[clang] Pragma vectorize_width() implies vectorize(enable)

[SjoerdMeijer]

Summary

Specifying the vectorization width was supposed to implicitly enable
vectorization, except that it wasn't really doing this. It was only
setting the vectorize.width metadata, but not vectorize.enable.

This should also fix PR27643.

[fhahn]

As also pointed out in the discussion on the cfe dev list, this is probably a bit
of a silly combination:

   
vectorize(enable) vectorize_width(1)

but it could still mean that the vectorizer interleaves. So, with this
simplification, disabled means disabled, and a width of 1 a width of 1.

It would be good to have a test case for that.

[SjoerdMeijer]

Thanks for looking again!
Good catch, feedback addressed.

(forgot to add this message when I uploaded the new diff)

[fhahn]

I think it would be slightly better to split off the change to disable vectorization via llvm.loop.vectorize.enable=false instead of width=1. This changes the behaviour from "disable vectorization, but allow interleaving in the vectoriser" to "disable the vectoriser". IMO this change makes sense, but it is probably better/safer to do it separately.

[SjoerdMeijer]

Thanks, and sorry for the delay. Back in the office now, and I am addressing this:

I think it would be slightly better to split off the change to disable vectorization via llvm.loop.vectorize.enable=false instead of width=1.

Yep, I agree

[SjoerdMeijer]

I think it would be slightly better to split off the change to disable vectorization via llvm.loop.vectorize.enable=false instead of width=1.

This is now D66796.

I will now start stripping it out from this patch.

[SjoerdMeijer]

Stripped out the functional change related to vectorize(disable).

[fhahn]

LGTM

[SjoerdMeijer]

Just uploading new diff for completeness; I only had to change a test-case, and thus thought that committing this is okay.

Many thanks again for reviewing and helping with the discussions!

clang/lib/CodeGen/CGLoopInfo.cpp

Metadata *Vals[] = {
         MDString::get(Ctx, "llvm.loop.vectorize.predicate.enable"),
         ConstantAsMetadata::get(ConstantInt::get(llvm::Type::getInt1Ty(Ctx),
                                                  IsVectorPredicateEnabled))};
     Args.push_back(MDNode::get(Ctx, Vals));
   }
 
   // Setting vectorize.width
   if (Attrs.VectorizeWidth > 0) {
+    // This implies vectorize.enable = true, but only add it when it is not
+    // already enabled.
+    if (Attrs.VectorizeEnable != LoopAttributes::Enable)
+      Args.push_back(
+          MDNode::get(Ctx, {MDString::get(Ctx, "llvm.loop.vectorize.enable"),
+                            ConstantAsMetadata::get(ConstantInt::get(
+                                llvm::Type::getInt1Ty(Ctx), 1))}));
+
     Metadata *Vals[] = {
         MDString::get(Ctx, "llvm.loop.vectorize.width"),
         ConstantAsMetadata::get(ConstantInt::get(llvm::Type::getInt32Ty(Ctx),
                                                  Attrs.VectorizeWidth))};
     Args.push_back(MDNode::get(Ctx, Vals));
   }
 
   // Setting interleave.count

clang/test/CodeGenCXX/pragma-loop-predicate.cpp

 // CHECK-LABEL: @{{.*}}test5{{.*}}(
 // CHECK: br label {{.*}}, !llvm.loop ![[LOOP5:.*]]
 
   #pragma clang loop vectorize(disable) vectorize_predicate(enable)
   for (int i = 0; i < Length; i++)
     List[i] = i * 2;
 }
 
-
 // CHECK:      ![[LOOP0]] = distinct !{![[LOOP0]], !3}
 // CHECK-NEXT: !3 = !{!"llvm.loop.vectorize.enable", i1 true}
 
 // CHECK-NEXT: ![[LOOP1]] = distinct !{![[LOOP1]], !5, !3}
 // CHECK-NEXT: !5 = !{!"llvm.loop.vectorize.predicate.enable", i1 true}
 
 // CHECK-NEXT: ![[LOOP2]] = distinct !{![[LOOP2]], !7, !3}
 // CHECK-NEXT: !7 = !{!"llvm.loop.vectorize.predicate.enable", i1 false}
 
