TWI499969B - Reordering buffer - Google Patents

Reordering buffer Download PDF

Info

Publication number
TWI499969B
TWI499969B TW103120494A TW103120494A TWI499969B TW I499969 B TWI499969 B TW I499969B TW 103120494 A TW103120494 A TW 103120494A TW 103120494 A TW103120494 A TW 103120494A TW I499969 B TWI499969 B TW I499969B
Authority
TW
Taiwan
Prior art keywords
memory
bit
group
bit data
memory module
Prior art date
Application number
TW103120494A
Other languages
Chinese (zh)
Other versions
TW201546711A (en
Inventor
Sau Gee Chen
Shen Jui Huang
Bo Wei Wang
Original Assignee
Univ Nat Chiao Tung
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Univ Nat Chiao Tung filed Critical Univ Nat Chiao Tung
Priority to TW103120494A priority Critical patent/TWI499969B/en
Application granted granted Critical
Publication of TWI499969B publication Critical patent/TWI499969B/en
Publication of TW201546711A publication Critical patent/TW201546711A/en

Links

Description

重新調序緩衝器Reorder buffer

本發明係有關一種重新調序緩衝器,尤指一種適用於連續流程快速傅立葉轉換運算之重新調序緩衝器。The present invention relates to a reordering buffer, and more particularly to a reordering buffer suitable for continuous flow fast Fourier transform operations.

目前快速傅利葉轉換(FFT)廣泛應用於有線及無線通訊系統及電子電機工程領域,具有極大之應用價值,但其運算複雜度極高。針對重新調序緩衝器之應用文獻上已有提出幾種能將資料流以自然順序讀出之方式。舉例來說,一種改良式多路徑延遲交換快速(Improved Multipath Delay Commutator)快速傅利葉轉換架構,其係將輸入資料流分為偶數索引(even-indexed)及奇數索引(odd-indexed)的資料流,再透過適當之排程進行每一級之蝶式運算(Butterfly Operation);為達成自然順序之輸出,於最後一級基底運算單元前,加入一重新調序緩衝器(Reordering buffer),將偶數索引資料流以位元反向寫入,自然順序讀出,而奇數索引資料流以自然順序寫入,位元反向讀出。惟,此架構只適用於單一資料流輸入及輸出,以致整體系統產出率(throughput)低。At present, Fast Fourier Transform (FFT) is widely used in wired and wireless communication systems and electronic motor engineering fields, and has great application value, but its computational complexity is extremely high. There are several ways in which the data stream can be read out in a natural order in the application literature for reordering buffers. For example, an improved Multipath Delay Commutator fast Fourier transform architecture divides an input data stream into an even-indexed and odd-indexed data stream. Then perform the Butterfly Operation for each level through appropriate scheduling; to achieve the natural order output, add a Reordering buffer to the even index data stream before the last level of the base unit. Write in reversed bits, read out in natural order, while odd indexed data streams are written in natural order and bits are read back. However, this architecture is only applicable to single data stream input and output, resulting in a low overall system throughput.

此外,在實際應用上(如OFDM系統),經常需要執行連續性(continuous-flow)之快速傅利葉轉換運算,以平行管線式快速傅利葉轉換處理器(Parallel Pipelined FFT Processor)為例,因具有高運算產出 (throughput),所以被許多高速通訊系統所採用,但是一般管線式快速傅利葉轉換處理器之輸出具有位元反向(bit-reverse)特性,如果要使其轉變成自然順序之輸出,則需加上額外之重新調序緩衝器(Reordering buffer)。然而目前卻沒有任何技術能夠將平行管線式快速傅利葉轉換處理器輸出的位元反相輸出轉成自然順序的輸出,不是具有延遲問題,就是只能單一路徑輸入及輸出之技術窘境。再者,為了維持無延遲之快速傅利葉轉換計算輸出,必須要有足夠之暫存記憶體空間及高效能之資料存取機制,然而所需記憶體面積必須相當龐大,實無法讓整體系統架構縮小化;因此,如何設計一種新穎的重新調序緩衝器能夠在支援連續流程快速傅利葉轉換運算時,利用最小記憶體面積作為資料存取機制,並將平行管線式快速傅利葉轉換處理器輸出的位元反相輸出轉成自然順序的輸出是亟待解決的問題。In addition, in practical applications (such as OFDM systems), it is often necessary to perform a continuous-flow fast Fourier transform operation, taking a parallel pipelined Fast Fourier Transform Processor (Parallel Pipelined FFT Processor) as an example, because of high computation output (throughput), so it is used by many high-speed communication systems, but the output of the general pipelined fast Fourier transform processor has bit-reverse characteristics. If you want to convert it into a natural order output, you need to add An additional reordering buffer. However, at present, there is no technology that can convert the inverted output of the bit line output of the parallel pipeline type fast Fourier transform processor into a natural order output, which is not a delay problem, or a technical dilemma of single path input and output. Furthermore, in order to maintain the fast Fourier transform calculation output without delay, it is necessary to have enough temporary memory space and high-performance data access mechanism. However, the required memory area must be quite large, which does not make the overall system architecture shrink. Therefore, how to design a novel reorder buffer can use the minimum memory area as the data access mechanism and the bit of the parallel pipeline fast Fourier transform processor when supporting continuous flow fast Fourier transform operation. The conversion of the inverted output to a natural sequence is an urgent problem to be solved.

有鑑於此,本發明遂針對上述先前技術的缺失,提出一種重新調序緩衝器,以有效克服上述的該等問題。In view of the above, the present invention has been directed to the lack of the prior art described above, and proposes a reordering buffer to effectively overcome the above problems.

本發明的主要目的在於提供一種重新調序緩衝器,其以最具面積效益之方式,使平行管線式快速傅利葉轉換處理器之位元反向輸出轉變成自然順序、連續、無延遲之輸出。It is a primary object of the present invention to provide a reordering buffer that converts the bit reverse output of a parallel pipelined fast Fourier transform processor into a natural sequential, continuous, delayless output in a manner that is most cost effective.

為達以上的目的,本發明提供一種重新調序緩衝器,包括一第一記憶體模組、一第二記憶體模組、一寫入交換器、一讀寫控制器及一讀取交換器。第一記憶體模組具有m個第一單埠記憶體單元,每一第一單埠記憶體單元儲存N/2m筆位元資料,N為傅立葉轉換之位元長度,m係為2的冪次方之整數;第二記憶體模組具有m個第二單埠記憶體單元,每一第二單 埠記憶體單元儲存N/2m筆位元資料。寫入交換器電性連接第一記憶體模組及第二記憶體模組,寫入交換器係接收m個傅立葉轉換之反向位元資料,並將反向位元資料交換為一預設排列方式之寫入位元資料,並交替寫入第一記憶體模組及該第二記憶體模組中。讀寫控制器電性連接寫入交換器、第一記憶體模組及第二記憶體模組,讀寫控制器係控制寫入位元資料寫入第一記憶體模組或第二記憶體模組中,或者是自第一記憶體模組或第二記憶體模組中讀出位元資料。讀取交換器電性連接第一記憶體模組、第二記憶體模組及讀寫控制器,讀取交換器係交替將第一記憶體模組及第二記憶體模組內的位元資料分別交換為自然順序之輸出位元資料,並以m個輸出路徑將輸出位元資料輸出。本發明所設計新穎的重新調序緩衝器架構,不僅可支援m倍平行快速傅利葉轉換輸入輸出,又能適用於連續流程快速傅利葉轉換運算(continuous-flow FFT operation),能使所有需要用到快速傅利葉轉換運算的通訊產品具有更高運算產出之應用價值。To achieve the above objective, the present invention provides a reordering buffer including a first memory module, a second memory module, a write switch, a read/write controller, and a read switch. . The first memory module has m first memory cells, each first memory cell stores N/2m pen bit data, N is a bit length of the Fourier transform, and m is a power of 2. The integer of the second power; the second memory module has m second memory units, each second single The memory unit stores N/2m pen bit data. The write switch is electrically connected to the first memory module and the second memory module, and the write switch receives the inverse bit data of the m Fourier transforms, and exchanges the reverse bit data into a preset. The bit data is written in the arrangement mode and alternately written into the first memory module and the second memory module. The read/write controller is electrically connected to the write switch, the first memory module and the second memory module, and the read/write controller controls the write bit data to be written into the first memory module or the second memory In the module, the bit data is read from the first memory module or the second memory module. The read switch is electrically connected to the first memory module, the second memory module and the read/write controller, and the read switch alternates the bits in the first memory module and the second memory module The data is exchanged for the natural order output bit data, and the output bit data is output in m output paths. The novel reordering buffer architecture designed by the invention not only supports m times parallel fast Fourier transform input and output, but also can be applied to continuous flow fast FFT operation, which can be used quickly. The communication products of Fourier transform operation have higher application value of computing output.

底下藉由具體實施例詳加說明,當更容易瞭解本發明的目的、技術內容、特點及其所達成的功效。The purpose, technical contents, features and effects achieved by the present invention will be more readily understood by the detailed description of the embodiments.

10‧‧‧第一記憶體模組10‧‧‧First memory module

12‧‧‧第二記憶體模組12‧‧‧Second memory module

14‧‧‧寫入交換器14‧‧‧Write switch

16‧‧‧讀寫控制器16‧‧‧Reading and writing controller

18‧‧‧讀取交換器18‧‧‧Reading switch

20‧‧‧平行管線式快速傅立葉轉換處理器20‧‧‧Parallel Pipeline Fast Fourier Transform Processor

22‧‧‧多工器22‧‧‧Multiplexer

第1圖為本發明的電路方塊圖。Figure 1 is a block diagram of the circuit of the present invention.

第2A~2E圖本發明以128位元長度交換為自然順序輸出之實施例示意圖。2A-2E is a schematic diagram of an embodiment of the present invention in which 128-bit lengths are exchanged for natural sequential output.

第3圖為本發明之寫入交換器的交換模式示意圖。Figure 3 is a schematic diagram of the switching mode of the write switch of the present invention.

第4A~4E圖為本發明以256位元長度交換為自然順序輸出之實施例 示意圖。4A to 4E are diagrams showing an embodiment in which the 256-bit length is exchanged for natural order output. schematic diagram.

