Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Reapply "[libc++] Explicitly convert to masks in SIMD code (#107983)" #121352

Conversation

vitalybuka
Copy link
Collaborator

@vitalybuka vitalybuka commented Dec 30, 2024

Created using spr 1.3.4
@vitalybuka vitalybuka requested a review from a team as a code owner December 30, 2024 18:49
@llvmbot llvmbot added the libc++ libc++ C++ Standard Library. Not GNU libstdc++. Not libc++abi. label Dec 30, 2024
@llvmbot
Copy link
Member

llvmbot commented Dec 30, 2024

@llvm/pr-subscribers-libcxx

Author: Vitaly Buka (vitalybuka)

Changes

This reverts commit 0ea40bf.


Full diff: https://github.com/llvm/llvm-project/pull/121352.diff

2 Files Affected:

  • (modified) libcxx/include/__algorithm/mismatch.h (+4-4)
  • (modified) libcxx/include/__algorithm/simd_utils.h (+50-27)
diff --git a/libcxx/include/__algorithm/mismatch.h b/libcxx/include/__algorithm/mismatch.h
index a6836792c05816..f5855379f68784 100644
--- a/libcxx/include/__algorithm/mismatch.h
+++ b/libcxx/include/__algorithm/mismatch.h
@@ -78,7 +78,7 @@ __mismatch_vectorized(_Iter __first1, _Iter __last1, _Iter __first2) {
       }
 
       for (size_t __i = 0; __i != __unroll_count; ++__i) {
-        if (auto __cmp_res = __lhs[__i] == __rhs[__i]; !std::__all_of(__cmp_res)) {
+        if (auto __cmp_res = std::__as_mask(__lhs[__i] == __rhs[__i]); !std::__all_of(__cmp_res)) {
           auto __offset = __i * __vec_size + std::__find_first_not_set(__cmp_res);
           return {__first1 + __offset, __first2 + __offset};
         }
@@ -90,7 +90,7 @@ __mismatch_vectorized(_Iter __first1, _Iter __last1, _Iter __first2) {
 
     // check the remaining 0-3 vectors
     while (static_cast<size_t>(__last1 - __first1) >= __vec_size) {
-      if (auto __cmp_res = std::__load_vector<__vec>(__first1) == std::__load_vector<__vec>(__first2);
+      if (auto __cmp_res = std::__as_mask(std::__load_vector<__vec>(__first1) == std::__load_vector<__vec>(__first2));
           !std::__all_of(__cmp_res)) {
         auto __offset = std::__find_first_not_set(__cmp_res);
         return {__first1 + __offset, __first2 + __offset};
@@ -107,8 +107,8 @@ __mismatch_vectorized(_Iter __first1, _Iter __last1, _Iter __first2) {
     if (static_cast<size_t>(__first1 - __orig_first1) >= __vec_size) {
       __first1 = __last1 - __vec_size;
       __first2 = __last2 - __vec_size;
-      auto __offset =
-          std::__find_first_not_set(std::__load_vector<__vec>(__first1) == std::__load_vector<__vec>(__first2));
+      auto __offset = std::__find_first_not_set(
+          std::__as_mask(std::__load_vector<__vec>(__first1) == std::__load_vector<__vec>(__first2)));
       return {__first1 + __offset, __first2 + __offset};
     } // else loop over the elements individually
   }
diff --git a/libcxx/include/__algorithm/simd_utils.h b/libcxx/include/__algorithm/simd_utils.h
index 4e3e4f2b9404e3..3ca79247bbd03c 100644
--- a/libcxx/include/__algorithm/simd_utils.h
+++ b/libcxx/include/__algorithm/simd_utils.h
@@ -116,42 +116,65 @@ template <class _VecT, class _Iter>
   }(make_index_sequence<__simd_vector_size_v<_VecT>>{});
 }
 
-template <class _Tp, size_t _Np>
-[[__nodiscard__]] _LIBCPP_HIDE_FROM_ABI bool __all_of(__simd_vector<_Tp, _Np> __vec) noexcept {
-  return __builtin_reduce_and(__builtin_convertvector(__vec, __simd_vector<bool, _Np>));
+template <size_t _Np>
+[[__nodiscard__]] _LIBCPP_HIDE_FROM_ABI bool __all_of(__simd_vector<bool, _Np> __vec) noexcept {
+  return __builtin_reduce_and(__vec);
 }
 
 template <class _Tp, size_t _Np>
-[[__nodiscard__]] _LIBCPP_HIDE_FROM_ABI size_t __find_first_set(__simd_vector<_Tp, _Np> __vec) noexcept {
-  using __mask_vec = __simd_vector<bool, _Np>;
+[[__nodiscard__]] _LIBCPP_HIDE_FROM_ABI auto __as_mask(__simd_vector<_Tp, _Np> __vec) noexcept {
+  static_assert(!is_same<_Tp, bool>::value, "vector type should not be a bool!");
+  return __builtin_convertvector(__vec, __simd_vector<bool, _Np>);
+}
 
