full memory mode support

[duggavo]

it would be nice to support RandomX full mode, mainly for testing, but it can be useful for faster hashing in Node.JS when a lot of memory is available.

[l1mey112]

This is currently not planned, because I am designing the library to be the most performant given the restrictions of the average browser.

Eventually yes, this could become a possibility. If enough people get behind a full speed RandomX implementation, completely implemented using web standards only (no FFI) I will budge.

[vExcess]

It is true that browsers won't let you allocate a 2 GB buffer, however you can allocate two 1GB buffers.
Running the following program in latest Chromium raised my RAM usage from 3.3GB to 5.5GB and in latest Firefox raises RAM usage from 2.9GB to 4.9GB. Both browsers run the program without erroring. This isn't even on a high end system. This was just run on my laptop with 6 GB of usable RAM (which is not a lot these days) running Linux Mint.

const GB = 1024 * 1024 * 1024;

class Buff2GB {
    arr1;
    arr2;
    
    constructor() {
        this.arr1 = new Uint8Array(GB);
        this.arr2 = new Uint8Array(GB);
    }
    
    read(idx) {
        return idx < GB ? this.arr1[idx] : this.arr2[idx - GB];
    }

    set(idx, val) {
        if (idx < GB) {
            this.arr1[idx] = val;
        } else {
            this.arr2[idx - GB] = val;
        }
    }
}

var mem = new Buff2GB();
for (let i = 0; i < 100000000; i++) {
    const idx = (Math.random() * GB * 2) | 0;
    const val = (Math.random() * 255) | 0;
    mem.set(idx, val);
}

console.log("DONE");

Also I should note thate Node JS does allow you to allocate a 2 GB buffer.