 // CHECK-NEXT: ![[LOOP3]] = distinct !{![[LOOP3]], !5, !3}
 
-// CHECK-NEXT: ![[LOOP4]] = distinct !{![[LOOP4]], !10}
+// CHECK-NEXT: ![[LOOP4]] = distinct !{![[LOOP4]], !3, !10}
 // CHECK-NEXT: !10 = !{!"llvm.loop.vectorize.width", i32 1}
 
-// CHECK-NEXT: ![[LOOP5]] = distinct !{![[LOOP5]], !10}
+// CHECK-NEXT: ![[LOOP5]] = distinct !{![[LOOP5]], !3, !10}

clang/test/CodeGenCXX/pragma-loop.cpp

 void template_test(double *List, int Length) {
   double Value = 10;
 
   for_template_test<double>(List, Length, Value);
   for_template_define_test<double, int>(List, Length, Value);
   for_template_constant_expression_test<double, 2, 4, 8>(List, Length);
 }
 
+void vec_width_1(int *List, int Length) {
+// CHECK-LABEL: @{{.*}}vec_width_1{{.*}}(
+// CHECK: br label {{.*}}, !llvm.loop ![[LOOP_15:.*]]
+
+  #pragma clang loop vectorize(enable) vectorize_width(1)
+  for (int i = 0; i < Length; i++)
+    List[i] = i * 2;
+}
+
+void width_1(int *List, int Length) {
+// CHECK-LABEL: @{{.*}}width_1{{.*}}(
+// CHECK: br label {{.*}}, !llvm.loop ![[LOOP_16:.*]]
+
+  #pragma clang loop vectorize_width(1)
+  for (int i = 0; i < Length; i++)
+    List[i] = i * 2;
+}
+
 // CHECK: ![[LOOP_1]] = distinct !{![[LOOP_1]], ![[UNROLL_FULL:.*]]}
 // CHECK: ![[UNROLL_FULL]] = !{!"llvm.loop.unroll.full"}
 
-// CHECK: ![[LOOP_2]] = distinct !{![[LOOP_2]], ![[UNROLL_DISABLE:.*]], ![[DISTRIBUTE_DISABLE:.*]], ![[WIDTH_8:.*]], ![[INTERLEAVE_4:.*]]}
+// CHECK: ![[LOOP_2]] = distinct !{![[LOOP_2]], ![[UNROLL_DISABLE:.*]], ![[DISTRIBUTE_DISABLE:.*]], ![[VECTORIZE_ENABLE:.*]], ![[WIDTH_8:.*]], ![[INTERLEAVE_4:.*]]}
 // CHECK: ![[UNROLL_DISABLE]] = !{!"llvm.loop.unroll.disable"}
 // CHECK: ![[DISTRIBUTE_DISABLE]] = !{!"llvm.loop.distribute.enable", i1 false}
+// CHECK: ![[VECTORIZE_ENABLE]] = !{!"llvm.loop.vectorize.enable", i1 true}
 // CHECK: ![[WIDTH_8]] = !{!"llvm.loop.vectorize.width", i32 8}
 // CHECK: ![[INTERLEAVE_4]] = !{!"llvm.loop.interleave.count", i32 4}
 
-// CHECK: ![[LOOP_3]] = distinct !{![[LOOP_3]], ![[INTERLEAVE_4:.*]], ![[INTENABLE_1:.*]], ![[FOLLOWUP_VECTOR_3:.*]]}
-// CHECK: ![[INTENABLE_1]] = !{!"llvm.loop.vectorize.enable", i1 true}
+// CHECK: ![[LOOP_3]] = distinct !{![[LOOP_3]], ![[INTERLEAVE_4:.*]], ![[VECTORIZE_ENABLE]], ![[FOLLOWUP_VECTOR_3:.*]]}
 // CHECK: ![[FOLLOWUP_VECTOR_3]] = !{!"llvm.loop.vectorize.followup_all", ![[AFTER_VECTOR_3:.*]]}
 // CHECK: ![[AFTER_VECTOR_3]] = distinct !{![[AFTER_VECTOR_3]], ![[ISVECTORIZED:.*]], ![[UNROLL_8:.*]]}
 // CHECK: ![[ISVECTORIZED]] = !{!"llvm.loop.isvectorized"}
 // CHECK: ![[UNROLL_8]] = !{!"llvm.loop.unroll.count", i32 8}
 
