The summary of EP2709017A1 states:

The invention provides a device for controlling the access to a cache structure comprising multiple cache sets during the execution of at least one computer program, the device comprising a module for generating seed values during the execution of the at least one computer program; a parametric hash function module for generating a cache set identifier to access the cache structure, the identifier being generated by combining a seed value generated by the module for generating seed values and predetermined bits of an address to access a main memory associated to the cache structure.

Find what appears to be a partial disclosure of the underlying principles of EP2709017A1 before the patent filing date of Sep 14, 2012 in the publication: F. J. Cazorla, E. Quin ̃ones, T. Vardanega, L. Cucu, B. Triquet, G. Bernat, E. Berger, J. Abella, F. Wartel, M. Houston, L. Santinelli, L. Kosmidis, C. Lo, and D. Maxim. PROARTIS: Probabilistically Analysable Real-Time Systems. Rapport de recherche INIRIA/RR-7869, INRIA, Jan 2012. In particular see the contents of Page 17 of PROARTIS: Probabilistically Analysable Real-Time Systems: "Fully-associative caches are complex to implement and energy-consuming. However, when using a set-associative (with random replacement) or a direct-map cache, the placement policy is deterministic. Deterministic placement policies create dependencies and lead to cache risk patterns and hence to time pathological cases. We devise designs based on set-associative or direct-mapped caches using random replacement algorithms, which exhibit a similar behaviour at a lower cost. At hardware level, we devise a solution where we can randomise the placement function at program start up. To that end we XOR the address accessing the cache with a randomly generated number and use the result as the new address to access the cache. Under every placement setup, the address colliding in the different sets are different. Hence, different runs will lead to different (random) colliding addresses in each set and hence different (random) conflicts and execution times. "

Also see Figure 10 of PROARTIS: Probabilistically Analysable Real-Time Systems.

Does the above earlier publication invalidate part or all of the EP patent application?


EP2709017 A1 is not a patent but a patent application, however intention to grant was communicated to applicant and the patent will be granted in designated European contracting states as taxes have been paid (actually depending on filed translations).

You mention a prior art which is a technical report authored by the inventors and other researchers. This prior art was disclosed in the patent application itself. This is fortunate for them regarding the US patent application, since it is definitely relevant and all inventors are authors of the report, they know about it. Failing to mention relevant prior art to the USPTO that an applicant knows about endangers the application.

Yet, without entering into the details, since they refer to this document and explain how the invention improves from what is described the report, and since examiner granted the application, I would say it does not invalidate the invention. It's totally different from a prior art you would have found that was unknown to examiner.

I would stress that patented inventions are limited to what is mentioned in independent claims, in this case, claims 1 and 7. From a quick look, it seems that the parametric hash function is part of the inventive step on top of the technical report you mention.

More generally, I wouldn't expect from a public European research institution to try to patent something that would be invalidated by a previously published technical report from the same authors. Authors probably know or were definitely told that they should patent before publication. The lack of argument with the European examiner is surprising (to me, at least) but might hint a very narrow patent.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.