請參閱第1圖,為本發明的電路方塊圖。重新調序緩衝器一第一記憶體模組10、一第二記憶體模組12、一寫入交換器14、一讀寫控制器16及一讀取交換器18。寫入交換器14電性連接第一記憶體模組10及第二記憶體模組12;讀寫控制器16電性連接寫入交換器14、第一記憶體模組10及第二記憶體模組12;讀取交換器18電性連接第一記憶體模組10、第二記憶體模組12及讀寫控制器16。第一記憶體模組10具有m個第一單埠記憶體單元,每一第一單埠記憶體單元儲存N/2m筆位元資料,N為傅立葉轉換之位元長度,m係為2的冪次方之整數。第二記憶體模組12具有m個第二單埠記憶體單元,每一第二單埠記憶體單元儲存N/2m筆位元資料。寫入交換器14係每次m個筆傅立葉轉換之反向位元資料,此反向位元資料係由一平行管線式快速傅立葉轉換處理器20輸出,且平行管線式快速傅立葉轉換處理器(parallel pipelined FFT processor)20電性連接寫入交換器14。其中,寫入交換器14係將反向位元資料交換為一預設排列方式之寫入位元資料,並交替寫入第一記憶體模組10及第二記憶體模組12中。讀寫控制器16係控制寫入位元資料寫入第一記憶體模組或第二記憶體模組中,或者是自第一記憶體模組10或第二記憶體模組12中讀出位元資料。讀取交換器18係交替將第一記憶體模組10及第二記憶體模組12內的位元資料分別交換為自然順序之輸出位元資料,並以m個輸出路徑將輸出位元資料輸出。重新調序緩衝器更包括一多工器22,電性連接第一記憶體模組10、第二記憶體模組12及讀取交換器18,第一記憶體模組10及第二記憶體模組12中的位元資料係由多工 器22選擇輸出,讀取交換器18係接收多工器22所輸出之位元資料以交換為自然順序之輸出位元資料。Please refer to FIG. 1 , which is a circuit block diagram of the present invention. The reordering buffers are a first memory module 10, a second memory module 12, a write switch 14, a read/write controller 16, and a read switch 18. The write switch 14 is electrically connected to the first memory module 10 and the second memory module 12; the read/write controller 16 is electrically connected to the write switch 14, the first memory module 10, and the second memory. The module 12 is electrically connected to the first memory module 10, the second memory module 12, and the read/write controller 16. The first memory module 10 has m first memory cells, each first memory cell stores N/2m pen bit data, N is the bit length of the Fourier transform, and m is 2 An integer of the power of the power. The second memory module 12 has m second memory cells, and each second memory cell stores N/2m pen metadata. The write switch 14 is an inverse bit data of m pen Fourier transforms each time, and the reverse bit data is output by a parallel pipeline type fast Fourier transform processor 20, and a parallel pipeline type fast Fourier transform processor ( The parallel pipelined FFT processor 20 is electrically connected to the write switch 14. The write switch 14 exchanges the reverse bit data into a preset arrangement of the bit data, and alternately writes the first memory module 10 and the second memory module 12. The read/write controller 16 controls the write bit data to be written into the first memory module or the second memory module, or is read from the first memory module 10 or the second memory module 12. Bit data. The read switch 18 alternately exchanges the bit data in the first memory module 10 and the second memory module 12 into natural bit output data, and outputs the bit data in m output paths. Output. The reordering buffer further includes a multiplexer 22 electrically connected to the first memory module 10, the second memory module 12, and the read switch 18, the first memory module 10 and the second memory. The bit data in module 12 is multiplexed The device 22 selects the output, and the read switch 18 receives the bit data output by the multiplexer 22 for exchange as a natural order output bit data.

為能進一步說明如何將平行管線式快速傅立葉轉換處理器輸出的反向位元資料交換為自然順序之輸出位元資料,本發明將交換模式根據快速傅立葉轉換(FFT)長度N分為兩種型式,且適用於任意2的冪次方點數,如後說明。寫入交換器14或讀取交換器18分別根據傅立葉轉換之位元長度N而選擇以交替公式將寫入位元資料寫入第一記憶體模組10或第二記憶體模組12,或自第一記憶體模組10或第二記憶體模組12中讀出位元資料,交替公式(1)如下所示:α =log2 (N /2m 2 ).....(1)In order to further explain how to exchange the inverse bit data outputted by the parallel pipeline type fast Fourier transform processor into the natural order output bit data, the present invention divides the switching mode into two types according to the fast Fourier transform (FFT) length N. And applies to any power of 2 power points, as explained later. The write switch 14 or the read switch 18 respectively selects to write the write bit data into the first memory module 10 or the second memory module 12 according to the bit length N of the Fourier transform, or The bit data is read from the first memory module 10 or the second memory module 12, and the alternating formula (1) is as follows: α = log 2 ( N /2 m 2 )..... (1 )

其中,α係為整數,N為傅立葉轉換之位元長度,m係為偶數,亦即為平行管線式快速傅立葉轉換處理器20輸出的平行度數目。Wherein, α is an integer, N is a bit length of the Fourier transform, and m is an even number, that is, the number of parallelisms output by the parallel pipeline type fast Fourier transform processor 20.

後續將詳細說明α 為偶數及奇數之兩種讀寫排程模式。首先,先以α 為偶數為例說明,請同時參閱第1圖、第2A~2E圖及第3圖,第2A~2E圖為本發明N為128位元長度交換為自然順序輸出之實施例示意圖,第3圖為本發明之寫入交換器的交換模式示意圖。當傅立葉轉換之位元長度N為128,其經過m為8倍平行的平行管線式快速傅立葉轉換處理器20後輸出128點的反向位元資料,α 係為偶數,則實施第一種交換模式;詳言之,寫入交換器14係將反向位元資料以各寫入一次之交替方式處理後,將寫入位元資料寫入第一記憶體模組10及第二記憶體模組12中,讀出交換器18係以各讀出一次之交替方式自第一記憶體模組10及第二記憶體模組12中讀出位元資料。首先,如第2A所示,為本發明之位元長度128點經8倍平行管線式 快速傅立葉轉換處理器20後之輸出示意圖。寫入交換器14係根據一預設排列方式選定之控制訊號進行交換反向位元資料,預設排列方式選定之控制訊號係由時脈計數器(clock cycle counter)來產生每一個記憶體之位址,預設排列方式與傅立葉轉換之位元長度N有關。傅立葉轉換之位元長度N為128,其分為16組八個位元參數,例如16組編號為C0~C15。再如第2B圖所示,為本發明之128點的反向位元資料經寫入交換器交換後之輸出;寫入交換器14係將反向位元資料中每兩組的八個位元參數以八種排列方式,例如八種排列的編號為CROT(0)~CROT(7),而每一種排列方式係遵循第3圖的CROT(0)~CROT(7)八種寫入交換模式進行交換,將反向位元資料交換為寫入位元資料,並將寫入位元資料交替寫入第一記憶體模組10及第二記憶體模組12中。再如第2C、第2D及2E圖所示,第2C、第2D分別為本發明之讀寫排程示意圖及本發明儲存寫入位元資料於記憶體中之示意圖;其中2D為奇數訊框(frame)時之儲存分怖,2E為偶數訊框時之儲存分怖,此處訊框係一周期為(N/m)之區間,並假設從一計數。第一記憶體模組10具有m個第一單埠記憶體單元,若m為8,則8組第一單埠記憶體單元為Bank_A0~Bank_A7,由於每一第一單埠記憶體單元儲存N/2m筆位元資料,因此每一個第一單埠記憶體單元具有8個儲存位址。同理,第二記憶體模組12具有m個第二單埠記憶體單元,若m為8,則8個第二單埠記憶體單元為Bank_B0~Bank_B7,由於每一第二單埠記憶體單元儲存N/2m筆位元資料,因此每一個第二單埠記憶體單元具有8個儲存位址。The following will explain in detail the two read and write scheduling modes in which α is even and odd. First, let α be an even number as an example. Please refer to FIG. 1 , 2A to 2E , and 3 , and 2A to 2E are examples in which N is 128-bit length exchanged to natural sequential output. Schematic diagram, FIG. 3 is a schematic diagram of an exchange mode of the write switch of the present invention. When the length N of the Fourier transform bit is 128, and after the parallel pipeline type fast Fourier transform processor 20 whose m is 8 times parallel, the inverted bit data of 128 points is output, and the α system is even, then the first exchange is implemented. Mode; in detail, the write switch 14 writes the write bit data into the first memory module 10 and the second memory module after processing the reverse bit data in an alternate manner. In the group 12, the read switch 18 reads the bit data from the first memory module 10 and the second memory module 12 in an alternate manner. First, as shown in FIG. 2A, the output of the present invention is 128 bits past the 8-time parallel pipeline type fast Fourier transform processor 20. The write switch 14 exchanges the reverse bit data according to a control signal selected by a preset arrangement. The control signal selected by the preset arrangement is generated by a clock cycle counter to generate each memory bit. The address, the preset arrangement is related to the bit length N of the Fourier transform. The Fourier transform bit length N is 128, which is divided into 16 groups of eight bit parameters, for example, 16 groups are numbered C0~C15. Further, as shown in FIG. 2B, the inverted bit data of 128 points of the present invention is outputted by the switch exchange; the write switch 14 is the eight bits of each of the two sets of reverse bit data. The meta-parameters are arranged in eight ways, for example, the eight arrays are numbered CROT(0)~CROT(7), and each arrangement follows the CROT(0)~CROT(7) eight write exchanges of Figure 3. The mode is exchanged, the reverse bit data is exchanged for writing the bit data, and the written bit data is alternately written into the first memory module 10 and the second memory module 12. 2C, 2D, and 2E, 2C and 2D are respectively a schematic diagram of reading and writing scheduling of the present invention, and a schematic diagram of storing the written bit data in the memory in the present invention; wherein 2D is an odd frame (frame) storage time division, 2E is the storage of the even frame, the frame is a period of (N / m), and is assumed to count from one. The first memory module 10 has m first memory cells. If m is 8, the first memory cells of the eight groups are Bank_A0~Bank_A7, and each first memory cell stores N. /2m pen bit data, so each first memory cell has 8 storage addresses. Similarly, the second memory module 12 has m second memory cells. If m is 8, the eight second memory cells are Bank_B0~Bank_B7, because each second memory is The unit stores N/2m pen bit data, so each second memory cell has 8 storage addresses.

暸解上述每一個元件之間的相對關係之後,接下來以實際運作為例詳細說明,寫入交換器14係將反向位元資料中每兩組的八個位元參 數以八種排列方式,其順序為CROT(0)、CROT(4)、CROT(2)、CROT(6)、CROT(1)、CROT(5)、CROT(3)、CROT(7)。詳言之,寫入交換器14將平行管線式快速傅立葉轉換處理器20之反向位元資料C0[0,64,32,96,16,80,48,112]及C1[8,72,40,104,24,88,56,120]以CROT(0)的排列方式進行交換,其交換模式係將第一組的八個位元參數[0,64,32,96,16,80,48,112]平行輸入至八組第一單埠記憶體單元Bank_A0~Bank_A7中的第一個位址A0,第二組的八個位元參數[8,72,40,104,24,88,56,120]平行輸入至八組第二單埠記憶體單元Bank_B0~Bank_B7中的第一個位址B0,且每一第一單埠記憶體單元及每一第二單埠記憶體單元中的第一個單埠記憶體單元係作為位移的初始單埠記憶體單元。After understanding the relative relationship between each of the above components, the following is a detailed description of the actual operation. The write switch 14 will associate eight bits of each of the two sets of reverse bit data. The order is in the order of CROT(0), CROT(4), CROT(2), CROT(6), CROT(1), CROT(5), CROT(3), CROT(7). In detail, the write switch 14 will reverse the bit data C0[0, 64, 32, 96, 16, 80, 48, 112] and C1 [8, 72, 40, 104, of the parallel pipelined fast Fourier transform processor 20, 24, 88, 56, 120] is exchanged in CROT (0) arrangement, the exchange mode is to input the eight bit parameters [0, 64, 32, 96, 16, 80, 48, 112] of the first group into eight in parallel. The first address A0 of the first memory cell Bank_A0~Bank_A7, the eight bit parameters of the second group [8, 72, 40, 104, 24, 88, 56, 120] are input in parallel to the eighth group of the second single第 The first address B0 of the memory cells Bank_B0~Bank_B7, and the first memory cell of each first memory cell and each second memory cell is displaced Initial memory unit.

