Instructions and Programs
CS 154: Computer Architecture Lecture #7
Winter 2020
Ziad Matni, Ph.D.
Dept. of Computer Science, UCSB
Administrative
I got nada
1/29/20 Matni, CS154, Wi20 2
Lecture Outline
Branch and Jump Addressing
Parallelism and Synchronization
Going from File to Machine Code
Relative Performance Comparisons
1/29/20 Matni, CS154, Wi20 3
Branch Addressing
I-Type of instruction (beq , bne)
Branch instructions specify:
Opcode + 2 registers + target address
Most branch targets are near the branch instruction in the text segment of memory
Either ahead or behind it
Addressing can be done relative to the value in PC Reg. (PC-Relative Addressing)
Target address = PC + offset (in words) x 4
PC is already incremented by 4 by this time
1/29/20 Matni, CS154, Wi20 4
Branching Far Away
If branch target is too far to encode with 16-bit offset, then assembler will rewrite the code
Example
beq $s0, $s1, L1
bne $s0, $s1, L2
j L1 L2:
# L1 is far away # rewritten
1/29/20
Matni, CS154, Wi20
5
Jump Addressing
J-Type of instruction (j , jal)
Jump (j and jal) targets could be anywhere in text segment Encode full address in instruction
Direct jump addressing
Target address = (address x 4 ) OR (PC[31: 28])
i.e. Take the 4 most sig. bits in PC
and concatenate the 26 bits in address field
and then concatenate another 00 (i.e x 4)
1/29/20 Matni, CS154, Wi20 6
Target Addressing Example
Assume Loop is at location 80000
1/29/20 Matni, CS154, Wi20 7
Examples:
addi $t0, $t0, 42
add $t0, $t1, $t3
lw $t0, 4($t1)
beq $t0, $t1, L1
j L1
1/29/20 Matni, CS154, Wi20 8
Addressing Mode Summary
Parallelism and Synchronization
Consider: 2 processors sharing an area of memory P1 writes, then P2 reads
There may be a data race if P1 and P2 dont synchronize Result depends of order of accesses
Hardware support required
Atomic read/write memory operation,
i.e. no other mem. access allowed between the read and write
Could be a single instruction
E.g., atomic swap of register memory
Or an atomic pair of instructions (like ll & sc)
1/29/20 Matni, CS154, Wi20 9
Synchronization in MIPS
Load link: ll rt, offset(rs)
Store conditional: sc rt, offset(rs)
Succeeds if location not changed since the ll: Returns 1 in rt Fails if location is changed: Returns 0 in rt
ll returns the current value of a memory location
A subsequent sc to the same memory location will store a new value there only if no updates have occurred to that location since the ll.
1/29/20 Matni, CS154, Wi20 10
Going From File to Machine Code
There are 4 steps in transforming a program in a file into a program running on a computer
1. Compiler
Takes a program in a HLL and translates to assembly language Some compilers have assemblers & linkers built-in
2. Assembler
Takes care of pseudoinstructions, number conversions (to hex)
Produces an object file (a combination of machine language instructions, data, and information needed to place instructions properly in memory)
This has to determine the addresses corresponding to all labels
1/29/20 Matni, CS154, Wi20 11
Producing an Object Module
Header: described contents of object module
Text segment: translated instructions
Static data segment: data allocated for the life of the program
Relocation info: for contents that depend on absolute location of loaded program
Symbol table: global definitions and external refs Debug info: for associating with source code
This may not have all the references/labels resolved yet
1/29/20 Matni, CS154, Wi20 12
Going From File to Machine Code (cont)
3. Linker
When a program comprises multiple object files, the linker combines these files into a unified executable program, resolving the symbols (references) as it goes along.
There are 3 steps for the linker:
1. Place code and data modules symbolically in memory.
2. Determine the addresses of data and instruction labels.
3. Patch both the internal and external references.
This produces one executable file with machine language instructions. 4. Loader
OS program that takes the executable code, sets up CPU memory for it, copies over the instructions to CPU memory, initializes all registers, jumps to the start-up routine (i.e. usually main:)
1/29/20 Matni, CS154, Wi20 13
4 steps in transforming a program in a file into a program running on a computer
Translation and Startup
1/29/20 Matni, CS154, Wi20 14
Dynamic Linking
Only finish linking a library procedure when it is called. Pros:
Often-used libraries need to be stored in only one location, not duplicated in every single executable file.
Saves memory and disk space
Updates/fixes to one library can be done modularly. Cuts down on
compiling time. Cons:
DLL hell: newer version of library is not backward compatible.
1/29/20 Matni, CS154, Wi20 15
Java
Java was invented to be different than C/C++
Intended to let application developers write once, run anywhere
Rather than compile to the assembly language of a target computer, Java is compiled first to the Java bytecode instruction set
These run on any Java virtual machine (JVM) regardless of the underlying computer architecture
JVM is a software interpreter that simulates an ISA
Advantage: portability
JVMs are found in hundreds of millions of devices (cell phones, Internet browsers, etc)
Performance can be enhanced with Just-in-Time compilation (JIT)
Java is very popular, but still generally slower than C/C++
1/29/20 Matni, CS154, Wi20 16
Program Performance:
Effect of Compiler Optimization on sort Program
* *
*
Ultimately, O3 runs the fastest.
1/29/20 Matni, CS154, Wi20 17
Instruction count and CPI are not good performance indicators in isolation
1. Compiler optimizations are sensitive to the algorithm
2. Java/JIT compiled code is significantly faster than JVM interpreted
3. Nothing can fix a dumb algorithm!
Program Performance:
Effect of Language and Algorithm
*
*
1/29/20
Matni, CS154, Wi20 18
*
YOUR TO-DOs for the Week
Readings! Work on Lab 4!
1/29/20 Matni, CS154, Wi20 19
1/29/20 Matni, CS154, Wi20 20
Reviews
There are no reviews yet.