A new generation of decimal abacus architecture computer technology system
Through a new generation of decimal abacus architecture, the problems of number system conversion loss and foreign architecture monopoly in traditional binary computers have been solved, realizing efficient and accurate calculation and an independent and controllable computer system, which is suitable for the independent and controllable needs of multiple fields.
Patent Information
- Application Number
- CN202610424898.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-04-02
- Publication Date
- 2026-06-26
AI Technical Summary
Existing computers use a traditional binary architecture, which leads to frequent number system conversion losses and hardware complexity. Foreign architecture monopolies and information security risks also prevent them from meeting the needs for independent control and high-precision computing.
Adopting a new generation of decimal abacus architecture, through underlying logic reconstruction and hardware innovation, the conversion between binary and decimal is eliminated, and information is defined, stored and applied in a unified manner according to decimal rules. The operation module is designed in combination with Chinese abacus rules to achieve unified decimal circulation throughout the system.
It has achieved a leapfrog improvement in computing efficiency and accuracy, reduced the difficulty of chip manufacturing, broken the foreign architecture monopoly, ensured information security, and is suitable for the independent and controllable needs of fields such as personal computing, professional design, artificial intelligence, supercomputing, military industry and finance.
Smart Images

Figure CN122284949A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of computer technology, specifically to a new generation of decimal abacus architecture computer technology system, applicable to all scenarios such as personal computing, professional design, artificial intelligence, supercomputing, military industry, finance, and scientific research. It fundamentally solves the technical pain points of existing computer CPU architecture monopoly, high-end manufacturing process, uncontrollable information security, large loss in number system conversion, and insufficient calculation accuracy, and provides an independent, controllable, efficient, accurate, and easy-to-manufacture computer system. Background Technology
[0002] Current computers all employ traditional binary architectures, with core computation, storage, and transmission processes relying on binary encoding. In practical applications, frequent decimal-to-binary conversions are necessary. This process not only incurs significant computational overhead and reduces operational efficiency but also leads to complex hardware circuits, stringent chip manufacturing requirements, and foreign monopolies on high-end processes, hindering the development of the domestic chip industry. Furthermore, existing computer CPU architectures largely depend on foreign architectures such as x86 and ARM, resulting in foreign control of core technologies, posing information security risks, and failing to meet the self-sufficiency and controllability requirements of critical sectors like government, finance, and the military. Multi-dimensional image processing is inefficient, and binary floating-point arithmetic inherently contains errors, failing to meet the computational demands of finance, scientific research, and high-precision scenarios. Existing improvements to computer architectures primarily focus on localized optimizations, without fundamentally reconstructing the computational logic and hardware architecture. This fails to completely eliminate base conversion overhead and overcome the constraints of foreign architecture monopolies and reliance on high-end processes. Therefore, there is an urgent need for a completely independent and original next-generation computer system with fundamentally reconstructed logic, capable of eliminating base conversion overhead, simplifying chip manufacturing, and achieving self-controllability. This would address the shortcomings of existing technologies and support the high-quality development of the national computer industry. Summary of the Invention
[0003] (I) Purpose of the Invention The purpose of this invention is to overcome the defects of existing binary computers and provide a new generation of decimal abacus architecture computer technology system. Through underlying logic reconstruction and hardware innovation, it breaks through the constraints of traditional binary architecture, simplifies chip manufacturing difficulty, reduces transcoding loss, and achieves a leapfrog improvement in computing efficiency, data accuracy, and architecture controllability. At the same time, it is compatible with the existing software ecosystem, breaks the monopoly of foreign CPU architecture, solves high-end process problems, strengthens the information security defense line, provides core technology support for the independent and controllable development of the national computer industry, and adapts to the needs of all application scenarios.