接續,寫入交換器14將平行管線式快速傅立葉轉換處理器20之反向位元資料C2[4,68,36,100,20,84,52,116]及C3[12,76,44,108,28,92,60,124]以CROT(4)的排列方式進行交換;其交換模式是將第三組的八個位元參數自初始單埠記憶體單元位移四個單埠記憶體單元後,八個位元參數排列為[20,84,52,116,4,68,36,100],並輸入至八組第一單埠記憶體單元Bank_A0~Bank_A7中的第二個位址A1;將第四組的八個位元參數自初始單埠記憶體單元位移四個單埠記憶體單元後,八個位元參數排列為[28,92,60,124,12,76,44,108],並輸入至八組第二單埠記憶體單元Bank_B0~Bank_B7中的第二個位址B1。In the continuation, the write switch 14 will reverse the bit data C2 [4, 68, 36, 100, 20, 84, 52, 116] and C3 [12, 76, 44, 108, 28, 92 of the parallel pipelined fast Fourier transform processor 20, 60,124] exchanged in CROT(4) arrangement; the exchange mode is to arrange eight bit parameters of the third group from the initial memory cell unit by four memory cells. It is [20, 84, 52, 116, 4, 68, 36, 100] and is input to the second address A1 of the eight groups of first memory cells Bank_A0~Bank_A7; the eight bit parameters of the fourth group are After the initial memory cell is shifted by four memory cells, the eight bit parameters are arranged as [28, 92, 60, 124, 12, 76, 44, 108] and input to the eight groups of second memory cells Bank_B0. The second address B1 in ~Bank_B7.

接續,寫入交換器14將平行管線式快速傅立葉轉換處理器20之反向位元資料C4[2,66,34,98,18,82,50,114]及C5[10,74,42,106,26,90,58,122] 以CROT(2)的排列方式進行交換;其交換模式是將第五組的八個位元參數自初始單埠記憶體單元位移兩個單埠記憶體單元後,八個位元參數排列為[50,114,2,66,34,98,18,82],並輸入至八組第一單埠記憶體單元Bank_A0~Bank_A7中的第三個位址A2;將第六組的八個位元參數自初始單埠記憶體單元位移兩個單埠記憶體單元後,八個位元參數排列為[58,122,10,74,42,106,26,90],並輸入至八組第二單埠記憶體單元Bank_B0~Bank_B7中的第三個位址B2。In the continuation, the write switch 14 will reverse the bit data C4 [2, 66, 34, 98, 18, 82, 50, 114] and C5 [10, 74, 42, 106, 26 of the parallel pipelined fast Fourier transform processor 20, 90,58,122] Exchanged in CROT(2) arrangement; the exchange mode is to shift the eight bit parameters of the fifth group from the initial memory cell to two memory cells, and the eight bit parameters are arranged as [ 50, 114, 2, 66, 34, 98, 18, 82], and input to the third address A2 of the eight groups of first memory cells Bank_A0~Bank_A7; the eight bit parameters of the sixth group are After the initial memory cell is shifted by two memory cells, the eight bit parameters are arranged as [58, 122, 10, 74, 42, 106, 26, 90] and input to the eight groups of second memory cells Bank_B0. The third address in the ~Bank_B7 is B2.

接續,寫入交換器14將平行管線式快速傅立葉轉換處理器20之反向位元資料C6[6,70,38,102,22,86,54,118]及C7[14,78,46,110,30,94,62,126]以CROT(6)的排列方式進行交換;其交換模式是將第七組的八個位元參數自初始單埠記憶體單元位移六個單埠記憶體單元後,八個位元參數排列為[38,102,22,86,54,118,6,70],並輸入至八組第一單埠記憶體單元Bank_A0~Bank_A7中的第四個位址A3;將第八組的八個位元參數自初始單埠記憶體單元位移六個單埠記憶體單元後,八個位元參數排列為[46,110,30,94,62,126,14,78],並輸入至八組第二單埠記憶體單元Bank_B0~Bank_B7中的第四個位址B3。In the continuation, the write switch 14 will reverse the bit data C6 [6, 70, 38, 102, 22, 86, 54, 118] and C7 [14, 78, 46, 110, 30, 94 of the parallel pipelined fast Fourier transform processor 20, 62,126] exchanged in CROT(6) arrangement; the exchange mode is to arrange eight bit parameters of the seventh group from the initial memory unit by six memory cells. [38,102,22,86,54,118,6,70], and input to the fourth address A3 of the eight groups of first memory cells Bank_A0~Bank_A7; the eight bit parameters of the eighth group are After the initial memory cell is shifted by six memory cells, the eight bit parameters are arranged as [46, 110, 30, 94, 62, 126, 14, 78] and input to the eight groups of second memory cells Bank_B0. The fourth address B3 in ~Bank_B7.

接續,寫入交換器14將平行管線式快速傅立葉轉換處理器20之反向位元資料C8[1,65,33,97,17,81,49,113]及C9[9,73,41,105,25,89,57,121]以CROT(1)的排列方式進行交換;其交換模式是將第九組的八個位元參數自初始單埠記憶體單元位移一個單埠記憶體單元後,八個位元參數排列為[113,1,65,33,97,17,81,49],並輸入至八組第一單埠記憶體單元Bank_A0~Bank_A7中的第五個位址A4;將第十組的八個位元參數自初始單埠記憶體 單元位移一個單埠記憶體單元後,八個位元參數排列為[121,9,73,41,105,25,89,57],並輸入至八組第二單埠記憶體單元Bank_B0~Bank_B7中的第五個位址B4。In the continuation, the write switch 14 will reverse the bit data C8 [1, 65, 33, 97, 17, 81, 49, 113] and C9 [9, 73, 41, 105, 25 of the parallel pipeline type fast Fourier transform processor 20, 89,57,121] exchanges in CROT(1) arrangement; the exchange mode is to shift the eight bit parameters of the ninth group from the initial memory cell to one memory cell, and eight bit parameters. Arranged as [113,1,65,33,97,17,81,49] and input to the fifth address A4 of the eight groups of first memory cells Bank_A0~Bank_A7; One bit parameter from initial memory After the cell is shifted by one memory cell, the eight bit parameters are arranged as [121, 9, 73, 41, 105, 25, 89, 57] and input into the eight groups of second memory cells Bank_B0~Bank_B7. The fifth address is B4.

接續,寫入交換器14將平行管線式快速傅立葉轉換處理器20之反向位元資料C10[5,69,37,101,21,85,53,117]及C11[13,77,45,109,29,93,61,125]以CROT(5)的排列方式進行交換;其交換模式是將第十一組的八個位元參數自初始單埠記憶體單元位移五個單埠記憶體單元後,八個位元參數排列為[101,21,85,53,117,5,69,37],並輸入至八組第一單埠記憶體單元Bank_A0~Bank_A7中的第六個位址A5;將第十二組的八個位元參數自初始單埠記憶體單元位移五個單埠記憶體單元後,八個位元參數排列為[109,29,93,61,125,13,77,45],並輸入至八組第二單埠記憶體單元Bank_B0~Bank_B7中的第六個位址B5。In the continuation, the write switch 14 will reverse the bit data C10 [5, 69, 37, 101, 21, 85, 53, 117] and C11 [13, 77, 45, 109, 29, 93 of the parallel pipelined fast Fourier transform processor 20, 61,125] exchanges in CROT(5) arrangement; the exchange mode is to shift the eight bit parameters of the eleventh group from the initial memory cell to five memory cells, and eight bit parameters. Arranged as [101, 21, 85, 53, 117, 5, 69, 37] and input to the sixth address A5 of the eight groups of first memory cells Bank_A0~Bank_A7; eight of the twelveth group After the bit parameter is shifted from the initial memory cell by five memory cells, the eight bit parameters are arranged as [109, 29, 93, 61, 125, 13, 77, 45] and input to the eighth group.第六 The sixth address B5 in the memory cells Bank_B0~Bank_B7.

接續,寫入交換器14將平行管線式快速傅立葉轉換處理器20之反向位元資料C12[3,67,35,99,19,83,51,115]及C13[11,75,43,107,27,91,59,123]以CROT(3)的排列方式進行交換;其交換模式是將第十三組的八個位元參數自初始單埠記憶體單元位移三個單埠記憶體單元後,八個位元參數排列為[83,51,115,3,67,35,99,19],並輸入至八組第一單埠記憶體單元Bank_A0~Bank_A7中的第七個位址A6;將第十四組的八個位元參數自初始單埠記憶體單元位移三個單埠記憶體單元後,八個位元參數排列為[91,59,123,11,75,43,107,27],並輸入至八組第二單埠記憶體單元Bank_B0~Bank_B7中的第七個位址B6。In the continuation, the write switch 14 converts the inverse bit data C12 [3, 67, 35, 99, 19, 83, 51, 115] and C13 [11, 75, 43, 107, 27 of the parallel pipelined fast Fourier transform processor 20, 91, 59, 123] exchanged in CROT (3) arrangement; the exchange mode is to shift the eight bit parameters of the thirteenth group from the initial memory cell to three memory cells, eight bits The meta-parameters are arranged as [83, 51, 115, 3, 67, 35, 99, 19] and input to the seventh address A6 of the eight sets of first memory cells Bank_A0~Bank_A7; After the eight bit parameters are shifted from the initial memory cell to three memory cells, the eight bit parameters are arranged as [91, 59, 123, 11, 75, 43, 107, 27] and input to the eighth group.第七 The seventh address B6 in the memory cells Bank_B0~Bank_B7.

接續,寫入交換器14將平行管線式快速傅立葉轉換處理器20 之反向位元資料C14[7,71,39,103,23,87,55,119]及C16[15,79,47,111,31,95,63,127]以CROT(7)的排列方式進行交換;其交換模式是將第十五組的八個位元參數自初始單埠記憶體單元位移七個單埠記憶體單元後,八個位元參數排列為[71,39,103,23,87,55,119,7],並輸入至八組第一單埠記憶體單元Bank_A0~Bank_A7中的第八個位址A7;將第十六組的八個位元參數自初始單埠記憶體單元位移七個單埠記憶體單元後,八個位元參數排列為[79,47,111,31,95,63,127,15],並輸入至八組第二單埠記憶體單元Bank_B0~Bank_B7中的第八個位址B7。Continuing, the write switch 14 will parallel the pipeline type fast Fourier transform processor 20 The reverse bit data C14[7,71,39,103,23,87,55,119] and C16[15,79,47,111,31,95,63,127] are exchanged in the CROT(7) arrangement; the exchange mode is After shifting the eight bit parameters of the fifteenth group from the initial memory cell to seven memory cells, the eight bit parameters are arranged as [71, 39, 103, 23, 87, 55, 119, 7], and Input to the eighth address A7 of the first group of memory cells Bank_A0~Bank_A7; and shift the eight bit parameters of the sixteenth group from the initial memory cell to seven memory cells The eight bit parameters are arranged as [79, 47, 111, 31, 95, 63, 127, 15] and input to the eighth address B7 of the eight sets of second memory cells Bank_B0~Bank_B7.

