15 213 The Class That Gives CMU Its Zip Introduction to Computer Systems Randal E Bryant January 15 2008 Topics Theme Five great realities of computer systems How this fits within CS curriculum class01a ppt 15 213 S 08 Course Theme Abstraction is good but don t forget reality Most CS courses emphasize abstraction Abstract data types Asymptotic analysis These abstractions have limits Especially in the presence of bugs Need to understand underlying implementations Useful outcomes Become more effective programmers Able to find and eliminate bugs efficiently Able to tune program performance Prepare for later systems classes in CS ECE Compilers Operating Systems Networks Computer Architecture 2 Embedded Systems 15 213 S 08 Great Reality 1 Int s are not Integers Float s are not Reals Examples Is x2 0 Float s Int s Yes 40000 40000 1600000000 50000 50000 Is x y z x y z Unsigned Signed Int s Float s Yes 1e20 1e20 3 14 3 14 1e20 1e20 3 14 3 15 213 S 08 Computer Arithmetic Does not generate random values Arithmetic operations have important mathematical properties Cannot assume usual properties Due to finiteness of representations Integer operations satisfy ring properties Commutativity associativity distributivity Floating point operations satisfy ordering properties Monotonicity values of signs Observation 4 Need to understand which abstractions apply in which contexts Important issues for compiler writers and serious application programmers 15 213 S 08 Great Reality 2 You ve got to know assembly Chances are you ll never write program in assembly Compilers are much better more patient than you are Understanding assembly key to machine level execution model Behavior of programs in presence of bugs High level language model breaks down Tuning program performance Understanding sources of program inefficiency Implementing system software Compiler has machine code as target Operating systems must manage process state 5 Creating fighting malware x86 assembly is the language of choice 15 213 S 08 Assembly Code Example Time Stamp Counter Special 64 bit register in Intel compatible machines Incremented every clock cycle Read with rdtsc instruction Application Measure time required by procedure In units of clock cycles double t start counter P t get counter printf P required f clock cycles n t 6 15 213 S 08 Code to Read Counter Write small amount of assembly code using GCC s asm facility Inserts assembly code into machine code generated by compiler static unsigned cyc hi 0 static unsigned cyc lo 0 Set hi and lo to the high and low order bits of the cycle counter void access counter unsigned hi unsigned lo asm rdtsc movl edx 0 movl eax 1 r hi r lo edx eax 7 15 213 S 08 Great Reality 3 Memory Matters Random Access Memory is an un physical abstraction Memory is not unbounded It must be allocated and managed Many applications are memory dominated Memory referencing bugs especially pernicious Effects are distant in both time and space Memory performance is not uniform 8 Cache and virtual memory effects can greatly affect program performance Adapting program to characteristics of memory system can lead to major speed improvements 15 213 S 08 Memory Referencing Bug Example double double fun int fun int i i volatile volatile double double d 1 d 1 3 14 3 14 volatile volatile long long int int a 2 a 2 a i 1073741824 a i 1073741824 Possibly Possibly out out of of bounds bounds return return d 0 d 0 fun 0 fun 0 fun 1 fun 1 fun 2 fun 2 fun 3 fun 3 fun 4 fun 4 9 3 14 3 14 3 14 3 14 3 1399998664856 3 1399998664856 2 00000061035156 2 00000061035156 3 14 3 14 then thensegmentation segmentationfault fault 15 213 S 08 Referencing Bug Explanation 10 Saved State 4 d7 d4 3 d3 d0 2 a 1 1 a 0 0 Location accessed by fun i C does not implement bounds checking Out of range write can affect other parts of program state 15 213 S 08 Memory Referencing Errors C and C do not provide any memory protection Out of bounds array references Invalid pointer values Abuses of malloc free Can lead to nasty bugs Whether or not bug has any effect depends on system and compiler Action at a distance Corrupted object logically unrelated to one being accessed Effect of bug may be first observed long after it is generated How can I deal with this 11 Program in Java or ML Understand what possible interactions may occur Use or develop tools to detect referencing errors 15 213 S 08 Memory System Performance Example void copyij int int int i j for i 0 i for j 0 dst i j src 2048 2048 dst 2048 2048 2048 i j 2048 j src i j 59 393 288 clock cycles void copyji int int int i j for j 0 j for i 0 dst i j src 2048 2048 dst 2048 2048 2048 j i 2048 i src i j 1 277 877 876 clock cycles 21 5 times slower Hierarchical memory organization Performance depends on access patterns Measured on 2GHz Intel Pentium 4 Including how step through multi dimensional array 12 15 213 S 08 The Memory Mountain Read throughput MB s 1200 Pentium III Xeon 550 MHz 16 KB on chip L1 d cache 16 KB on chip L1 i cache 512 KB off chip unified L2 cache copyij 1000 L1 800 600 copyji 400 xe L2 200 13 2k 8k 32k 128k 512k 2m 8m s15 s13 s9 s11 Stride words Mem s7 s5 s3 s1 0 Working set size bytes 15 213 S 08 Great Reality 4 There s more to performance than asymptotic complexity Constant factors matter too Easily see 10 1 performance range depending on how code written Must optimize at multiple levels algorithm data representations procedures and loops Must understand system to optimize performance 14 How programs compiled and executed How to measure program performance and identify bottlenecks How to improve performance without destroying code modularity and generality 15 213 S 08 Code Performance Example Compute product of array elements double product double d int n double result 1 int i for i 0 i n i result result d i return result Multiply all elements of array Performance on class machines 7 0 clock cycles per element Latency of floating point multiplier 15 15 213 S 08 Loop Unrollings Unroll by 2 Assume n is even double product u2 double d int n double result 1 int i for i 0 i n i 2 result result d i d i 1 return result Unroll by 2 Assume n is even double product u2r double d int n double result 1 int i for i 0 i n i 2 result result d i d i 1 return result Do two loop elements per iteration Reduces overhead Cycles per element u2 7 0 u2r 3 6 16 15 213 S 08 1 d0 u2 Serial Computation d1 Computation length 12 d2 1 d 0 d 1 d 2 d 3 d 4 d 5 d 6 d 7 d 8 d 9 d 10 d 11 …
View Full Document
Unlocking...