MIPS32-MSA-NEXT:    fill.w $w2, $4
+; MIPS32-MSA-NEXT:    addvi.w $w0, $w0, 2
+; MIPS32-MSA-NEXT:    addvi.w $w1, $w1, 1
+; MIPS32-MSA-NEXT:    slli.w $w2, $w2, 31
+; MIPS32-MSA-NEXT:    srai.w $w2, $w2, 31
+; MIPS32-MSA-NEXT:    xor.v $w1, $w1, $w0
 ; MIPS32-MSA-NEXT:    and.v $w1, $w1, $w2
-; MIPS32-MSA-NEXT:    or.v $w0, $w0, $w1
+; MIPS32-MSA-NEXT:    xor.v $w0, $w0, $w1
 ; MIPS32-MSA-NEXT:    jr $ra
 ; MIPS32-MSA-NEXT:    st.w $w0, 0($7)
   %a = load <4 x i32>, ptr %p1, align 16
@@ -709,21 +659,19 @@ define void @test_ctselect_v4i32_mixed(i1 %cond, ptr %p1, ptr %p2, ptr %out) {
 define <4 x i32> @test_ctselect_v4i32_args(i1 %cond, <4 x i32> %a, <4 x i32> %b) nounwind {
 ; MIPS64-MSA-LABEL: test_ctselect_v4i32_args:
 ; MIPS64-MSA:       # %bb.0:
-; MIPS64-MSA-NEXT:    insert.d $w2[0], $7
+; MIPS64-MSA-NEXT:    insert.d $w0[0], $7
+; MIPS64-MSA-NEXT:    insert.d $w1[0], $5
 ; MIPS64-MSA-NEXT:    sll $1, $4, 0
-; MIPS64-MSA-NEXT:    ldi.b $w0, -1
-; MIPS64-MSA-NEXT:    fill.w $w1, $1
-; MIPS64-MSA-NEXT:    insert.d $w2[1], $8
-; MIPS64-MSA-NEXT:    slli.w $w1, $w1, 31
-; MIPS64-MSA-NEXT:    srai.w $w1, $w1, 31
-; MIPS64-MSA-NEXT:    shf.w $w2, $w2, 177
-; MIPS64-MSA-NEXT:    xor.v $w0, $w1, $w0
-; MIPS64-MSA-NEXT:    and.v $w0, $w0, $w2
-; MIPS64-MSA-NEXT:    insert.d $w2[0], $5
-; MIPS64-MSA-NEXT:    insert.d $w2[1], $6
-; MIPS64-MSA-NEXT:    shf.w $w2, $w2, 177
+; MIPS64-MSA-NEXT:    fill.w $w2, $1
+; MIPS64-MSA-NEXT:    insert.d $w0[1], $8
+; MIPS64-MSA-NEXT:    insert.d $w1[1], $6
+; MIPS64-MSA-NEXT:    slli.w $w2, $w2, 31
+; MIPS64-MSA-NEXT:    xor.v $w1, $w1, $w0
+; MIPS64-MSA-NEXT:    srai.w $w2, $w2, 31
+; MIPS64-MSA-NEXT:    shf.w $w0, $w0, 177
+; MIPS64-MSA-NEXT:    shf.w $w1, $w1, 177
 ; MIPS64-MSA-NEXT:    and.v $w1, $w1, $w2
-; MIPS64-MSA-NEXT:    or.v $w0, $w1, $w0
+; MIPS64-MSA-NEXT:    xor.v $w0, $w0, $w1
 ; MIPS64-MSA-NEXT:    shf.w $w0, $w0, 177
 ; MIPS64-MSA-NEXT:    copy_s.d $2, $w0[0]
 ; MIPS64-MSA-NEXT:    jr $ra
@@ -733,26 +681,24 @@ define <4 x i32> @test_ctselect_v4i32_args(i1 %cond, <4 x i32> %a, <4 x i32> %b)
 ; MIPS32-MSA:       # %bb.0:
 ; MIPS32-MSA-NEXT:    lw $2, 24($sp)
 ; MIPS32-MSA-NEXT:    lw $1, 28($sp)
+; MIPS32-MSA-NEXT:    insert.w $w1[0], $6
 ; MIPS32-MSA-NEXT:    fill.w $w2, $4
-; MIPS32-MSA-NEXT:    ldi.b $w1, -1
 ; MIPS32-MSA-NEXT:    insert.w $w0[0], $2
+; MIPS32-MSA-NEXT:    insert.w $w1[1], $7
 ; MIPS32-MSA-NEXT:    slli.w $w2, $w2, 31
 ; MIPS32-MSA-NEXT:    srai.w $w2, $w2, 31
 ; MIPS32-MSA-NEXT:    insert.w $w0[1], $1
 ; MIPS32-MSA-NEXT:    lw $1, 32($sp)
-; MIPS32-MSA-NEXT:    xor.v $w1, $w2, $w1
 ; MIPS32-MSA-NEXT:    insert.w $w0[2], $1
 ; MIPS32-MSA-NEXT:    lw $1, 36($sp)
 ; MIPS32-MSA-NEXT:    insert.w $w0[3], $1
 ; MIPS32-MSA-NEXT:    lw $1, 16($sp)
-; MIPS32-MSA-NEXT:    and.v $w0, $w1, $w0