其中,寫入交換器14將128點的反向位元資料全部交換為寫入位元資料,並交替寫入第一記憶體模組10及第二記憶體模組12中之後,由多工器選擇輸出第一記憶體模組10及第二記憶體模組12中的位元資料,例如,選擇將Bank_A0~Bank_A7中的位元資料0~7輸出(如第2D圖的虛框所示),再由讀取交換器18將0~7位元資料交換為自然順序由0至7排列之輸出位元資料;接著交替選擇將Bank_B0~Bank_B7中的位元資料8~15輸出(如第2D圖的虛框所示),再由讀取交換器18將8~15位元資料交換為自然順序由8至15排列之輸出位元資料。再如第2E圖,為讀取交換器18將128位元資料交換為自然順序輸出的輸出位元資料,並呈現於另兩組記憶體模組GroupA、GroupB中,例如GroupA中的Bank_A0~Bank_A7的第一個位址會依序排列為0~7的位元,而GroupB中的Bank_B0~Bank_B7的第一個位址會依序排列為8~15的位元;Bank_A0~Bank_A7第二個位址會依序16~23的位元,而GroupB中的Bank_B0~Bank_B7的第二個位址會依序排列為24~31的位元,以此類推依序將128點位元資料全部以自然順序讀出。此時,當將 GroupA中的0~7位元輸出至下一級的功能端,空出的位址即能由寫入交換器14繼續寫入下一筆128點的反向位元資料。由於只要空出位址即能寫入下一筆位元資料,因此第一記憶體模組10寫入時,第二記憶體模組12則讀出,以交替讀/寫方式就能夠使平行管線式快速傅利葉轉換處理器之位元反向輸出轉變成自然順序、連續、無延遲之輸出,進而提高輸出效率。The write switch 14 exchanges all the 128-bit reverse bit data into the write bit data, and alternately writes the first memory module 10 and the second memory module 12, and then the multiplexer The device selects and outputs the bit data in the first memory module 10 and the second memory module 12, for example, selects to output the bit data 0~7 in Bank_A0~Bank_A7 (as shown by the dashed box in the 2D picture) Then, the read switch 18 exchanges the 0~7 bit data into the output bit data in the natural order from 0 to 7; then alternately selects the bit data 8~15 in Bank_B0~Bank_B7 to output (such as As shown by the dashed box of the 2D map, the read switch 18 then exchanges 8 to 15 bits of data into output bit data in natural order from 8 to 15. Further, as shown in FIG. 2E, the 128-bit data is exchanged for the natural-sequential output output bit data for the read switch 18, and is presented in the other two groups of memory modules GroupA and GroupB, for example, Bank_A0~Bank_A7 in GroupA. The first address will be sequentially arranged as 0~7 bits, and the first address of Bank_B0~Bank_B7 in GroupB will be sequentially arranged as 8~15 bits; Bank_A0~Bank_A7 will be the second bit. The address will be in the order of 16~23 bits, and the second address of Bank_B0~Bank_B7 in GroupB will be arranged in the order of 24~31 bits, and the 128-bit data will be all natural in this order. Read out sequentially. At this time, when The 0~7 bits in GroupA are output to the function end of the next stage, and the vacated address can be written by the write switch 14 to write the next 128-bit reverse bit data. Since the next bit data can be written as long as the address is vacated, the second memory module 12 is read when the first memory module 10 is written, and the parallel pipeline can be alternately read/written. The bit-inverted output of the fast Fourier transform processor transforms into a natural sequential, continuous, delay-free output, which in turn increases output efficiency.

接著,以α 為奇數為例說明,請同時參閱第1圖、第3圖及4A~4E圖,第4A~4E圖為本發明以256位元長度交換為自然順序輸出之實施例示意圖。當傅立葉轉換之位元長度N為256,其經過m為8倍平行的平行管線式快速傅立葉轉換處理器20後輸出256點的反向位元資料,由於α 係為奇數,故實施第二種交換模式;詳言之,寫入交換器14首次先將寫入位元資料寫入於第一記憶體模組10一次,接著以各寫入兩次之交替方式將寫入位元資料寫入第二記憶體模組12及第一記憶體模組10中,讀出交換器18係以各讀出一次之交替方式自第一記憶體模組10及第二記憶體模組12中讀出位元資料。如第4A所示,為本發明之位元長度256點經8倍平行管線式快速傅立葉轉換處理器20後之輸出示意圖。寫入交換器14係根據預設排列方式選定之控制訊號進行交換該反向位元資料,傅立葉轉換之位元長度為256,其分為32組八個位元參數,例如32組編號為C0~C31。再如第4B圖所示,為本發明之256點的反向位元資料經寫入交換器交換後之輸出;寫入交換器14係將反向位元資料中每四組的八個位元參數以八種排列方式,例如八種排列的編號為CROT(0)~CROT(7),而每一種排列方式係遵循第3圖的CROT(0)~CROT(7)八種寫入交換模式進行交換,由寫入交換器14將反向位元資料交換為預設排列方式之寫入位元資料,並將寫入位元資料交替寫入 第一記憶體模組10及第二記憶體模組12中。再如第4C、第4D及4E圖所示,第4C、第4D分別為本發明之讀寫排程示意圖及本發明儲存寫入位元資料於至記憶體中之示意圖;第一記憶體模組10具有m個第一單埠記憶體單元,若m為8,則8組第一單埠記憶體單元為Bank_A0~Bank_A7,由於每一第一單埠記憶體單元儲存N/2m筆位元資料,因此每一個第一單埠記憶體單元具有16個儲存位址。同理,第二記憶體模組12具有m個第二單埠記憶體單元,若m為8,則8個第二單埠記憶體單元為Bank_B0~Bank_B7,由於每一第二單埠記憶體單元儲存N/2m筆位元資料,因此每一個第二單埠記憶體單元具有16個儲存位址。Next, taking α as an odd number as an example, please refer to FIG. 1 , FIG. 3 , and FIG. 4A to FIG. 4E simultaneously. FIGS. 4A to 4E are schematic diagrams showing an embodiment in which the 256-bit length is exchanged as a natural sequential output. When the length N of the Fourier transform bit is 256, the output is 256 points of reverse bit data after the parallel pipeline type fast Fourier transform processor 20 whose m is 8 times parallel. Since the α system is an odd number, the second type is implemented. Switch mode; in detail, the write switch 14 first writes the write bit data to the first memory module 10 once, and then writes the write bit data in an alternate manner of each write twice. In the second memory module 12 and the first memory module 10, the read switch 18 is read out from the first memory module 10 and the second memory module 12 in an alternate manner. Bit data. As shown in FIG. 4A, it is a schematic diagram of the output of the present invention having a bit length of 256 points and an 8-time parallel pipeline type fast Fourier transform processor 20. The write switch 14 exchanges the reverse bit data according to the control signal selected by the preset arrangement. The length of the Fourier transform bit is 256, which is divided into 32 groups of eight bit parameters, for example, 32 groups are numbered as C0. ~C31. As shown in FIG. 4B, the 256-point reverse bit data of the present invention is outputted by the switch exchange; the write switch 14 is the eight bits of each of the four sets of reverse bit data. The meta-parameters are arranged in eight ways, for example, the eight arrays are numbered CROT(0)~CROT(7), and each arrangement follows the CROT(0)~CROT(7) eight write exchanges of Figure 3. The mode is exchanged, and the reverse bit data is exchanged by the write switch 14 into the write bit data of the preset arrangement mode, and the write bit data is alternately written into the first memory module 10 and the second memory. In the body module 12. 4C, 4D, and 4E, FIG. 4C and FIG. 4D are respectively a schematic diagram of reading and writing scheduling of the present invention, and a schematic diagram of storing the writing bit data in the memory in the memory; the first memory model Group 10 has m first memory cells. If m is 8, then the first memory cells of the 8 groups are Bank_A0~Bank_A7, and each N^m pen cell is stored for each first memory cell. Data, so each first memory unit has 16 storage addresses. Similarly, the second memory module 12 has m second memory cells. If m is 8, the eight second memory cells are Bank_B0~Bank_B7, because each second memory is The unit stores N/2m pen bit data, so each second memory cell has 16 storage addresses.

暸解上述每一個元件之間的相對關係之後,接下來以實際運作為例詳細說明,寫入交換器14係將反向位元資料以八種排列方式,其順序為CROT(0)、CROT(4)、CROT(2)、CROT(6)、CROT(1)、CROT(5)、CROT(3)、CROT(7),如後說明。After understanding the relative relationship between each of the above components, the actual operation will be described in detail as an example. The write switch 14 will arrange the reverse bit data in eight ways, in the order of CROT (0), CROT ( 4), CROT (2), CROT (6), CROT (1), CROT (5), CROT (3), CROT (7), as will be described later.

首先,寫入交換器14將平行管線式快速傅立葉轉換處理器20之反向位元資料C0[0,128,64,192,32,160,96,224]、C1[16,144,80,208,48,176,112,240]、C2[8,136,72,200,40,168,104,232]及C3[24,152,88,216,56,184,120,248]以CROT(0)的排列方式進行交換,其交換模式係將第一組的八個位元參數[0,128,64,192,32,160,96,224]及第四組的八個位元參數[24,152,88,216,56,184,120,248]分別平行輸入至八組第一單埠記憶體單元Bank_A0~Bank_A7中的第一個位址A0及第四個位址A1,第二組的八個位元參數[8,136,72,200,40,168,104,232]及第三組的八個位元參數[24,152,88,216,56,184,120,248]分別平行輸入至八組第二單埠記憶體單元 Bank_B0~Bank_B7中的第一個位址B0及第二個位址B1,且每一第一單埠記憶體單元及每一第二單埠記憶體單元中的第一個單埠記憶體單元係作為位移的初始單埠記憶體單元。First, the write switch 14 will reverse the bit data C0[0,128,64,192,32,160,96,224], C1[16,144,80,208,48,176,112,240], C2[8,136,72,200,40,168,104,232 of the parallel pipelined fast Fourier transform processor 20. ] and C3 [24, 152, 88, 216, 56, 184, 120, 248] are exchanged in CROT (0) arrangement, the exchange mode is the first group of eight bit parameters [0, 128, 64, 192, 32, 160, 96, 224] and the fourth group of eight The bit parameters [24, 152, 88, 216, 56, 184, 120, 248] are input in parallel to the first address A0 and the fourth address A1 of the first group of memory cells Bank_A0~Bank_A7, respectively, and the eight bits of the second group. The meta-parameters [8, 136, 72, 200, 40, 168, 104, 232] and the third set of eight bit parameters [24, 152, 88, 216, 56, 184, 120, 248] are input in parallel to the eight sets of second memory cells, respectively. The first address B0 and the second address B1 in Bank_B0~Bank_B7, and the first memory cell in each first memory cell and each second memory cell As the initial memory unit of the displacement.

接續,寫入交換器14將平行管線式快速傅立葉轉換處理器20之反向位元資料C4[4,132,68,196,36,164,100,228]、C5[20,148,84,212,52,180,116,244]、C6[12,140,76,204,44,172,108,236]及C7[28,156,92,220,60,188,124,252]以CROT(4)的排列方式進行交換;其交換模式是將第五組及第八組的八個位元參數自初始單埠記憶體單元位移四個單埠記憶體單元後,第五組的八個位元參數排列為[36,164,100,228,4,132,68,196],並輸入至八組第一單埠記憶體單元Bank_A0~Bank_A7中的第三個位址A2;第八組的八個位元參數排列為[60,188,124,252,28,156,92,220],並輸入至八組第一單埠記憶體單元Bank_A0~Bank_A7中的第四個位址A3。將第六組及第七組的八個位元參數自初始單埠記憶體單元位移四個單埠記憶體單元後,第六組的八個位元參數排列為[52,180,116,244,20,148,84,212],並輸入至八組第二單埠記憶體單元Bank_B0~Bank_B7中的第三個位址B2;第七組的八個位元參數排列為[44,172,108,236,12,140,76,204],並輸入至八組第二單埠記憶體單元Bank_B0~Bank_B7中的第四個位址B3。In the continuation, the write switch 14 will reverse the bit data C4 [4, 132, 68, 196, 36, 164, 100, 228], C5 [20, 148, 84, 212, 52, 180, 116, 244], C6 [12, 140, 76, 204, 44, 172, 108, 236] of the parallel pipelined fast Fourier transform processor 20 and C7[28,156,92,220,60,188,124,252] is exchanged in CROT(4) arrangement; the exchange mode is to shift the eight bit parameters of the fifth and eighth groups from the initial memory cell by four memory. After the body unit, the eight bit parameters of the fifth group are arranged as [36, 164, 100, 228, 4, 132, 68, 196] and input to the third address A2 of the eight groups of first memory cells Bank_A0~Bank_A7; The eight bit parameters are arranged as [60, 188, 124, 252, 28, 156, 92, 220] and input to the fourth address A3 of the eight sets of first memory cells Bank_A0~Bank_A7. After shifting the eight bit parameters of the sixth group and the seventh group from the initial memory cell to four memory cells, the eight bit parameters of the sixth group are arranged as [52, 180, 116, 244, 20, 148, 84, 212], And input to the third address B2 of the eight groups of the second memory cells Bank_B0~Bank_B7; the eight bit parameters of the seventh group are arranged as [44, 172, 108, 236, 12, 140, 76, 204] and input to the eighth group of the second group.第 The fourth address B3 in the memory cells Bank_B0~Bank_B7.

