[AArch64] prevent (shl (srl x, c1), c2) -> (and (shift x, c3)) when load

Currently, process of replacing bitwise operations consisting of
`(shl (srl x, c1), c2)` with `And` is performed by `DAGCombiner`.

However, in certain case like `(shl (srl, x, c1) 2)` is do not
need to transform to `AND` if it was used to `Load` Target.

Consider following case:
```
        lsr x8, x8, #56
        and x8, x8, #0xfc
        ldr w0, [x2, x8]
        ret
```

In this case, we can remove the `AND` by changing the target of `LDR`
to `[X2, X8, LSL #2]` and right-shifting amount change to 56 to 58.

after changed:
```
        lsr x8, x8, #58
        ldr w0, [x2, x8, lsl #2]
        ret
```

This patch checks to see if the `(shl (srl x, c1) 2)` operation on
`load` target can be prevent transform to `And`.

Author: ParkHanbum

llvm/lib/Target/AArch64/AArch64ISelLowering.cpp

@@ -18012,6 +18012,23 @@ bool AArch64TargetLowering::shouldFoldConstantShiftPairToMask(
     return (!C1 || !C2 || C1->getZExtValue() >= C2->getZExtValue());
   }
 
+  // We do not need to fold when this shifting used in specific load case:
+  // (ldr x, (add x, (shl (srl x, c1) 2)))
+  if (N->getOpcode() == ISD::SHL && N->hasOneUse()) {
+    if (auto C2 = dyn_cast<ConstantSDNode>(N->getOperand(1))) {
+      unsigned ShlAmt = C2->getZExtValue();
+      if (auto ShouldADD = *N->use_begin();
+          ShouldADD->getOpcode() == ISD::ADD && ShouldADD->hasOneUse()) {
+        if (auto ShouldLOAD = dyn_cast<LoadSDNode>(*ShouldADD->use_begin())) {
+          unsigned ByteVT = ShouldLOAD->getMemoryVT().getSizeInBits() / 8;
+          if ((1ULL << ShlAmt) == ByteVT &&
+              isIndexedLoadLegal(ISD::PRE_INC, ShouldLOAD->getMemoryVT()))
+            return false;
+        }
+      }
+    }
+  }
+
   return true;
 }
 

