VLIW

Question 1

What is VLIW?

Answer 1

Very Long Instruction Word Processor. The compiler is responsible for ILP and the processor just executes the code

Question 2

How does a VLIW processor work?

Answer 2

each instruction is equivalent to N independent instructions. It executes in order

Question 3

What are the pros of a VLIW?

Answer 3

1. Compiler does the heavy lifting ==> don’t need to waste time scheduling while we are executing
2. Simpler hardware
3. Can be more energy efficient
4. Works well on loops and “regular” code (multiplying matrices, etc)

Question 4

What are the cons of a VLIW?

Answer 4

1. Latencies are not always the same => hard for compiler to know in advance how long an instruction will take (load is slower if we have a cache miss)
2. Most applications don’t fit under the ‘regular’ code model
3. Not good with pointers or decision making
4. High code bloat / lots of nops

Question 5

When do we want to use VLIW?

Answer 5

Specific use cases such as signal processing where there are lots of loops with many iterations