Claims
- 1. A method for writing a scalar value to a predetermined position in a first vector stored in a storage device in a computer system, the method comprising:
- generating a second vector containing plural copies of said scalar value;
- generating a signal representing said predetermined position;
- applying said second vector and said signal to said storage device; and
- in response to said signal, the storage device writing a portion of said second vector to said predetermined position in the first vector while leaving the rest of the first vector unchanged;
- wherein said signal is a function of said data type of said scalar value.
- 2. A method for writing a scalar value to a predetermined position in a first vector stored in a storage device in a computer system, the method comprising:
- generating a second vector containing plural copies of said scalar value;
- generating a signal representing said predetermined position;
- applying said second vector and said signal to said storage device; and
- in response to said signal, the storage device writing a portion of said second vector to said predetermined position in the first vector while leaving the rest of the first vector unchanged;
- wherein as a result of writing a portion of the second vector to the predetermined position of the first vector, the storage device obtains a third vector in place of the first vector, and the method further comprises:
- before a portion of the second vector is written to the predetermined position in the first vector, reading the first vector from the storage device to a first circuit, the first circuit writing a portion of the second vector to said predetermined position in the copy of the first vector to obtain a copy of the third vector, and providing the copy of the third vector to a vector processing circuit, the vector processing circuit thus receiving a copy of the third vector before the third vector could be read from the storage device.
- 3. A method for writing a scalar value to a predetermined position in a first vector stored in a storage device in a computer system the method comprising:
- generating a second vector containing plural copies of said scalar value;
- generating a signal representing said predetermined position;
- applying said second vector and said signal to said storage device;
- in response to said signal, the storage device writing a portion of said second vector to said predetermined position in the first vector while leaving the rest of the first vector unchanged; and
- applying said second vector comprises applying all of the second vector to an input of the storage device in a single clock cycle, and the storage device writes the portion of the second vector to the predetermined position of the first vector in the same clock cycle.
- 4. A structure comprising:
- a storage device for storing one or more vectors;
- a circuit for generating a signal representing a predetermined position in a vector, wherein a scalar value is to be written to said predetermined position in a first vector stored in the storage device; and
- a circuit for generating a second vector containing plural copies of the scalar value;
- wherein when the second vector and said signal are applied to the storage device, the storage device writes a portion of the second vector to said predetermined position in the first vector while leaving the rest of the first vector unchanged; and
- wherein said signal is a function of a data type of said scalar value.
- 5. A structure comprising:
- a storage device for storing one or more vectors;
- a circuit for generating a signal representing a predetermined position in a vector, wherein a scalar value is to be written to said predetermined position in a first vector stored in the storage device; and
- a circuit for generating a second vector containing plural copies of the scalar value;
- wherein when the second vector and said signal are applied to the storage device, the storage device writes a portion of the second vector to said predetermined position in the first vector while leaving the rest of the first vector unchanged;
- wherein as a result of writing a portion of the second vector to the predetermined position of the first vector, the storage device obtains a third vector in place of the first vector; and
- the structure further comprises means for reading the first vector from the storage device to a first circuit before a portion of the second vector is written to the predetermined position in the first vector, the first circuit thus obtaining a copy of the first vector;
- wherein the first circuit is for writing a portion of the second vector to said predetermined position in the copy of the first vector to obtain a copy of the third vector, and for providing the copy of the third vector to a vector processing circuit, the vector processing circuit thus receiving a copy of the third vector before the third vector could be read from the storage device.
- 6. A structure comprising:
- a storage device for storing one or more vectors;
- a circuit for generating a signal representing a predetermined position in a vector, wherein a scalar value is to be written to said predetermined position in a first vector stored in the storage device; and
- a circuit for generating a second vector containing plural copies of the scalar value;
- wherein when the second vector and said signal are applied to the storage device, the storage device writes a portion of the second vector to said predetermined position in the first vector while leaving the rest of the first vector unchanged; and
- wherein the application of the second vector to the storage device and writing a portion of the second vector to said predetermined position in the first vector occur in a single clock cycle.
CROSS-REFERENCE TO RELATED APPLICATIONS
This application is related to U.S. patent application Ser. No. 08/733,907, (now U.S. Pat. No. 5,832,288) entitled "Element Select Mechanism for a Vector Processor" by Roney S. Wong, U.S. Pat. No. 5,832,288, issued on Nov. 3, 1998.
US Referenced Citations (6)