-; MIPS32-MSA-NEXT:    insert.w $w1[0], $6
-; MIPS32-MSA-NEXT:    insert.w $w1[1], $7
 ; MIPS32-MSA-NEXT:    insert.w $w1[2], $1
 ; MIPS32-MSA-NEXT:    lw $1, 20($sp)
 ; MIPS32-MSA-NEXT:    insert.w $w1[3], $1
-; MIPS32-MSA-NEXT:    and.v $w1, $w2, $w1
-; MIPS32-MSA-NEXT:    or.v $w0, $w1, $w0
+; MIPS32-MSA-NEXT:    xor.v $w1, $w1, $w0
+; MIPS32-MSA-NEXT:    and.v $w1, $w1, $w2
+; MIPS32-MSA-NEXT:    xor.v $w0, $w0, $w1
 ; MIPS32-MSA-NEXT:    copy_s.w $2, $w0[0]
 ; MIPS32-MSA-NEXT:    copy_s.w $3, $w0[1]
 ; MIPS32-MSA-NEXT:    copy_s.w $4, $w0[2]
@@ -766,21 +712,19 @@ define <4 x i32> @test_ctselect_v4i32_args(i1 %cond, <4 x i32> %a, <4 x i32> %b)
 define <4 x i32> @test_ctselect_v4i32_multi_use(i1 %cond, <4 x i32> %a, <4 x i32> %b) {
 ; MIPS64-MSA-LABEL: test_ctselect_v4i32_multi_use:
 ; MIPS64-MSA:       # %bb.0:
-; MIPS64-MSA-NEXT:    insert.d $w2[0], $7
+; MIPS64-MSA-NEXT:    insert.d $w0[0], $7
+; MIPS64-MSA-NEXT:    insert.d $w1[0], $5
 ; MIPS64-MSA-NEXT:    sll $1, $4, 0
-; MIPS64-MSA-NEXT:    ldi.b $w0, -1
-; MIPS64-MSA-NEXT:    fill.w $w1, $1
-; MIPS64-MSA-NEXT:    insert.d $w2[1], $8
-; MIPS64-MSA-NEXT:    slli.w $w1, $w1, 31
-; MIPS64-MSA-NEXT:    srai.w $w1, $w1, 31
-; MIPS64-MSA-NEXT:    shf.w $w2, $w2, 177
-; MIPS64-MSA-NEXT:    xor.v $w0, $w1, $w0
-; MIPS64-MSA-NEXT:    and.v $w0, $w0, $w2
-; MIPS64-MSA-NEXT:    insert.d $w2[0], $5
-; MIPS64-MSA-NEXT:    insert.d $w2[1], $6
-; MIPS64-MSA-NEXT:    shf.w $w2, $w2, 177
+; MIPS64-MSA-NEXT:    fill.w $w2, $1
+; MIPS64-MSA-NEXT:    insert.d $w0[1], $8
+; MIPS64-MSA-NEXT:    insert.d $w1[1], $6
+; MIPS64-MSA-NEXT:    slli.w $w2, $w2, 31
+; MIPS64-MSA-NEXT:    xor.v $w1, $w1, $w0
+; MIPS64-MSA-NEXT:    srai.w $w2, $w2, 31
+; MIPS64-MSA-NEXT:    shf.w $w0, $w0, 177
+; MIPS64-MSA-NEXT:    shf.w $w1, $w1, 177
 ; MIPS64-MSA-NEXT:    and.v $w1, $w1, $w2
-; MIPS64-MSA-NEXT:    or.v $w0, $w1, $w0
+; MIPS64-MSA-NEXT:    xor.v $w0, $w0, $w1
 ; MIPS64-MSA-NEXT:    addv.w $w0, $w0, $w0
 ; MIPS64-MSA-NEXT:    shf.w $w0, $w0, 177
 ; MIPS64-MSA-NEXT:    copy_s.d $2, $w0[0]
@@ -791,26 +735,24 @@ define <4 x i32> @test_ctselect_v4i32_multi_use(i1 %cond, <4 x i32> %a, <4 x i32
 ; MIPS32-MSA:       # %bb.0:
 ; MIPS32-MSA-NEXT:    lw $2, 24($sp)
 ; MIPS32-MSA-NEXT:    lw $1, 28($sp)
+; MIPS32-MSA-NEXT:    insert.w $w1[0], $6
 ; MIPS32-MSA-NEXT:    fill.w $w2, $4
-; MIPS32-MSA-NEXT:    ldi.b $w1, -1
 ; MIPS32-MSA-NEXT:    insert.w $w0[0], $2
+; MIPS32-MSA-NEXT:    insert.w $w1[1], $7
 ; MIPS32-MSA-NEXT:    slli.w $w2, $w2, 31
 ; MIPS32-MSA-NEXT:    srai.w $w2, $w2, 31
 ; MIPS32-MSA-NEXT:    insert.w $w0[1], $1
 ; MIPS32-MSA-NEXT:    lw $1, 32($sp)
-; MIPS32-MSA-NEXT:    xor.v $w1, $w2, $w1
 ; MIPS32-MSA-NEXT:    insert.w $w0[2], $1
 ; MIPS32-MSA-NEXT:    lw $1, 36($sp)
 ; MIPS32-MSA-NEXT:    insert.w $w0[3], $1
