參數(shù)資料
型號: IBM32NPR101EPXCAC133
英文描述: Microprocessor
中文描述: 微處理器
文件頁數(shù): 39/131頁
文件大?。?/td> 1679K
代理商: IBM32NPR101EPXCAC133
PRS28.4G
IBM Packet Routing Switch
prs28.03.fm
August 31, 2000
Functional Description
Page 39 of 131
3.5.2.2 Anti Streaming
An error can also occur if an adapter transmits packets regardless of the output queue grant and/or memory
grant being deasserted. When an input receives a packet destined to an output for which the output queue
grant or memory grant have not been given in the past 8 LU cycles, a Flow Control Violation interrupt is set
and, if not masked, the main interrupt is asserted and the packet is discarded.
When the incoming packet is multicast, this anti streaming mechanism is only based on the memory grant,
and does not take the output queue grants into considerations.
3.6 Output Queues and Output Queue Priorities
The addresses of packets in the packet memory are stored in one output queue (or in multiple output queues
when the packet is a multicast packet). There are sixteen output queues, one queue for each output. Each
output queue can store the maximum number of addresses.
In addition, each output queue is logically divided into four independent queues, one queue for each packet
priority. When reading the output queue, addresses in a higher priority queue have precedence over
addresses in a lower priority queue. Consequently, higher priority packets will overtake lower priority packets
that are stored in the PRS28.4G.
When the output SND_GRANT is active, the corresponding output queue is emptied at the rate at which
packets are transmitted. This operation is also performed when the output port is disabled or when Sync
Packets are transmitted on that output (slow flush), regardless of the SND_GRANT value.
3.7 Shared Memory
3.7.1 Organization
The shared memory is organized as two dual port large RAMs, each containing 512 rows of 20 bytes. The
two RAMs operate in parallel. The first RAM stores all the bytes in the master stream and the second RAM
stores all the bytes in the slave stream.
Each input collects two rows of a packet, and then passes these rows to the packet memory. Similarly, each
output reads two rows from the packet memory and transmits it to the output.
3.7.2 Shared Memory Access by Local Processor
The local processor has full read and write access to the entire packet memory through the application regis-
ters, via the Memory Row register and the Table Pointer and Data Register. For example, this function is
used when creating or reading control packets. The access of the local processor to the shared memory has
the lowest priority and will only occur when there is at least one idle input (for write access) or one idle output
(for read access), or when the sequencer has an idle slot (for LU of 17 to 20, or 34 to 40).
For LU of length 17 to 20 and 34 to 40, the access time to the shared memory is guarantee to be at the most
one LU cycle. For LU of length 16 and 32, the access time can only be guaranteed if the Control Packet
Access Priority bit is set in the Configuration Register 0. In this case, the access time is guaranteed to be no
greater than 3 LU cycles.
相關(guān)PDF資料
PDF描述
IBM37RGB524CF17A Video DAC with Color Palette (RAMDAC)
IBM37RGB524CF22A Video DAC with Color Palette (RAMDAC)
IBM39MPEGS420PBA18C
IBM39MPEGS422PBA17C
IBM39MPEGSIPBA18C
相關(guān)代理商/技術(shù)參數(shù)
參數(shù)描述
IBM35CPC945C03C-2 制造商:IBM 功能描述:
IBM37RGB524CF17A 制造商:未知廠家 制造商全稱:未知廠家 功能描述:Video DAC with Color Palette (RAMDAC)
IBM37RGB524CF22A 制造商:未知廠家 制造商全稱:未知廠家 功能描述:Video DAC with Color Palette (RAMDAC)
IBM39ENV422DLL00C 制造商:IBM 功能描述:
IBM39ENV422PBA17C 制造商:IBM Microelectronics 功能描述:VID ENCODER 420PIN HPBGA - Trays 制造商:IBM 功能描述:IBM IBM39ENV422PBA17C Encoders - Decoders