Extends:
Kilo-instruction Multiprocessors

Description:

Report of research results

Firstly, we have proposed the Kilo-Instruction Multiprocessors as an environment where thousands of instruction can be in flight, and we showed that it can effectively hide the memory wall problem that multiprocessors systems unfortunately magnify. Now, we have detailed what we believe a correct design for the Kilo-Instruction Multiprocessors where we introduce for the first time the “implicit transaction” term. This specific environment tries to leverage the Kilo-Instruction Processors mechanisms making the system behaving as other current transactional memory systems but transparently to the programmer. We think it is a complexity-effective approach that future multiprocessors must be aware of. Therefore, we plan in the near term to employ research efforts in this direction.

 

Current and future work includes:

• Working with the simulator environment in order to build up the detailed infrastructure that will allow us to evaluate in detail the Kilo-Instruction Multiprocessors and the Implicit Transactional Memory system. Then, we will extensively evaluate such proposals using scientific and commercial workloads.
• Transactional memory systems open a new research line in the cache coherence protocols field. Of course, it is especially applicable to our proposal due to its particular idiosyncrasy, where an implicit transaction-based behaviour is present. Then, we want to work in optimizations of the cache coherence protocols, both for a snoopy bus-based and a directory-based multiprocessor/multi-core system.
• Our proposal is based on the Kilo-Instruction Processors, so thousands of in-flight instructions can be present in each processor. The basic design proposed for these processors do not describe a clear solution to manage the appearing huge amount of memory operations. Some recent works propose new helpful configurations for solving this problem with the loads and store queues. However, none of these works actually match all the needs of our multiprocessor proposal. There is a need for proposing a solution that, on the one hand, solves the inconvenient of managing many memory operations and, on the other hand, helps our system in other specific tasks like “store packaging” and “silent stores elimination”.
• Transactional memory systems are often related to speculative mechanisms that allow the concurrent execution of critical sections or the execution beyond a locked barrier. So it seems obvious that we can do research in this direction. Therefore, we also want to take care of these speculative mechanisms to incorporate them to our final design, to improve existent mechanisms or even to design new specific ones.
• Thread-level speculation can be seamlessly integrated within the Kilo-Instruction Multiprocessor framework. In order to support transactional memory programming models efficiently, it becomes important to study policies and mechanisms for reduction of miss-speculation and load imbalance overhead. This topic will also be studied within this cluster.

The project will mainly involve the collaboration between three European universities: the UPC in Barcelona (Spain), the University of Cantabria in Santander (Spain), and the Chalmers University of Technology in Gothenburg (Sweden). Furthermore, the collaboration will also include other important researchers around the world, being one of them James E. Smith from the University of Wisconsin-Madison (USA).

Nature:
Research cluster

Total funding:
Requested: € 86920
Granted: € 50000

Including fellowship funding:
Requested: € 13200
Granted: € 13200

Description of how the funding will be used:

Based on the following prices:

 

- Travel: round trip GOT-BCN/person (800 Euros)

- Travel: round trip Santander-BCN/person (400 Euros)

- Travel: round trip GOT-Santander/person (800 Euros)

- Stay: Barcelona 1 day/person (120 Euros)

- Stay: Göteborg 1 day/person (200 Euros)

- Stay: Santander 1 day/person (100 Euros)

 

We will need:

2 meetings in Barcelona (3 days/each): (2800 + 1800) x 2 = 9200 euros

- Travel: 2 x 800 + 3 x 400 = 2800 euros

- Stay: 3 x 5 x 120 = 1800 euros

 

2 meetings in Santander (3 days/each): (2800 + 1500) x 2 = 8600 euros

- Travel: 2 x 800 + 3 x 400 = 2800 euros

- Stay: 3 x 5 x 100 = 1500 euros

 

2 meetings in Göteborg (3 days/each): (4800 + 3600) x 2 = 16800 euros

- Travel: 6 x 800 = 4800 euros

- Stay: 3 x 6 x 200 = 3600 euros

 

4 HiPEAC Cluster meetings (3 days/each): (4800 + 2880) x 4 = 30720 euros

- Travel: 8 x 600 = 4800 euros approx. (depending on where the meeting is located)

- Stay: 3 x 8 x 120 = 2880 euros approx. (depending on where the meeting is located)

 

4 weeks stay – M.M. Waliullah/Mafijul Islam in Barcelona: 3600 euros

- Travel: 1600 euros

- Stay: 2000 euros

 

4 weeks stay – Marco Galluzzi in Santander: 2400 euros

- Travel: 400 euros

- Stay: 2000 euros

 

4 weeks stay – Enrique Vallejo in Barcelona: 2400 euros

- Travel: 400 euros

- Stay: 2000 euros

 

Fellowship requested at UPC, Barcelona (Spain)

12 months, 1100 euros/month. Total = 13200 euros

Duration of the funding:
Requested: 12 month(s)
Granted: 12 month(s), starting on: Tue, January 1, 1980

Participating members:
STENSTROM Per (Chalmers University of Technology) (--member--)
VALERO Mateo (UPC) (--member--)
BEIVIDE Ramon (University of Cantabria) (--member--)

Other people collaborating:

M. M. Waliullah, Chalmers University of Technology (Sweden)

Mafijul Islam, Chalmers University of Technology (Sweden)

Adrian Cristal, UPC (Spain)

Marco Galluzzi, UPC (Spain)

Valentin Puente, University of Cantabria (Spain)

Fernando Vallejo, University of Cantabria (Spain)

Enrique Vallejo, University of Cantabria (Spain)