-// CHECK: ![[LOOP_4]] = distinct !{![[LOOP_4]], ![[WIDTH_2:.*]], ![[INTERLEAVE_2:.*]]}
+// CHECK: ![[LOOP_4]] = distinct !{![[LOOP_4]], ![[VECTORIZE_ENABLE]], ![[WIDTH_2:.*]], ![[INTERLEAVE_2:.*]]}
 // CHECK: ![[WIDTH_2]] = !{!"llvm.loop.vectorize.width", i32 2}
 // CHECK: ![[INTERLEAVE_2]] = !{!"llvm.loop.interleave.count", i32 2}
 
 // CHECK: ![[LOOP_5]] = distinct !{![[LOOP_5]], ![[UNROLL_DISABLE:.*]], ![[DISTRIBUTE_DISABLE:.*]], ![[WIDTH_1:.*]]}
 // CHECK: ![[WIDTH_1]] = !{!"llvm.loop.vectorize.width", i32 1}
 
 // CHECK: ![[LOOP_6]] = distinct !{![[LOOP_6]], ![[WIDTH_2:.*]], ![[INTERLEAVE_2:.*]], ![[FOLLOWUP_VECTOR_6:.*]]}
 // CHECK: ![[FOLLOWUP_VECTOR_6]] = !{!"llvm.loop.vectorize.followup_all", ![[AFTER_VECTOR_6:.*]]}
 // CHECK: ![[AFTER_VECTOR_6]] = distinct !{![[AFTER_VECTOR_6]], ![[ISVECTORIZED:.*]], ![[UNROLL_8:.*]]}
 
-// CHECK: ![[LOOP_7]] = distinct !{![[LOOP_7]], ![[WIDTH_5:.*]]}
+// CHECK: ![[LOOP_7]] = distinct !{![[LOOP_7]], ![[VECTORIZE_ENABLE]], ![[WIDTH_5:.*]]}
 // CHECK: ![[WIDTH_5]] = !{!"llvm.loop.vectorize.width", i32 5}
 
 // CHECK: ![[LOOP_8]] = distinct !{![[LOOP_8]], ![[WIDTH_5:.*]]}
 
 // CHECK: ![[LOOP_9]] = distinct !{![[LOOP_9]], ![[WIDTH_8:.*]], ![[INTERLEAVE_8:.*]], ![[FOLLOWUP_VECTOR_9:.*]]}
 // CHECK: ![[FOLLOWUP_VECTOR_9]] = !{!"llvm.loop.vectorize.followup_all", ![[AFTER_VECTOR_9:.*]]}
 // CHECK: ![[AFTER_VECTOR_9]] = distinct !{![[AFTER_VECTOR_9]], ![[ISVECTORIZED:.*]], ![[UNROLL_8:.*]]}
 

 // CHECK: ![[UNROLL_24]] = !{!"llvm.loop.unroll.count", i32 24}
 
 // CHECK: ![[LOOP_13]] = distinct !{![[LOOP_13]], ![[WIDTH_8:.*]], ![[INTERLEAVE_16:.*]], ![[FOLLOWUP_VECTOR_13:.*]]}
 // CHECK: ![[INTERLEAVE_16]] = !{!"llvm.loop.interleave.count", i32 16}
 // CHECK: ![[FOLLOWUP_VECTOR_13]] = !{!"llvm.loop.vectorize.followup_all", ![[AFTER_VECTOR_13:.*]]}
 // CHECK: ![[AFTER_VECTOR_13]] = distinct !{![[AFTER_VECTOR_13]], ![[ISVECTORIZED:.*]], ![[UNROLL_32:.*]]}
 // CHECK: ![[UNROLL_32]] = !{!"llvm.loop.unroll.count", i32 32}
 
-// CHECK: ![[LOOP_14]] = distinct !{![[LOOP_14]], ![[WIDTH_10:.*]]}
+// CHECK: ![[LOOP_14]] = distinct !{![[LOOP_14]], ![[VECTORIZE_ENABLE]], ![[WIDTH_10:.*]]}
 // CHECK: ![[WIDTH_10]] = !{!"llvm.loop.vectorize.width", i32 10}
+
+// CHECK:      ![[LOOP_15]] = distinct !{![[LOOP_15]], ![[WIDTH_1]], ![[VECTORIZE_ENABLE]]}
+
+// CHECK-NEXT: ![[LOOP_16]] = distinct !{![[LOOP_16]], ![[VECTORIZE_ENABLE]], ![[WIDTH_1]]}