• atzanteol@sh.itjust.works
    link
    fedilink
    arrow-up
    15
    ·
    10 months ago

    It’s not a “language” issue it’s a “computer” issue. This math is being done on the CPU.

    IEEE 754

    Some languages do provide for “arbitrary precision math” (Java’s BigDecimal for example) but it’s slower to do that. Not what you want if you’re multiplying a 4k matrix every millisecond.