[llvm] 87a3f1a - [Hexagon] Better detection of impossible completions to perfect shuffles

[Krzysztof Parzyszek via llvm-commits]

Author: Krzysztof Parzyszek
Date: 2022-12-05T12:40:08-08:00
New Revision: 87a3f1ab3b03923f069af3c6d3c0a92871de511e

URL: https://github.com/llvm/llvm-project/commit/87a3f1ab3b03923f069af3c6d3c0a92871de511e
DIFF: https://github.com/llvm/llvm-project/commit/87a3f1ab3b03923f069af3c6d3c0a92871de511e.diff

LOG: [Hexagon] Better detection of impossible completions to perfect shuffles

If there is an entry P that has C bits set, it could become one of C
different possibilities. If P occurs more than C times, then there are
no valid completions.

Added: 
    llvm/test/CodeGen/Hexagon/autohvx/isel-shuffle-no-perfect-completion.ll

Modified: 
    llvm/lib/Target/Hexagon/HexagonISelDAGToDAGHVX.cpp

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/Hexagon/HexagonISelDAGToDAGHVX.cpp b/llvm/lib/Target/Hexagon/HexagonISelDAGToDAGHVX.cpp
index b529fae64cc14..eb67b67063d58 100644
--- a/llvm/lib/Target/Hexagon/HexagonISelDAGToDAGHVX.cpp
+++ b/llvm/lib/Target/Hexagon/HexagonISelDAGToDAGHVX.cpp
@@ -1957,6 +1957,27 @@ SmallVector<uint32_t, 8> HvxSelector::getPerfectCompletions(ShuffleMask SM,
   // are exactly the same as in s1, which means that they are exactly the
   // same as in P. This implies that P == Q.
 
+  // There can be a situation where there are more entries with the same
+  // bits set than there are set bits (e.g. value 9 occuring more than 2
+  // times). In such cases it will be impossible to complete this to a
+  // perfect shuffle.
+  SmallVector<uint32_t, 8> Sorted(Worklist);
+  llvm::sort(Sorted.begin(), Sorted.end());
+
+  for (unsigned I = 0, E = Sorted.size(); I != E;) {
+    unsigned P = Sorted[I], Count = 1;
+    while (++I != E && P == Sorted[I])
+      ++Count;
+    if (countPopulation(P) < Count) {
+      // Reset all occurences of P, if there are more occurrences of P
+      // than there are bits in P.
+      for_each(Worklist, [P](unsigned &Q) {
+        if (Q == P)
+          Q = 0;
+      });
+    }
+  }
+
   return Worklist;
 }
 
@@ -1981,6 +2002,17 @@ HvxSelector::completeToPerfect(ArrayRef<uint32_t> Completions, unsigned Width) {
     Comps[I] = T;
   }
 
+#ifndef NDEBUG
+  // Check that we have generated a valid completion.
+  uint32_t OrAll = 0;
+  for (unsigned I = 0, E = Comps.size(); I != E; ++I) {
+    uint32_t C = Comps[I];
+    assert(isPowerOf2_32(C));
+    OrAll |= C;
+  }
+  assert(OrAll == (1u << Width) -1);
+#endif
+
   return Comps;
 }
 
@@ -2280,7 +2312,9 @@ OpRef HvxSelector::perfect(ShuffleMask SM, OpRef Va, ResultStack &Results) {
   if (llvm::any_of(Comps, [](uint32_t P) { return P == 0; }))
     return OpRef::fail();
 
+dbgs() << "Comps:["; for (unsigned I : Comps) dbgs() << ' ' << I; dbgs() << " ]\n";
   auto Pick = completeToPerfect(Comps, LogLen);
+dbgs() << "Pick:["; for (unsigned I : Pick) dbgs() << ' ' << I; dbgs() << " ]\n";
   for (unsigned I = 0; I != LogLen; ++I)
     Perm[I] = Log2_32(Pick[I]);
 

diff  --git a/llvm/test/CodeGen/Hexagon/autohvx/isel-shuffle-no-perfect-completion.ll b/llvm/test/CodeGen/Hexagon/autohvx/isel-shuffle-no-perfect-completion.ll
new file mode 100644
index 0000000000000..b606608f81f55
--- /dev/null
+++ b/llvm/test/CodeGen/Hexagon/autohvx/isel-shuffle-no-perfect-completion.ll
@@ -0,0 +1,11 @@
+; RUN: llc -march=hexagon < %s | FileCheck %s
+
+; Check that this doesn't end up being an entirely perfect shuffle.
+; CHECK: vshuff
+; CHECK-NOT: vdeal
+define <32 x i32> @f0(<32 x i32> %a0, <32 x i32> %a1) #0 {
+  %v0 = shufflevector <32 x i32> %a0, <32 x i32> %a1, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+  ret <32 x i32> %v0
+}
+
+attributes #0 = { nounwind memory(none) "target-features"="+hvxv62,+hvx-length128b" }