接續,寫入交換器14將平行管線式快速傅立葉轉換處理器20之反向位元資料C8[2,130,66,194,34,162,98,226]、C9[18,146,82,210,50,178,114,242]、C10[10,138,74,202,42,170,106,234]及C11[26,154,90,218,58,186,122,250]以CROT(2)的排列方式進行交換;其交 換模式是將第九組及第十二組的八個位元參數自初始單埠記憶體單元位移兩個單埠記憶體單元後,第九組的八個位元參數排列為[98,226,2,130,66,194,34,162],並輸入至八組第一單埠記憶體單元Bank_A0~Bank_A7中的第五個位址A4;第十二組的八個位元參數排列為[122,250,26,154,90,218,58,186],並輸入至八組第一單埠記憶體單元Bank_A0~Bank_A7中的第六個位址A5。將第十組及第十一組的八個位元參數自初始單埠記憶體單元位移兩個單埠記憶體單元後,第十組的八個位元參數排列為[114,242,18,146,82,210,50,178],並輸入至八組第二單埠記憶體單元Bank_B0~Bank_B7中的第五個位址B4;第十一組的八個位元參數排列為[106,234,10,138,74,202,42,170],並輸入至八組第二單埠記憶體單元Bank_B0~Bank_B7中的第六個位址B5。Next, the write switch 14 will reverse the bit data C8 [2, 130, 66, 194, 34, 162, 98, 226], C9 [18, 146, 82, 210, 50, 178, 114, 242], C10 [10, 138, 74, 202, 42, 170, 106, 234 of the parallel pipelined fast Fourier transform processor 20. ] and C11 [26, 154, 90, 218, 58, 186, 122, 250] exchanged in CROT (2) arrangement; The change mode is to shift the eight bit parameters of the ninth group and the twelfth group from the initial memory cell to two memory cells, and the eight bit parameters of the ninth group are arranged as [98, 226, 2, 130 , 66, 194, 34, 162], and input to the fifth address A4 of the eight groups of first memory cells Bank_A0~Bank_A7; the eight bit parameters of the twelfth group are arranged as [122, 250, 26, 154, 90, 218, 58, 186 ], and input to the sixth address A5 of the eight groups of first memory cells Bank_A0~Bank_A7. After shifting the eight bit parameters of the tenth and eleventh groups from the initial memory cell to two memory cells, the eight bit parameters of the tenth group are arranged as [114, 242, 18, 146, 82, 210, 50, 178], and input to the fifth address B4 of the eight sets of second memory cells Bank_B0~Bank_B7; the eight bit parameters of the eleventh group are arranged as [106, 234, 10, 138, 74, 202, 42, 170], and Input to the sixth address B5 of the eight sets of second memory cells Bank_B0~Bank_B7.

接續,寫入交換器14將平行管線式快速傅立葉轉換處理器20之反向位元資料C12[6,134,70,198,38,166,102,230]、C13[22,150,86,214,54,182,118,246]、C14[14,142,78,206,46,174,110,238]及C15[30,158,94,222,62,190,126,254]以CROT(6)的排列方式進行交換;其交換模式是將第十三組及第十六組的八個位元參數自初始單埠記憶體單元位移六個單埠記憶體單元後,第十三組的八個位元參數排列為[70,198,38,166,102,230,6,134],並輸入至八組第一單埠記憶體單元Bank_A0~Bank_A7中的第七個位址A6;第十六組的八個位元參數排列為[94,222,62,190,126,254,30,158],並輸入至八組第一單埠記憶體單元Bank_A0~Bank_A7中的第八個位址A7。將第十四組及第十五組的八個位元參數自初始單埠記憶體單元位移六個單埠記憶體單元後,第十四組的八個 位元參數排列為[86,214,54,182,118,246,22,150],並輸入至八組第二單埠記憶體單元Bank_B0~Bank_B7中的第七個位址B6;第十五組的八個位元參數排列為[78,206,46,174,110,238,14,142],並輸入至八組第二單埠記憶體單元Bank_B0~Bank_B7中的第八個位址B7。In the continuation, the write switch 14 will reverse the bit data C12 [6, 134, 70, 198, 38, 166, 102, 230], C13 [22, 150, 86, 214, 54, 182, 118, 246], C14 [14, 142, 78, 206, 46, 174, 110, 238] of the parallel pipelined fast Fourier transform processor 20 and C15[30,158,94,222,62,190,126,254] is exchanged in the arrangement of CROT(6); the exchange mode is to shift the eight bit parameters of the thirteenth group and the sixteenth group from the initial memory cell unit by six singles. After the memory unit, the eight bit parameters of the thirteenth group are arranged as [70, 198, 38, 166, 102, 230, 6, 134], and are input to the seventh address A6 of the eight groups of first memory cells Bank_A0~Bank_A7; The eight bit parameters of the sixteenth group are arranged as [94, 222, 62, 190, 126, 254, 30, 158] and input to the eighth address A7 of the eight sets of first memory cells Bank_A0~Bank_A7. After shifting the eight bit parameters of the fourteenth and fifteenth groups from the initial memory cell to six memory cells, the eightteenth group of eight The bit parameters are arranged as [86, 214, 54, 182, 118, 246, 22, 150] and input to the seventh address B6 of the eight sets of second memory cells Bank_B0~Bank_B7; the eight bit parameters of the fifteenth group are arranged as [ 78, 206, 46, 174, 110, 238, 14, 142], and input to the eighth address B7 of the eight sets of second memory cells Bank_B0~Bank_B7.

接續,寫入交換器14將平行管線式快速傅立葉轉換處理器20之反向位元資料C16[1,129,65,193,33,161,97,225]、C17[17,145,81,209,49,177,113,241]、C18[9,137,73,201,41,169,105,233]及C19[25,153,89,217,57,185,121,249]以CROT(1)的排列方式進行交換;其交換模式是將第十七組及第二十組的八個位元參數自初始單埠記憶體單元位移一個單埠記憶體單元後,第十七組的八個位元參數排列為[225,1,129,65,193,33,161,97],並輸入至八組第一單埠記憶體單元Bank_A0~Bank_A7中的第九個位址A8;第二十組的八個位元參數排列為[249,25,153,89,217,57,185,121],並輸入至八組第一單埠記憶體單元Bank_A0~Bank_A7中的第十個位址A9。將第十八組及第十九組的八個位元參數自初始單埠記憶體單元位移一個單埠記憶體單元後,第十八組的八個位元參數排列為[241,17,145,81,209,49,177,113],並輸入至八組第二單埠記憶體單元Bank_B0~Bank_B7中的第九個位址B8;第十九組的八個位元參數排列為[233,9,137,73,201,41,169,105],並輸入至八組第二單埠記憶體單元Bank_B0~Bank_B7中的第十個位址B9。Next, the write switch 14 will reverse the bit data C16 [1, 129, 65, 193, 33, 161, 97, 225], C17 [17, 145, 81, 209, 49, 177, 113, 241], C18 [9, 137, 73, 201, 41, 169, 105, 233 of the parallel pipeline type fast Fourier transform processor 20. And C19 [25, 153, 89, 217, 57, 185, 121, 249] are exchanged in the arrangement of CROT (1); the exchange mode is to shift the eight bit parameters of the seventeenth and twentieth groups from the initial memory cell. After the memory unit, the eight-bit parameters of the seventeenth group are arranged as [225, 1, 129, 65, 193, 33, 161, 97] and input to the ninth of the eight groups of first memory cells Bank_A0~Bank_A7. Address A8; the octet parameters of the twentieth group are arranged as [249, 25, 153, 89, 217, 57, 185, 121] and input to the tenth address A9 of the eight groups of first memory cells Bank_A0~Bank_A7. . After shifting the eight bit parameters of the eighteenth and nineteenth groups from the initial memory cell to one memory cell, the eight bit parameters of the eighteenth group are arranged as [241, 17, 145, 81, 209 , 49, 177, 113], and input to the ninth address B8 of the eight sets of second memory cells Bank_B0~Bank_B7; the eight bit parameters of the nineteenth group are arranged as [233, 9, 137, 73, 201, 41, 169, 105], And input to the tenth address B9 of the eight groups of the second memory cells Bank_B0~Bank_B7.

接續,寫入交換器14將平行管線式快速傅立葉轉換處理器20之反向位元資料C20[5,133,69,197,37,165,101,229]、C21[21,149,85,213,53,181,117,245]、C22[13,141,77,205,45,173,109,237]及 C23[29,157,93,221,61,189,125,253]以CROT(5)的排列方式進行交換;其交換模式是將第二十一組及第二十四組的八個位元參數自初始單埠記憶體單元位移五個單埠記憶體單元後,第二十一組的八個位元參數排列為[197,37,165,101,229,5,133,69],並輸入至八組第一單埠記憶體單元Bank_A0~Bank_A7中的第十一個位址A10;第二十四組的八個位元參數排列為[221,61,189,125,253,29,157,93],並輸入至八組第一單埠記憶體單元Bank_A0~Bank_A7中的第十二個位址A11。將第二十二組及第二十三組的八個位元參數自初始單埠記憶體單元位移五個單埠記憶體單元後,第二十二組的八個位元參數排列為[213,53,181,117,245,21,149,85],並輸入至八組第二單埠記憶體單元Bank_B0~Bank_B7中的第十一個位址B10;第二十三組的八個位元參數排列為[205,45,173,109,237,13,141,77],並輸入至八組第二單埠記憶體單元Bank_B0~Bank_B7中的第十二個位址B11。In the continuation, the write switch 14 converts the inverse bit data C20 [5, 133, 69, 197, 37, 165, 101, 229], C21 [21, 149, 85, 213, 53, 181, 117, 245], C22 [13, 141, 77, 205, 45, 173, 109, 237] of the parallel pipeline type fast Fourier transform processor 20 and C23[29,157,93,221,61,189,125,253] is exchanged in the arrangement of CROT(5); the exchange mode is to shift the eight bit parameters of the twenty-first group and the twenty-fourth group from the initial memory unit by five. After the memory unit, the eight bit parameters of the twenty-first group are arranged as [197, 37, 165, 101, 229, 5, 133, 69] and input to the tenth of the eight groups of the first memory cells Bank_A0~Bank_A7. One address A10; the eight bit parameters of the twenty-fourth group are arranged as [221, 61, 189, 125, 253, 29, 157, 93] and input to the twelfth of the eight groups of first memory cells Bank_A0~Bank_A7 Address A11. After the eight bit parameters of the twenty-second group and the twenty-third group are shifted from the initial memory cell unit by five memory cells, the eight-bit parameter of the twenty-second group is arranged as [213] , 53, 181, 117, 245, 21, 149, 85], and input to the eleventh address B10 of the eight sets of second memory cells Bank_B0~Bank_B7; the eight bit parameters of the twenty-third group are arranged as [205, 45, 173, 109, 237 , 13, 141, 77], and input to the twelfth address B11 of the eight sets of second memory cells Bank_B0~Bank_B7.

