ACA Unit 8 Hardware and Software for VLIW and EPIC Notes — Unit 8 – Download as PDF File .pdf), Text File .txt) or read online. G-2 Appendix G Hardware and Software for VLIW and EPIC. In this chapter we discuss compiler technology for increasing the amount of par- allelism that we. View Notes – from ENG at BGS Institute of Technology. | Website for.

Author: Meztikora Diran
Country: Bhutan
Language: English (Spanish)
Genre: Politics
Published (Last): 2 June 2014
Pages: 491
PDF File Size: 8.35 Mb
ePub File Size: 19.8 Mb
ISBN: 778-1-48050-255-5
Downloads: 42472
Price: Free* [*Free Regsitration Required]
Uploader: Dikora

Most modern CPUs guess which branch will be taken even before the calculation is complete, so that they can load the instructions sortware the branch, or in some architectures even start to compute them speculatively.

The Loop Buffer Fetch column shows the fetch of instructions from the MLB, and the Execute column shows the instructions that are actually executed. Since determining the order of execution of operations including which operations can execute simultaneously is handled by the abd, the processor does not need the scheduling hardware that the three methods described above require.

Very long instruction word

Torczon, Engineering a Compiler. These bits are set at compile timethus relieving the hardware from calculating this dependency information. There is a distinct difference in the results between control- and loop-oriented benchmarks. The total schedule length TL is the number of cycles to complete one loop iteration. Each instruction on the C6X-1 processors is bit. Along with the above systems, during the same time —Intel implemented VLIW in the Intel itheir first bit microprocessor, and the first processor to implement VLIW on one chip.

For his dissertation Bulldog: Therefore, it is critical that a VLIW processor be a good compiler target. International Symposium on Computer Architecture. For code size reduction, smaller is better, and for speed improvement, larger is better. Certain branch instructions appearing in header-based fetch packets can reach half-word program addresses. Because the C6X compiler often produces execute packets with multiple instructions, swapping instructions within an execute packet increases the conversion rate of potential bit instructions.


In the above schedule, very little parallelism has been exploited because ins1ins2and ins3 must execute in order within the given loop iteration. The 84 benchmarks used for this analysis are organized into the groups enumerated below.

Co-design of Compiler and Hardware Techniques to Reduce Program Code Size on a VLIW Processor

Because VLIWs typically represent instructions scheduled in parallel with a longer instruction word that incorporates the individual instructions, this results in a much longer opcode termed very long to specify what executes on a given cycle.

The program size parameter measures softawre code-size reduction in the entire program, and the loop size parameter measures relative code-size reduction in the software-pipelined loops.

If the p-bit of instruction isthen instruction is part of the same execute packet as instruction. Only the kernel code is explicitly represented. Execution of a software-pipelined loop using the modulo loop buffer. A latency is the number of cycles it takes for the effect of an instruction to complete. Single-core Multi-core Manycore Heterogeneous architecture. Software-pipelined loops that do not use the MLB still benefit from loop collapsing. The p-bits are scanned from lower anx higher addresses.

In practice, however, epilog stages are usually much eoftware. Execute packet are padded with explicit parallel NOP instructions to prevent subsequent execute packets from spanning a fetch packet boundary. He also developed region scheduling methods to identify parallelism beyond basic blocks. In contrast, VLIW executes operations in parallel, based on a xoftware schedule, determined when programs are compiled. If the branch takes an unexpected way, the compiler has already generated compensating code to discard speculative results to preserve program semantics.

In contrast dpic superscalar processors, which have dedicated hardware to dynamically find ILP at run-time, VLIW architectures rely completely on the compiler to find ILP before program execution. Execution of a modulo scheduled loop. Due to sodtware design requirements of a high performance VLIW processor, bit instructions must be kept on a bit boundary. To accommodate these operation fields, VLIW instructions are usually at least 64 bits wide, and far wider on some architectures.

Example of software-pipelined loop with one epilog stage collapsed.

We presented the code-size reduction and performance impact of using these techniques to compile a set of 84 benchmarks. Each instruction in an execute packet must use a different wnd unit.


Instruction scheduling is used to fill the latency or delay slots with other useful instructions. For example, because of their long latency, branch instructions are often eppic by a multi-cycle NOP instruction. On the C6X-1 processors, execute packets cannot span a fetch packet boundary. These principles made it easier for compilers to softsare fast code.

Benchmark code size reduction and speedup performance improvement are measured on the following four configurations: In a VLIW, the compiler uses heuristics or profile information to guess the direction of a branch. Finally, the saturation bit bit 14 indicates whether basic arithmetic operations saturate on overflow and underflow.

For most superscalar designs, the instruction width is 32 bits or fewer. Reducing code size improves system performance by allowing space for more code in on-chip memory and program caches. Many of these benchmarks are complete applications. From Wikipedia, the free encyclopedia. Morgan Kaufmann Publishers Softaare.

Superscalar CPUs use hardware to decide which operations can run in parallel at runtime, while VLIW CPUs annd software the compiler to decide which operations can run in parallel hardwafe advance.

However, EPIC architecture is sometimes distinguished from a pure VLIW architecture, since EPIC advocates full instruction predication, rotating register files, and a very long instruction word that can encode non-parallel instruction groups.

The compressor does not swap or move instructions outside of execute packets, nor change registers of instructions in order to improve compression. Except for a few special case instructions such as the NOP, each instruction has a predicate encoded in the first four bits.

In this example, the cost of eliminating the second epilog stage the addition of two instructions outweighs the benefit eliminating one instruction. We call such instruction specialization tailoring.