[0004] (II) Core Design Principles and Logical Framework The core design of this system follows the principle of "binary as the base, decimal as the core, and abacus as the bridge," constructing a complete computer system with three layers of logic. This upgrades the system from "machine language" to "human cognitive language" at the bottom layer, unifying the definition, storage, editing, and application of all final information. The three-layer logical framework is as follows: Bottom Physical Layer: Transistor switches generate binary 0 / 1 basic signals, serving as the sole carrier for temporary computation and transmission, providing basic signal support for the entire system. This utilizes existing transistor hardware, eliminating the need for additional transistor development. Middle Computational Layer: Based on Chinese abacus rules, the design uses abacus logic as its framework, employing 0 / 1 as beads (0 for the lower bead, 1 for the upper bead, signals such as breaking five and breaking ten) to perform addition, subtraction, multiplication, and division operations. Top Application Layer: Using decimal codes as the unified information carrier, this layer completes all data storage, function implementation, content creation, and interactive display, aligning with human cognitive habits and enabling intuitive data management and application without requiring users to concern themselves with the underlying encoding logic.
[0005] The core logic boundaries are clearly defined: binary 0 / 1 signals are only responsible for low-level temporary calculations and transmission, and do not participate in the definition, storage and application of final information; decimal numbers govern the definition, storage, editing and application of all final information, ensuring the uniformity and efficiency of data flow.
[0006] (III) Modify the existing CPU by removing the arithmetic and register modules. The cache, controller, communication interface, functional modules, and power management remain largely unchanged; only the signal rules are modified to unify decimal input and output, reducing multi-signal transcoding latency. (IV) Operation Module Design: Designed according to the rules of Chinese abacus calculation. Addition: The input structures of the addend and augend are different. Gate circuits are used to detect the "breaking five" signal. Based on the abacus calculation rules, gate circuits are used to generate the "breaking ten and carrying over" signal. Subtraction: The input structures of the subtrahend and minuend are the same. Gate circuits are used to detect the "breaking five" signal. Based on the abacus calculation rules, gate circuits are used to generate the "breaking ten and borrowing one" signal. Multiplication: The input structures of the multiplier and multiplicand are the same. The addition module generates multiplication tables (Method 1), and the matrix module generates multiplication tables (Method 2). Combined adders solve multiplication operations from one digit to multiple digits. Division: The input structures of the divisor and dividend are the same. The multiplication tables generated by the multiplication method 1 module are connected to the subtraction module group to realize the quotient and remainder. A latch is used to copy the remainder and combine it with the input of the next digit dividend, then input it again into the subtraction group to realize the next digit quotient and remainder. This cycle continues until the calculation is complete and the quotient is obtained. Decimal and zero-padding signals implement decimal logic and shifting.
[0007] (V) Signal Design: Numbers are represented by 4 binary bits, Pinyin and English letters by 6 binary bits (5 for Pinyin + 0 / 1 for case sensitivity), instructions by 8 binary bits, and text by 20 binary bits. Text is represented by 5 decimal digits or 4 letters to enable input method functionality. Images are represented by 9 decimal digits (9*4=36 binary digits), with 3 bits representing a primary color (red, blue, green). Sound is also represented by decimal digits (one decimal digit corresponds to 4 binary digits), and voltage, current, and resistance are all decimal variables. This ensures a unified decimal circulation, binary storage, and operation across the entire system. Attached Figure Description
[0008] Figure 1 Decimal Operating System Architecture Diagram Figure 2 Addition Logic Diagram Figure 3 Subtraction Logic Diagram Figure 4. Method 1 of the multiplication module Figure 5. Method 2 of the multiplication module Figure 6 Division Logic Diagram Figure 7. Multiplication module combination method diagram Figure 8. Operational number input array diagram (the 6 empty lines in the figure represent the start / end / decimal / zero padding / greater than / less than signals respectively). Detailed Implementation
