[PATCH] D108763: Use type sizes when determining dependence

[Malhar via Phabricator via llvm-commits]

malharJ added inline comments.


================
Comment at: llvm/test/Transforms/LoopVectorize/depend_diff_types.ll:40
+  %indvars.iv = phi i64 [ 1, %L.LB1_319.preheader ], [ %indvars.iv.next, %L.LB1_319 ]
+  %.dY0001_321.0 = phi i32 [ %1, %L.LB1_319.preheader ], [ %25, %L.LB1_319 ]
+  %5 = getelementptr i64, i64* %3, i64 %indvars.iv
----------------
Looks like this isn't being used anywhere and can be removed ?

If so, can also remove line 65 
```
%25 = add nsw i32 %.dY0001_321.0, -1
```

Then also change


```
%26 = icmp sgt i32 %.dY0001_321.0, 1
```

to something like


```
%26 = icmp slt i32 %indvars.iv.next, M
```


================
Comment at: llvm/test/Transforms/LoopVectorize/depend_diff_types.ll:58-63
+  %20 = add nsw i64 %indvars.iv, -1824
+  %21 = getelementptr i64, i64* %4, i64 %20
+  %22 = load i64, i64* %21, align 8, !tbaa !6
+  %23 = add nuw i64 %indvars.iv, 15296
+  %24 = getelementptr i64, i64* %3, i64 %23
+  store i64 %22, i64* %24, align 8, !tbaa !4
----------------
perhaps this can be removed ?

It looks like the IR corresponds to 


```
1) v3[i+31] = (v1[1903+i] * v3[1023+i]);   // ---> all accesses are double*

2) v3[i+1007] =  v1[i+47];   // ---> all accesses are i64*

3) v3[i+16319] = v1[i+79];   // ---> all accesses are i64* (could remove ?)
```

Perhaps we are only concerned with the lines 1 and 2, as they are of different types ?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D108763/new/

https://reviews.llvm.org/D108763