接續,寫入交換器14將平行管線式快速傅立葉轉換處理器20之反向位元資料C24[3,131,67,195,35,163,99,227]、C25[19,147,83,211,51,179,115,243]、C26[11,139,75,203,43,171,107,235]及C27[27,155,91,219,59,187,123,251]以CROT(3)的排列方式進行交換;其交換模式是將第二十五組及第二十八組的八個位元參數自初始單埠記憶體單元位移三個單埠記憶體單元後,第二十五組的八個位元參數排列為[163,99,227,3,131,67,195,35],並輸入至八組第一單埠記憶體單元Bank_A0~Bank_A7中的第十三個位址A12;第二十八組的八個位元參數排列為[187,123,251,27,155,91,219,59],並輸入至八組第一單埠記憶體單元Bank_A0~Bank_A7中的第十四個位址A13。將第二十六組及第二十七組的 八個位元參數自初始單埠記憶體單元位移三個單埠記憶體單元後,第二十六組的八個位元參數排列為[179,115,243,19,147,83,211,51],並輸入至八組第二單埠記憶體單元Bank_B0~Bank_B7中的第十三個位址B12;第二十七組的八個位元參數排列為[171,107,235,11,139,75,203,43],並輸入至八組第二單埠記憶體單元Bank_B0~Bank_B7中的第十四個位址B13。In the continuation, the write switch 14 will reverse the bit data C24 [3, 131, 67, 195, 35, 163, 99, 227], C25 [19, 147, 83, 211, 51, 179, 115, 243], C26 [11, 139, 75, 203, 43, 171, 107, 235 of the parallel pipeline type fast Fourier transform processor 20. ] and C27 [27, 155, 91, 219, 59, 187, 123, 251] are exchanged in CROT (3) arrangement; the exchange mode is to convert the eight bit parameters of the twenty-fifth group and the twenty-eighth group from the initial memory unit. After shifting three memory cells, the eight bit parameters of the twenty-fifth group are arranged as [163, 99, 227, 3, 131, 67, 195, 35] and input to the eight groups of first memory cells Bank_A0~Bank_A7. The thirteenth address in the A12; the eight bit parameters of the twenty-eighth group are arranged as [187, 123, 251, 27, 155, 91, 219, 59] and input to the eight groups of the first memory cells Bank_A0~Bank_A7 The fourteenth address is A13. The twenty-sixth group and the twenty-seventh group After the eight bit parameters are shifted from the initial memory cell by three memory cells, the eight bit parameters of the twenty-sixth group are arranged as [179, 115, 243, 19, 147, 83, 211, 51] and input to eight groups. The thirteenth address B12 of the second memory cell Bank_B0~Bank_B7; the eight bit parameters of the twenty-seventh group are arranged as [171,107,235,11,139,75,203,43] and input to the eighth group of the second group. The fourteenth address B13 in the memory cells Bank_B0~Bank_B7.

接續,寫入交換器14將平行管線式快速傅立葉轉換處理器20之反向位元資料C28[7,135,71,199,39,167,103,231]、C29[23,151,87,215,55,183,119,247]、C30[15,143,79,207,47,175,111,239]及C31[31,159,95,223,63,191,127,255]以CROT(7)的排列方式進行交換;其交換模式是將第二十九組及第三十二組的八個位元參數自初始單埠記憶體單元位移七個單埠記憶體單元後,第二十九組的八個位元參數排列為[135,71,199,39,167,103,231,7],並輸入至八組第一單埠記憶體單元Bank_A0~Bank_A7中的第十五個位址A14;第三十二組的八個位元參數排列為[159,95,223,63,191,127,255,31],並輸入至八組第一單埠記憶體單元Bank_A0~Bank_A7中的第十六個位址A15。將第三十組及第三十一組的八個位元參數自初始單埠記憶體單元位移七個單埠記憶體單元後,第三十組的八個位元參數排列為[151,87,215,55,183,119,247,23],並輸入至八組第二單埠記憶體單元Bank_B0~Bank_B7中的第十五個位址B14;第三十一組的八個位元參數排列為[143,79,207,47,175,111,239,15],並輸入至八組第二單埠記憶體單元Bank_B0~Bank_B7中的第十六個位址B15。In the continuation, the write switch 14 will reverse the bit data C28 [7, 135, 71, 199, 39, 167, 103, 231], C29 [23, 151, 87, 215, 55, 183, 119, 247], C30 [15, 143, 79, 207, 47, 175, 111, 239] of the parallel pipeline type fast Fourier transform processor 20 and C31[31,159,95,223,63,191,127,255] is exchanged in the arrangement of CROT(7); the exchange mode is to shift the eight bit parameters of the twenty-ninth group and the thirty-third group from the initial memory cell. After the memory unit, the eight bit parameters of the twenty-ninth group are arranged as [135, 71, 199, 39, 167, 103, 231, 7] and input to the tenth of the eight groups of first memory cells Bank_A0~Bank_A7. Five address A14; the eight-bit eight-bit parameter is arranged as [159, 95, 223, 63, 191, 127, 255, 31] and is input to the sixteenth of the eight groups of first memory cells Bank_A0~Bank_A7. Address A15. After the eight bit parameters of the thirtieth and thirty-first groups are shifted from the initial memory cell by seven memory cells, the eight bit parameters of the thirtieth group are arranged as [151, 87, 215 , 55, 183, 119, 247, 23], and input to the fifteenth address B14 of the eight sets of second memory cells Bank_B0~Bank_B7; the eight bit parameters of the thirty-first group are arranged as [143, 79, 207, 47, 175, 111, 239 , 15], and input to the sixteenth address B15 of the eight sets of second memory cells Bank_B0~Bank_B7.

其中,寫入交換器14將256點的反向位元資料全部交換為寫入位元資料,由時脈計數器(clock cycle counter)將每4個clock cycle變換一 次交換模式,並交替寫入第一記憶體模組10及第二記憶體模組12中;因此,平行管線式快速傅立葉轉換處理器20輸出的256點反向位元資料與128點反向位元資料的讀寫排程不同,主要是根據快速傅立葉轉換的長度與平行度來決定。之後,由多工器選擇輸出第一記憶體模組10及第二記憶體模組12中的位元資料,例如於反向位元資料C30時,則選擇第一記憶體模組10中的Bank_A0~Bank_A7開始將第一筆資料輸出(如第4D圖的虛框所示),再由讀取交換器18將0~7位元資料交換為自然順序由0至7排列之輸出位元資料;接著選擇將Bank_B0~Bank_B7中的位元資料8~15及16~23輸出(如第4D圖的虛框所示),再由讀取交換器18將8~15及16~23位元資料交換為自然順序由8至15排列及16至23排列之輸出位元資料;再連續自Bank_A0~Bank_A7中輸出兩筆位元資料24~31及32~39,由讀取交換器18將24~31及32~39位元資料交換為自然順序由24至31排列及32至39排列之輸出位元資料,依此規律持續進行,直到將256點的位元資料全部輸出為止。其中,寫入的交換模式與讀取的交換模式是類似的。再如第2E圖,為讀取交換器18將256位元資料交換為自然順序輸出的輸出位元資料,並呈現於另兩組記憶體模組GroupA、GroupB中,例如GroupA中的Bank_A0~Bank_A7的第一個位址會依序排列為0~7的位元,而GroupB中的Bank_B0~Bank_B7的第一個位址會依序排列為8~15的位元,第二個位址會依序排列為16~23的位元;Bank_A0~Bank_A7第二個位址會依序24~31的位元及第三個位址會依序32~39的位元,以此類推依序將256點位元資料全部以自然順序讀出。此時,當將GroupA中的0~7位元輸出至下一級的功能端,空出的位址即能由寫入交換器14繼續寫入下一筆256點的反向位元資料。由於只要空出位址即 能寫入下一筆位元資料,因此第一記憶體模組10寫入時,第二記憶體模組12則讀出,以交替讀/寫方式就能夠使平行管線式快速傅利葉轉換處理器之位元反向輸出轉變成自然順序、連續、無延遲之輸出,進而提高輸出效率。The write switch 14 exchanges all 256-point reverse bit data into write bit data, and converts every 4 clock cycles by a clock cycle counter. The secondary switching mode is alternately written into the first memory module 10 and the second memory module 12; therefore, the 256-point inverted bit data output by the parallel pipelined fast Fourier transform processor 20 is inverted with 128 points. The reading and writing schedule of the bit data is different, which is mainly determined according to the length and parallelism of the fast Fourier transform. Then, the multiplexer selects and outputs the bit data in the first memory module 10 and the second memory module 12, for example, when the reverse bit data C30 is selected, the first memory module 10 is selected. Bank_A0~Bank_A7 starts to output the first data (as shown by the dashed box in Figure 4D), and then the read switch 18 exchanges 0~7 bits of data into output bit data with natural order from 0 to 7. Then choose to output the bit data 8~15 and 16~23 in Bank_B0~Bank_B7 (as shown by the dashed box in Figure 4D), and then read the switch 18 to 8~15 and 16~23 bit data. The exchange is the output bit data of the natural order from 8 to 15 and 16 to 23; and then continuously output two bit data 24~31 and 32~39 from Bank_A0~Bank_A7, and the read switch 18 will be 24~ The 31 and 32-39-bit data exchanges are output bit data of natural order from 24 to 31 and 32 to 39, and continue to proceed according to this rule until all 256-bit bit data is output. Among them, the exchange mode of writing is similar to the exchange mode of reading. As shown in FIG. 2E, the 256-bit data is exchanged for the natural-sequential output output bit data for the read switch 18, and is presented in the other two groups of memory modules GroupA and GroupB, for example, Bank_A0~Bank_A7 in GroupA. The first address will be sequentially arranged as 0~7 bits, and the first address of Bank_B0~Bank_B7 in GroupB will be sequentially arranged as 8~15 bits, and the second address will be The order is arranged as a bit of 16~23; the second address of Bank_A0~Bank_A7 will be in the order of 24-31 bits and the third address will be in the order of 32~39 bits, and 256 in this order. The point bit data is all read in natural order. At this time, when 0~7 bits in GroupA are output to the function end of the next stage, the vacated address can be written by the write switch 14 to write the next 256-point reverse bit data. As long as the address is vacated The next bit data can be written. Therefore, when the first memory module 10 is written, the second memory module 12 is read, and the parallel pipeline type fast Fourier transform processor can be alternately read/written. The bit reverse output is converted to a natural sequential, continuous, delay-free output, which in turn increases output efficiency.

其中,本發明是針對平行管線式傅立葉轉換處理器的位元反向輸出,轉成自然順序輸出之重新調序緩衝器架構的新設計,因此以交替公式(1)計算出α 為偶數或奇數,皆能利用上述特殊的讀寫排程來支援快速傅立葉轉換的運算,並搭配兩群單埠第一記憶體模組與第二記液體模組完成自然順序輸出功能。Wherein, the present invention is directed to a new design of a reordering buffer architecture of a parallel pipelined Fourier transform processor, which is converted into a natural sequential output, so that α is an even or odd number by an alternating formula (1). The special read and write schedules can be used to support the fast Fourier transform operation, and the natural sequence output function is completed by using two groups of first memory modules and second liquid modules.