[0009] This system implements a single architecture that removes the traditional arithmetic and register modules from the CPU, retaining the original basic units such as cache, controller, communication interface, functional modules, and power management. The main improvements are enhanced signal conversion speed and expanded signal scheduling platform. The overall system only redefines the signal rules, without altering the original execution steps of transmission, addressing, and feedback signals. The core improvements are twofold: First, it completely eliminates the traditional binary-to-decimal conversion operations, eliminating the need for calculations between the two number systems; second, the entire system uniformly follows decimal rules. When a set value needs to be output, it directly reads the predefined binary code to complete the corresponding operation, eliminating the need for a number system conversion process and fundamentally solving the latency and complexity problems caused by traditional number system conversions. Signal transmission uses start and end markers within the digital signal as intervals to ensure the integrity of file transmission order. It can flexibly support subsequent expansion to ultra-wide bit transmission links such as 128-bit, 256-bit, and 512-bit. Instruction signals, module signals, address signals, function signals, file signals, format signals, and input / output signals are transmitted continuously and in an integrated manner. The system uses decimal as its overall framework, employing a fixed mapping method where every 4 binary bits correspond to one decimal bit. Different types of information are distinguished by different bit widths: numbers use 4 binary bits, letters use 6 binary bits (5 for pinyin + one 0 / 1 for case sensitivity), instructions use 8 binary bits, text uses 20 binary bits, images use 36 binary bits, and sound is encoded using 4 binary bits corresponding to decimal. Text is either a 5-digit decimal number or a 4-letter combination, enabling the use of an input method. Physical variables such as voltage, current, and resistance in the real system are also directly transmitted in decimal variable form without conversion, while the entire system's underlying storage and operation are still based on binary. Information at each level flows in an orderly manner with the cooperation of the decoder and crystal. The address is searched layer by layer according to the start and end signals to determine the corresponding input and output information. After reaching the designated position, the corresponding operation is executed. The end signal is sent back to the CPU to complete one information execution instruction. The transmission channel can adopt the existing multi-bandwidth hub structure without additional technical upgrades. The whole system achieves a high-efficiency system architecture that can directly read the code without base conversion by using a unified decimal rule and a fixed binary mapping relationship.
[0010] The two operation modules are implemented according to the Chinese abacus architecture. 1 represents the upper bead and 0 represents the lower bead. The start / end signals realize the operation / clearing function. The decimal number is written independently. The zero-padding signal solves the shift function. The size signal solves the logical judgment.
[0011] Addition uses two sets of fixed signal inputs (01000---1 00001---1) with different floating gate transistors. The interleaved signal at bit 3, combined with other bit signals, uses logic gates to generate a "break five" or "not break five" signal, solving the problem of reading the entire signal (e.g., converting 010100 to the result of whether it represents the number 2 or 8). The signal at bit 6 changes, combined with other bit signals, using logic gates to generate a carry signal (e.g., in 00101000, counting from the end to the beginning, the 7th bit is 0, indicating that each of the 1st to 5th bits being 0 represents 1, which is equivalent to the lower 5 beads in abacus calculation; the 6th bit being 0 represents 5, which is equivalent to one bead in the upper two codes of abacus calculation; the 7th bit is the execution code of the entire signal; and the 8th bit is the top bead of the upper two codes of abacus calculation, representing a carry. According to this rule, 00101000 represents 4 with no carry). A 5-to-32 decoder is used to read the lower 5 bead signals, and logic gates are used to combine the bit information to read the upper 2 bead signals, generating a 4-bit binary code. The carry signal is sent back to the input port to determine whether to use a fixed carry floating-gate transistor signal or a general floating-gate transistor fixed signal to implement the carry logic. This solves the problem of 4-bit binary code input and output using the same mapping table logic. Decimal point alignment is maintained during input; when two input devices simultaneously produce decimal signals, the decimal signal is triggered to directly input the result, solving the problem of decimal synchronization. Ultimately, a zipper-style addition operation is implemented.
[0012] Subtraction uses two sets of general-purpose floating-gate transistor fixed signals (00001-1) as inputs. Logic gate circuits are used to combine the bit information to generate the 7th bit execution code and the 8th bit borrow code. A 4-to-16 decoder reads the lower 5 bead signals, and logic gate circuits are used to combine the bit signals to read the upper 2 bead signals, combining them to generate a 4-bit binary code. The carry signal is sent back to the input port to determine whether to use the borrow floating-gate transistor fixed signal or the general-purpose floating-gate transistor fixed signal to implement the carry logic. This solves the problem of 4-bit binary code input and 4-bit binary code output using the same mapping table logic. After the end signal is transmitted, if there is still a signal in the borrow latch, it indicates a less than signal; otherwise, it indicates a greater than or equal to signal. If the result is 0, it indicates an equal to signal. The decimal point is kept aligned during input. When both inputs simultaneously produce decimal signals, the decimal signal is triggered to directly input the result, solving the problem of decimal synchronization and ultimately realizing a zipper-style subtraction operation.