 ; MIPS32-MSA-NEXT:    lw $1, 16($sp)
-; MIPS32-MSA-NEXT:    and.v $w0, $w1, $w0
-; MIPS32-MSA-NEXT:    insert.w $w1[0], $6
-; MIPS32-MSA-NEXT:    insert.w $w1[1], $7
 ; MIPS32-MSA-NEXT:    insert.w $w1[2], $1
 ; MIPS32-MSA-NEXT:    lw $1, 20($sp)
 ; MIPS32-MSA-NEXT:    insert.w $w1[3], $1
-; MIPS32-MSA-NEXT:    and.v $w1, $w2, $w1
-; MIPS32-MSA-NEXT:    or.v $w0, $w1, $w0
+; MIPS32-MSA-NEXT:    xor.v $w1, $w1, $w0
+; MIPS32-MSA-NEXT:    and.v $w1, $w1, $w2
+; MIPS32-MSA-NEXT:    xor.v $w0, $w0, $w1
 ; MIPS32-MSA-NEXT:    addv.w $w0, $w0, $w0
 ; MIPS32-MSA-NEXT:    copy_s.w $2, $w0[0]
 ; MIPS32-MSA-NEXT:    copy_s.w $3, $w0[1]
diff --git a/llvm/test/CodeGen/Mips/ctselect-fallback.ll b/llvm/test/CodeGen/Mips/ctselect-fallback.ll
index d89d7fc69871..6a61412367f7 100644
--- a/llvm/test/CodeGen/Mips/ctselect-fallback.ll
+++ b/llvm/test/CodeGen/Mips/ctselect-fallback.ll
@@ -11,7 +11,7 @@ define i8 @test_ctselect_i8(i1 %cond, i8 %a, i8 %b) {
 ; M32-NEXT:    negu $2, $2
 ; M32-NEXT:    and $1, $1, $2
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    xor $2, $1, $6
+; M32-NEXT:    xor $2, $6, $1
 ;
 ; M64-LABEL: test_ctselect_i8:
 ; M64:       # %bb.0:
@@ -23,7 +23,7 @@ define i8 @test_ctselect_i8(i1 %cond, i8 %a, i8 %b) {
 ; M64-NEXT:    and $1, $2, $1
 ; M64-NEXT:    sll $2, $6, 0
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    xor $2, $1, $2
+; M64-NEXT:    xor $2, $2, $1
   %result = call i8 @llvm.ct.select.i8(i1 %cond, i8 %a, i8 %b)
   ret i8 %result
 }
@@ -36,7 +36,7 @@ define i16 @test_ctselect_i16(i1 %cond, i16 %a, i16 %b) {
 ; M32-NEXT:    negu $2, $2
 ; M32-NEXT:    and $1, $1, $2
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    xor $2, $1, $6
+; M32-NEXT:    xor $2, $6, $1
 ;
 ; M64-LABEL: test_ctselect_i16:
 ; M64:       # %bb.0:
@@ -48,7 +48,7 @@ define i16 @test_ctselect_i16(i1 %cond, i16 %a, i16 %b) {
 ; M64-NEXT:    and $1, $2, $1
 ; M64-NEXT:    sll $2, $6, 0
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    xor $2, $1, $2
+; M64-NEXT:    xor $2, $2, $1
   %result = call i16 @llvm.ct.select.i16(i1 %cond, i16 %a, i16 %b)
   ret i16 %result
 }
@@ -56,26 +56,24 @@ define i16 @test_ctselect_i16(i1 %cond, i16 %a, i16 %b) {
 define i32 @test_ctselect_i32(i1 %cond, i32 %a, i32 %b) {
 ; M32-LABEL: test_ctselect_i32:
 ; M32:       # %bb.0:
-; M32-NEXT:    andi $1, $4, 1
-; M32-NEXT:    negu $2, $1
-; M32-NEXT:    addiu $1, $1, -1
-; M32-NEXT:    and $2, $2, $5
-; M32-NEXT:    and $1, $1, $6
+; M32-NEXT:    andi $2, $4, 1
+; M32-NEXT:    xor $1, $5, $6
+; M32-NEXT:    negu $2, $2
+; M32-NEXT:    and $1, $1, $2
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    or $2, $2, $1
+; M32-NEXT:    xor $2, $6, $1
 ;
 ; M64-LABEL: test_ctselect_i32:
 ; M64:       # %bb.0:
 ; M64-NEXT:    sll $1, $4, 0
-; M64-NEXT:    sll $3, $5, 0
