Compiler optimizations for low power

العناوين الأخرى

تحسينات المترجمات لتقليل استهلاك الطاقة

مقدم أطروحة جامعية

Abu Rummalah, Najm al-Din Mahmud S.

مشرف أطروحة جامعية

Shubaki, Ghassan

أعضاء اللجنة

al-Qarallah, Isam A.
Jabri, Riyad
Sarhan, Samil

الجامعة

جامعة الأميرة سمية للتكنولوجيا

الكلية

كلية الملك الحسين لعلوم الحوسبة

دولة الجامعة

الأردن

الدرجة العلمية

ماجستير

تاريخ الدرجة العلمية

2016

الملخص الإنجليزي

Energy efficiency is becoming more important with the proliferation of modern embedded systems and handheld devices.

Energy consumption does not only depend on the hardware design, but it also depends on the software being executed on the processor.

Being the program that generates and optimizes the machine executable code for nearly all software applications, the compiler plays an important role in determining a program’s energy consumption.

Multiple compiler optimization algorithms have been proposed for reducing energy consumption.

However, the impact of many of these algorithms on energy consumption has not been studied using real hardware measurements.

In most previous work, only simulation results have been reported.

In this work we study the impact of one compiler optimization, namely instruction scheduling, on CPU energy consumption using direct hardware measurements.

The study is done for two target processors: an ARM processor and an Intel x86 processor.

We propose a new hardware experimental setup to measure energy consumption on the ARM processor while Intel’s built-in energy monitor is used for the Intel x86 processor.

The experimental results show that the impact of a compiler on energy consumption is significant and measurable.

Yet the results suggest that it is quite challenging to develop an energy-aware instruction scheduling algorithm, because minimizing energy-related factors, such as bit switching and the degree of instruction-level parallelism, may conflict with execution time and execution time has a higher impact on energy consumption than other energy-related factors.

التخصصات الرئيسية

تكنولوجيا المعلومات وعلم الحاسوب

عدد الصفحات

71

قائمة المحتويات

• APA
• MLA
• AMA

لغة النص

الإنجليزية

نوع البيانات

رسائل جامعية

رقم السجل

BIM-720827