Improving Cache Performance

Transcription

1 Improving Cache Performance Tuesday 27 October 15 Many slides adapted from: and Design, Patterson & Hennessy 5th Edition, 2014, MK and from Prof. Mary Jane Irwin, PSU

2 Summary Previous Class Memory hierarchy Caches Today: Improving Cache Performance 2

3 Terminology Block (or line): the minimum unit of information that is present (or not) in a cache Hit Rate: the fraction of memory accesses found in a level of the memory hierarchy Miss Rate: the fraction of memory accesses not found in a level of the memory hierarchy 1 - (Hit Rate) Miss Penalty: Time to replace a block in that level with the corresponding block from a lower 3

4 Measuring Cache Performance Components of CPU time Program execution cycles Includes cache hit time Memory stall cycles Mainly from cache misses Assuming cache hit costs are included as part of the normal CPU execution cycle, then: CPU time = IC CPI CC = IC (CPI ideal + Memory-stall cycles) CC CPI stall Assuming a combined read/write miss rate: Memory-stall cycles = accesses/program miss rate miss penalty 4

5 Cache Performance Example Given I-cache miss rate = 2% D-cache miss rate = 4% Miss penalty = 100 cycles Base CPI (ideal cache) = 2 Load & stores are 36% of instructions Miss cycles per instruction I-cache: = 2 D-cache: = 1.44 Actual CPI = = 5.44 Ideal CPU is 5.44/2 =2.72 times faster 5

6 Average Access Time Hit time is also important for performance Average memory access time (AMAT) AMAT = Hit time + Miss rate Miss penalty Example CPU with 1ns clock, hit time = 1 cycle, miss penalty = 20 cycles, I-cache miss rate = 5% Ø AMAT = = 2ns 2 cycles per instruction 6

7 Performance Summary When CPU performance increases Miss penalty becomes more significant Decreasing based on CPI Greater proportion of time spent on memory stalls Increasing clock rate Memory stalls account for more CPU cycles Can t neglect cache behavior when evaluating system performance 7

8 Reducing Cache Miss Rates: Associative Cache Allow for more flexible block placement: In a direct mapped cache a memory block maps to exactly one cache block At the other extreme, we can allow a memory block be mapped to any cache block fully associative cache A compromise is to divide the cache into sets each of which consists of n ways (n-way set associative). A memory block maps to a unique set (specified by the index field) and can be placed in any way of that set (so there are n choices) 8

9 Associative Caches Fully associative Allow a given block to go in any cache entry Requires all entries to be searched at once Comparator per entry (expensive) n-way set associative Each set contains n entries Block number determines which set (Block number) modulo (#Sets in cache) Search all entries in a given set at once n comparators (less expensive) 9

10 Associative Cache Example 10

11 Spectrum of Associativity For a cache with 8 entries: 11

12 Direct mapped - Example Consider the main memory word reference string Start with an empty cache - all blocks initially marked as not valid 0 miss 4 miss 0 miss 4 miss Mem(0) 00 Mem(0) 01 Mem(4) Mem(0) 4 0 miss 4 miss 0 miss 4 miss Mem(4) Mem(0) 01 Mem(4) 0 00 Mem(0) 8 requests, 8 misses Ping pong effect due to conflict misses - two memory locations that map into the same cache block 12

13 2-way set associative - Example Consider the main memory word reference string Start with an empty cache - all blocks initially marked as not valid miss 4 miss 0 hit 4 hit 000 Mem(0) 000 Mem(0) 000 Mem(0) 000 Mem(0) 010 Mem(4) 010 Mem(4) 010 Mem(4) 8 requests, 2 misses Solves the ping pong effect in a direct mapped cache due to conflict misses since now two memory locations that map into the same cache set can co-exist! 13

14 Set Associative Cache Organization 14

15 How Much Associativity The choice of direct mapped or set associative depends on the cost of a miss versus the cost of implementation Miss Rate KB 8KB 16KB 32KB 64KB 128KB 256KB 512KB 0 1-way 2-way 4-way 8-way Associativity Data from Hennessy & Patterson, Computer Architecture, 2003 Largest gains are in going from direct mapped to 2- way (20%+ reduction in miss rate) 15

16 Replacement Policy Direct mapped: no choice Set associative Prefer non-valid entry, if there is one Otherwise, choose among entries in the set Least-recently used (LRU) Choose the one unused for the longest time Simple for 2-way, manageable for 4-way, too hard beyond that Random Gives approximately the same performance as LRU for high associativity 16

