Transcript Slide 1
Computing Systems Assessing and Understanding Performance [email protected] 1 Performance Measure, Report, and Summarize performance Make intelligent choices See through the marketing hype Key to understanding underlying organizational motivation Why is some hardware better than others for different programs? What factors of system performance are hardware related? (e.g., Do we need a new machine, or a new operating system?) How does the machine's instruction set affect performance? We are looking for metrics for measuring performance from the viewpoint of both a computer user and a designer [email protected] 2 Defining performance Airplane Pass. capacity Range (miles) Speed (m.p.h.) Pass. throughput (pass. x m.p.h.) Boeing 777 375 4630 610 228750 Boeing 747 470 4150 610 286700 Concorde 132 4000 1350 178200 Douglas DC 8-50 146 8720 544 79424 • How much faster is the Concorde compared to the 747 ? • Is the Concorde faster compared to the 747 ? • How much bigger is the 747 than the Douglas DC-8? • Which of these airplanes has the best performance ? [email protected] 3 Understanding performance The performance of a program depends on: - the algorithm, - the language, - the compiler, - the architecture - the actual hardware [email protected] 4 Computer performance: Time, Time, Time !!! Response Time = Execution Time = Latency - The time between the start and completion of a task Throughput - Total amount of work completed in a given time If we upgrade a machine with a new faster processor what do we increase? If we add a new processor to a system that uses multiple processors what do we increase? [email protected] 5 Execution Time Execution Time (response time or elapsed time) total time to complete a program, it counts everything (disk accesses, memory accesses, input/output activities) a useful number, but often not good for comparison purposes CPU (execution) time doesn't count time spent waiting for I/O or time spent running other programs can be broken up into system time (CPU time spent in the OS), and user time (CPU time spent in the program) Our focus: user CPU time time spent executing the lines of code that are "in" our program [email protected] 6 Book’s definition of performance - For some program running on machine X, 1 Performance X ExecutionTime x - "X is n times faster than Y" Performance X ExecutionTime Y n Performance Y ExecutionTime X Problem (Relative Performance) : – machine A runs a program in 10 seconds – machine B runs the same program in 25 seconds – How much faster is machine A compared to B ? [email protected] 7 Measuring Time Instead of reporting execution time in seconds, we often use cycles (= clock cycles = ticks = clock ticks = clocks = clock periods) seconds cycles seconds program program cycle Clock “ticks” indicate when to start activities: time cycle time = time between ticks = seconds per cycle clock rate (frequency) = cycles per second (1 Hz. = 1 cycle/sec) Example: a 4 GHz. clock has a 250 ps. cycle time [email protected] 8 How to improve performance CPU _execution_time _ for_a_program CPU_clock_ cycles_for_a_program clock_cycle_time Time seconds cycles seconds program program cycle So, to improve performance (everything else being equal) you can either (increase or decrease?) ________ the # of required cycles for a program, or ________ the clock cycle time or, said another way, ________ the clock rate. [email protected] 9 Example - improving performance Our favorite program runs in 10 seconds on computer A, which has a 4 GHz. clock. We are trying to help a computer designer build a new machine B, that will run this program in 6 seconds. The designer can use new (or perhaps more expensive) technology to substantially increase the clock rate, but has informed us that this increase will affect the rest of the CPU design, causing machine B to require 1.2 times as many clock cycles as machine A for the same program. What clock rate should we tell the designer to target?" [email protected] 10 Cycles required for a program Can we assume that # cycles = # instructions ? This assumption is incorrect. Different instructions take different amounts of time. Why? remember that these are machine instructions, not lines of C code Multiplication takes more time than addition Floating point operations take longer than integer ones Accessing memory takes more time than accessing registers Important point: changing the cycle time often changes the number of cycles required for various instructions (more later) [email protected] 11 Clock cycles per instruction It is clear that the execution time of a program must depends on the number of machine instructions generated by the compiler: CPU_clock_ cycles instructions_for_a_program average_clock_cycles_per_instruction the average number of clock cycles each instruction takes to execute is often abbreviated CPI CPI provides a way of comparing two different implementations of the same ISA (since the IC required for a program will be the same) [email protected] 12 The “performance equation” A given program will require: some number of instructions (machine instructions) some number of cycles per each instruction some number of seconds per cycle CPU_t ime instruction_ count CPI clock_cycle_time This useful formula separate the 3 key factors that affect performance. [email protected] 13 Performance - the BIG picture The only complete and reliable measure of performance is determined by execution time Time Seconds Instructions Clock_cycles Seconds Program Program Instruction Clock_ cycle NO! Do any of the other variables equal performance? # of cycles to execute program? # of instructions in program? # of cycles per second? average # of cycles per instruction (CPI)? average # of instructions per second (MIPS)? Common pitfall: thinking one of the variables is indicative of performance when it really isn’t. [email protected] 14 Example - CPI Suppose we have two implementations of the same instruction set architecture (ISA). For some program: Machine A has a clock cycle time of 250 ps and a CPI of 2.0 Machine B has a clock cycle time of 500 ps and a CPI of 1.2 What machine is faster for this program, and by how much? If two machines have the same ISA which of the following quantities will always be identical? - clock rate, CPI, execution time, # of instructions, # of cycles, MIPS [email protected] 15 Example - Number of Instructions A compiler designer is trying to decide between two code sequences for a particular machine. Based on the hardware implementation, there are three different classes of instructions: Class A, Class B, and Class C, and they require one, two, and three cycles (respectively). The first code sequence has 5 instructions: 