綜上所述,本發明之重新調序緩衝器的記憶體寫入/讀取位址之產生具有相當規律性,利用一個時脈計數器來產生每一個記憶體之位址,而寫入位址之產生方式是依據當時快速傅立葉轉換的位元長度以交替公式(1)計算出為是奇數或偶數而有所不同,而每一快速傅立葉轉換的位元長度的周期為N/m。假設快速傅立葉轉換的位元長度是從1開始計數,則於奇數符元期間,經由寫入交換器交換後之各路資料便會開始填入第一記憶體模組或是第二記憶體模組,而且從位址0開始依序寫入,每次寫入後,第一記憶體模組或第二記憶體模組之記憶體位址便加1。至於在每一個時脈時是寫入第一記憶體模組或是第二記憶體模組,則須依據快速傅利葉轉換之長度與平行度來決定。而於偶數符元時,此時經由寫入交換器交換後之各路資料會寫入在前一個快速傅立葉轉換的位元長度與其呈現位元反向之資料所在之位置。更進一步而言,本發明設計一種能夠執行連續性的快速傅立葉轉換之運算,且在支援連續流程快速傅立葉轉換運算時,僅需要大小為N的單埠記憶體,對於整體的系統架構設計來說,極具最小面積使用效 益,且將多倍的資料平行輸入與輸出,具有極高的運算產出功效,對於應用於有線及無線通訊系統及電子電機工程領域具有極大的應用價值。In summary, the memory write/read address of the reorder buffer of the present invention is relatively regular, and a clock counter is used to generate the address of each memory, and the address is written. The generation manner is different according to the bit length of the fast Fourier transform at that time, which is calculated as an odd or even number by the alternating formula (1), and the period of the bit length of each fast Fourier transform is N/m. Assuming that the bit length of the fast Fourier transform is counted from 1, the data exchanged by the write switch will start to fill in the first memory module or the second memory mode during the odd symbol period. The group is written sequentially from address 0. After each write, the memory address of the first memory module or the second memory module is incremented by one. As for the writing of the first memory module or the second memory module at each clock, it is determined according to the length and parallelism of the fast Fourier transform. In the case of even symbols, the data exchanged by the write switch at this time is written at the position where the length of the bit of the previous fast Fourier transform is opposite to the data of the presentation bit. Furthermore, the present invention designs a fast Fourier transform operation capable of performing continuity, and only requires a size N memory when supporting continuous flow fast Fourier transform operations, for the overall system architecture design. Very small area effect Benefits, and multiple times the data input and output in parallel, has a very high operational output, and has great application value in the field of wired and wireless communication systems and electronic motor engineering.

唯以上所述者,僅為本發明的較佳實施例而已,並非用來限定本發明實施的範圍。故凡依本發明申請範圍所述的特徵及精神所為的均等變化或修飾,均應包括於本發明的申請專利範圍內。The above is only the preferred embodiment of the present invention and is not intended to limit the scope of the present invention. Equivalent changes or modifications of the features and spirits of the inventions are intended to be included within the scope of the invention.

10‧‧‧第一記憶體模組10‧‧‧First memory module

12‧‧‧第二記憶體模組12‧‧‧Second memory module

14‧‧‧寫入交換器14‧‧‧Write switch

16‧‧‧讀寫控制器16‧‧‧Reading and writing controller

18‧‧‧讀取交換器18‧‧‧Reading switch

20‧‧‧平行管線式快速傅立葉轉換處理器20‧‧‧Parallel Pipeline Fast Fourier Transform Processor

22‧‧‧多工器22‧‧‧Multiplexer

Claims (8)

一種重新調序緩衝器,包括:一第一記憶體模組,具有m個第一單埠記憶體單元,每一該第一單埠記憶體單元儲存N/2m筆位元資料,N為傅立葉轉換之位元長度,m係為2的冪次方;一第二記憶體模組,具有m個第二單埠記憶體單元,每一該第二單埠記憶體單元儲存N/2m筆位元資料;一寫入交換器,電性連接該第一記憶體模組及該第二記憶體模組,該寫入交換器係接收2的冪次方傅立葉轉換之反向位元資料,並將該反向位元資料交換為一預設排列方式之寫入位元資料,並交替寫入該第一記憶體模組及該第二記憶體模組中;一讀寫控制器,電性連接該寫入交換器、該第一記憶體模組及該第二記憶體模組,該讀寫控制器係控制該寫入位元資料寫入該第一記憶體模組或該第二記憶體模組中,或者是自該第一記憶體模組或該第二記憶體模組中讀出該位元資料;及一讀取交換器,電性連接該第一記憶體模組、該第二記憶體模組及該讀寫控制器,該讀取交換器係交替將該第一記憶體模組及該第二記憶體模組內的該位元資料分別交換為自然順序之輸出位元資料,並以m個輸出路徑將該輸出位元資料輸出。A reordering buffer includes: a first memory module having m first memory cells, each of the first memory cells storing N/2m pen data, N is Fourier The length of the converted bit, m is a power of 2; a second memory module having m second memory cells, each of which stores N/2m pens Metadata; a write switch electrically connected to the first memory module and the second memory module, wherein the write switch receives the inverse bit data of the power-of-fourth Fourier transform of 2, and The reverse bit data is exchanged into a preset arrangement of the bit data, and is alternately written into the first memory module and the second memory module; a read/write controller, electrical Connecting the write switch, the first memory module and the second memory module, the read/write controller controlling the write bit data to be written into the first memory module or the second memory In the body module, or reading the bit data from the first memory module or the second memory module; and reading The switch is electrically connected to the first memory module, the second memory module, and the read/write controller, and the read switch alternates the first memory module and the second memory module The bit data in the group is exchanged for the natural order output bit data, and the output bit data is output by m output paths. 如請求項1所述的重新調序緩衝器,其中該寫入交換器或該讀取交換器分別根據該傅立葉轉換之位元長度N而選擇以交替公式將該寫入位元資料寫入該第一記憶體模組或該第二記憶體模組,或自該第一記憶體模組 或該第二記憶體模組中讀出該位元資料,該交替公式如下所示:α =log2 (N /2m 2 )其中,α 係為整數,N為傅立葉轉換之位元長度,m係為2的冪次方。The reordering buffer of claim 1, wherein the write switch or the read switch selects to write the write bit data in an alternating formula according to the bit length N of the Fourier transform, respectively. The first memory module or the second memory module, or reading the bit data from the first memory module or the second memory module, the alternating formula is as follows: α = log 2 ( N /2 m 2 ) where α is an integer, N is the length of the Fourier transform bit, and m is the power of 2. 如請求項2所述的重新調序緩衝器,其中該α 係為偶數,則該寫入交換器係以各寫入一次之交替方式將該寫入位元資料寫入該第一記憶體模組及該第二記憶體模組中,該讀出交換器係以各讀出一次之交替方式自該第一記憶體模組及該第二記憶體模組中讀出該位元資料。The reordering buffer of claim 2, wherein the alpha is an even number, the write switch writes the write bit data to the first memory module in an alternating manner of writing once. In the group and the second memory module, the read switch reads the bit data from the first memory module and the second memory module in an alternate manner. 如請求項3所述的重新調序緩衝器,其中該寫入交換器係根據該預設排列方式選定之控制訊號進行交換該反向位元資料,該傅立葉轉換之位元長度為128時,其分為16組八個位元參數,將該反向位元資料中每兩組的八個位元參數以八種排列方式,交換為該寫入位元資料,並將該寫入位元資料交替寫入該第一記憶體模組及該第二記憶體模組中;其中該八種排列方式的順序為:第一組及第二組的八個位元參數分別平行輸入至八組該第一單埠記憶體單元及該第二單埠記憶體單元中,且每一該第一單埠記憶體單元及每一該第二單埠記憶體單元中的第一個單埠記憶體單元係作為位移的初始單埠記憶體單元;第三組及第四組的八個位元參數分別自該初始單埠記憶體單元位移四個單埠記憶體單元並輸入至該八組第一單埠記憶體單元及該第二單埠記憶體單元中;第五組及第六組的八個位元參數分別自該初始單埠記憶體單元位移兩個單埠記憶體單元並輸入至該八組第一單埠記憶體單元及該第二單埠記 憶體單元中;第七組及第八組的八個位元參數分別自該初始單埠記憶體單元位移六個單埠記憶體單元並輸入至該八組第一單埠記憶體單元及該第二單埠記憶體單元中;第九組及第十組的八個位元參數分別自該初始單埠記憶體單元位移一個單埠記憶體單元並輸入至該八組第一單埠記憶體單元及該第二單埠記憶體單元中;第十一組及第十二組的八個位元參數分別自該初始單埠記憶體單元位移五個單埠記憶體單元並輸入至該八組第一單埠記憶體單元及該第二單埠記憶體單元中;第十三組及第十四組的八個位元參數分別自該初始單埠記憶體單元位移三個單埠記憶體單元並輸入至該八組第一單埠記憶體單元及該第二單埠記憶體單元中;及第十五組及第十六組的八個位元參數分別自該初始單埠記憶體單元位移七個單埠記憶體單元並輸入至該八組第一單埠記憶體單元及該第二單埠記憶體單元中。The reordering buffer of claim 3, wherein the write switch exchanges the reverse bit data according to the control signal selected by the preset arrangement, and the length of the Fourier transform bit is 128. It is divided into 16 groups of eight bit parameters, and the eight bit parameters of each two groups in the reverse bit data are exchanged into the bit data in eight arrangements, and the bit is written. The data is alternately written into the first memory module and the second memory module; wherein the order of the eight arrangements is: the eight bit parameters of the first group and the second group are input into the eight groups in parallel The first memory cell and the second memory cell, and each of the first memory cells and the first one of each of the second memory cells The unit system is used as the initial memory unit of the displacement; the eight bit parameters of the third group and the fourth group are respectively displaced from the initial memory unit by four memory units and input to the first group of the eight groups.單埠 memory unit and the second memory unit; fifth group Parameter eight bytes from the sixth group were the initial displacement of the single-port memory unit of two single-port memory cell and input to the first set of eight single-port memory unit and the second single-port note In the memory unit, the eight bit parameters of the seventh group and the eighth group are respectively shifted from the initial memory unit to six memory cells and input to the eight first memory units and In the second memory unit, the eight bit parameters of the ninth group and the tenth group are respectively shifted from the initial memory cell to one memory cell and input to the eight first memory. In the unit and the second memory unit; the eight bit parameters of the eleventh group and the twelfth group are respectively displaced from the initial memory unit by five memory cells and input to the eight groups The first memory cell and the second memory cell; the eight bit parameters of the thirteenth and fourteenth groups are respectively displaced from the initial memory cell by three memory cells And inputting to the eight sets of the first memory unit and the second memory unit; and the eight bit parameters of the fifteenth and sixteenth groups are respectively displaced from the initial memory unit Seven memory cells are input to the eight first memory Element and the second single-port memory unit. 如請求項2所述的重新調序緩衝器,其中該α 係為奇數,則該寫入交換器首次先寫入該寫入位元資料於該第一記憶體模組一次,接著以各寫入兩次之交替方式將該寫入位元資料寫入該第二記憶體模組及該第一記憶體模組中,該讀出交換器係以各讀出一次之交替方式自該第一記憶體模組及該第二記憶體模組中讀出該位元資料。The reordering buffer of claim 2, wherein the alpha is an odd number, the write switch first writes the write bit data to the first memory module for the first time, and then writes each Writing the bit data into the second memory module and the first memory module in an alternate manner of two times, the read switch is from the first in an alternate manner of reading each time The bit data is read out in the memory module and the second memory module. 如請求項5所述的重新調序緩衝器,其中該寫入交換器係根據該預設排列 方式選定之控制訊號進行交換該反向位元資料,該傅立葉轉換之位元長度為256時,其分為32組八個位元參數,將該反向位元資料中每四組的八個位元參數以八種排列方式,交換為該預設排列方式之該寫入位元資料,並將該寫入位元資料交替寫入該第一記憶體模組及該第二記憶體模組中;其中該八種排列方式的順序為:第一組及第四組的八個位元參數係平行輸入至八組該第一單埠記憶體單元,第二組及第三組的八個位元參數係平行輸入至八組該第二單埠記憶體單元中,且該八組第一單埠記憶體單元及該八組第二單埠記憶體單元中的每一組第一個單埠記憶體單元係作為位移的初始單埠記憶體單元;第五組及第八組的八個位元參數分別自該初始單埠記憶體單元位移四個單埠記憶體單元並輸入至該八組第一單埠記憶體單元中,第六組及第七組的八個位元參數分別自該初始單埠記憶體單元位移四個單埠記憶體單元並輸入至該八組第二單埠記憶體單元中;第九組及第十二組的八個位元參數分別自該初始單埠記憶體單元位移兩個單埠記憶體單元並輸入至該八組第一單埠記憶體單元,第十組及第十一組的八個位元參數分別自該初始單埠記憶體單元位移兩個單埠記憶體單元並輸入至該八組第二單埠記憶體單元中;第十三組及第十六組的八個位元參數分別自該初始單埠記憶體單元位移六個單埠記憶體單元並輸入至該八組第一單埠記憶體單元,第十四組及第十五組的八個位元參數分別自該初始單埠記憶體單元位移六個單埠記憶體單元並輸入至該八組第二單埠記憶體單元中; 第十七組及第二十組的八個位元參數分別自該初始單埠記憶體單元位移一個單埠記憶體單元並輸入至該八組第一單埠記憶體單元,第十八組及第十九組的八個位元參數分別自該初始單埠記憶體單元位移一個單埠記憶體單元並輸入至該八組第二單埠記憶體單元中;第二十一組及第二十四組的八個位元參數分別自該初始單埠記憶體單元位移五個單埠記憶體單元並輸入至該八組第一單埠記憶體單元,第二十二組及第二十三組的八個位元參數分別自該初始單埠記憶體單元位移五個單埠記憶體單元並輸入至該八組第二單埠記憶體單元中;第二十五組及第二十八組的八個位元參數分別自該初始單埠記憶體單元位移三個單埠記憶體單元並輸入至該八組該第一單埠記憶體單元,第二十四組及第二十七組的八個位元參數分別自該初始單埠記憶體單元位移三個單埠記憶體單元位址並輸入至該八組第二單埠記憶體單元中;及第二十九組及第三十二組的八個位元參數分別自該初始單埠記憶體單元位移七個單埠記憶體單元並輸入至該八組第一單埠記憶體單元,第三十組及第三十一組的八個位元參數分別自該初始單埠記憶體單元位移七個單埠記憶體單元並輸入至該八組第二單埠記憶體單元中。The reordering buffer of claim 5, wherein the write switch is arranged according to the preset The selected control signal exchanges the reverse bit data. When the bit length of the Fourier transform is 256, it is divided into 32 groups of eight bit parameters, and eight bits of each of the four sets of the inverted bit data are The bit parameter is exchanged for the write bit data of the preset arrangement in eight arrangements, and the write bit data is alternately written into the first memory module and the second memory module. The order of the eight arrangements is that the eight bit parameters of the first group and the fourth group are input in parallel to eight groups of the first memory unit, and the second group and the third group are eight. The bit parameters are input in parallel to the eight groups of the second memory unit, and the first group of the first group of the first group of memory cells and the group of the second group of memory cells The memory cell is used as the initial memory cell of the displacement; the eight bit parameters of the fifth and eighth groups are respectively displaced from the initial memory cell by four memory cells and input to the eight Among the first memory cells in the group, eight bits in the sixth and seventh groups Four memory cells are respectively displaced from the initial memory cell and input into the eight second memory cells; the eight bit parameters of the ninth group and the twelfth group are respectively from the initial The memory unit shifts two memory cells and inputs to the eight first memory cells, and the eight bit parameters of the tenth and eleventh groups respectively from the initial memory cell Displaces two memory cells and inputs them into the eight second memory cells; the eight bit parameters of the thirteenth and sixteenth groups are respectively shifted from the initial memory cell by six單埠 memory unit is input to the eight sets of first memory cells, and eight bit parameters of the fourteenth and fifteenth groups are respectively displaced from the initial memory cell by six memory Units are input to the eight sets of second memory cells; The eight bit parameters of the seventeenth group and the twentieth group are respectively shifted from the initial memory cell to one memory cell and input to the eight first memory cells, the eighteenth group and The eight-bit parameters of the nineteenth group are respectively shifted from the initial memory unit to one memory unit and input to the eight-group second memory unit; the twenty-first group and the twentieth The eight bit parameters of the four groups are respectively displaced from the initial memory cell by five memory cells and input to the eight first memory cells, the twenty-second group and the twenty-third group The eight bit parameters are respectively shifted from the initial memory cell by five memory cells and input into the eight second memory cells; the twenty-fifth and twenty-eighth groups The eight bit parameters are respectively displaced from the initial memory cell by three memory cells and input to the eight groups of the first memory cells, the twenty-fourth group and the twenty-seventh group of eight The bit parameters are shifted from the initial memory cell by three memory The unit address is input to the eight sets of second memory cells; and the eight bit parameters of the twenty-ninth and thirty-second groups are respectively displaced from the initial memory cell by seven單埠The memory unit is input to the eight sets of first memory cells, and the eight bit parameters of the thirtieth and thirty-first groups are respectively displaced from the initial memory cell by seven memory cells. And input to the eight groups of second memory cells. 如請求項1所述的重新調序緩衝器,其中該反向位元資料係由一平行管線式快速傅立葉轉換處理器輸出,該平行管線式快速傅立葉轉換處理器電性連接該寫入交換器。The reordering buffer of claim 1, wherein the reverse bit data is output by a parallel pipelined fast Fourier transform processor electrically connected to the write switch . 如請求項1所述的重新調序緩衝器,更包括一多工器,電性連接該第一記憶體模組、該第二記憶體模組及該讀取交換器,該第一記憶體模組及該 第二記憶體模組中的該位元資料係由該多工器選擇輸出,該讀取交換器係接收該多工器所輸出之該位元資料以交換為自然順序之該輸出位元資料。The reordering buffer of claim 1, further comprising a multiplexer electrically connected to the first memory module, the second memory module, and the read switch, the first memory Module and the The bit data in the second memory module is selected and output by the multiplexer, and the read switch receives the bit data output by the multiplexer to exchange the output bit data in a natural order. .
TW103120494A 2014-06-13 2014-06-13 Reordering buffer TWI499969B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
TW103120494A TWI499969B (en) 2014-06-13 2014-06-13 Reordering buffer