17 Reducing Cache Miss Rates: Multilevel Caches Use multiple levels of caches With advancing technology have more than enough room on the die for bigger L1 caches or for a second level of caches normally a unified L2 cache i.e., holds both instructions and data Many high-end systems already include unified L3 cache Design considerations for L1 and L2 caches are very different Primary cache attached to CPU focus on minimizing hit time (i.e. small, but fast)» Smaller with smaller block sizes Level-2 cache services misses from primary cache focus on reducing miss rate (i.e. large, slower than L1) to reduce the penalty of long main memory access times» Larger with larger block sizes» Higher levels of associativity 17

18 Multilevel Cache Considerations Primary cache Focus on minimal hit time L-2 cache Focus on low miss rate to avoid main memory access Hit time has less overall impact Results L-1 cache usually smaller than a single cache L-1 block size smaller than L-2 block size t access = t hitl1 + p missl1 t penaltyl1 t penaltyl1 = t hitl2 + p missl2 t penaltyl2 t access = t hitl1 + p missl1 (t hitl2 + p missl2 t penaltyl2 ) 18

19 Multilevel Cache Example Given CPU base CPI = 1, clock rate = 4GHz Miss rate/instruction = 2% Main memory access time = 100ns With just primary cache Miss penalty = 100ns/0.25ns = 400 cycles Effective CPI = = 9 19

20 Example (cont.) Now add L-2 cache Access time = 5ns Global miss rate to main memory = 0.5% Primary miss with L-2 hit Penalty = 5ns/0.25ns = 20 cycles Primary miss with L-2 miss Extra penalty = 400 cycles CPI = = 3.4 Performance ratio = 9/3.4 =

21 Two Machines Cache Parameters Characteristic ARM Cortex-A8 Intel Nehalem L1 cache organization Split instruction and data caches Split instruction and data caches L1 cache size 32 KiB each for instructions/data 32 KiB each for instructions/data per core L1 cache associativity 4-way (I), 4-way (D) set associative 4-way (I), 8-way (D) set associative L1 replacement Random Approximated LRU L1 block size 64 bytes 64 bytes L1 write policy Write-back, Write-allocate(?) Write-back, No-write-allocate L1 hit time (load-use) 1 clock cycle 4 clock cycles, pipelined L2 cache organization Unified (instruction and data) Unified (instruction and data) per core L2 cache size 128 KiB to 1 MiB 256 KiB (0.25 MiB) L2 cache associativity 8-way set associative 8-way set associative L2 replacement Random(?) Approximated LRU L2 block size 64 bytes 64 bytes L2 write policy Write-back, Write-allocate (?) Write-back, Write-allocate L2 hit time 11 clock cycles 10 clock cycles L3 cache organization -- L3 cache size L3 cache associativity L3 replacement L3 block size L3 write policy L3 hit time Unified (instruction and data) 8 MiB, shared 16-way set associative Approximated LRU 64 bytes Write-back, Write-allocate 35 clock cycles 21

22 Interactions with Advanced CPUs Out-of-order CPUs can execute instructions during cache miss Pending store stays in load/store unit Dependent instructions wait in reservation stations Independent instructions continue Effect of miss depends on program data flow Much harder to analyze Use system simulation 22

23 Victim Cache Instead of completely discarding each block when it has to be replaced, temporarily keep it in a victim buffer. Rather than stalling on a subsequent cache miss, the contents of the buffer are checked on a subsequent miss to see if they have the desired data before going to the next lower-level of memory. Small cache (e.g., 4 to 16 positions) Fully associative Particularly efficient for small direct mapped caches (more than 25% reduction of the miss rate in a 4kB cache). 23

24 Interactions with Software Misses depend on memory access patterns Algorithm behavior Compiler optimization for memory access 24

25 Summary: Improving Cache Performance 1. Reduce the time to hit in the cache smaller cache direct mapped cache smaller blocks for writes no write allocate no hit on cache, just write to write buffer write allocate to avoid two cycles (first check for hit, then write) pipeline writes via a delayed write buffer to cache 2. Reduce the miss rate bigger cache more flexible placement (increase associativity) larger blocks (16 to 64 bytes typical) victim cache small buffer holding most recently discarded blocks 25

26 Summary: Improving Cache Performance 3. Reduce the miss penalty smaller blocks use a write buffer to hold dirty blocks being replaced so don t have to wait for the write to complete before reading check write buffer (and/or victim cache) on read miss may get lucky for large blocks fetch critical word first use multiple cache levels L2 cache not tied to CPU clock rate faster backing store/improved memory bandwidth wider buses memory interleaving, DDR SDRAMs 26

27 Summary: The Cache Design Space Several interacting dimensions cache size block size associativity replacement policy write-through vs write-back write allocation Cache Size Associativity Block Size 27

28 Next Class Optimizing cache usage 28

29 Improving Cache Performance Tuesday 27 October 15 Many slides adapted from: and Design, Patterson & Hennessy 5th Edition, 2014, MK and from Prof. Mary Jane Irwin, PSU