+; M64-NEXT:    xor $2, $5, $6
 ; M64-NEXT:    andi $1, $1, 1
-; M64-NEXT:    negu $2, $1
-; M64-NEXT:    addiu $1, $1, -1
-; M64-NEXT:    and $2, $2, $3
-; M64-NEXT:    sll $3, $6, 0
-; M64-NEXT:    and $1, $1, $3
+; M64-NEXT:    sll $2, $2, 0
+; M64-NEXT:    negu $1, $1
+; M64-NEXT:    and $1, $2, $1
+; M64-NEXT:    sll $2, $6, 0
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $2, $1
+; M64-NEXT:    xor $2, $2, $1
   %result = call i32 @llvm.ct.select.i32(i1 %cond, i32 %a, i32 %b)
   ret i32 %result
 }
@@ -88,22 +86,21 @@ define i64 @test_ctselect_i64(i1 %cond, i64 %a, i64 %b) {
 ; M32-NEXT:    negu $3, $3
 ; M32-NEXT:    xor $2, $6, $1
 ; M32-NEXT:    and $2, $2, $3
-; M32-NEXT:    xor $2, $2, $1
+; M32-NEXT:    xor $2, $1, $2
 ; M32-NEXT:    lw $1, 20($sp)
 ; M32-NEXT:    xor $4, $7, $1
 ; M32-NEXT:    and $3, $4, $3
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    xor $3, $3, $1
+; M32-NEXT:    xor $3, $1, $3
 ;
 ; M64-LABEL: test_ctselect_i64:
 ; M64:       # %bb.0:
-; M64-NEXT:    andi $1, $4, 1
-; M64-NEXT:    dnegu $2, $1
-; M64-NEXT:    daddiu $1, $1, -1
-; M64-NEXT:    and $2, $2, $5
-; M64-NEXT:    and $1, $1, $6
+; M64-NEXT:    andi $2, $4, 1
+; M64-NEXT:    xor $1, $5, $6
+; M64-NEXT:    dnegu $2, $2
+; M64-NEXT:    and $1, $1, $2
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $2, $1
+; M64-NEXT:    xor $2, $6, $1
   %result = call i64 @llvm.ct.select.i64(i1 %cond, i64 %a, i64 %b)
   ret i64 %result
 }
@@ -111,23 +108,21 @@ define i64 @test_ctselect_i64(i1 %cond, i64 %a, i64 %b) {
 define ptr @test_ctselect_ptr(i1 %cond, ptr %a, ptr %b) {
 ; M32-LABEL: test_ctselect_ptr:
 ; M32:       # %bb.0:
-; M32-NEXT:    andi $1, $4, 1
-; M32-NEXT:    negu $2, $1
-; M32-NEXT:    addiu $1, $1, -1
-; M32-NEXT:    and $2, $2, $5
-; M32-NEXT:    and $1, $1, $6
+; M32-NEXT:    andi $2, $4, 1
+; M32-NEXT:    xor $1, $5, $6
+; M32-NEXT:    negu $2, $2
+; M32-NEXT:    and $1, $1, $2
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    or $2, $2, $1
+; M32-NEXT:    xor $2, $6, $1
 ;
 ; M64-LABEL: test_ctselect_ptr:
 ; M64:       # %bb.0:
-; M64-NEXT:    andi $1, $4, 1
-; M64-NEXT:    dnegu $2, $1
-; M64-NEXT:    daddiu $1, $1, -1
-; M64-NEXT:    and $2, $2, $5
-; M64-NEXT:    and $1, $1, $6
+; M64-NEXT:    andi $2, $4, 1
+; M64-NEXT:    xor $1, $5, $6
+; M64-NEXT:    dnegu $2, $2
+; M64-NEXT:    and $1, $1, $2
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $2, $1
+; M64-NEXT:    xor $2, $6, $1
   %result = call ptr @llvm.ct.select.p0(i1 %cond, ptr %a, ptr %b)
   ret ptr %result
 }
@@ -151,13 +146,12 @@ define i32 @test_ctselect_const_false(i32 %a, i32 %b) {
 ; M32-LABEL: test_ctselect_const_false:
 ; M32:       # %bb.0:
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    or $2, $zero, $5
+; M32-NEXT:    move $2, $5