-  // This has MSan disabled du to https://github.com/llvm/llvm-project/issues/85876
-  auto __impl = [&]<class _MaskT>(_MaskT) _LIBCPP_NO_SANITIZE("memory") noexcept {
-#  if defined(_LIBCPP_BIG_ENDIAN)
-    return std::min<size_t>(
-        _Np, std::__countl_zero(__builtin_bit_cast(_MaskT, __builtin_convertvector(__vec, __mask_vec))));
-#  else
-    return std::min<size_t>(
-        _Np, std::__countr_zero(__builtin_bit_cast(_MaskT, __builtin_convertvector(__vec, __mask_vec))));
-#  endif
-  };
-
-  if constexpr (sizeof(__mask_vec) == sizeof(uint8_t)) {
-    return __impl(uint8_t{});
-  } else if constexpr (sizeof(__mask_vec) == sizeof(uint16_t)) {
-    return __impl(uint16_t{});
-  } else if constexpr (sizeof(__mask_vec) == sizeof(uint32_t)) {
-    return __impl(uint32_t{});
-  } else if constexpr (sizeof(__mask_vec) == sizeof(uint64_t)) {
-    return __impl(uint64_t{});
+// This uses __builtin_convertvector around the __builtin_shufflevector to work around #107981.
+template <size_t _Np>
+[[__nodiscard__]] _LIBCPP_HIDE_FROM_ABI __simd_vector<bool, 8>
+__extend_vector(__simd_vector<bool, _Np> __vec) noexcept {
+  using _VecT = __simd_vector<bool, _Np>;
+  if constexpr (_Np == 4) {
+    return __builtin_convertvector(
+        __builtin_shufflevector(__vec, _VecT{}, 0, 1, 2, 3, 4, 5, 6, 7), __simd_vector<bool, 8>);
+  } else if constexpr (_Np == 2) {
+    return std::__extend_vector(
+        __builtin_convertvector(__builtin_shufflevector(__vec, _VecT{}, 0, 1, 2, 3), __simd_vector<bool, 4>));
+  } else if constexpr (_Np == 1) {
+    return std::__extend_vector(
+        __builtin_convertvector(__builtin_shufflevector(__vec, _VecT{}, 0, 1), __simd_vector<bool, 2>));
   } else {
-    static_assert(sizeof(__mask_vec) == 0, "unexpected required size for mask integer type");
+    static_assert(sizeof(_VecT) == 0, "Unexpected vector size");
+  }
+}
+
+template <size_t _Np>
+[[__nodiscard__]] _LIBCPP_HIDE_FROM_ABI auto __to_int_mask(__simd_vector<bool, _Np> __vec) {
+  if constexpr (_Np < 8) {
+    return std::__bit_cast<uint8_t>(std::__extend_vector(__vec));
+  } else if constexpr (_Np == 8) {
+    return std::__bit_cast<uint8_t>(__vec);
+  } else if constexpr (_Np == 16) {
+    return std::__bit_cast<uint16_t>(__vec);
+  } else if constexpr (_Np == 32) {
+    return std::__bit_cast<uint32_t>(__vec);
+  } else if constexpr (_Np == 64) {
+    return std::__bit_cast<uint64_t>(__vec);
+  } else {
+    static_assert(sizeof(__simd_vector<bool, _Np>) == 0, "Unexpected vector size");
     return 0;
   }
 }
 
-template <class _Tp, size_t _Np>
-[[__nodiscard__]] _LIBCPP_HIDE_FROM_ABI size_t __find_first_not_set(__simd_vector<_Tp, _Np> __vec) noexcept {
+template <size_t _Np>
+[[__nodiscard__]] _LIBCPP_HIDE_FROM_ABI size_t __find_first_set(__simd_vector<bool, _Np> __vec) noexcept {
+#  if defined(_LIBCPP_BIG_ENDIAN)
+  return std::min<size_t>(_Np, std::__countl_zero(std::__to_int_mask(__vec)));
+#  else
+  return std::min<size_t>(_Np, std::__countr_zero(std::__to_int_mask(__vec)));
+#  endif
+}
+
+template <size_t _Np>
+[[__nodiscard__]] _LIBCPP_HIDE_FROM_ABI size_t __find_first_not_set(__simd_vector<bool, _Np> __vec) noexcept {
   return std::__find_first_set(~__vec);
 }
 

Copy link
Contributor

@philnik777 philnik777 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Well, that was an easy fix for me :D

@philnik777 philnik777 merged commit ed572f2 into main Jan 1, 2025
62 checks passed
@philnik777 philnik777 deleted the users/vitalybuka/spr/reapply-libc-explicitly-convert-to-masks-in-simd-code-107983 branch January 1, 2025 10:30
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
libc++ libc++ C++ Standard Library. Not GNU libstdc++. Not libc++abi.
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants