3d_gpu/arch_comparison.mdwn

   1 # Architectural Comparisons
   2
   3 This table compares important factors for a decision in which ISA to use
   4 and support
   5
   6 [[!table  data="""
   7 Feature/Need        | RISC-V | MIPS    | OpenPOWER  | OpenRISC |
   8 Open                | NO (1) | NO (2)  | WIP (3)    | Yes      |
   9 Libre-friendly      | NO (1) | unknown | Yes        | Yes      |
  10 Trademarked         | Yes    | Yes     | Yes        | No       |
  11 Extensible          | NO (4) | unknown | Potentially (5)        | Yes      |
  12 gcc                 | Yes    | Yes     | Yes        | Yes      |
  13 llvm                | Yes    | Yes     | Yes        | Yes      |
  14 llvm upstream       | Yes    | Yes     | Yes        | No       |
  15 distros (debian)    | Yes    | Yes     | Yes        | No       |
  16 distros (ubuntu)    | No     | No      | Yes        | No       |
  17 distros (fedora)    | Yes    | Yes     | Yes        | No       |
  18 buildroot           | Yes    | Yes     | Yes        | Yes      |
  19 linux kernel        | Yes    | Yes     | Yes        | Yes      |
  20 64-bit linux        | Yes    | Yes     | Yes        | No       |
  21 virtualizable       | Yes    | unknown | Yes        | No (6)   |
  22 64-bit asics        | many   | many    | many       | none     |
  23 sw community size   | large  | medium  | very large | tiny (7) |
  24 little-endian asics | Yes    | Yes     | Yes        | No       |
  25 Java (OpenJDK)      | no JIT | unknown | Yes        | No       |
  26 Java (Android)      | No?    | Yes     | No         | No       |
  27 V8 node.js/chromium | no JIT | Yes     | Yes        | No       |
  28 PyPy                | No     | No      | Yes        | No       |
  29 Mono (.net)         | No?    | 32-bit  | Yes        | No       |
  30 Spidermonkey (Firefox JS) | no JIT | Yes (JIT Unsupported) | no JIT | No |
  31 """]]
  32
  33 * (1) too numerous reasons to count, for well over 4 years.
  34 * (2) efforts to be "open" have stalled, website non-functional
  35 * (3) OpenPower Foundation making a huge effort, really appreciated
  36 * (4) repeated in-good-faith requests for participation and inclusion in
  37   extending RISC-V have met with stone-cold silence for over 18 months,
  38   in direct violation of Trademark Law's clear-cut responsibilities.
  39 * (5) provisionally and in principle: yes.  however as of 2019jan21 this is
  40   yet to be tested and an agreement reached.
  41 * (6) no mentions of virtualization found in spec
  42 * (7) dozens to hundreds of people, several companies