Fix vectorization remarks.

This patch changes the vectorization remarks to also inform when
vectorization is possible but not beneficial.

Added tests to exercise some loop remarks.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207574 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Transforms/Vectorize/LoopVectorize.cpp b/lib/Transforms/Vectorize/LoopVectorize.cpp
index 586c220162fb5e9d7cd3d5ecc9777667a9d03689..af0e686a521667444702b3d618f3cd1117773afa 100644 (file)
--- a/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -1209,6 +1209,13 @@ struct LoopVectorize : public FunctionPass {
       if (UF == 1)
         return false;
       DEBUG(dbgs() << "LV: Trying to at least unroll the loops.\n");
+
+      // Report the unrolling decision.
+      F->getContext().emitOptimizationRemark(
+          DEBUG_TYPE, *F, L->getStartLoc(),
+          Twine("unrolled with interleaving factor " + Twine(UF) +
+                " (vectorization not beneficial)"));
+
       // We decided not to vectorize, but we may want to unroll.
       InnerLoopUnroller Unroller(L, SE, LI, DT, DL, TLI, UF);
       Unroller.vectorize(&LVL);
@@ -1217,17 +1224,17 @@ struct LoopVectorize : public FunctionPass {
       InnerLoopVectorizer LB(L, SE, LI, DT, DL, TLI, VF.Width, UF);
       LB.vectorize(&LVL);
       ++LoopsVectorized;
+
+      // Report the vectorization decision.
+      F->getContext().emitOptimizationRemark(
+          DEBUG_TYPE, *F, L->getStartLoc(),
+          Twine("vectorized loop (vectorization factor: ") + Twine(VF.Width) +
+              ", unrolling interleave factor: " + Twine(UF) + ")");
     }
 
     // Mark the loop as already vectorized to avoid vectorizing again.
     Hints.setAlreadyVectorized(L);
 
-    // Report the vectorization decision.
-    F->getContext().emitOptimizationRemark(
-        DEBUG_TYPE, *F, L->getStartLoc(),
-        Twine("vectorized loop (vectorization factor: ") + Twine(VF.Width) +
-            ", unroll factor: " + Twine(UF) + ")");
-
     DEBUG(verifyFunction(*L->getHeader()->getParent()));
     return true;
   }

diff --git a/test/Transforms/LoopUnroll/loop-remarks.ll b/test/Transforms/LoopUnroll/loop-remarks.ll
new file mode 100644 (file)
index 0000000..ff3ac17
--- /dev/null
+++ b/test/Transforms/LoopUnroll/loop-remarks.ll
@@ -0,0 +1,25 @@
+; RUN: opt < %s -S -loop-unroll -pass-remarks=loop-unroll -unroll-count=16 2>&1 | FileCheck -check-prefix=COMPLETE-UNROLL %s
+; RUN: opt < %s -S -loop-unroll -pass-remarks=loop-unroll -unroll-count=4 2>&1 | FileCheck -check-prefix=PARTIAL-UNROLL %s
+
+; COMPLETE-UNROLL: remark: {{.*}}: completely unrolled loop with 16 iterations
+; PARTIAL-UNROLL: remark: {{.*}}: unrolled loop by a factor of 4
+
+define i32 @sum() {
+entry:
+  br label %for.body
+
+for.body:                                         ; preds = %for.body, %entry
+  %s.06 = phi i32 [ 0, %entry ], [ %add1, %for.body ]
+  %i.05 = phi i32 [ 0, %entry ], [ %inc, %for.body ]
+  %add = add nsw i32 %i.05, 4
+  %call = tail call i32 @baz(i32 %add) #2
+  %add1 = add nsw i32 %call, %s.06
+  %inc = add nsw i32 %i.05, 1
+  %exitcond = icmp eq i32 %inc, 16
+  br i1 %exitcond, label %for.end, label %for.body
+
+for.end:                                          ; preds = %for.body
+  ret i32 %add1
+}
+
+declare i32 @baz(i32)

