Low Level Programming

Introduction

  • Model of computation is a set of basic operations and their respective costs.
  • Most models of computations are also abstract machines. i.e church Lambda calculus, turing machines, von neumann
  • Computer Architecture describes the functionality, organization and implementation of computer systems.
  • Von Neumann was robust and easy to program hence its popularity.
  • Desc:
    • This consists of one processor, one memory bank connected to a common bus
    • CPU execute instructions fetched from memory by a control unit.
    • ALU perfroms the needed calcs.
    • Memory stores only bits and stores both encoded instructions and data to operate on.
    • One byte = 8 bits
  • Assembly Language for a chosen processor is a pl consisting of mnemonics for each possible binary encoded instruction(machine code).
  • An architecture does not always define a precise instruction set unlike a model of computation.

Introduction 2(Electronics)

  • Electric charge - causes matter to experience a force - Coulombs

  • Electric current - Flow of charge - Amps

  • Voltage - electric potential difference between two points, voltage increases current, current pressure - Volts

  • Resistance - Difficulty of current flow through a material - Ohms

  • One Amp corresponds to one coulomb per second.

  • Ohms Law - Current flowing from one point to another is equal to voltage across the points divided by the resistance between them. I =V/R.

  • AC vs DC

  • Kirchhoff's Voltage Law - Sum of volatages around a circuit is zero.

  • Resistance serially connected , total resistance is just the sum.

  • Light Emitting Diodes (LED) - lights up when current flows via it.

  • Diode - electronic component that allows current to flow through it in one direction, low resistance in one direction(allow) and high in the other(impede)

  • Forward voltage - voltage amount dropped across LED when current flows through it.

  • Transistor - device used to switch or amplify current.

  • Types: BJTs - Bipolar Junction Transistor, FETs - Field Effective Transistors.

  • BJTs have three terminals: base, collector and emitter.

  • Two types of BJTs: NPN and PNP. Differ in how they respond when current is applied at base terminal.

  • NPN - applying a small current at the base allows a larger current to flow from the collector to the emitter.

  • Logic gates - circuit elements that implement logical functions where inputs and outputs are high and low voltages.

  • Combinational Logic circuit - combine logic gates in such a way that output is a function of the present inputs.

  • Sequential logic - output is a function of both present and past inputs.

  • Integrated circuits

  • Ics in Dual in-line package

  • Resistor-Transistor Logic circuits, Diode-Transistor Logic, Transistor-transistor Logic

  • Pinout diagram.

Von Neumann Arch Extensions

  • Registers:

    • Memory cells placed directly on the CPU chip.
    • Faster but more complicated and expensive.
    • Don't use the bus.
    • Registers are based on transistors while main memory uses condensers.
    • Could implement main memory as registers are but:
      • Registers are more expensive.
      • Instructions encode the register number as part of their code hence more registers equals bigger instruction sizes.
      • Registers add complexity to circuits to address them, more complex circuits are harder to speed up, not easy to set up a large register file to work on 5ghz
    • Locality of reference
      • Temporal locality - accesses to one address are likely to be close in time.
      • Spatial locality - accesses X, the next one wil likely be close to X
    • General purpose registers
      • They are interchangeable and can be used in many different commands.
      • Names can be r1,r2,r3 ..... r15, represent meaning they bear for some special instruction i.e rc1 - cycle, rcx - cycle counter, rax- accumulator
      • registers do not have addresses as they are implemented differently to main memory
      • Some registers have system wide importance hence only modified by the OS. i.e rip register - stores address of next instruction. rflags - current program state.
      • Registers for floating point numbers and specialised parallel instructions(SIMD). i.e multimedia xmm0 - xmm15.
      • SIMD(Single Instruction Multiple Data) allow a single instruction to be applied simultaneously to multiple data items.
    • System registers
      • designed to be used specifically by the OS.
      • They don't hold values in computations instead store information required by system-wide data structures.
      • i.e cr2,cr3 - virtual memory, cr8 - interrupts.

  • Hardware Stack:

    • Stack is a ds in general that supports pushing and popping.
    • Used in computations and store of local variables and implement function call sequence.
    • There is hardware support for such data structure.
    • Useful to implement function calls in higher-level languages, save the context of computations.

  • Interrupts:

    • Allows one to change execution order based on events external to the program itself.
    • Execution suspended, Registers saved and CPU executes corresponding handling routine.
    • i.e signal from external device, zero division, invalid instruction, privilege instruction in non-privileged environment.

  • Protection rings:

    • Mechanism designed to limit the applications capabilities for security and robustness reasons.
    • Invented for Multics OS predecessor to Unix.
    • Each ring corresponds to certain privilege level.
    • Each instruction type is linked with one or more privilege levels and is not executable on others.
    • A CPU is always in a state corresponding to one of the so-called protection rings.
    • Each defines a set of allowed instructions.

  • Virtual Memory:

    • This is an abstraction over physical memory, which helps distribute it between programs in a safe and effective way.
    • Also isolates programs from one another.

- Parallelism will be key and exploitation of locality of data.
- CPU speeds increased faster than memory speeds.
- Memory keeps getting further with the introduction of new levels of caches.
 
- CPUs with caches...as transistor density increased, cache capabilities were integrated onto CPUs.
- GPUs grew out of need to provide hardware support for displays as use of graphics use grew.
- GPU speed also increase faster tha memory speed.
- GPU also integrated onto CPUs as transistor density increased.
- NIC(Network Integration Controller) capabilities also integrated onto CPUs.

- Work/Time = Work/Instructions(Path Length) * Instruction/Cycle(IPC) * Cycle/Time(Frequency)
- Better Algorithm - Same work with fewer instructions
- Compiler can optimize for fewer instructions, choose those with better IPC.
- Cache efficient algorithms: Higher IPC.
- Vectorization: same work with fewer instructions.
- Parallelization: more instructions per cycle.

Unix

  • Unix postulates that everything is a file in the sense that looks like a stream of bytes.
  • Abstract things like:
    • data access on a hard drive/ssd.
    • data exchange between programs.
    • interaction with external devices.
  • OS purpose is to abstract and manage resources via a set of routines to handle communication with devices, files, programs.
  • A programs cannot bypass the OS and interact with resources directly.
  • Communication happens via system calls provided to user applications.
  • Unix identifies a file with its descriptor as soon as it is opened by a program...this is nothing but an integer value.
  • File is opened by invoking the open system call however 3 other files opened stdin, stdout, stderror. 0, 1, 2.
  • Write system call invoked for stdout - writes a given amount of bytes from memory starting at a given address to a file with a given descriptor