9-1
CHAPTER 9
PERFORMANCE CONSIDERATIONS
9.1
INTRODUCTION
System performance is a key attribute of any embedded computer system. How quickly a pro-
gram is run is the common measure of performance. Program performance is a function of many
parameters: CPU speed, clock speed, memory latency, memory data transfer rate, memory size,
disk access time, disk data transfer rate, video access time, compiler efficiency, operating system
efficiency, program algorithms, etc. This chapter focuses on the memory system parameters that
affect performance. External caches are also examined as a means of improving memory system
performance.
Chapters 5
and
6
give specific examples of memory and cache designs.
Memory system design is important. The Intel486™ processor is faster than any practical mem-
ory system. It contains a significant amount of logic (e.g., caches, write buffers, prefetcher) to
allow the execution logic to keep operating even with slow external memories. The on-chip cach-
es and data bandwidth requirements of the Intel486 processor are different than earlier micropro-
cessors. Memory system design should be approached differently as well. This chapter describes
the memory requirements and bus usage characteristics of the Intel486 processor.
9.1.1
Memory Performance Factors
The ideal memory subsystem would operate without wait states. All bus cycles on the Intel486
processor would complete in only two clocks for single access and five clocks for cache fill. This
is impractical for almost all applications since they would require huge amounts of 15 ns memory
to run at 33 MHz. Practical systems use DRAM of 60-100 ns access times. The Intel486 proces-
sor is designed to effectively use DRAM. This chapter examines memory system design using
DRAM.
There are many different performance options in the design of the memory subsystem for the
Intel486 processor. The CPU clock speed sets the maximum possible performance. Higher is fast-
er, but it then requires faster memories to keep the whole system performance scaling at the fre-
quency rate. The Intel486 processor is designed to allow overall performance to increase up to a
point with higher clock speed and constant memory speed.
The most common attribute of memory design is the number of wait states, if any, that are re-
quired to read a data item. At 33 MHz, a read operation requires 15 ns SRAM. For slower DRAM
or Flash access, at 33 MHz add 30 ns access time for each wait state. Wait states exist in practical
memory system design. This chapter examines how they affect Intel486 processor performance.
The Intel486 processor adds a new metric to memory design: read transfer rate. It is important for
filling the internal cache of the Intel486 processor. The Intel486 processor can transfer data from
memory on every clock for most read transfers. This is twice the rate of individual memory cy-
cles. Memory systems supporting this high speed transfer rate increase performance 10-20% over
those without.
A third important attribute is write cycle time. The Intel486 processor write-through cache gen-
erates approximately twice as many writes as reads. Write performance is especially important
Содержание Embedded Intel486
Страница 16: ......
Страница 18: ......
Страница 26: ......
Страница 28: ......
Страница 42: ......
Страница 44: ......
Страница 62: ......
Страница 64: ......
Страница 138: ......
Страница 139: ...5 Memory Subsystem Design Chapter Contents 5 1 Introduction 5 1 5 2 Processor and Cache Feature Overview 5 1 ...
Страница 140: ......
Страница 148: ......
Страница 150: ......
Страница 170: ......
Страница 172: ......
Страница 226: ......
Страница 228: ......
Страница 264: ......
Страница 282: ......
Страница 284: ......