 ;
 ; M64-LABEL: test_ctselect_const_false:
 ; M64:       # %bb.0:
-; M64-NEXT:    sll $1, $5, 0
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $zero, $1
+; M64-NEXT:    sll $2, $5, 0
   %result = call i32 @llvm.ct.select.i32(i1 false, i32 %a, i32 %b)
   ret i32 %result
 }
@@ -166,29 +160,27 @@ define i32 @test_ctselect_const_false(i32 %a, i32 %b) {
 define i32 @test_ctselect_icmp_eq(i32 %x, i32 %y, i32 %a, i32 %b) {
 ; M32-LABEL: test_ctselect_icmp_eq:
 ; M32:       # %bb.0:
-; M32-NEXT:    xor $1, $4, $5
-; M32-NEXT:    sltu $1, $zero, $1
-; M32-NEXT:    addiu $1, $1, -1
-; M32-NEXT:    and $2, $1, $6
-; M32-NEXT:    not $1, $1
-; M32-NEXT:    and $1, $1, $7
+; M32-NEXT:    xor $2, $4, $5
+; M32-NEXT:    xor $1, $6, $7
+; M32-NEXT:    sltiu $2, $2, 1
+; M32-NEXT:    negu $2, $2
+; M32-NEXT:    and $1, $1, $2
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    or $2, $2, $1
+; M32-NEXT:    xor $2, $7, $1
 ;
 ; M64-LABEL: test_ctselect_icmp_eq:
 ; M64:       # %bb.0:
 ; M64-NEXT:    sll $1, $5, 0
 ; M64-NEXT:    sll $2, $4, 0
-; M64-NEXT:    sll $3, $7, 0
 ; M64-NEXT:    xor $1, $2, $1
-; M64-NEXT:    sll $2, $6, 0
-; M64-NEXT:    sltu $1, $zero, $1
-; M64-NEXT:    addiu $1, $1, -1
-; M64-NEXT:    and $2, $1, $2
-; M64-NEXT:    not $1, $1
-; M64-NEXT:    and $1, $1, $3
+; M64-NEXT:    xor $2, $6, $7
+; M64-NEXT:    sltiu $1, $1, 1
+; M64-NEXT:    sll $2, $2, 0
+; M64-NEXT:    negu $1, $1
+; M64-NEXT:    and $1, $2, $1
+; M64-NEXT:    sll $2, $7, 0
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $2, $1
+; M64-NEXT:    xor $2, $2, $1
   %cond = icmp eq i32 %x, %y
   %result = call i32 @llvm.ct.select.i32(i1 %cond, i32 %a, i32 %b)
   ret i32 %result
@@ -197,29 +189,27 @@ define i32 @test_ctselect_icmp_eq(i32 %x, i32 %y, i32 %a, i32 %b) {
 define i32 @test_ctselect_icmp_ne(i32 %x, i32 %y, i32 %a, i32 %b) {
 ; M32-LABEL: test_ctselect_icmp_ne:
 ; M32:       # %bb.0:
-; M32-NEXT:    xor $1, $4, $5
-; M32-NEXT:    sltiu $1, $1, 1
-; M32-NEXT:    addiu $1, $1, -1
-; M32-NEXT:    and $2, $1, $6
-; M32-NEXT:    not $1, $1
-; M32-NEXT:    and $1, $1, $7
+; M32-NEXT:    xor $2, $4, $5
+; M32-NEXT:    xor $1, $6, $7
+; M32-NEXT:    sltu $2, $zero, $2
+; M32-NEXT:    negu $2, $2
+; M32-NEXT:    and $1, $1, $2
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    or $2, $2, $1
+; M32-NEXT:    xor $2, $7, $1
 ;
 ; M64-LABEL: test_ctselect_icmp_ne:
 ; M64:       # %bb.0:
 ; M64-NEXT:    sll $1, $5, 0
 ; M64-NEXT:    sll $2, $4, 0
-; M64-NEXT:    sll $3, $7, 0
 ; M64-NEXT:    xor $1, $2, $1
-; M64-NEXT:    sll $2, $6, 0
-; M64-NEXT:    sltiu $1, $1, 1
-; M64-NEXT:    addiu $1, $1, -1
-; M64-NEXT:    and $2, $1, $2
-; M64-NEXT:    not $1, $1
-; M64-NEXT:    and $1, $1, $3
+; M64-NEXT:    xor $2, $6, $7
+; M64-NEXT:    sltu $1, $zero, $1
