[llvm] [CodeGen] Use 128bits for LaneBitmask. (PR #111157)

[Jay Foad via llvm-commits]

================
@@ -29,72 +29,120 @@
 #ifndef LLVM_MC_LANEBITMASK_H
 #define LLVM_MC_LANEBITMASK_H
 
+#include "llvm/ADT/APInt.h"
+#include "llvm/ADT/SmallString.h"
 #include "llvm/Support/Compiler.h"
 #include "llvm/Support/Format.h"
 #include "llvm/Support/MathExtras.h"
 #include "llvm/Support/Printable.h"
 #include "llvm/Support/raw_ostream.h"
+#include <utility>
 
 namespace llvm {
 
-  struct LaneBitmask {
-    // When changing the underlying type, change the format string as well.
-    using Type = uint64_t;
-    enum : unsigned { BitWidth = 8*sizeof(Type) };
-    constexpr static const char *const FormatStr = "%016llX";
+struct LaneBitmask {
+  static constexpr unsigned int BitWidth = 128;
 
-    constexpr LaneBitmask() = default;
-    explicit constexpr LaneBitmask(Type V) : Mask(V) {}
-
-    constexpr bool operator== (LaneBitmask M) const { return Mask == M.Mask; }
-    constexpr bool operator!= (LaneBitmask M) const { return Mask != M.Mask; }
-    constexpr bool operator< (LaneBitmask M)  const { return Mask < M.Mask; }
-    constexpr bool none() const { return Mask == 0; }
-    constexpr bool any()  const { return Mask != 0; }
-    constexpr bool all()  const { return ~Mask == 0; }
-
-    constexpr LaneBitmask operator~() const {
-      return LaneBitmask(~Mask);
-    }
-    constexpr LaneBitmask operator|(LaneBitmask M) const {
-      return LaneBitmask(Mask | M.Mask);
-    }
-    constexpr LaneBitmask operator&(LaneBitmask M) const {
-      return LaneBitmask(Mask & M.Mask);
-    }
-    LaneBitmask &operator|=(LaneBitmask M) {
-      Mask |= M.Mask;
-      return *this;
-    }
-    LaneBitmask &operator&=(LaneBitmask M) {
-      Mask &= M.Mask;
-      return *this;
+  explicit LaneBitmask(APInt V) {
+    switch (V.getBitWidth()) {
+    case BitWidth:
+      Mask[0] = V.getRawData()[0];
+      Mask[1] = V.getRawData()[1];
+      break;
+    default:
+      llvm_unreachable("Unsupported bitwidth");
     }
+  }
+  constexpr explicit LaneBitmask(uint64_t Lo = 0, uint64_t Hi = 0) : Mask{Lo, Hi} {}
 
-    constexpr Type getAsInteger() const { return Mask; }
+  constexpr bool operator==(LaneBitmask M) const {
+    return Mask[0] == M.Mask[0] && Mask[1] == M.Mask[1];
+  }
+  constexpr bool operator!=(LaneBitmask M) const {
+    return Mask[0] != M.Mask[0] || Mask[1] != M.Mask[1];
+  }
+  constexpr bool operator<(LaneBitmask M) const {
+    return Mask[1] < M.Mask[1] || Mask[0] < M.Mask[0];
+  }
+  constexpr bool none() const { return Mask[0] == 0 && Mask[1] == 0; }
+  constexpr bool any() const { return Mask[0] != 0 || Mask[1] != 0; }
+  constexpr bool all() const { return ~Mask[0] == 0 && ~Mask[1] == 0; }
 
-    unsigned getNumLanes() const { return llvm::popcount(Mask); }
-    unsigned getHighestLane() const {
-      return Log2_64(Mask);
-    }
+  constexpr LaneBitmask operator~() const { return LaneBitmask(~Mask[0], ~Mask[1]); }
+  constexpr LaneBitmask operator|(LaneBitmask M) const {
+    return LaneBitmask(Mask[0] | M.Mask[0], Mask[1] | M.Mask[1]);
+  }
+  constexpr LaneBitmask operator&(LaneBitmask M) const {
+    return LaneBitmask(Mask[0] & M.Mask[0], Mask[1] & M.Mask[1]);
+  }
+  LaneBitmask &operator|=(LaneBitmask M) {
+    Mask[0] |= M.Mask[0];
+    Mask[1] |= M.Mask[1];
+    return *this;
+  }
+  LaneBitmask &operator&=(LaneBitmask M) {
+    Mask[0] &= M.Mask[0];
+    Mask[1] &= M.Mask[1];
+    return *this;
+  }
 
-    static constexpr LaneBitmask getNone() { return LaneBitmask(0); }
-    static constexpr LaneBitmask getAll() { return ~LaneBitmask(0); }
-    static constexpr LaneBitmask getLane(unsigned Lane) {
-      return LaneBitmask(Type(1) << Lane);
-    }
+  APInt getAsAPInt() const { return APInt(BitWidth, {Mask[0], Mask[1]}); }
+  constexpr std::pair<uint64_t, uint64_t> getAsPair() const { return {Mask[0], Mask[1]}; }
 
-  private:
-    Type Mask = 0;
-  };
+  unsigned getNumLanes() const {
+    return Mask[1] ? llvm::popcount(Mask[1]) + llvm::popcount(Mask[0])
----------------
jayfoad wrote:

I would hope you could rely on modern host machines having a fast popcount instruction so you don't need to special-case `Mask[1]` here, but I'm not sure.

https://github.com/llvm/llvm-project/pull/111157