[0013] Multiplication uses two sets of general-purpose floating-gate transistor fixed signals (00001-1) as input. The method for product of a single multiplier and multiplicand is as follows: Method 1 combines nine adders and one 0 signal to achieve the product of the multiplicand and multipliers from 0 to 9. A 4-to-16 decoder selects the first multiplier. Method 2 uses two 4-to-16 decoders to generate a multiplication table array. Each unit is an 8-bit floating-gate transistor fixed signal (i.e., 8 bits corresponding to two decimal digits). When one multiplier is input, the multiplicand input generates an eight-bit binary sequence. The first four bits and the last four bits are input to the addend and multiplier ports of the adder, respectively, for chained addition. Note that the first four bits of the eight bits are the tens digit of the multiplication table; a latch is needed to achieve staggered addition. This method achieves the product of a single multiplier and multiplicand. Methods for summing multi-digit multipliers and multiplicands: Method 1: A single-digit multiplier and multiplicand product module is combined with an adder to achieve the summing of multi-digit multipliers and multiplicands (note that the second multiplier input triggers a zero-padding signal first, the third triggers two zero-padding signals, and so on), ultimately achieving a zipper-style multiplication operation. Method 2: A single-digit multiplier and multiplicand product module is followed by a buffer module to record each digit of the multiplier and multiplicand product. As soon as two results are generated, they are immediately input into the adder for addition. When only one result remains, it is the final result (note that the second multiplier input triggers a zero-padding signal first, the third triggers two zero-padding signals, and so on), ultimately achieving a zipper-style multiplication operation. Decimal point alignment is not maintained during input. Counter B starts counting when the multiplier and multiplicand begin input. When the first decimal signal is generated in the multiplier and multiplicand, counter A starts counting. When the second decimal signal is generated, counter B stops counting. When A=B, the decimal signal is triggered and written to the result decimal signal.
[0014] Division uses two sets of general-purpose floating-gate transistor fixed signals (00001-1) as inputs. The divisor combination inputs nine adders to achieve the product of the divisor and 0-9, corresponding to the subtrahend inputs of nine subtractors. The minuend input inputs the dividend with the corresponding number of digits as the divisor. When a less than signal appears in the 0-9 subtractor, the output output is the previous combination index of the subtractor 0-9, which is the quotient. The remainder is stored in the buffer module. After switching to the minuend input, one digit of the dividend is input. This cycle continues until the end, yielding the quotient and remainder, ultimately achieving a zipper-style division operation. (Note: If division continues after the dividend ends, a decimal point needs to be added to the quotient). Decimal point alignment is not maintained during input. When the number of decimal places in the divisor > the number of decimal places in the dividend, zeros are added to the end of the dividend. <被除数位数 → 商的小数位数="它们的差(注意:这时的补点需把商的结果从前向后数补点)<br">In this embodiment, the computer system uses a mature 28nm process to manufacture chips, achieving the computing performance of a traditional 7nm high-end process with 100% accuracy and no binary floating-point error. Its operating speed is improved compared to traditional binary computers, completely eliminating number system conversion losses. The core hardware architecture is entirely independently developed, breaking the monopoly of foreign x86 and ARM architectures, eliminating information security risks, and meeting the self-controllable needs of key fields such as military, finance, and scientific research. External hardware retains existing forms, reducing industrial upgrading costs. It can be widely applied in personal computing, professional design, supercomputing, military, and other scenarios, possessing extremely high practicality, versatility, and industrialization value.< / 被除数位数>
Claims
1. A high performance computer system architecture without the need for radix conversion, characterized in that, The proposed architecture removes the traditional arithmetic and register modules from the CPU, retaining only the cache, controller, communication interface, functional modules, and power management basic unit. It enhances signal conversion speed and expands the signal scheduling platform, redefining only the signal rules without altering the original execution steps of transmission, addressing, and feedback signals. This system adopts a three-layer logical framework with a full-link design, using abacus rules as the computational core, binary as the information carrier, and decimal as the information core. It completely reconstructs the CPU core, simplifying chip manufacturing while remaining compatible with the existing software ecosystem. The core design follows the principle of "binary as the base, decimal as the core, and abacus as the bridge." Binary handles underlying temporary computation and data transmission, while decimal uniformly manages the definition, storage, editing, and application of all final information. It comprehensively covers applications across personal computing, professional design, artificial intelligence, supercomputing, and military applications, achieving a leap forward in computational efficiency, data accuracy, and architectural controllability, providing core technological support for the independent and controllable development of the national computer industry.