+; M64-NEXT:    sll $2, $2, 0
+; M64-NEXT:    negu $1, $1
+; M64-NEXT:    and $1, $2, $1
+; M64-NEXT:    sll $2, $7, 0
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $2, $1
+; M64-NEXT:    xor $2, $2, $1
   %cond = icmp ne i32 %x, %y
   %result = call i32 @llvm.ct.select.i32(i1 %cond, i32 %a, i32 %b)
   ret i32 %result
@@ -228,29 +218,25 @@ define i32 @test_ctselect_icmp_ne(i32 %x, i32 %y, i32 %a, i32 %b) {
 define i32 @test_ctselect_icmp_slt(i32 %x, i32 %y, i32 %a, i32 %b) {
 ; M32-LABEL: test_ctselect_icmp_slt:
 ; M32:       # %bb.0:
-; M32-NEXT:    slt $1, $4, $5
-; M32-NEXT:    xori $1, $1, 1
-; M32-NEXT:    addiu $1, $1, -1
-; M32-NEXT:    and $2, $1, $6
-; M32-NEXT:    not $1, $1
-; M32-NEXT:    and $1, $1, $7
+; M32-NEXT:    slt $2, $4, $5
+; M32-NEXT:    xor $1, $6, $7
+; M32-NEXT:    negu $2, $2
+; M32-NEXT:    and $1, $1, $2
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    or $2, $2, $1
+; M32-NEXT:    xor $2, $7, $1
 ;
 ; M64-LABEL: test_ctselect_icmp_slt:
 ; M64:       # %bb.0:
 ; M64-NEXT:    sll $1, $5, 0
 ; M64-NEXT:    sll $2, $4, 0
-; M64-NEXT:    sll $3, $7, 0
 ; M64-NEXT:    slt $1, $2, $1
-; M64-NEXT:    sll $2, $6, 0
-; M64-NEXT:    xori $1, $1, 1
-; M64-NEXT:    addiu $1, $1, -1
-; M64-NEXT:    and $2, $1, $2
-; M64-NEXT:    not $1, $1
-; M64-NEXT:    and $1, $1, $3
+; M64-NEXT:    xor $2, $6, $7
+; M64-NEXT:    negu $1, $1
+; M64-NEXT:    sll $2, $2, 0
+; M64-NEXT:    and $1, $2, $1
+; M64-NEXT:    sll $2, $7, 0
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $2, $1
+; M64-NEXT:    xor $2, $2, $1
   %cond = icmp slt i32 %x, %y
   %result = call i32 @llvm.ct.select.i32(i1 %cond, i32 %a, i32 %b)
   ret i32 %result
@@ -259,29 +245,25 @@ define i32 @test_ctselect_icmp_slt(i32 %x, i32 %y, i32 %a, i32 %b) {
 define i32 @test_ctselect_icmp_ult(i32 %x, i32 %y, i32 %a, i32 %b) {
 ; M32-LABEL: test_ctselect_icmp_ult:
 ; M32:       # %bb.0:
-; M32-NEXT:    sltu $1, $4, $5
-; M32-NEXT:    xori $1, $1, 1
-; M32-NEXT:    addiu $1, $1, -1
-; M32-NEXT:    and $2, $1, $6
-; M32-NEXT:    not $1, $1
-; M32-NEXT:    and $1, $1, $7
+; M32-NEXT:    sltu $2, $4, $5
+; M32-NEXT:    xor $1, $6, $7
+; M32-NEXT:    negu $2, $2
+; M32-NEXT:    and $1, $1, $2
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    or $2, $2, $1
+; M32-NEXT:    xor $2, $7, $1
 ;
 ; M64-LABEL: test_ctselect_icmp_ult:
 ; M64:       # %bb.0:
 ; M64-NEXT:    sll $1, $5, 0
 ; M64-NEXT:    sll $2, $4, 0
-; M64-NEXT:    sll $3, $7, 0
 ; M64-NEXT:    sltu $1, $2, $1
-; M64-NEXT:    sll $2, $6, 0
-; M64-NEXT:    xori $1, $1, 1
-; M64-NEXT:    addiu $1, $1, -1
-; M64-NEXT:    and $2, $1, $2
-; M64-NEXT:    not $1, $1
-; M64-NEXT:    and $1, $1, $3
+; M64-NEXT:    xor $2, $6, $7
+; M64-NEXT:    negu $1, $1
+; M64-NEXT:    sll $2, $2, 0
+; M64-NEXT:    and $1, $2, $1
