SLIDE 1
Lecture 07 - Multicore Computation
Lecture 07 - Multicore Computation
This was thinking mid-90s.
Lecture 07 Multicore Computation Lecture based on notes from John - - PowerPoint PPT Presentation
Lecture 07 Multicore Computation Lecture based on notes from John - - PowerPoint PPT Presentation
Lecture 07 - Multicore Computation Lecture 07 Multicore Computation Lecture based on notes from John Mellor-Crummey Department of Computer Science Rice University & Jernej Barbic Lecture 07 - Multicore Computation This was thinking
SLIDE 2
SLIDE 3
Lecture 07 - Multicore Computation
SLIDE 4
Lecture 07 - Multicore Computation
Circuit complexity and interconnect delay limit practicality of support structures for larger issue width
SLIDE 5
Lecture 07 - Multicore Computation
SLIDE 6
Lecture 07 - Multicore Computation
SLIDE 7
Lecture 07 - Multicore Computation
SLIDE 8
Lecture 07 - Multicore Computation
SLIDE 9
Lecture 07 - Multicore Computation
SLIDE 10
10 10 Lecture 07 - Multicore Computation
SLIDE 11
11 11 Lecture 07 - Multicore Computation
SLIDE 12
12 12 Lecture 07 - Multicore Computation
Some important points
- Technology alone is not driving push to multi-core
– What was state of the art - more issue, superscalar - provides diminishing performance returns b/c of program properties
- Still, performance gains possible with scaling
- If CCs/instruction performance gains tapped out +
scaling performance inhibited (b/c of lower Vdd, lower clock rates), where does performance come from?
SLIDE 13
13 13 Lecture 07 - Multicore Computation
Some important points
- Performance must come from combination of
parallelism + previously ignored HW optimizations
– E.g. instead of getting 2x from technology, get 10% from A, 5% from B, etc.
SLIDE 14
14 14 Lecture 07 - Multicore Computation
SLIDE 15
15 15 Lecture 07 - Multicore Computation
SLIDE 16
16 16 Lecture 07 - Multicore Computation
SLIDE 17
17 17 Lecture 07 - Multicore Computation
SLIDE 18
18 18 Lecture 07 - Multicore Computation
SLIDE 19
19 19 Lecture 07 - Multicore Computation
SLIDE 20
20 20 Lecture 07 - Multicore Computation
The cores fit on a single processor socket (also called CMP - chip multiprocessor)
SLIDE 21
21 21 Lecture 07 - Multicore Computation
SLIDE 22
22 22 Lecture 07 - Multicore Computation
SLIDE 23
23 23 Lecture 07 - Multicore Computation
Back to case study…
SLIDE 24
24 24 Lecture 07 - Multicore Computation
(standard benchmarks parallelized for comparison)
SLIDE 25
25 25 Lecture 07 - Multicore Computation
SLIDE 26
26 26 Lecture 07 - Multicore Computation
SLIDE 27
27 27 Lecture 07 - Multicore Computation
(If CPU time constant, performance comes from parallelism)
SLIDE 28
28 28 Lecture 07 - Multicore Computation
Take Aways
SLIDE 29
29 29 Lecture 07 - Multicore Computation
SLIDE 30
30 30 Lecture 07 - Multicore Computation
SLIDE 31
31 31 Lecture 07 - Multicore Computation
SLIDE 32
32 32 Lecture 07 - Multicore Computation
Multi-core flavors
- Cores need not be the same
– (If they are, we talk about symmetric core machines) – (If not, asymmetric)
- Imagine FPGA + GP processor?
SLIDE 33
33 33 Lecture 07 - Multicore Computation
Other issues: (Amdahl’s Law and Parallelization)
SLIDE 34
34 34 Lecture 07 - Multicore Computation
SLIDE 35
35 35 Lecture 07 - Multicore Computation
SLIDE 36
36 36 Lecture 07 - Multicore Computation
SLIDE 37
37 37 Lecture 07 - Multicore Computation
SLIDE 38
38 38 Lecture 07 - Multicore Computation
SLIDE 39
39 39 Lecture 07 - Multicore Computation
SLIDE 40
40 40 Lecture 07 - Multicore Computation
SLIDE 41
41 41 Lecture 07 - Multicore Computation
Other issues: Core-to-core communication
Must factor in communication costs in processing time too…
SLIDE 42
42 42 Lecture 07 - Multicore Computation