Support for RISC-V architecture #163

Activity

1. ColinH

   self-assigned this

   on Sep 1, 2026
2. braydonf commented on Sep 1, 2026

   

   braydonf

   on Sep 1, 2026

   Contributor

   More actions

   Based on the comment in double.hpp, RISC-V (RVA23) gives the result of Div_double(89255.0, 1e22) as equal to 89255e-22. ^{[1] [2]} Thus, adding __riscv alongside __x86_64__ and __aarch64__ should be all that is needed, or perhaps defined(__riscv) && __riscv_xlen == 64 to check for 64 bits?

   • [1] https://github.com/taocpp/json/blob/main/include/tao/json/external/double.hpp#L49
   • [2] https://gist.github.com/braydonf/a969415344291e761d937c5fe48df0aa

   👍React with 👍1Reacted by Doug Hoyte
3. braydonf commented on Sep 2, 2026

   

   braydonf

   on Sep 2, 2026

   Contributor

   More actions

   With this change, all tests passed running make test from the cmake build:

   diff --git a/include/tao/json/external/double.hpp b/include/tao/json/external/double.hpp
   index 0219c27b..7cff5903 100644
   --- a/include/tao/json/external/double.hpp
   +++ b/include/tao/json/external/double.hpp
   @@ -63,7 +63,8 @@
       defined(__SH4__) || defined(__alpha__) ||                            \
       defined(_MIPS_ARCH_MIPS32R2) ||                                      \
       defined(__AARCH64EL__) || defined(__aarch64__) ||                    \
   -   defined(__EMSCRIPTEN__) || defined(__loongarch64)
   +   defined(__EMSCRIPTEN__) || defined(__loongarch64) ||                 \
   +   (defined(__riscv) && __riscv_xlen == 64)
    #define TAO_JSON_DOUBLE_CONVERSION_CORRECT_DOUBLE_OPERATIONS 1
    #elif defined(__mc68000__)
    #undef TAO_JSON_DOUBLE_CONVERSION_CORRECT_DOUBLE_OPERATIONS

   PR: #164
4. ColinH

   closed this as completedin #164on Sep 20, 2026