+; M64-NEXT:    sll $2, $7, 0
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $2, $1
+; M64-NEXT:    xor $2, $2, $1
   %cond = icmp ult i32 %x, %y
   %result = call i32 @llvm.ct.select.i32(i1 %cond, i32 %a, i32 %b)
   ret i32 %result
@@ -291,28 +273,26 @@ define i32 @test_ctselect_icmp_ult(i32 %x, i32 %y, i32 %a, i32 %b) {
 define i32 @test_ctselect_load(i1 %cond, ptr %p1, ptr %p2) {
 ; M32-LABEL: test_ctselect_load:
 ; M32:       # %bb.0:
-; M32-NEXT:    andi $1, $4, 1
+; M32-NEXT:    lw $2, 0($6)
 ; M32-NEXT:    lw $3, 0($5)
-; M32-NEXT:    negu $2, $1
-; M32-NEXT:    addiu $1, $1, -1
-; M32-NEXT:    and $2, $2, $3
-; M32-NEXT:    lw $3, 0($6)
-; M32-NEXT:    and $1, $1, $3
+; M32-NEXT:    andi $1, $4, 1
+; M32-NEXT:    negu $1, $1
+; M32-NEXT:    xor $3, $3, $2
+; M32-NEXT:    and $1, $3, $1
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    or $2, $2, $1
+; M32-NEXT:    xor $2, $2, $1
 ;
 ; M64-LABEL: test_ctselect_load:
 ; M64:       # %bb.0:
-; M64-NEXT:    sll $1, $4, 0
-; M64-NEXT:    lw $3, 0($5)
-; M64-NEXT:    andi $1, $1, 1
-; M64-NEXT:    negu $2, $1
-; M64-NEXT:    addiu $1, $1, -1
+; M64-NEXT:    sll $3, $4, 0
+; M64-NEXT:    lw $1, 0($6)
+; M64-NEXT:    lw $2, 0($5)
+; M64-NEXT:    andi $3, $3, 1
+; M64-NEXT:    xor $2, $2, $1
+; M64-NEXT:    negu $3, $3
 ; M64-NEXT:    and $2, $2, $3
-; M64-NEXT:    lw $3, 0($6)
-; M64-NEXT:    and $1, $1, $3
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $2, $1
+; M64-NEXT:    xor $2, $1, $2
   %a = load i32, ptr %p1
   %b = load i32, ptr %p2
   %result = call i32 @llvm.ct.select.i32(i1 %cond, i32 %a, i32 %b)
@@ -323,41 +303,37 @@ define i32 @test_ctselect_load(i1 %cond, ptr %p1, ptr %p2) {
 define i32 @test_ctselect_nested(i1 %cond1, i1 %cond2, i32 %a, i32 %b, i32 %c) {
 ; M32-LABEL: test_ctselect_nested:
 ; M32:       # %bb.0:
-; M32-NEXT:    andi $1, $5, 1
+; M32-NEXT:    andi $2, $5, 1
+; M32-NEXT:    xor $1, $6, $7
 ; M32-NEXT:    andi $3, $4, 1
-; M32-NEXT:    negu $2, $1
-; M32-NEXT:    addiu $1, $1, -1
-; M32-NEXT:    negu $4, $3
-; M32-NEXT:    and $2, $2, $6
-; M32-NEXT:    and $1, $1, $7
-; M32-NEXT:    or $1, $2, $1
-; M32-NEXT:    addiu $2, $3, -1
-; M32-NEXT:    lw $3, 16($sp)
-; M32-NEXT:    and $1, $4, $1
-; M32-NEXT:    and $2, $2, $3
+; M32-NEXT:    negu $2, $2
+; M32-NEXT:    negu $3, $3
+; M32-NEXT:    and $1, $1, $2
+; M32-NEXT:    lw $2, 16($sp)
+; M32-NEXT:    xor $1, $7, $1
+; M32-NEXT:    xor $1, $1, $2
+; M32-NEXT:    and $1, $1, $3
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    or $2, $1, $2
+; M32-NEXT:    xor $2, $2, $1
 ;
 ; M64-LABEL: test_ctselect_nested:
 ; M64:       # %bb.0:
 ; M64-NEXT:    sll $1, $5, 0
-; M64-NEXT:    sll $3, $6, 0
-; M64-NEXT:    sll $4, $4, 0
+; M64-NEXT:    xor $2, $6, $7
+; M64-NEXT:    sll $3, $4, 0
 ; M64-NEXT:    andi $1, $1, 1
