[ARM] Mark 255 and 65535 as cheap for Thumb1 "And"

[dmgreen]

Summary

This prevents Constant Hoisting from pulling the constant out of the block, allowing
us to still produce LDRH/UXTH nodes. LDRB/UXTB (255) is already cheap by the
default getIntImmCost, but I've added it for clarity.

[samparker]

LGTM

[dmgreen]

Thanks Sam

llvm/trunk/lib/Target/ARM/ARMTargetTransformInfo.cpp

int ARMTTIImpl::getIntImmCost(unsigned Opcode, unsigned Idx, const APInt &Imm,
   // know it's constant. So it's not so much that the immediate is cheap (it's
   // not), but that the alternative is worse.
   // FIXME: this is probably unneeded with GlobalISel.
   if ((Opcode == Instruction::SDiv || Opcode == Instruction::UDiv ||
        Opcode == Instruction::SRem || Opcode == Instruction::URem) &&
       Idx == 1)
     return 0;
 
-  if (Opcode == Instruction::And)
-      // Conversion to BIC is free, and means we can use ~Imm instead.
-      return std::min(getIntImmCost(Imm, Ty), getIntImmCost(~Imm, Ty));
+  if (Opcode == Instruction::And) {
+    // UXTB/UXTH
+    if (Imm == 255 || Imm == 65535)
+      return 0;
+  }
 
   if (Opcode == Instruction::Add)
     // Conversion to SUB is free, and means we can use -Imm instead.
     return std::min(getIntImmCost(Imm, Ty), getIntImmCost(-Imm, Ty));
 
   if (Opcode == Instruction::ICmp && Imm.isNegative() &&
       Ty->getIntegerBitWidth() == 32) {
     int64_t NegImm = -Imm.getSExtValue();

llvm/trunk/test/CodeGen/Thumb/uxth.ll

for.body:
   br i1 %exitcond, label %for.cond.cleanup, label %for.body
 
 for.cond.cleanup:
   ret void
 }
 
 define void @uxth_loop_2(i32* %a, i32 %n) {
 ; V6M-LABEL: uxth_loop_2:
-; V6M:         .save {r4, r5, r7, lr}
-; V6M-NEXT:    push {r4, r5, r7, lr}
-; V6M-NEXT:    movs r3, #0
-; V6M-NEXT:    ldr r2, .LCPI5_0
+; V6M:         .save {r4, lr}
+; V6M-NEXT:    push {r4, lr}
+; V6M-NEXT:    movs r2, #0
 ; V6M-NEXT:  .LBB5_1:
-; V6M-NEXT:    adds r4, r3, #1
-; V6M-NEXT:    ands r3, r2
-; V6M-NEXT:    ldr r5, [r0]
-; V6M-NEXT:    ands r5, r2
-; V6M-NEXT:    muls r5, r3, r5
-; V6M-NEXT:    stm r0!, {r5}
-; V6M-NEXT:    cmp r1, r4
-; V6M-NEXT:    mov r3, r4
+; V6M-NEXT:    uxth r3, r2
+; V6M-NEXT:    ldrh r4, [r0]
+; V6M-NEXT:    muls r4, r3, r4
+; V6M-NEXT:    stm r0!, {r4}
+; V6M-NEXT:    adds r2, r2, #1
+; V6M-NEXT:    cmp r1, r2
 ; V6M-NEXT:    bne .LBB5_1
-; V6M-NEXT:    pop {r4, r5, r7, pc}
-; V6M-NEXT:    .p2align 2
-; V6M-NEXT:  .LCPI5_0:
-; V6M-NEXT:    .long 65535
+; V6M-NEXT:    pop {r4, pc}
 entry:
   br label %for.body
 
 for.body:
   %i = phi i32 [ 0, %entry ], [ %inc, %for.body ]
   %arrayidx = getelementptr inbounds i32, i32* %a, i32 %i
   %x = load i32, i32* %arrayidx
   %xn = and i32 %x, 65535