IP Library Granted Patent US 9,104,432
Granted Patent B2
US 9,104,432 · App. 13/925,083 · Granted Aug 11, 2015

Extracting stream graph structure in a computer language by pre-executing a deterministic subset

Inventors: Joshua S. Auerbach (Ridgefield, CT); David F. Bacon (Sleepy Hollow, NY); Perry S. Cheng (Cambridge, MA); Stephen Fink (Yorktown Heights, NY); Rodric Rabbah (Yonkers, NY)
Assignee: International Business Machines Corporation
G06F8/443
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 9,104,432
App. No.
13/925,083
Granted
Aug 11, 2015
Kind
B2
Abstract

Compile-time recognition of graph structure where graph has arbitrary connectivity and is constructed using recursive computations is provided. In one aspect, the graph structure recognized at compile time may be duplicated at runtime and can then operate on runtime values not known at compile time.

Claims (27)

1. A method for extracting a stream graph for a hardware circuit from a program written in a programming language, comprising:

receiving a program;

analyzing the program, by a processor, for one or more constructs in the programming language that builds the stream graph for programming the hardware circuit, a node in the stream graph representing a computation to be performed, the computation expressed in the programming language, an edge in the stream graph representing data flowing from the node to another node in the stream graph; and

outputting one or more artifacts representing the stream graph,

wherein the stream graph can be compiled for the hardware circuit to execute and produce a result with different node initialization values and different streaming data values that are presented in an execution of the program not known at compile time, and

wherein the stream graph can be built using recursive computation and the connectivity of the stream graph can be arbitrary,

the analyzing comprising identifying one or more stream graph constructions that are relocatable,

wherein the one or more stream graph constructions are identified as relocatable if the one or more stream graphs constructions are built from one or more immutable nodes and one or more connections that do not alter nodes of the one or more stream graphs.

2. The method of claim 1 , wherein the analyzing comprises:

identifying one or more expressions that are repeatable.

3. The method of claim 2 , wherein the one or more expressions are identified as being repeatable if the expressions have immutable datatypes.

4. The method of claim 2 , wherein the one or more expressions are identified as being repeatable if the expressions are local methods.

5. The method of claim 1 , wherein the one or more stream graph constructions are identified as relocatable if computations associated with the one or more stream graph constructions are repeatable.

6. The method of claim 1 , wherein the one or more stream graph constructions are identified as relocatable if computations associated with the one or more stream graph constructions are repeatable, except for a presence of one or more task initializations.

7. The method of claim 1 , wherein a compiler automatically identifies all relocatable stream graph constructions in the program.

8. The method of claim 1 , wherein one or more of relocatable stream graph constructions are identified in the program manually.

9. The method of claim 1 , wherein the analyzing comprises creating a modified form of a task creation expression, wherein one or more unrepeatable initializations are replaced by one or more marker values.

10. The method of claim 9 , wherein the analyzing further comprises executing the modified form of the task creation expression at compile time within an environment that mimics a runtime environment, and capturing a stream graph that is produced from the executing.

11. The method of claim 10 , wherein the analyzing further comprises converting the captured stream graph into the artifacts, the artifacts comprising at least hardware description language to synthesize for the hardware circuit.

12. A method for extracting a stream graph for a hardware circuit from a program written in a programming language, comprising:

receiving a program;

analyzing the program, by a processor, for one or more constructs in the programming language that builds the stream graph for programming the hardware circuit, a node in the stream graph representing a computation to be performed, the computation expressed in the programming language, an edge in the stream graph representing data flowing from the node to another node in the stream graph; and

outputting one or more artifacts representing the stream graph,

wherein the stream graph can be compiled for the hardware circuit to execute and produce a result with different node initialization values and different streaming data values that are presented in an execution of the program not known at compile time, and

wherein the stream graph can be built using recursive computation and the connectivity of the stream graph can be arbitrary,

the analyzing comprising identifying one or more stream graph constructions that are relocatable,

wherein the one or more stream graph constructions are identified as relocatable even if the one or more stream graph constructions include one or more calls to one or more unrepeatable local methods, provided that an unrepeatable part is only passed as one or more task initialization parameters.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Jun 24, 2013
From: AUERBACH, JOSHUA S.; BACON, DAVID F.; CHENG, PERRY S.; FINK, STEPHEN; RABBAH, RODRIC
To: INTERNATIONAL BUSINESS MACHINES CORPORATION
Reel/Frame 030672/0277 →
Continuity (1)
Related Publication 20140380290A1 · Dec 25, 2014