欢迎访问ic37.com |
会员登录 免费注册
发布采购

RM5271-300S 参数 Datasheet PDF下载

RM5271-300S图片预览
型号: RM5271-300S
PDF下载: 下载PDF文件 查看货源
内容描述: 64位微处理器\n [64-Bit Microprocessor ]
分类和应用: 微处理器
文件页数/大小: 24 页 / 366 K
品牌: ETC [ ETC ]
 浏览型号RM5271-300S的Datasheet PDF文件第2页浏览型号RM5271-300S的Datasheet PDF文件第3页浏览型号RM5271-300S的Datasheet PDF文件第4页浏览型号RM5271-300S的Datasheet PDF文件第5页浏览型号RM5271-300S的Datasheet PDF文件第7页浏览型号RM5271-300S的Datasheet PDF文件第8页浏览型号RM5271-300S的Datasheet PDF文件第9页浏览型号RM5271-300S的Datasheet PDF文件第10页  
with a new mapping. However, the processor also provides  
a mechanism whereby a system specific number of map-  
pings can be locked into the TLB, thereby avoiding random  
replacement. This mechanism allows the operating system  
to guarantee that certain pages are always mapped for per-  
formance reasons and for deadlock avoidance. This mech-  
anism also facilitates the design of real-time systems by  
allowing deterministic access to critical software.  
0xFFFFFFFF Kernel virtual address space  
(kseg3)  
Mapped, 0.5GB  
0xE0000000  
0xDFFFFFFF Supervisor virtual address space  
(ksseg)  
Mapped, 0.5GB  
0xC0000000  
0xBFFFFFFF Uncached kernel physical address space  
The JTLB also contains information that controls the cache  
coherency protocol for each page. Specifically, each page  
has attribute bits to determine whether the coherency algo-  
rithm is one of the following:  
(kseg1)  
Unmapped, 0.5GB  
0xA0000000  
0x9FFFFFFF Cached kernel physical address space  
(kseg0)  
Unmapped, 0.5GB  
uncached  
non-coherent write-back]  
non-coherent write-through with write-allocate  
non-coherent write-through without  
write-allocate  
sharable  
exclusive  
update  
0x80000000  
0x7FFFFFFF User virtual address space  
(kuseg)  
Mapped, 2.0GB  
Note that both of the write-through protocols bypass the  
secondary cache since the secondary does not support  
writes of less than a complete cache line. The non-coher-  
ent protocols are used for both code and data on the  
RM5271 with data using write-back or write-through  
depending on the application.  
0x00000000  
Figure 4 Kernel Mode Virtual Addressing (32-bit)  
When the RM5271 is configured as a 64-bit microproces-  
sor, the virtual address space layout is an upward compati-  
ble extension of the 32-bit virtual address space layout.  
The coherency attributes generate coherent transaction  
types on the system interface. However, cache coherency  
is not supported in the RM5271 and therefore the coher-  
ency attributes should never be used.  
Joint TLB  
For fast virtual-to-physical address translation, the RM5271  
uses a large, fully associative TLB that maps 96 virtual  
pages to their corresponding physical addresses. As indi-  
cated by its name, the joint TLB, or JTLB, is used for both  
instruction and data translations. The JTLB is organized as  
48 pairs of even-odd entries, and maps a virtual address  
and address space identifier into the large, 64GB physical  
address space.  
Instruction TLB  
The RM5271 implements a 2-entry instruction TLB (ITLB)  
to minimize contention for the JTLB, eliminate the timing  
critical path of translating through a large associative array,  
and save power. Each ITLB entry maps a 4KB page. The  
ITLB improves performance by allowing instruction address  
translation to occur in parallel with data address translation.  
When a miss occurs on an instruction address translation  
by the ITLB, the least-recently used ITLB entry is filled from  
the JTLB. The operation of the ITLB is completely transpar-  
ent to the user.  
Two mechanisms are provided to assist in controlling the  
amount of mapped space and the replacement characteris-  
tics of various memory regions. First, the page size can be  
configured, on a per-entry basis, to use page sizes in the  
range of 4KB to 16MB (in multiples of 4). The CP0 Page-  
Mask register is loaded with the desired page size of a  
mapping, and that size is stored into the TLB along with the  
virtual address when a new entry is written. Thus, operat-  
ing systems can create special purpose maps. For exam-  
ple, a entire frame buffer can be memory mapped using  
only one TLB entry.  
Data TLB  
The RM5271 implements a 4-entry data TLB (DTLB). Each  
DTLB entry maps a 4KB page. The DTLB improves perfor-  
mance by allowing data address translation to occur in par-  
allel with instruction address translation. When a miss  
occurs on a data address translation by the DTLB, the  
DTLB is filled from the JTLB. The DTLB refill is pseudo-  
LRU: the least recently used entry of the least recently  
The second mechanism controls the replacement algorithm  
when a TLB miss occurs. The RM5271 provides a random  
replacement algorithm to select the TLB entry to be written  
6
RM5271 Microprocessor, Document Rev. 1.3  
Quantum Effect Devices  
www.qedinc.com