Hi,
I tried fast forward with SMT processor. I get a fatal: Workload Size (2) > Max Supported Threads (1) on This CPU. I think that is because atomic mode doesn't support SMT. Is there any other method to do fast forward in SMT mode? Thanks! Best regards Fangfei ________________________________ From: gem5-users-boun...@gem5.org [gem5-users-boun...@gem5.org] on behalf of Tao Zhang [tao.zhang.0...@gmail.com] Sent: Thursday, November 08, 2012 12:06 AM To: 'gem5 users mailing list' Subject: Re: [gem5-users] fast forward in SMT processor I never use SMT mode after the fast-forward. But I think there is no difference between O3CPU with/without SMT ( I did use ordinary O3CPU after the fast-forward). You can refer to configs/common/Simulation.py to see how fast-forward works. Basically, the system maintains two cpu lists, one atomic and the other one your timing cpu model. The fast-forward part is done by atomicCPU. It has nothing to do with the timing cpu model. Why not use “-F” to see whether it works or not? -Tao From: gem5-users-boun...@gem5.org [mailto:gem5-users-boun...@gem5.org] On Behalf Of Fangfei Liu Sent: Wednesday, November 07, 2012 11:58 PM To: gem5 users mailing list Subject: Re: [gem5-users] fast forward in SMT processor Hi, In my understanding, fast-forward is done by running the program in atomic cpu mode. That means for the first X instructions, gem5 simulates SMT processor in atomic mode, but it's supposed not to be able to simulate SMT in atomic mode. Have you ever tried to do fast-forward for SMT processor? Thanks! Fangfei ________________________________ From: gem5-users-boun...@gem5.org<mailto:gem5-users-boun...@gem5.org> [gem5-users-boun...@gem5.org] on behalf of Tao Zhang [tao.zhang.0...@gmail.com] Sent: Wednesday, November 07, 2012 11:51 PM To: 'gem5 users mailing list' Subject: Re: [gem5-users] fast forward in SMT processor I didn’t get your concern. What do you worry about? The atomicCPU can quickly fast-forward the benchmark to your ROI while it will cost much more time if detailed or inorder cpu are used to complete the fast-forward. -Tao From: gem5-users-boun...@gem5.org<mailto:gem5-users-boun...@gem5.org> [mailto:gem5-users-boun...@gem5.org] On Behalf Of Fangfei Liu Sent: Wednesday, November 07, 2012 11:49 PM To: gem5 users mailing list Subject: Re: [gem5-users] fast forward in SMT processor Thanks for your reply. Since before the switch, it runs in atomic mode with SMT. Does it matter? ________________________________ From: gem5-users-boun...@gem5.org<mailto:gem5-users-boun...@gem5.org> [gem5-users-boun...@gem5.org] on behalf of Tao Zhang [tao.zhang.0...@gmail.com] Sent: Wednesday, November 07, 2012 11:45 PM To: 'gem5 users mailing list' Subject: Re: [gem5-users] fast forward in SMT processor After the fast-forward, gem5 will switch the cpu to detailed or inorder mode, which can satisfy your requirement. -Tao From: gem5-users-boun...@gem5.org<mailto:gem5-users-boun...@gem5.org> [mailto:gem5-users-boun...@gem5.org] On Behalf Of Fangfei Liu Sent: Wednesday, November 07, 2012 11:32 PM To: gem5-users@gem5.org<mailto:gem5-users@gem5.org> Subject: [gem5-users] fast forward in SMT processor Hello, I want to simulate two SPEC 2006 benchmarks in SMT processor. (SE mode) I noticed that only detailed and inorder cpu type support SMT. Since SPEC 2006 is too large to run to completion, I have to fast forward to some region of interest. But fast forward runs in atomic mode. I was wondering whether there is a way to do fast forward on SMT processor. Thanks in advance! Best regards Fangfei
_______________________________________________ gem5-users mailing list gem5-users@gem5.org http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
