Computer Startup and System Organization

Computer Startup

• Bootstrap program is loaded during power-up or reboot.
• It is typically stored in ROM or EPROM, also known as firmware.
• The bootstrap program initializes all system aspects, loads the operating system kernel, and starts its execution.
• The bootstrap program is loaded from the disk by the system's ROM during the first stage of the loading process.
• The bootstrap program initializes the CPU, installs banks, and launches the debug agent.
• It discovers available RAM and builds the kernel's virtual address space and device tree.
• The kernel is loaded by the bootstrap program and performs memory tests to determine the available RAM.
• The kernel identifies and configures devices, initializes the system, and starts system processes.

Computer System Organization

• One or more CPUs and device controllers connect through a common bus, providing access to shared memory.
• CPUs and devices run concurrently, competing for memory cycles.

Kernel and System Programs

• Once loaded and running, the kernel provides services to the system and its users.
• System programs are loaded into memory at boot time, becoming system processes or system daemons that run continuously while the kernel is active.
• On Unix-like systems, the first system process is "init", which starts many other daemons.
• The system is fully booted after this phase and waits for events.

I/O Devices and CPU Interaction

• I/O devices and the CPU can execute concurrently.
• Device controllers manage specific device types.
• Each device controller has a local buffer.
• The CPU moves data between main memory and local buffers.
• I/O occurs from the device to the controller's local buffer.
• The device controller informs the CPU about the completion of an operation using an interrupt.
• Interrupts are triggered by events from hardware or software.
• Hardware interrupts occur by sending signals to the CPU, usually through the system bus.
• Software interrupts occur by executing a system call, also known as a monitor call.

I/O Structure

• To start an I/O operation, the device driver loads registers in the device controller.
• The device controller analyzes these registers to determine actions, like reading from the keyboard.
• The controller transfers data between the device and its local buffer.
• The device controller notifies the driver via an interrupt when the data transfer is complete.
• The device driver returns control to the operating system, potentially returning the data or a pointer to it if the operation was a read.
• For other operations, the device driver returns status information.
• Interrupt-driven I/O is suitable for small data transfers but can be inefficient for large amounts of data, like disk I/O.

Direct Memory Access (DMA)

• DMA is a technique used to reduce overhead for bulk data transfers.
• The device controller, after setting up pointers and counters, transfers data directly between its buffer and memory, without CPU intervention.
• This method reduces the number of interrupts to one per block, instead of one per byte for low-speed devices.
• While the device controller performs DMA, the CPU handles other tasks.

Multiprocessing

• Two types of multiprocessing:
  • Asymmetric Multiprocessing (AMP): Each processor is assigned a specific task. A central processor manages the system, while other processors either receive instructions from it or have predefined tasks. This creates a boss-worker relationship, and the boss processor schedules and allocates work.
  • Symmetric Multiprocessing (SMP): All processors perform all tasks equally, forming a peer-to-peer relationship. Each processor has its own registers and a private cache, but they share physical memory.

Multiprocessor Architecture

• Multiprocessing aims to boost computing power by adding CPUs.
• Adding CPUs can also increase the addressable memory if the CPU has an integrated memory controller.
• Multiprocessing can alter the memory access model from UMA (uniform memory access) to NUMA (non-uniform memory access).

Multicore Systems

• Multicore systems are a recent trend where multiple computing cores reside on a single chip.
• They are more efficient than multi-chip designs because on-chip communication is faster than between-chip communication.
• Multicore systems consume less power compared to multiple single-core chips.
• Each core has its own register set and local cache; some designs use shared caches or a combination of local and shared caches.

Processes

• A program is a passive entity, while a process is an active entity.
• A process requires resources, such as CPU, memory, I/O, files, and initialization data, to accomplish tasks.
• Terminating a process requires reclaiming reusable resources.
• A single-threaded process has a single program counter pointing to the next instruction to execute.
• It executes instructions sequentially, one after another, until completion.
• A multi-threaded process has one program counter per thread.
• Typically, a system has many processes (user and operating-system) running concurrently on one or more CPUs.
• Concurrency is achieved by multiplexing the CPUs among the processes/threads.

Process management activities

• The operating system manages processes by:
  • Creating and deleting user and system processes.
  • Suspending and resuming processes.
  • Providing synchronization mechanisms for processes.
  • Providing communication mechanisms for processes.
  • Providing deadlock handling mechanisms.

Memory Management

• To execute a program, all or parts of its instructions and data must be in memory.