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
TW103120494A TWI499969B (en) 2014-06-13 2014-06-13 Reordering buffer

Publications (2)

Publication Number Publication Date
TWI499969B true TWI499969B (en) 2015-09-11
TW201546711A TW201546711A (en) 2015-12-16

Family

ID=54608120

Family Applications (1)

Application Number Title Priority Date Filing Date
TW103120494A TWI499969B (en) 2014-06-13 2014-06-13 Reordering buffer

Country Status (1)

Country Link
TW (1) TWI499969B (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030131032A1 (en) * 2001-12-28 2003-07-10 Gil Vinitzky Bit-reversed indexing in a modified harvard DSP architecture
TW200727637A (en) * 2006-01-09 2007-07-16 Via Tech Inc A FFT processor, and implement method thereof and an OFDM processor using the same
WO2008023253A2 (en) * 2006-08-22 2008-02-28 Nokia Corporation Fft-based signal processing with reduced latency
TW201025033A (en) * 2008-12-31 2010-07-01 Univ Nat Sun Yat Sen Pipelined FFT circuit architecture

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030131032A1 (en) * 2001-12-28 2003-07-10 Gil Vinitzky Bit-reversed indexing in a modified harvard DSP architecture
TW200727637A (en) * 2006-01-09 2007-07-16 Via Tech Inc A FFT processor, and implement method thereof and an OFDM processor using the same
WO2008023253A2 (en) * 2006-08-22 2008-02-28 Nokia Corporation Fft-based signal processing with reduced latency
TW201025033A (en) * 2008-12-31 2010-07-01 Univ Nat Sun Yat Sen Pipelined FFT circuit architecture

Also Published As

Publication number Publication date
TW201546711A (en) 2015-12-16

Similar Documents

Publication Publication Date Title
US11755474B2 (en) Tile based interleaving and de-interleaving for digital signal processing
CN102163141B (en) Addressing module structure for realizing digital signal processor
US9317481B2 (en) Data access method and device for parallel FFT computation
GB2505446A (en) DMA controller with address generator for interleaving and deinterleaving operations
CN103218348B (en) Fast Fourier Transform (FFT) disposal route and system
TWI647568B (en) Distributed serialized data buffer and a memory module, method for routing data and memory system for a cascadable and extended memory subsystem
CN101776988B (en) Restructurable matrix register file with changeable block size
US20140089369A1 (en) Multi-granularity parallel fft computation device
JPWO2015011874A1 (en) Digital filter device, digital filter processing method, and storage medium storing digital filter program
TWI499969B (en) Reordering buffer
JPWO2011036918A1 (en) Data rearrangement circuit, variable delay circuit, fast Fourier transform circuit, and data rearrangement method
US9268744B2 (en) Parallel bit reversal devices and methods
CN103034621B (en) The address mapping method of base 2 × K parallel FFT framework and system
CN102411557A (en) Multi-granularity parallel FFT (Fast Fourier Transform) computing device
CN110888622A (en) Method, device, equipment and medium for realizing asynchronous FIFO (first in first out) with any depth
CN115033293A (en) Zero-knowledge proof hardware accelerator, generating method, electronic device and storage medium
JP6451647B2 (en) Fast Fourier transform apparatus, fast Fourier transform method, and fast Fourier transform program
Zhou et al. Extended Abstract for Evaluating Asynchrony in Gibraltar RAID's GPU Reed-Solomon Coding Library
JP6943283B2 (en) Fast Fourier Transform Equipment, Data Sorting Equipment, Fast Fourier Transform Processing Methods and Programs
CN102622318B (en) Storage controlling circuit and vector data addressing method controlled by same
CN111814675A (en) Convolutional neural network characteristic diagram assembling system based on FPGA supporting dynamic resolution
Long et al. A low-complexity generalized memory addressing scheme for continuous-flow fast Fourier transform
CN114741352B (en) FPGA-based bilinear interpolation resampling implementation method and device
RU197098U1 (en) RECONFIGURABLE QUICK FOURIER TRANSFORMER OF EXTRA LARGE TRANSFORMATION LENGTH
JP2022152001A (en) High-speed fourier transformation device and digital filter device

Legal Events

Date Code Title Description
MM4A Annulment or lapse of patent due to non-payment of fees