In file included from hb-face.cc:35:
hb-ot-cmap-table.hh: In member function 'void OT::CmapSubtableFormat4::_compiles_assertion_on_line_388() const':
hb-ot-cmap-table.hh:388: error: ISO C++ says that these are ambiguous, even though the worst conversion for the first is better than the worst conversion for the second:
hb-open-type.hh:354: note: candidate 1: const Type& OT::UnsizedArrayOf<Type>::operator[](unsigned int) const [with Type = OT::IntType<short unsigned int, 2u>]
hb-ot-cmap-table.hh:388: note: candidate 2: operator[](const T*, int) <built-in>
hb-ot-cmap-table.hh: In member function 'void OT::CmapSubtableFormat4::_instance_assertion_on_line_388() const':
hb-ot-cmap-table.hh:388: error: ISO C++ says that these are ambiguous, even though the worst conversion for the first is better than the worst conversion for the second:
hb-open-type.hh:354: note: candidate 1: const Type& OT::UnsizedArrayOf<Type>::operator[](unsigned int) const [with Type = OT::IntType<short unsigned int, 2u>]
hb-ot-cmap-table.hh:388: note: candidate 2: operator[](const T*, int) <built-in>
hb-face.cc: In function 'hb_blob_t* _hb_face_builder_data_reference_blob(hb_face_builder_data_t*)':
hb-face.cc:650: error: ISO C++ says that these are ambiguous, even though the worst conversion for the first is better than the worst conversion for the second:
hb-vector.hh:81: note: candidate 1: Type& hb_vector_t<Type, PreallocedCount>::operator[](unsigned int) [with Type = hb_face_builder_data_t::table_entry_t, unsigned int PreallocedCount = 32u]
hb-face.cc:650: note: candidate 2: operator[](T*, int) <built-in>
hb-face.cc:650: error: ISO C++ says that these are ambiguous, even though the worst conversion for the first is better than the worst conversion for the second:
hb-vector.hh:81: note: candidate 1: Type& hb_vector_t<Type, PreallocedCount>::operator[](unsigned int) [with Type = hb_face_builder_data_t::table_entry_t, unsigned int PreallocedCount = 32u]
hb-face.cc:650: note: candidate 2: operator[](const T*, int) <built-in>
hb-face.cc:651: error: ISO C++ says that these are ambiguous, even though the worst conversion for the first is better than the worst conversion for the second:
hb-vector.hh:81: note: candidate 1: Type& hb_vector_t<Type, PreallocedCount>::operator[](unsigned int) [with Type = hb_face_builder_data_t::table_entry_t, unsigned int PreallocedCount = 32u]
hb-face.cc:651: note: candidate 2: operator[](T*, int) <built-in>
hb-face.cc:651: error: ISO C++ says that these are ambiguous, even though the worst conversion for the first is better than the worst conversion for the second:
hb-vector.hh:81: note: candidate 1: Type& hb_vector_t<Type, PreallocedCount>::operator[](unsigned int) [with Type = hb_face_builder_data_t::table_entry_t, unsigned int PreallocedCount = 32u]
hb-face.cc:651: note: candidate 2: operator[](const T*, int) <built-in>
Profiling showed that type conversions were adding considerable cycles in time
spent doing text shaping.
The idea is to optimize it using native processor instructions to help Blink
layout performance.
Doing further investigation revelead that compilers may not use the
proper instruction on ARM 32bits builds (i.e. REV16).
One way to insure that the generated ASM was ideal for both gcc/clang
was using __builtin_bswap16.
Added bonus is that we no longer need to test for CPU architecture.
Behdad Esfahbod
Ebrahim Byagowi
Adenilson Cavalcanti