2. The architecture of claim 1, wherein, The three-layer logical framework is as follows: Bottom Physical Layer: Transistor switches generate binary 0 / 1 basic signals, serving as the sole carrier for temporary computation and transmission, providing basic signal support for the entire system. It utilizes existing transistor hardware, eliminating the need for developing new transistors. Middle Computation Layer: Based on Chinese abacus rules, the design uses abacus logic as its framework, employing 0 / 1 as beads (0 for the lower bead, 1 for the upper bead, signals such as breaking five, breaking ten, etc.) to perform addition, subtraction, multiplication, and division operations. Top Application Layer: Using decimal codes as the unified information carrier, it completes all data storage, function implementation, content creation, and interactive display, aligning with human cognitive habits to achieve intuitive data management and application without requiring users to concern themselves with the underlying encoding logic.
3. The architecture of claim 1, wherein, Signal transmission uses start and end markers within the digital signal as intervals to ensure the integrity of file transmission order. It supports 128-bit, 256-bit, 512-bit, and higher ultra-wide bit transmission link expansion. Instruction signals, module signals, address signals, function signals, file signals, format signals, and input / output signals are transmitted continuously and in an integrated manner. Information at each level flows in an orderly manner with the cooperation of the decoder and crystal. According to the start and end signals, the address is found layer by layer, the corresponding input / output information is determined, the corresponding operation is executed, and the end signal is sent back to the CPU to complete one information execution instruction. The transmission channel adopts the existing multi-bandwidth hub structure, requiring no additional technical upgrades.
4. The architecture of claim 1, wherein, The architecture is based on the decimal system, employing a fixed mapping of 4 binary bits to 1 decimal bit. Different types of information are distinguished by different bit widths. Specifically, numbers use 4 binary bits, letters use 6 binary bits, instructions use 8 binary bits, text uses 20 binary bits, images use 36 binary bits, and sound uses 4 binary bits corresponding to decimal values. Physical variables such as voltage, current, and resistance in the real system are transmitted as decimal variables without conversion, while the entire system is stored and operated in binary at the underlying level. Size, parity, multiples, divisibility, equality, intervals, positive and negative signs, and overflow are all inherent logical attributes of the decimal system.
5. The architecture of claim 1, wherein, During information transmission, audio can be embedded, and simultaneous on-screen transmission can be performed covertly. The 0-9 mapping or language mapping code can be changed to edit the transmission, thereby achieving secure transmission of military information.
6. A high performance computer system architecture without the need for endianness conversion, characterized in that, The calculation modules are all implemented according to the Chinese abacus architecture. 1 represents the upper bead and 0 represents the lower bead. The start / end signals realize the calculation / clearing functions. The decimal number is written independently. The zero-padding signal solves the shift function. The size signal solves the logical judgment.
7. The architecture of claim 6, wherein, The addition operation uses two sets of fixed floating-gate transistor signals (01000---1 00001---1) with different signals. Logic gates are used to generate signals indicating whether the number exceeds five or not, and to generate a carry signal. A decoder reads the lower five-bead signal, and logic gates combine the bit information to read the upper two-bead signal, summing them to generate a 4-bit binary code. The carry signal is sent back to the input port to determine whether to use a fixed carry floating-gate transistor signal or a general floating-gate transistor signal to implement the carry logic. This achieves a 4-bit binary code input and output, and the decimal number is represented using the same system mapping table. Decimal point alignment is maintained during input. When both inputs simultaneously produce decimal signals, the decimal signal is triggered, directly inputting the result, thus solving the problem of decimal synchronization. Ultimately, a zipper-style addition operation is implemented.
