Optimising Compilers

EDA230 Optimizing Compilers

 7.5 points, LP1 2011

•  Course responsible: Jonas Skeppstedt
• Course ombudsman: Fredrik Strandin 
• Litteratur: J. Skeppstedt, An Introduction to Theory of Optimising Compilers 

Schedule of the 12 lectures

The lectures are Wednesdays at 13-15 in E:1406 and Fridays 10-12 in different locations. See below.

                                        

F1	31/8	Introduction to optimising compilers	E:1406
F2	2/9	Control flow graphs and dominance analysis	E:1406
F3	7/9	Translation to and from SSA Form	V:B
F4	9/9	Register allocation	A:B
F5	14/9	Constant Propagation	E:1406
F6	16/9	Value numbering	E:1406
F7	21/9	Partial redundancy elimination	E:1406
F8	23/9	Operator strength reduction	E:1406
F9	28/9	Dead code elimination	E:1406
F10	30/9	Data dependence analysis	E:1406
F11	5/10	SIMD vectorization	E:1406
F12	7/10	Modulo scheduling	E:1406
F13	12/10	Repetition and reserved time for presentation of article	E:1406
R	14/10	Reserve time for presentation of article	E:1406

 

Exercises

There are two occasions to select from: Mondays 10-12 and Thursdays 13-15 both in E:2116. 

The exercises are in weeks 2-7 and are compulsory, and you are allowed to skip one exercise.

The reason they are compulsory is for the presentation of scientific articles. If nobody will present any article, you can leave. You need to present one article during the course.

 

 
Competition
 

                  

Year	Position	Group	Faculty/Programme	Clock cycles
2011	1	Linus Åkesson	CS Dept/PhD Student	112805
	2	Rasmus Barringer	CS Dept/PhD Student	449009
	3	Tobias Alex-Petersen	Engineering/E	831832
2010	1	Joakim Andersson/Jon Sten	Engineering/D	126616
	2	Carl Cristian Arlock/Erik Pettersson/Mikael Rudner	Engineering/D, LTH, LiTH, LiTH	149909
	3	Dan Kvelstad	Engineering/D	150443
2009	1	Manfred Dellkrantz/Jesper Öqvist	Engineering/D	950
	2	Fredrik Persson/Andreas Larsson	Engineering/D	17320
	3	Niklas Fors/Erik Linge	Engineering/D	20680
2008	1	Jonas Paulsson	Engineering/D	18977
	2	Hans Wennborg	Engineering/D	22437
	3	Johan Gunnarsson	Engineering/D	32698
2007	1	Bjorn Carlin (Pi)/Hans Gylling (D)	Engineering/Pi, D	1047
	2	Magnus Hojer	Engineering/D	7865
	3	Monthadar Al-Jaberi/Mario Boikov	Engineering/D	150367
	3	Torbjorn Andersson	Engineering/E	150367
	3	Philip Carnstam/Jan-Ola Svensson	Engineering/D	150367
2006	1	Fredrik Nilsson	Engineering/D	3388
	2	David Henningsson	Science/CS	10229
	3	Erik Andre/Mitch Selander/Anders Sunnermark	Engineering/D	43756
2005	1	Mats Mattsson	Engineering/Pi	5738
	2	Magnus Malmborn/Johan Heander	Engineering/F	16844
	3	Mikael Collin	KTH/Doktorand	63477
2004	1	Jonas Åström	Engineering/Pi	21538
	2	Anders Hugosson	Science/CS	30660
	3	Fredrik Redgård	Engineering/D	41863
2003	1	Alexander Malmberg	Engineering/D	3744
	2	Rickard Möller	Science/CS	6551
	3	Tobias Johansson	Engineering/D	65088
2002	1	Richard Johansson	Engineering/D	7930
	2	Fredrik Brozén and Anders Gustafsson	Engineering/D	13517
	3	Jon Hasselgren	Engineering/D	21842
2001	1	Bo Do and Per Fransson	Science/CS	65776
	2	Karl Torpel	Engineering/D	65998
	3	Bartek Kania	Science/CS	71874
2000	1	Per Cederberg	Robotics/Doktorand	10768
	2	Ola Nilsson and Hans Wachtmeister	Engineering/D	61071
	3	Klas Lindberg	Science/CS	64935

 

Goals with the course
 The compiler is the programmer's most important tool. To be able to implement programs which are correct, efficient, and easy to maintain, it is important that the programmer knows what must be optimized by hand, and which work the compiler performs faster and better. Programmers become more productive with this knowledge. 

 The goals with the course is that the student after the course will: 

•  have a thorough understanding of the possibilities and limitations of optimizing compilers, and 
•  have a thorough knowledge about which algorithms and data structures are used in modern optimizing compilers. 

 Contents
 Control-flow analysis, data-flow analysis, dependence analysis, alias analysis, redundancy elimination, loop optimisations, procedure optimisations, register allocation, code scheduling, optimisation for object oriented languages, optimisation of the memory hierarchy, and vectorization for processors with SIMD instructions, such as PowerPC/AltiVec. An introduction to parallelisation for shared memory multiprocessors is also given. 

 

 

• Print version