Method and device for detecting a race condition and a computer program product
A method is provided for detecting a race condition of a parallel task when accessing a shared resource in a multi-core processing system. The method requires that a core requires only a read access to the data set of another core, thereby ensuring better decoupling of the tasks. In an initialisation phase, initial values of global variables are assigned, in an activation phase, each core determines if the other core has written new values to the variables and if so, detects a race condition. Initial values are restored for each variable in a deactivation phase.
1. A method for detecting a race condition, comprising: writing a pre-defined value to a copy of a first global variable D 1 , wherein the first global variable is write-accessible, in a first processing core; identifying a race condition when a value of a copy of a third global variable A 2 , wherein the third global variable is read-accessible, in a second processing core does not equal an initial value, writing a value to a copy of a second global variable A 1 , wherein the second global variable is write-accessible, in the first processing core; identifying a race condition when a value of a copy of a fourth global variable D 2 , wherein the fourth global variable is read-accessible, in the second processing core does not equal an initial value; accessing a shared resource; and writing an initial value to the copy of the first global variable D 1 , and writing an initial value to the copy of the second global variable A 1 .
2. The method of claim 1 further comprising writing initial values to the copies of the first and second global variables D 1 , A 1 , prior to writing the pre-defined value to the copy of the first global variable D 1 .
3. The method of claim 1 further comprising generating a race detected message and/or terminating when a race condition is detected.
4. The method of claim 1 further comprising at least a first thread and a second thread calling the method when accessing the shared resource.
5. A device for detecting a race condition, said device comprising: a processing unit arranged to write a pre-defined value to a copy of a first global variable D 1 , wherein the first global variable is write-accessible, in a first processing core, identify a race condition when a value of a copy of a third global variable A 2 , wherein the third global variable is read-accessible, in a second processing core does not equal an initial value; write a value to a copy of a second global variable A 1 , wherein the second global variable is write-accessible, in the first processing core; identify a race condition when a value of a copy of a fourth global variable D 2 , wherein the fourth global variable is read-accessible, does not equal an initial value; access a shared resource; write an initial value to the copy of the first global variable D 1 , and write an initial value to the copy of the second global variable A 1 .
6. The device of claim 5 wherein the processing unit is further arranged to write initial values to the copies of the first and second global variables D 1 , A 1 , prior to writing the pre-defined value to the copy of the first global variable D 1 .
7. The device of claim 5 wherein the processing unit is further arranged to generate a race detected message and/or terminate when a race condition is identified.
8. The device of claim 5 wherein the device is implemented in an integrated circuit.