Add ADD and SUB to the predicable ARM instructions.

It is not my plan to duplicate the entire ARM instruction set with
predicated versions. We need a way of representing predicated
instructions in SSA form without requiring a separate opcode.

Then the pseudo-instructions can go away.

llvm-svn: 162061

3 files changed, 35 insertions, 21 deletions

diff --git a/llvm/test/CodeGen/ARM/select.ll b/llvm/test/CodeGen/ARM/select.ll
index 418d4f31ee2..55755666289 100644
--- a/llvm/test/CodeGen/ARM/select.ll
+++ b/llvm/test/CodeGen/ARM/select.ll
@@ -76,12 +76,11 @@ define double @f7(double %a, double %b) {
 ; block generated, odds are good that we have close to the ideal code for this:
 ;
 ; CHECK-NEON:      _f8:
+; CHECK-NEON:      movw    [[R3:r[0-9]+]], #1123
 ; CHECK-NEON:      adr     [[R2:r[0-9]+]], LCPI7_0
-; CHECK-NEON-NEXT: movw    [[R3:r[0-9]+]], #1123
-; CHECK-NEON-NEXT: adds    {{r.*}}, [[R2]], #4
 ; CHECK-NEON-NEXT: cmp     r0, [[R3]]
-; CHECK-NEON-NEXT: it      ne
-; CHECK-NEON-NEXT: movne   {{r.*}}, [[R2]]
+; CHECK-NEON-NEXT: it      eq
+; CHECK-NEON-NEXT: addeq.w {{r.*}}, [[R2]]
 ; CHECK-NEON-NEXT: ldr
 ; CHECK-NEON:      bx
 
diff --git a/llvm/test/CodeGen/ARM/select_xform.ll b/llvm/test/CodeGen/ARM/select_xform.ll
index d98d7a628e2..26f7cb68901 100644
--- a/llvm/test/CodeGen/ARM/select_xform.ll
+++ b/llvm/test/CodeGen/ARM/select_xform.ll
@@ -4,13 +4,13 @@
 
 define i32 @t1(i32 %a, i32 %b, i32 %c) nounwind {
 ; ARM: t1:
-; ARM: sub r0, r1, #-2147483647
-; ARM: movgt r0, r1
+; ARM: suble r1, r1, #-2147483647
+; ARM: mov r0, r1
 
 ; T2: t1:
 ; T2: mvn r0, #-2147483648
-; T2: add r0, r1
-; T2: movgt r0, r1
+; T2: addle.w r1, r1
+; T2: mov r0, r1
   %tmp1 = icmp sgt i32 %c, 10
   %tmp2 = select i1 %tmp1, i32 0, i32 2147483647
   %tmp3 = add i32 %tmp2, %b
@@ -19,12 +19,12 @@ define i32 @t1(i32 %a, i32 %b, i32 %c) nounwind {
 
 define i32 @t2(i32 %a, i32 %b, i32 %c, i32 %d) nounwind {
 ; ARM: t2:
-; ARM: sub r0, r1, #10
-; ARM: movgt r0, r1
+; ARM: suble r1, r1, #10
+; ARM: mov r0, r1
 
 ; T2: t2:
-; T2: sub.w r0, r1, #10
-; T2: movgt r0, r1
+; T2: suble.w r1, r1, #10
+; T2: mov r0, r1
   %tmp1 = icmp sgt i32 %c, 10
   %tmp2 = select i1 %tmp1, i32 0, i32 10
   %tmp3 = sub i32 %b, %tmp2
@@ -164,3 +164,18 @@ define i32 @t11(i32 %a, i32 %b) nounwind {
   %tmp1 = select i1 %cond, i32 %x, i32 %a
   ret i32 %tmp1
 }
+
+; Fold ADDri12 into movcc
+define i32 @t12(i32 %a, i32 %b) nounwind {
+; ARM: t12:
+; ARM: cmp r0, r1
+; ARM: addge r0, r1,
+
+; T2: t12:
+; T2: cmp r0, r1
+; T2: addwge r0, r1, #3000
+  %x = add i32 %b, 3000
+  %cond = icmp slt i32 %a, %b
+  %tmp1 = select i1 %cond, i32 %a, i32 %x
+  ret i32 %tmp1
+}
diff --git a/llvm/test/CodeGen/Thumb2/thumb2-select_xform.ll b/llvm/test/CodeGen/Thumb2/thumb2-select_xform.ll
index 74729fd4150..ead198f2162 100644
--- a/llvm/test/CodeGen/Thumb2/thumb2-select_xform.ll
+++ b/llvm/test/CodeGen/Thumb2/thumb2-select_xform.ll
@@ -4,9 +4,9 @@ define i32 @t1(i32 %a, i32 %b, i32 %c) nounwind {
 ; CHECK: t1
 ; CHECK: mvn r0, #-2147483648
 ; CHECK: cmp r2, #10
-; CHECK: add r0, r1
-; CHECK: it  gt
-; CHECK: movgt r0, r1
+; CHECK: it  le
+; CHECK: addle.w r1, r1, r0
+; CHECK: mov r0, r1
         %tmp1 = icmp sgt i32 %c, 10
         %tmp2 = select i1 %tmp1, i32 0, i32 2147483647
         %tmp3 = add i32 %tmp2, %b
@@ -15,10 +15,10 @@ define i32 @t1(i32 %a, i32 %b, i32 %c) nounwind {
 
 define i32 @t2(i32 %a, i32 %b, i32 %c) nounwind {
 ; CHECK: t2
-; CHECK: add.w r0, r1, #-2147483648
 ; CHECK: cmp r2, #10
-; CHECK: it  gt
-; CHECK: movgt r0, r1
+; CHECK: it  le
+; CHECK: addle.w r1, r1, #-2147483648
+; CHECK: mov r0, r1
 
         %tmp1 = icmp sgt i32 %c, 10
         %tmp2 = select i1 %tmp1, i32 0, i32 2147483648
@@ -28,10 +28,10 @@ define i32 @t2(i32 %a, i32 %b, i32 %c) nounwind {
 
 define i32 @t3(i32 %a, i32 %b, i32 %c, i32 %d) nounwind {
 ; CHECK: t3
-; CHECK: sub.w r0, r1, #10
 ; CHECK: cmp r2, #10
-; CHECK: it  gt
-; CHECK: movgt r0, r1
+; CHECK: it  le
+; CHECK: suble.w r1, r1, #10
+; CHECK: mov r0, r1
         %tmp1 = icmp sgt i32 %c, 10
         %tmp2 = select i1 %tmp1, i32 0, i32 10
         %tmp3 = sub i32 %b, %tmp2