8. The architecture of claim 6, wherein, Subtraction uses two sets of general-purpose floating-gate transistor fixed signals (00001-1) as inputs. Logic gate circuits are used to combine the bit information to generate the execution code and borrow code. A decoder reads the lower 5 bead signals, and logic gate circuits are used to combine the bit signals to read the upper 2 bead signals, combining them to generate a 4-bit binary code. The carry signal is sent back to the input port to determine whether to use the borrow floating-gate transistor fixed signal or the general-purpose floating-gate transistor fixed signal to implement the carry logic. This achieves a 4-bit binary code input and output, and the decimal number is represented using the same system mapping table. After the end signal is transmitted, if there is still a signal in the borrow latch, it indicates a less than signal; otherwise, it indicates a greater than or equal to signal. If the result is 0, it indicates an equal to signal. The decimal point is kept aligned during input. When both inputs simultaneously produce a decimal signal, the decimal signal is triggered to directly input the result, solving the problem of decimal synchronization and ultimately achieving a zipper-style subtraction operation.
9. The architecture of claim 6, wherein, Multiplication uses two sets of general-purpose floating-gate transistor fixed signals (00001-1) as input. The method for product of a single multiplier and multiplicand is as follows: Method 1 combines nine adders and one 0 signal to achieve the product of the multiplicand and multipliers from 0 to 9. A 4-to-16 decoder selects the first multiplier. Method 2 uses two 4-to-16 decoders to generate a multiplication table array. Each unit is an 8-bit floating-gate transistor fixed signal (i.e., 8 bits corresponding to two decimal digits). When one multiplier is input, the multiplicand input generates an eight-bit binary sequence. The first four bits and the last four bits are input to the addend and multiplier ports of the adder, respectively, for chained addition. Note that the first four bits of the eight bits are the tens digit of the multiplication table; a latch is needed to achieve staggered addition. This method achieves the product of a single multiplier and multiplicand. Methods for summing multi-digit multipliers and multiplicands: Method 1: A single-digit multiplier and multiplicand product module is combined with an adder to achieve the summing of multi-digit multipliers and multiplicands (note that the second multiplier input triggers a zero-padding signal first, the third triggers two zero-padding signals, and so on), ultimately achieving a zipper-style multiplication operation. Method 2: A single-digit multiplier and multiplicand product module is followed by a buffer module to record each digit of the multiplier and multiplicand product. As soon as two results are generated, they are immediately input into the adder for addition. When only one result remains, it is the final result (note that the second multiplier input triggers a zero-padding signal first, the third triggers two zero-padding signals, and so on), ultimately achieving a zipper-style multiplication operation. Decimal point alignment is not maintained during input. Counter B starts counting when the multiplier and multiplicand begin input. When the first decimal signal is generated in the multiplier and multiplicand, counter A starts counting. When the second decimal signal is generated, counter B stops counting. When A=B, the decimal signal is triggered and written to the result decimal signal.
10. The architecture of claim 6, wherein, Division uses two sets of general-purpose floating-gate transistor fixed signals (00001---1) as inputs. The divisor combination inputs nine adders to achieve the product of the divisor and numbers from 0 to 9. Correspondingly, nine subtractor inputs are used for the subtrahend input. The minuend input is the dividend with the corresponding number of digits as the divisor. When a less than signal appears in the 0-9 subtractor, the output of the previous combination number of the subtractor (0-9) is the quotient. The remainder is stored in a buffer module. Then, the minuend input is switched to a single digit of the dividend. This process is repeated until the end, yielding the quotient and remainder, thus achieving a zipper-style division operation. (Note: If division continues after the dividend has ended, a decimal point needs to be added to the quotient). Decimal point alignment is not maintained during input. When the number of decimal places in the divisor > the number of decimal places in the dividend, zeros are added to the end of the dividend. When the number of decimal places in the divisor < the number of decimal places in the dividend, the number of decimal places in the quotient equals their difference (Note: in this case, the decimal point is added from the beginning of the quotient).
