ARM has bit-banding specifically for this. I think it’s limited to M-profile CPUs (e.g. v7-M) but I’ve definitely used this before. It basically creates a 4-byte virtual address for every bit in a region. So the CPU itself can’t “address” a bit but it can access an address backed by only 1 bit of SRAM or registers (this is also useful to atomically access certain bits in registers without needing to use SW atomics).