Merge pull request #2158 from sycloid/revert-gh-2122-fix-for-gh-2121

Revert gh 2122 fix for gh 2121

Author: ndgrigorian

dpctl/tensor/_copy_utils.py

@@ -742,9 +742,6 @@ def astype(
             order=copy_order,
             buffer_ctor_kwargs={"queue": usm_ary.sycl_queue},
         )
-    # see #2121
-    if ary_dtype == dpt.bool:
-        usm_ary = dpt.not_equal(usm_ary, 0, order=copy_order)
     _copy_from_usm_ndarray_to_usm_ndarray(R, usm_ary)
     return R
 

dpctl/tensor/libtensor/include/kernels/copy_and_cast.hpp

@@ -101,7 +101,7 @@ class GenericCopyFunctor
         const ssize_t &src_offset = offsets.get_first_offset();
         const ssize_t &dst_offset = offsets.get_second_offset();
 
-        CastFnT fn{};
+        static constexpr CastFnT fn{};
         dst_[dst_offset] = fn(src_[src_offset]);
     }
 };
@@ -237,9 +237,9 @@ class ContigCopyFunctor
 
         static constexpr std::uint8_t elems_per_wi = n_vecs * vec_sz;
 
-        using dpctl::tensor::type_utils::is_complex;
-        if constexpr (!enable_sg_loadstore || is_complex<srcT>::value ||
-                      is_complex<dstT>::value)
+        using dpctl::tensor::type_utils::is_complex_v;
+        if constexpr (!enable_sg_loadstore || is_complex_v<srcT> ||
+                      is_complex_v<dstT>)
         {
             std::uint16_t sgSize = ndit.get_sub_group().get_local_range()[0];
             const std::size_t gid = ndit.get_global_linear_id();

dpctl/tensor/libtensor/include/utils/type_utils.hpp

@@ -25,6 +25,7 @@
 #pragma once
 #include <complex>
 #include <cstddef>
+#include <cstdint>
 #include <stdexcept>
 #include <sycl/sycl.hpp>
 #include <type_traits>
@@ -55,26 +56,45 @@ template <typename T> inline constexpr bool is_complex_v = is_complex<T>::value;
 
 template <typename dstTy, typename srcTy> dstTy convert_impl(const srcTy &v)
 {
-    if constexpr (std::is_same<dstTy, srcTy>::value) {
+    if constexpr (std::is_same_v<dstTy, srcTy>) {
         return v;
     }
-    else if constexpr (std::is_same_v<dstTy, bool> && is_complex<srcTy>::value)
-    {
-        // bool(complex_v) == (complex_v.real() != 0) && (complex_v.imag() !=0)
-        return (convert_impl<bool, typename srcTy::value_type>(v.real()) ||
-                convert_impl<bool, typename srcTy::value_type>(v.imag()));
+    else if constexpr (std::is_same_v<dstTy, bool>) {
+        if constexpr (is_complex_v<srcTy>) {
+            // bool(complex_v) ==
+            //     (complex_v.real() != 0) && (complex_v.imag() !=0)
+            return (convert_impl<bool, typename srcTy::value_type>(v.real()) ||
+                    convert_impl<bool, typename srcTy::value_type>(v.imag()));
+        }
+        else {
+            return static_cast<dstTy>(v != srcTy{0});
+        }
+    }
+    else if constexpr (std::is_same_v<srcTy, bool>) {
+        // C++ interprets a byte of storage behind bool by only
+        // testing is least significant bit, leading to both
+        // 0x00 and 0x02 interpreted as False, while 0x01 and 0xFF
+        // interpreted as True. NumPy's interpretation of underlying
+        // storage is different: any bit set is interpreted as True,
+        // no bits set as False, see gh-2121
+        const std::uint8_t &u = sycl::bit_cast<std::uint8_t>(v);
+        if constexpr (is_complex_v<dstTy>) {
+            return (u == 0) ? dstTy{} : dstTy{1, 0};
+        }
+        else {
+            return (u == 0) ? dstTy{} : dstTy{1};
+        }
     }
-    else if constexpr (is_complex<srcTy>::value && !is_complex<dstTy>::value) {
+    else if constexpr (is_complex_v<srcTy> && !is_complex_v<dstTy>) {
         // real_t(complex_v) == real_t(complex_v.real())
         return convert_impl<dstTy, typename srcTy::value_type>(v.real());
     }
-    else if constexpr (!std::is_integral<srcTy>::value &&
-                       !std::is_same<dstTy, bool>::value &&
-                       std::is_integral<dstTy>::value &&
-                       std::is_unsigned<dstTy>::value)
+    else if constexpr (!std::is_integral_v<srcTy> &&
+                       !std::is_same_v<dstTy, bool> &&
+                       std::is_integral_v<dstTy> && std::is_unsigned_v<dstTy>)
     {
         // first cast to signed variant, the cast to unsigned one
-        using signedT = typename std::make_signed<dstTy>::type;
+        using signedT = typename std::make_signed_t<dstTy>;
         return static_cast<dstTy>(convert_impl<signedT, srcTy>(v));
     }
     else {