Thanks for Georgios and Crio's comments!
I want to test the performance, and wish the result could be close to real devices.
I have searched for several SoC (ARMv8) from the link (https://en.wikipedia.org/wiki/Comparison_of_ARMv8-A_cores).
While the frequency and cache/TLB size are easy to figure out, I can not found official specs to describe the Cache access latency.
I found a website, which records a tested Cortex-A57 result in (https://www.7-cpu.com/cpu/Cortex-A57.html).
I put the result here (just as a reference for others may need):
L1 Data Cache Latency = 4 cycles for simple access via pointer
L1 Data Cache Latency = 5 cycles for access with complex address calculation (size_t n, *p; n = p[n]).
L2 Cache Latency = 18 cycles
----- Original Message -----
From: "Georgios S. Bousdras" <***@ulb.ac.be>
To: "gem5 users mailing list" <email@example.com>
Sent: Monday, December 3, 2018 8:55:47 PM
Subject: Re: [gem5-users] Any recommendation of ARM HPI configuration
You need to know what are you going to test (perfomance/power).
Take in to account that different architectures support different applications.
Define the demands of the system and then do what Ciro said, find an existed SoC which match your requirements and create your processor similar to it.
PhD Researchers - Embedded Systems
BEAMS - ULB
Post by 杜东
I am using GEM5 ARM with the Minor CPU (specifically, HPI mode).
TLB: 256 entries I/D TLB
L1 I/D cache: 1 cycle (data/tag/response)
L2 I/D cache: 13 cycles for data and tag access
memory type: DDR3_1600_8x8.
This default configuration seems too great?
I am not sure whether real devices can achieve the performance, especially 1 cycle L1 cache access.
So do you have any recommendation on the configurations?
Any comments are welcomed!
gem5-users mailing list