diff --git a/test/Transforms/LoopVectorize/X86/vectorization-remarks.ll b/test/Transforms/LoopVectorize/X86/vectorization-remarks.ll
new file mode 100644 (file)
index 0000000..685d034
--- /dev/null
+++ b/test/Transforms/LoopVectorize/X86/vectorization-remarks.ll
@@ -0,0 +1,67 @@
+; RUN: opt < %s -loop-vectorize -mtriple=x86_64-unknown-linux -S -pass-remarks='loop-vectorize' 2>&1 | FileCheck -check-prefix=VECTORIZED %s
+; RUN: opt < %s -loop-vectorize -force-vector-width=1 -force-vector-unroll=4 -mtriple=x86_64-unknown-linux -S -pass-remarks='.*vectorize.*' 2>&1 | FileCheck -check-prefix=UNROLLED %s
+
+; VECTORIZED: remark: {{.*}}.c:17:8: vectorized loop (vectorization factor: 4, unrolling interleave factor: 1)
+; UNROLLED: remark: {{.*}}.c:17:8: unrolled with interleaving factor 4 (vectorization not beneficial)
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+
+define i32 @foo(i32 %n) #0 {
+entry:
+  %diff = alloca i32, align 4
+  %cb = alloca [16 x i8], align 16
+  %cc = alloca [16 x i8], align 16
+  store i32 0, i32* %diff, align 4, !dbg !10, !tbaa !11
+  br label %for.body, !dbg !15
+
+for.body:                                         ; preds = %for.body, %entry
+  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
+  %add8 = phi i32 [ 0, %entry ], [ %add, %for.body ], !dbg !19
+  %arrayidx = getelementptr inbounds [16 x i8]* %cb, i64 0, i64 %indvars.iv, !dbg !19
+  %0 = load i8* %arrayidx, align 1, !dbg !19, !tbaa !21
+  %conv = sext i8 %0 to i32, !dbg !19
+  %arrayidx2 = getelementptr inbounds [16 x i8]* %cc, i64 0, i64 %indvars.iv, !dbg !19
+  %1 = load i8* %arrayidx2, align 1, !dbg !19, !tbaa !21
+  %conv3 = sext i8 %1 to i32, !dbg !19
+  %sub = sub i32 %conv, %conv3, !dbg !19
+  %add = add nsw i32 %sub, %add8, !dbg !19
+  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1, !dbg !15
+  %exitcond = icmp eq i64 %indvars.iv.next, 16, !dbg !15
+  br i1 %exitcond, label %for.end, label %for.body, !dbg !15
+
+for.end:                                          ; preds = %for.body
+  store i32 %add, i32* %diff, align 4, !dbg !19, !tbaa !11
+  call void @ibar(i32* %diff) #2, !dbg !22
+  ret i32 0, !dbg !23
+}
+
+declare void @ibar(i32*) #1
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!7, !8}
+!llvm.ident = !{!9}
+
+!0 = metadata !{i32 786449, metadata !1, i32 12, metadata !"clang version 3.5.0 ", i1 true, metadata !"", i32 0, metadata !2, metadata !2, metadata !3, metadata !2, metadata !2, metadata !"", i32 2} ; [ DW_TAG_compile_unit ] [./vectorization-remarks.c] [DW_LANG_C99]
+!1 = metadata !{metadata !"vectorization-remarks.c", metadata !"."}
+!2 = metadata !{}
+!3 = metadata !{metadata !4}
+!4 = metadata !{i32 786478, metadata !1, metadata !5, metadata !"foo", metadata !"foo", metadata !"", i32 5, metadata !6, i1 false, i1 true, i32 0, i32 0, null, i32 256, i1 true, i32 (i32)* @foo, null, null, metadata !2, i32 6} ; [ DW_TAG_subprogram ] [line 5] [def] [scope 6] [foo]
+!5 = metadata !{i32 786473, metadata !1}          ; [ DW_TAG_file_type ] [./vectorization-remarks.c]
+!6 = metadata !{i32 786453, i32 0, null, metadata !"", i32 0, i64 0, i64 0, i64 0, i32 0, null, metadata !2, i32 0, null, null, null} ; [ DW_TAG_subroutine_type ] [line 0, size 0, align 0, offset 0] [from ]
+!7 = metadata !{i32 2, metadata !"Dwarf Version", i32 4}
+!8 = metadata !{i32 1, metadata !"Debug Info Version", i32 1}
+!9 = metadata !{metadata !"clang version 3.5.0 "}
+!10 = metadata !{i32 8, i32 3, metadata !4, null} ; [ DW_TAG_imported_declaration ]
+!11 = metadata !{metadata !12, metadata !12, i64 0}
+!12 = metadata !{metadata !"int", metadata !13, i64 0}
+!13 = metadata !{metadata !"omnipotent char", metadata !14, i64 0}
+!14 = metadata !{metadata !"Simple C/C++ TBAA"}
+!15 = metadata !{i32 17, i32 8, metadata !16, null}
+!16 = metadata !{i32 786443, metadata !1, metadata !17, i32 17, i32 8, i32 2, i32 3} ; [ DW_TAG_lexical_block ] [./vectorization-remarks.c]
+!17 = metadata !{i32 786443, metadata !1, metadata !18, i32 17, i32 8, i32 1, i32 2} ; [ DW_TAG_lexical_block ] [./vectorization-remarks.c]
+!18 = metadata !{i32 786443, metadata !1, metadata !4, i32 17, i32 3, i32 0, i32 0} ; [ DW_TAG_lexical_block ] [./vectorization-remarks.c]
+!19 = metadata !{i32 18, i32 5, metadata !20, null}
+!20 = metadata !{i32 786443, metadata !1, metadata !18, i32 17, i32 27, i32 0, i32 1} ; [ DW_TAG_lexical_block ] [./vectorization-remarks.c]
+!21 = metadata !{metadata !13, metadata !13, i64 0}
+!22 = metadata !{i32 20, i32 3, metadata !4, null}
+!23 = metadata !{i32 21, i32 3, metadata !4, null}