Fixed-point and floating-point arithmetic operator circuits in specialized processing blocks
The present embodiments relate to circuitry that efficiently performs floating-point arithmetic operations and fixed-point arithmetic operations. Such circuitry may be implemented in specialized processing blocks. If desired, the specialized processing blocks may include configurable interconnect circuitry to support a variety of different use modes. For example, the specialized processing block may efficiently perform a fixed-point or floating-point addition operation or a portion thereof, a fixed-point or floating-point multiplication operation or a portion thereof, a fixed-point or floating-point multiply-add operation or a portion thereof, just to name a few. In some embodiments, two or more specialized processing blocks may be arranged in a cascade chain and perform together more complex operations such as a recursive mode dot product of two vectors of floating-point numbers or a Radix-2 Butterfly circuit, just to name a few.
1. Processing circuitry that is configurable to perform fixed-point arithmetic operations and floating-point arithmetic operations, comprising:
a plurality of input ports;
an dynamic multiplication circuit that selectively performs fixed-point multiplication or floating-point multiplication, wherein a product of the dynamic multiplication circuit is based at least in part on a subset of the plurality of input ports;
a plurality of adder circuits configured to perform addition on the product based at least in part on a mode of processing circuitry; and
selection circuitry configured to select an output for the processing circuitry based at least in part on the product and on control signals received by the processing circuitry, wherein selecting the output comprises selecting a bypass of the plurality of adder circuits in a no-add mode.
2. The processing circuitry of claim 1 , comprising a floating-point adder circuit of the plurality of adder circuits that receives the product when the control signals indicate that a floating-point operation is to be performed using the product.
3. The processing circuitry of claim 2 , wherein the selection circuitry is configured to select the output from the floating-point adder circuit based at least in part on the control signals.
4. The processing circuitry of claim 3 , comprising a floating-point rounding circuit configured to:
receive the product from the dynamic multiplication circuit;
adjust the product by performing an operation on the product; and
output the adjusted product to the selection circuitry.
5. The processing circuitry of claim 4 , wherein the operation comprises rounding the product.
6. The processing circuitry of claim 5 , wherein rounding the product comprises a round up operation, a round down operation, or a round-to-nearest-number operation.
7. The processing circuitry of claim 4 , wherein the operation comprises a normalization operation including shifting a mantissa to restore a value and increment or decrement an exponent accordingly.
8. The processing circuitry of claim 1 , comprising a fixed-point adder circuit of the plurality of adder circuits that receives the product when the control signals indicate that a fixed-point operation is to be performed using the product.
9. The processing circuitry of claim 8 , wherein the selection circuitry is configured to select the output from the fixed-point adder circuit based at least in part on the control signals.
10. The processing circuitry of claim 9 , comprising a logic block circuit that receives the product when the control signals indicate that a fixed-point operation is to be performed using the product.
11. The processing circuitry of claim 10 , wherein the logic block circuit is configured to perform a unary operation on the product.
12. The processing circuitry of claim 11 , wherein the unary operation comprises an identity operation or an inversion operation.
13. The processing circuitry of claim 10 , wherein the logic block circuit is configured to perform a binary operation between the product and an additional input to the processing circuitry.
14. The processing circuitry of claim 13 , wherein the binary operation comprises a logical AND operation, a logical NAND operation, a logical OR operation, a logical NOR operation, a logical XOR operation, a left-shift operation, or a right-shift operation.
15. A method, comprising:
receiving a plurality of inputs at a plurality of ports;
selecting a mode of multiplication to be applied by a dynamic multiplication circuit, wherein the selection of the mode of multiplication comprises selecting between a fixed-point multiplication mode and a floating-point multiplication mode;
multiplying a subset of the plurality of inputs together using the selected mode of multiplication to generate a product;
selectively transmitting, using selection circuitry, the product to an adder circuit based at least in part on a control signal indicating an adder mode or a no-adder mode, wherein selectively transmitting comprises:
transmitting the product to a fixed-point adder circuit when the adder mode comprises a fixed-point adder mode;
transmitting the product to a floating-point adder circuit when the adder mode comprises a floating-point adder mode; and
bypassing the fixed-point adder circuit and the floating-point adder circuit when the adder mode comprises the no-adder mode; and
outputting an output value based at least in part on the mode of multiplication and the adder mode.
16. The method of claim 15 , comprising:
receiving the product from the dynamic multiplication circuit at floating-point rounding circuit in the floating-point adder mode;
adjusting the product by performing an operation on the product; and
outputting the adjusted product to the selection circuitry.
17. The method of claim 16 , wherein the operation comprises rounding the product or normalizing the product.
18. An integrated circuit device, comprising:
a plurality of input ports;
an dynamic multiplication circuit that selectively performs fixed-point multiplication or floating-point multiplication, wherein a product of the dynamic multiplication circuit is based at least in part on a subset of the plurality of input ports;
adder selection circuitry that selectively routes the product of the dynamic multiplication circuit;
a plurality of adder circuits that each selectively receives the product based at least in part on adder control signals used to configure the adder selection circuitry in an adder mode; and
selection circuitry that selects an output for the integrated circuit device based at least in part on the product and routing of the product by the adder selection circuitry, wherein selectively routing comprises bypassing the plurality of adder circuits and transmitting the product to the selection circuitry based at least in part on the adder mode comprising a no-add mode.
19. The integrated circuit device of claim 18 , wherein selectively routing comprises:
transmitting the product to a fixed-point adder circuit of the plurality of adder circuits when the adder mode comprises a fixed-point adder mode; and
transmitting the product to a floating-point adder circuit of the plurality of adder circuits when the adder mode comprises a floating-point adder mode.
20. The integrated circuit device of claim 18 , comprising a floating-point rounding circuit configured to:
receive the product from the dynamic multiplication circuit;
adjust the product by performing rounding or normalizing the product; and
output the adjusted product to the selection circuitry.