[llvm] 0360d03 - [SCEV] Add coverage for howFarToZero with stride=VF*vscale

[Philip Reames via llvm-commits]

Author: Philip Reames
Date: 2024-06-04T15:38:50-07:00
New Revision: 0360d03240f59e6b26a16769943f905c9520fe04

URL: https://github.com/llvm/llvm-project/commit/0360d03240f59e6b26a16769943f905c9520fe04
DIFF: https://github.com/llvm/llvm-project/commit/0360d03240f59e6b26a16769943f905c9520fe04.diff

LOG: [SCEV] Add coverage for howFarToZero with stride=VF*vscale

Added: 
    

Modified: 
    llvm/test/Analysis/ScalarEvolution/scalable-vector.ll

Removed: 
    


################################################################################
diff  --git a/llvm/test/Analysis/ScalarEvolution/scalable-vector.ll b/llvm/test/Analysis/ScalarEvolution/scalable-vector.ll
index 13648ebdaf868..81434d1bf064c 100644
--- a/llvm/test/Analysis/ScalarEvolution/scalable-vector.ll
+++ b/llvm/test/Analysis/ScalarEvolution/scalable-vector.ll
@@ -73,4 +73,54 @@ define i64 @vscale_exact_range() vscale_range(2) {
   ret i64 %vscale
 }
 
+define void @vscale_step_ne_tripcount(i64 %N) vscale_range(2, 1024) {
+; CHECK-LABEL: 'vscale_step_ne_tripcount'
+; CHECK-NEXT:  Classifying expressions for: @vscale_step_ne_tripcount
+; CHECK-NEXT:    %0 = sub i64 -1, %N
+; CHECK-NEXT:    --> (-1 + (-1 * %N)) U: full-set S: full-set
+; CHECK-NEXT:    %1 = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    --> vscale U: [2,1025) S: [2,1025)
+; CHECK-NEXT:    %2 = mul i64 %1, 4
+; CHECK-NEXT:    --> (4 * vscale)<nuw><nsw> U: [8,4097) S: [8,4097)
+; CHECK-NEXT:    %4 = sub i64 %2, 1
+; CHECK-NEXT:    --> (-1 + (4 * vscale)<nuw><nsw>)<nsw> U: [7,4096) S: [7,4096)
+; CHECK-NEXT:    %n.rnd.up = add i64 %N, %4
+; CHECK-NEXT:    --> (-1 + (4 * vscale)<nuw><nsw> + %N) U: full-set S: full-set
+; CHECK-NEXT:    %n.mod.vf = urem i64 %n.rnd.up, %2
+; CHECK-NEXT:    --> (-1 + (vscale * (4 + (-4 * ((-1 + (4 * vscale)<nuw><nsw> + %N) /u (4 * vscale)<nuw><nsw>))<nsw>)<nsw>) + %N) U: full-set S: full-set
+; CHECK-NEXT:    %n.vec = sub i64 %n.rnd.up, %n.mod.vf
+; CHECK-NEXT:    --> (4 * vscale * ((-1 + (4 * vscale)<nuw><nsw> + %N) /u (4 * vscale)<nuw><nsw>)) U: [0,-3) S: [-9223372036854775808,9223372036854775805)
+; CHECK-NEXT:    %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
+; CHECK-NEXT:    --> {0,+,(4 * vscale)<nuw><nsw>}<nuw><%vector.body> U: [0,-3) S: [-9223372036854775808,9223372036854775805) Exits: <<Unknown>> LoopDispositions: { %vector.body: Computable }
+; CHECK-NEXT:    %index.next = add nuw i64 %index, %2
+; CHECK-NEXT:    --> {(4 * vscale)<nuw><nsw>,+,(4 * vscale)<nuw><nsw>}<nuw><%vector.body> U: [8,-3) S: [-9223372036854775808,9223372036854775805) Exits: <<Unknown>> LoopDispositions: { %vector.body: Computable }
+; CHECK-NEXT:  Determining loop execution counts for: @vscale_step_ne_tripcount
+; CHECK-NEXT:  Loop %vector.body: Unpredictable backedge-taken count.
+; CHECK-NEXT:  Loop %vector.body: Unpredictable constant max backedge-taken count.
+; CHECK-NEXT:  Loop %vector.body: Unpredictable symbolic max backedge-taken count.
+;
+entry:
+  %0 = sub i64 -1, %N
+  %1 = call i64 @llvm.vscale.i64()
+  %2 = mul i64 %1, 4
+  %3 = icmp ult i64 %0, %2
+  br i1 %3, label %loop.exit, label %vector.ph
+
+vector.ph:                                        ; preds = %entry
+  %8 = sub i64 %2, 1
+  %n.rnd.up = add i64 %N, %8
+  %n.mod.vf = urem i64 %n.rnd.up, %2
+  %n.vec = sub i64 %n.rnd.up, %n.mod.vf
+  br label %vector.body
+
+vector.body:                                      ; preds = %vector.body, %vector.ph
+  %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
+  %index.next = add nuw i64 %index, %2
+  %22 = icmp eq i64 %index.next, %n.vec
+  br i1 %22, label %loop.exit, label %vector.body
+
+loop.exit:
+  ret void
+}
+
 declare i64 @llvm.vscale.i64()