llvm/test/CodeGen/AArch64/arm64-fold-lshr.ll

@@ -6,9 +6,8 @@ define i16 @load16_shr63(i64 %a, i64 %b, ptr %table) {
 ; CHECK-LABEL: load16_shr63:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    mul x8, x1, x0
-; CHECK-NEXT:    lsr x8, x8, #62
-; CHECK-NEXT:    and x8, x8, #0x2
-; CHECK-NEXT:    ldrh w0, [x2, x8]
+; CHECK-NEXT:    lsr x8, x8, #63
+; CHECK-NEXT:    ldrh w0, [x2, x8, lsl #1]
 ; CHECK-NEXT:    ret
 entry:
   %mul = mul i64 %b, %a
@@ -22,9 +21,8 @@ define i16 @load16_shr2(i64 %a, i64 %b, ptr %table) {
 ; CHECK-LABEL: load16_shr2:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    mul x8, x1, x0
-; CHECK-NEXT:    lsr x8, x8, #1
-; CHECK-NEXT:    and x8, x8, #0x7ffffffffffffffe
-; CHECK-NEXT:    ldrh w0, [x2, x8]
+; CHECK-NEXT:    lsr x8, x8, #2
+; CHECK-NEXT:    ldrh w0, [x2, x8, lsl #1]
 ; CHECK-NEXT:    ret
 entry:
   %mul = mul i64 %b, %a
@@ -38,8 +36,8 @@ define i16 @load16_shr1(i64 %a, i64 %b, ptr %table) {
 ; CHECK-LABEL: load16_shr1:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    mul x8, x1, x0
-; CHECK-NEXT:    and x8, x8, #0xfffffffffffffffe
-; CHECK-NEXT:    ldrh w0, [x2, x8]
+; CHECK-NEXT:    lsr x8, x8, #1
+; CHECK-NEXT:    ldrh w0, [x2, x8, lsl #1]
 ; CHECK-NEXT:    ret
 entry:
   %mul = mul i64 %b, %a
@@ -53,9 +51,8 @@ define i32 @load32_shr63(i64 %a, i64 %b, ptr %table) {
 ; CHECK-LABEL: load32_shr63:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    mul x8, x1, x0
-; CHECK-NEXT:    lsr x8, x8, #61
-; CHECK-NEXT:    and x8, x8, #0x4
-; CHECK-NEXT:    ldr w0, [x2, x8]
+; CHECK-NEXT:    lsr x8, x8, #63
+; CHECK-NEXT:    ldr w0, [x2, x8, lsl #2]
 ; CHECK-NEXT:    ret
 entry:
   %mul = mul i64 %b, %a
@@ -69,8 +66,8 @@ define i32 @load32_shr2(i64 %a, i64 %b, ptr %table) {
 ; CHECK-LABEL: load32_shr2:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    mul x8, x1, x0
-; CHECK-NEXT:    and x8, x8, #0xfffffffffffffffc
-; CHECK-NEXT:    ldr w0, [x2, x8]
+; CHECK-NEXT:    lsr x8, x8, #2
+; CHECK-NEXT:    ldr w0, [x2, x8, lsl #2]
 ; CHECK-NEXT:    ret
 entry:
   %mul = mul i64 %b, %a
@@ -84,9 +81,8 @@ define i32 @load32_shr1(i64 %a, i64 %b, ptr %table) {
 ; CHECK-LABEL: load32_shr1:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    mul x8, x1, x0
-; CHECK-NEXT:    lsl x8, x8, #1
-; CHECK-NEXT:    and x8, x8, #0xfffffffffffffffc
-; CHECK-NEXT:    ldr w0, [x2, x8]
+; CHECK-NEXT:    lsr x8, x8, #1
+; CHECK-NEXT:    ldr w0, [x2, x8, lsl #2]
 ; CHECK-NEXT:    ret
 entry:
   %mul = mul i64 %b, %a
@@ -100,9 +96,8 @@ define i64 @load64_shr63(i64 %a, i64 %b, ptr %table) {
 ; CHECK-LABEL: load64_shr63:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    mul x8, x1, x0
-; CHECK-NEXT:    lsr x8, x8, #60
-; CHECK-NEXT:    and x8, x8, #0x8
-; CHECK-NEXT:    ldr x0, [x2, x8]
+; CHECK-NEXT:    lsr x8, x8, #63
+; CHECK-NEXT:    ldr x0, [x2, x8, lsl #3]
 ; CHECK-NEXT:    ret
 entry:
   %mul = mul i64 %b, %a
@@ -116,9 +111,8 @@ define i64 @load64_shr2(i64 %a, i64 %b, ptr %table) {
 ; CHECK-LABEL: load64_shr2:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    mul x8, x1, x0
-; CHECK-NEXT:    lsl x8, x8, #1
-; CHECK-NEXT:    and x8, x8, #0xfffffffffffffff8
-; CHECK-NEXT:    ldr x0, [x2, x8]
+; CHECK-NEXT:    lsr x8, x8, #2
+; CHECK-NEXT:    ldr x0, [x2, x8, lsl #3]
 ; CHECK-NEXT:    ret
 entry:
   %mul = mul i64 %b, %a
@@ -132,9 +126,8 @@ define i64 @load64_shr1(i64 %a, i64 %b, ptr %table) {
 ; CHECK-LABEL: load64_shr1:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    mul x8, x1, x0
-; CHECK-NEXT:    lsl x8, x8, #2
-; CHECK-NEXT:    and x8, x8, #0xfffffffffffffff8
-; CHECK-NEXT:    ldr x0, [x2, x8]
+; CHECK-NEXT:    lsr x8, x8, #1
+; CHECK-NEXT:    ldr x0, [x2, x8, lsl #3]
 ; CHECK-NEXT:    ret
 entry:
   %mul = mul i64 %b, %a