2 of A, 1 of B, and 2 of C. The second sequence has 6 instructions: 4 of A, 1 of B, and 1 of C. Which sequence will be faster? How much? What is the average CPI for each sequence? N Hint: CPU_clock_ cycles (CPI i Ci ) i1 N = number of instruction classes, Ci = count of the # of instructions of class i executed [email protected] 16 MIPS Million instructions per second MIPS Instruction count Execution time 106 Problems using MIPS for comparing computers MIPS specifies the instruction execution rate but does not take into account that instructions may have different capabilities MIPS varies between programs on the same computer; thus a computer cannot have a single MIPS rating for all programs MIPS can vary inversely with performance !!! [email protected] 17 Example – MIPS Two different compilers are being tested for a 4 GHz. machine with three different classes of instructions: Class A, Class B, and Class C, which require one, two, and three cycles (respectively). Both compilers are used to produce code for a large piece of software. The first compiler's code uses: - 5 million Class A instructions, - 1 million Class B instructions, - 1 million Class C instructions. The second compiler's code uses: - 10 million Class A instructions, - 1 million Class B instructions, - 1 million Class C instructions. Which sequence will be faster according to MIPS? Which sequence will be faster according to execution time? [email protected] 18 Benchmarks Performance best determined by running a real application Use programs typical of expected workload or typical of expected class of applications (e.g., compilers/editors, scientific applications, graphics, etc.) Small benchmarks nice for architects and designers easy to standardize can be abused SPEC (System Performance Evaluation Cooperative) companies have agreed on a set of real program and inputs valuable indicator of performance (and compiler technology) can still be abused (see Intel’s benchmark ) [email protected] 19 Benchmark “games” An embarrassed Intel Corp. acknowledged Friday that a bug in a software program known as a compiler had led the company to overstate the speed of its microprocessor chips on an industry benchmark by 10 percent. However, industry analysts said the coding error…was a sad commentary on a common industry practice of “cheating” on standardized performance tests…The error was pointed out to Intel two days ago by a competitor, Motorola …came in a test known as SPECint92…Intel acknowledged that it had “optimized” its compiler to improve its test scores. The company had also said that it did not like the practice but felt to compelled to make the optimizations because its competitors were doing the same thing…At the heart of Intel’s problem is the practice of “tuning” compiler programs to recognize certain computing problems in the test and then substituting special handwritten pieces of code… Saturday, January 6, 1996 New York Times [email protected] 20 Benchmarks Different classes and applications of computers requires different types of benchmark suites SPEC CPU2000 SPECweb99 EEMBC The execution time measurements are normalized by dividing the execution time on a Sun Ultra 5_10 with a 300 MHz processor by the execution time on a measured computer (this measure is called SPEC ratio) The guiding principle in reporting performance measurements should be reproducibility (an important aspect of reproducibility is the choice of input) [email protected] 21 Spec’89 – Compiler enhancement and performance 800 700 SPEC performance ratio 600 500 400 300 200 100 0 gcc espresso spice doduc nasa7 li eqntott matrix300 fpppp tomcatv Benchmark Compiler [email protected] Enhanced compiler 22 SPEC CPU2000 [email protected] 23 SPEC2000 Ratio PIII CINT2000/clock rate in MHz 0.47 CFP2000/clock rate in MHz 0.34 P4 0.36 0.39 Does doubling the clock rate double the performance? 1400 1200 Pentium 4 CFP2000 1000 Pentium 4 CINT2000 800 600 Pentium III CINT2000 400 Pentium III CFP2000 200 0 500 1000 1500 2000 2500 3000 3500 Clock rate in MHz [email protected] 24 SPEC2000 Can a machine with a slower clock rate have better performance? 1.6 Pentium M @ 1.6/0.6 GHz Pentium 4-M @ 2.4/1.2 GHz 1.4 Pentium III-M @ 1.2/0.8 GHz 1.2 1.0 0.8 0.6 0.4 0.2 0.0 SPECINT2000 SPECFP2000 SPECINT2000 SPECFP2000 SPECINT2000 SPECFP2000 Always on/maximum clock Laptop mode/adaptive clock Minimum power/minimum clock Benchmark and power mode [email protected] 25 Performance, power, and energy efficiency Power is increasingly becoming a key limitation in processor performance (especially for embedded processors) In CMOS technology the primary source of power dissipation is: sw itching_pow er capacitive_load Voltage2 sw itching_frequency For power limited application, the most important metric is energy efficiency, which is computed by taking performance and dividing by average power consumption when running the benchmark [email protected] 26 Summarizing performance Although summarizing measurements result in less information, marketers and even users often prefer to have a single number to compare performance Arithmetic mean of the execution times (underlying assumption that the programs in the workload are each run an equal number of times) 1 n AM Time i n i1 Weighted arithmetic mean (wi frequency of the program in the workload) n WAM w i Time i i1 [email protected] 27 Amdahl’s law The performance enhancement possible with a given improvement is limited by the amount that the improved feature is used Execution time after improvement = Execution time affected + Execution time unaffected Amount of improvement Principle: make the common case fast [email protected] 28 Example – Amdahl’s law Suppose a program runs in 100 seconds on a machine, with multiply responsible for 80 seconds of this time. How much do we have to improve the speed of multiplication if we want the program to run 4 times faster? How about making it 5 times faster? [email protected] 29 Summary Performance is specific to a particular program Execution time is the only valid and unimpeachable measure of performance For a given ISA, increases in CPU performance come from three sources Increases in clock rate Improvement in processor organization that lower the CPI Compiler enhancements that lower the instruction count and/or the average CPI (e.g. by using simpler instructions) Pitfalls Using a subset of the “performance equation” as performance metric Expecting the improvement of one aspect of a machine’s performance to increase total performance by an amount proportional to the size of the partial improvement Designing only for performance without considering cost, functionality and other requirements is unrealistic [email protected] 30