-; M64-NEXT:    andi $4, $4, 1
-; M64-NEXT:    negu $2, $1
-; M64-NEXT:    addiu $1, $1, -1
-; M64-NEXT:    negu $5, $4
-; M64-NEXT:    and $2, $2, $3
-; M64-NEXT:    sll $3, $7, 0
-; M64-NEXT:    and $1, $1, $3
-; M64-NEXT:    addiu $3, $4, -1
-; M64-NEXT:    or $1, $2, $1
+; M64-NEXT:    sll $2, $2, 0
+; M64-NEXT:    andi $3, $3, 1
+; M64-NEXT:    negu $1, $1
+; M64-NEXT:    negu $3, $3
+; M64-NEXT:    and $1, $2, $1
+; M64-NEXT:    sll $2, $7, 0
+; M64-NEXT:    xor $1, $2, $1
 ; M64-NEXT:    sll $2, $8, 0
-; M64-NEXT:    and $1, $5, $1
-; M64-NEXT:    and $2, $3, $2
+; M64-NEXT:    xor $1, $1, $2
+; M64-NEXT:    and $1, $1, $3
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $1, $2
+; M64-NEXT:    xor $2, $2, $1
   %inner = call i32 @llvm.ct.select.i32(i1 %cond2, i32 %a, i32 %b)
   %result = call i32 @llvm.ct.select.i32(i1 %cond1, i32 %inner, i32 %c)
   ret i32 %result
diff --git a/llvm/test/CodeGen/Mips/ctselect-side-effects.ll b/llvm/test/CodeGen/Mips/ctselect-side-effects.ll
index 6cfa07afdd51..069100e2d2a7 100644
--- a/llvm/test/CodeGen/Mips/ctselect-side-effects.ll
+++ b/llvm/test/CodeGen/Mips/ctselect-side-effects.ll
@@ -38,26 +38,24 @@ define i32 @test_constant_fold() {
 define i32 @test_protected_no_branch(i1 %cond, i32 %a, i32 %b) {
 ; M32-LABEL: test_protected_no_branch:
 ; M32:       # %bb.0:
-; M32-NEXT:    andi $1, $4, 1
-; M32-NEXT:    negu $2, $1
-; M32-NEXT:    addiu $1, $1, -1
-; M32-NEXT:    and $2, $2, $5
-; M32-NEXT:    and $1, $1, $6
+; M32-NEXT:    andi $2, $4, 1
+; M32-NEXT:    xor $1, $5, $6
+; M32-NEXT:    negu $2, $2
+; M32-NEXT:    and $1, $1, $2
 ; M32-NEXT:    jr $ra
-; M32-NEXT:    or $2, $2, $1
+; M32-NEXT:    xor $2, $6, $1
 ;
 ; M64-LABEL: test_protected_no_branch:
 ; M64:       # %bb.0:
 ; M64-NEXT:    sll $1, $4, 0
-; M64-NEXT:    sll $3, $5, 0
+; M64-NEXT:    xor $2, $5, $6
 ; M64-NEXT:    andi $1, $1, 1
-; M64-NEXT:    negu $2, $1
-; M64-NEXT:    addiu $1, $1, -1
-; M64-NEXT:    and $2, $2, $3
-; M64-NEXT:    sll $3, $6, 0
-; M64-NEXT:    and $1, $1, $3
+; M64-NEXT:    sll $2, $2, 0
+; M64-NEXT:    negu $1, $1
+; M64-NEXT:    and $1, $2, $1
+; M64-NEXT:    sll $2, $6, 0
 ; M64-NEXT:    jr $ra
-; M64-NEXT:    or $2, $2, $1
+; M64-NEXT:    xor $2, $2, $1
   %result = call i32 @llvm.ct.select.i32(i1 %cond, i32 %a, i32 %b)
   ret i32 %result
 }
author	wizardengineer <juliuswoosebert@gmail.com>	2026-03-07 15:38:15 -0500
committer	wizardengineer <juliuswoosebert@gmail.com>	2026-03-07 16:39:57 -0500
commit	9d45e799ab7c4dea36c4c8df5ed757a92886c814 (patch)
tree	1d58dc1b5a08bf10e6732d72b8021280f759ff2a
parent	89b18c0965ccec867ad8e234c77ac7f83c7edfc5 (diff)
download	llvm-users/wizardengineer/ct-select-mips.tar.gz llvm-users/wizardengineer/ct-select-mips.tar.bz2 llvm-users/wizardengineer/ct-select-mips.zip