Mul­ti­pro­cessors are a core com­pon­ent in all types of com­put­ing in­fra­struc­ture, from phones to data­cen­ters. This course will build on the pre­requis­ites of pro­cessor design and con­cur­rency to in­tro­duce the es­sen­tial tech­no­lo­gies re­quired to com­bine mul­tiple pro­cessing ele­ments in­to a single com­puter.

Course con­tent:

• Forms of par­al­lel­ism
• Par­al­lel pro­gram­ming mod­els
• Cache co­her­ence
• Memory con­sist­ency
• Syn­chron­iz­a­tion
• In­ter­con­nec­tion net­works
• Soft­ware ef­fi­ciency & op­tim­iz­a­tion
• GPU ar­chi­tec­ture & pro­gram­ming

Learn­ing out­comes

By the end of the course, the stu­dent must be able to:

• De­tect and ad­dress in­ef­fi­cien­cies in par­al­lel soft­ware
• Design and eval­u­ate soft­ware for mul­tiple par­al­lel plat­forms
• Design and eval­u­ate hard­ware for shared memory
• Com­pare and con­trast hard­ware design choices in par­al­lel plat­forms
• Demon­strate and de­scribe the op­er­a­tion of snoop­ing and dir­ect­ory co­her­ence pro­to­cols

Learn­ing pre­requis­ites

Re­quired courses

• CS-206 Con­cur­rency
• CS-208 Com­puter ar­chi­tec­ture / Ar­chi­tec­ture des Or­din­ateurs

Of­fice hours

Grad­ing

A tent­at­ive break­down of grade is giv­en be­low. Grades will be pos­ted on Moodle.