[AArch64] Support merging of narrow zero stores

Author: guy-david

llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp

@@ -1676,10 +1676,12 @@ static bool areCandidatesToMergeOrPair(MachineInstr &FirstMI, MachineInstr &MI,
   if (!PairIsValidLdStrOpc)
     return false;
 
-  // FIXME: We don't support merging narrow stores with mixed scaled/unscaled
-  // offsets.
+  // Narrow stores do not have a matching pair opcodes, so constrain their
+  // merging to zero stores.
   if (isNarrowStore(OpcA) || isNarrowStore(OpcB))
-    return false;
+    return getLdStRegOp(FirstMI).getReg() == AArch64::WZR &&
+           getLdStRegOp(MI).getReg() == AArch64::WZR &&
+           TII->getMemScale(FirstMI) == TII->getMemScale(MI);
 
   // The STR<S,D,Q,W,X>pre - STR<S,D,Q,W,X>ui and
   // LDR<S,D,Q,W,X,SW>pre-LDR<S,D,Q,W,X,SW>ui

llvm/test/CodeGen/AArch64/str-narrow-zero-merge.mir

@@ -29,8 +29,7 @@ name:  merge_scaled_str_with_unscaled_8
 body:             |
   bb.0.entry:
     ; CHECK-LABEL: name: merge_scaled_str_with_unscaled_8
-    ; CHECK: STRBBui $wzr, $x0, 4 :: (store (s8))
-    ; CHECK-NEXT: STURBBi $wzr, $x0, 5 :: (store (s8))
+    ; CHECK: STRHHui $wzr, $x0, 2 :: (store (s8))
     ; CHECK-NEXT: RET undef $lr
     STRBBui $wzr, $x0, 4 :: (store (s8))
     STURBBi $wzr, $x0, 5 :: (store (s8))
@@ -41,14 +40,35 @@ name:  merge_unscaled_str_with_scaled_8
 body:             |
   bb.0.entry:
     ; CHECK-LABEL: name: merge_unscaled_str_with_scaled_8
-    ; CHECK: STURBBi $wzr, $x0, 4 :: (store (s8))
-    ; CHECK-NEXT: STRBBui $wzr, $x0, 5 :: (store (s8))
+    ; CHECK: STURHHi $wzr, $x0, 4 :: (store (s8))
     ; CHECK-NEXT: RET undef $lr
     STURBBi $wzr, $x0, 4 :: (store (s8))
     STRBBui $wzr, $x0, 5 :: (store (s8))
     RET undef $lr
 ...
 ---
+name:  merge_unscaled_str_with_scaled_8_lower_address_second
+body:             |
+  bb.0.entry:
+    ; CHECK-LABEL: name: merge_unscaled_str_with_scaled_8_lower_address_second
+    ; CHECK: STURHHi $wzr, $x0, 0 :: (store (s8))
+    ; CHECK-NEXT: RET undef $lr
+    STURBBi $wzr, $x0, 1 :: (store (s8))
+    STRBBui $wzr, $x0, 0 :: (store (s8))
+    RET undef $lr
+...
+---
+name:  merge_scaled_str_with_unscaled_8_lower_address_second
+body:             |
+  bb.0.entry:
+    ; CHECK-LABEL: name: merge_scaled_str_with_unscaled_8_lower_address_second
+    ; CHECK: STRHHui $wzr, $x0, 0 :: (store (s8))
+    ; CHECK-NEXT: RET undef $lr
+    STRBBui $wzr, $x0, 1 :: (store (s8))
+    STURBBi $wzr, $x0, 0 :: (store (s8))
+    RET undef $lr
+...
+---
 name:  merge_unscaled_str_with_unscaled_str_16
 body:             |
   bb.0:
@@ -75,8 +95,7 @@ name:  merge_scaled_str_with_unscaled_16
 body:             |
   bb.0.entry:
     ; CHECK-LABEL: name: merge_scaled_str_with_unscaled_16
-    ; CHECK: STRHHui $wzr, $x0, 2 :: (store (s16))
-    ; CHECK-NEXT: STURHHi $wzr, $x0, 6 :: (store (s16))
+    ; CHECK: STRWui $wzr, $x0, 1 :: (store (s16))
     ; CHECK-NEXT: RET undef $lr
     STRHHui $wzr, $x0, 2 :: (store (s16))
     STURHHi $wzr, $x0, 6 :: (store (s16))
@@ -87,14 +106,35 @@ name:  merge_unscaled_str_with_scaled_16
 body:             |
   bb.0.entry:
     ; CHECK-LABEL: name: merge_unscaled_str_with_scaled_16
-    ; CHECK: STURHHi $wzr, $x0, 4 :: (store (s16))
-    ; CHECK-NEXT: STRHHui $wzr, $x0, 3 :: (store (s16))
+    ; CHECK: STURWi $wzr, $x0, 4 :: (store (s16))
     ; CHECK-NEXT: RET undef $lr
     STURHHi $wzr, $x0, 4 :: (store (s16))
     STRHHui $wzr, $x0, 3 :: (store (s16))
     RET undef $lr
 ...
 ---
+name:  merge_unscaled_str_with_scaled_16_lower_address_second
+body:             |
+  bb.0.entry:
+    ; CHECK-LABEL: name: merge_unscaled_str_with_scaled_16_lower_address_second
+    ; CHECK: STURWi $wzr, $x0, 2 :: (store (s16))
+    ; CHECK-NEXT: RET undef $lr
+    STURHHi $wzr, $x0, 4 :: (store (s16))
+    STRHHui $wzr, $x0, 1 :: (store (s16))
+    RET undef $lr
+...
+---
+name:  merge_scaled_str_with_unscaled_16_lower_address_second
+body:             |
+  bb.0.entry:
+    ; CHECK-LABEL: name: merge_scaled_str_with_unscaled_16_lower_address_second
+    ; CHECK: STRWui $wzr, $x0, 0 :: (store (s16))
+    ; CHECK-NEXT: RET undef $lr
+    STRHHui $wzr, $x0, 1 :: (store (s16))
+    STURHHi $wzr, $x0, 0 :: (store (s16))
+    RET undef $lr
+...
+---
 name:  merge_unscaled_str_with_unscaled_32
 body:             |
   bb.0.entry: