Re: [gem5-dev] Review Request 3090: cpu: Support virtual addr in elastic traces

[Nilay Vaish]

---
This is an automatically generated e-mail. To reply, visit:
http://reviews.gem5.org/r/3090/#review7696
---

Ship it!


Ship It!

- Nilay Vaish


On Dec. 3, 2015, 1:31 a.m., Curtis Dunham wrote:
> 
> ---
> This is an automatically generated e-mail. To reply, visit:
> http://reviews.gem5.org/r/3090/
> ---
> 
> (Updated Dec. 3, 2015, 1:31 a.m.)
> 
> 
> Review request for Default.
> 
> 
> Repository: gem5
> 
> 
> Description
> ---
> 
> This patch adds support to optionally capture the virtual address and asid
> for load/store instructions in the elastic traces. If they are present in
> the traces, Trace CPU will set those fields of the request during replay.
> 
> This patch adds to the "Elastic Traces" patch series of RB3026-3031,3089.
> 
> 
> Diffs
> -
> 
>   src/cpu/o3/probe/elastic_trace.hh PRE-CREATION 
>   src/cpu/o3/probe/elastic_trace.cc PRE-CREATION 
>   src/cpu/trace/trace_cpu.hh PRE-CREATION 
>   src/cpu/trace/trace_cpu.cc PRE-CREATION 
>   src/mem/request.hh 135c16fa409d814dfbda979e59583534601dc1c7 
>   src/proto/inst_dep_record.proto PRE-CREATION 
>   util/decode_inst_dep_trace.py PRE-CREATION 
>   util/encode_inst_dep_trace.py PRE-CREATION 
>   src/cpu/o3/probe/ElasticTrace.py PRE-CREATION 
> 
> Diff: http://reviews.gem5.org/r/3090/diff/
> 
> 
> Testing
> ---
> 
> 
> Thanks,
> 
> Curtis Dunham
> 
>

___
gem5-dev mailing list
gem5-dev@gem5.org
http://m5sim.org/mailman/listinfo/gem5-dev

Re: [gem5-dev] Review Request 3090: cpu: Support virtual addr in elastic traces

[Nilay Vaish]

> On Dec. 1, 2015, 5:58 p.m., Nilay Vaish wrote:
> > src/cpu/o3/probe/elastic_trace.hh, lines 357-359
> > 
> >
> > Do we really need this?  I would rather always trace the virtual 
> > address as well.  I don't think there are any code correctness or clarity 
> > or performance gains to be made from this check.
> 
> Andreas Sandberg wrote:
> I haven't done the comparison, but my guess is that it'd grow the trace 
> by about 10%. AFAIK, the protobuf format should be pretty good at handling 
> unused fields. Besides, there are cases where the virtual address isn't 
> meaningful (e.g., when tracing device accesses).
> 
> Also, making virtual addresses mandatory would only really affect the 
> probe. We would still have to handle the case where old traces don't contain 
> them.
> 
> Radhika Jagtap wrote:
> The decision to make tracing virtual addresses as optional is because it 
> is not used in majority of memory-system studies and as Andreas S pointed out 
> it increases the size of the traces significantly.

Ok.


- Nilay


---
This is an automatically generated e-mail. To reply, visit:
http://reviews.gem5.org/r/3090/#review7665
---


On Dec. 3, 2015, 1:31 a.m., Curtis Dunham wrote:
> 
> ---
> This is an automatically generated e-mail. To reply, visit:
> http://reviews.gem5.org/r/3090/
> ---
> 
> (Updated Dec. 3, 2015, 1:31 a.m.)
> 
> 
> Review request for Default.
> 
> 
> Repository: gem5
> 
> 
> Description
> ---
> 
> This patch adds support to optionally capture the virtual address and asid
> for load/store instructions in the elastic traces. If they are present in
> the traces, Trace CPU will set those fields of the request during replay.
> 
> This patch adds to the "Elastic Traces" patch series of RB3026-3031,3089.
> 
> 
> Diffs
> -
> 
>   src/cpu/o3/probe/elastic_trace.hh PRE-CREATION 
>   src/cpu/o3/probe/elastic_trace.cc PRE-CREATION 
>   src/cpu/trace/trace_cpu.hh PRE-CREATION 
>   src/cpu/trace/trace_cpu.cc PRE-CREATION 
>   src/mem/request.hh 135c16fa409d814dfbda979e59583534601dc1c7 
>   src/proto/inst_dep_record.proto PRE-CREATION 
>   util/decode_inst_dep_trace.py PRE-CREATION 
>   util/encode_inst_dep_trace.py PRE-CREATION 
>   src/cpu/o3/probe/ElasticTrace.py PRE-CREATION 
> 
> Diff: http://reviews.gem5.org/r/3090/diff/
> 
> 
> Testing
> ---
> 
> 
> Thanks,
> 
> Curtis Dunham
> 
>

___
gem5-dev mailing list
gem5-dev@gem5.org
http://m5sim.org/mailman/listinfo/gem5-dev

Re: [gem5-dev] Review Request 3090: cpu: Support virtual addr in elastic traces

[Radhika Jagtap]

> On Dec. 1, 2015, 5:58 p.m., Nilay Vaish wrote:
> > src/cpu/o3/probe/elastic_trace.hh, lines 357-359
> > 
> >
> > Do we really need this?  I would rather always trace the virtual 
> > address as well.  I don't think there are any code correctness or clarity 
> > or performance gains to be made from this check.
> 
> Andreas Sandberg wrote:
> I haven't done the comparison, but my guess is that it'd grow the trace 
> by about 10%. AFAIK, the protobuf format should be pretty good at handling 
> unused fields. Besides, there are cases where the virtual address isn't 
> meaningful (e.g., when tracing device accesses).
> 
> Also, making virtual addresses mandatory would only really affect the 
> probe. We would still have to handle the case where old traces don't contain 
> them.

The decision to make tracing virtual addresses as optional is because it is not 
used in majority of memory-system studies and as Andreas S pointed out it 
increases the size of the traces significantly.


- Radhika


---
This is an automatically generated e-mail. To reply, visit:
http://reviews.gem5.org/r/3090/#review7665
---


On Sept. 3, 2015, 4:17 p.m., Curtis Dunham wrote:
> 
> ---
> This is an automatically generated e-mail. To reply, visit:
> http://reviews.gem5.org/r/3090/
> ---
> 
> (Updated Sept. 3, 2015, 4:17 p.m.)
> 
> 
> Review request for Default.
> 
> 
> Repository: gem5
> 
> 
> Description
> ---
> 
> This patch adds support to optionally capture the virtual address and asid
> for load/store instructions in the elastic traces. If they are present in
> the traces, Trace CPU will set those fields of the request during replay.
> 
> This patch adds to the "Elastic Traces" patch series of RB3026-3031,3089.
> 
> 
> Diffs
> -
> 
>   src/cpu/o3/probe/ElasticTrace.py PRE-CREATION 
>   src/cpu/o3/probe/elastic_trace.hh PRE-CREATION 
>   src/cpu/o3/probe/elastic_trace.cc PRE-CREATION 
>   src/cpu/trace/trace_cpu.hh PRE-CREATION 
>   src/cpu/trace/trace_cpu.cc PRE-CREATION 
>   src/proto/inst_dep_record.proto PRE-CREATION 
>   util/decode_inst_dep_trace.py PRE-CREATION 
>   util/encode_inst_dep_trace.py PRE-CREATION 
> 
> Diff: http://reviews.gem5.org/r/3090/diff/
> 
> 
> Testing
> ---
> 
> 
> Thanks,
> 
> Curtis Dunham
> 
>

___
gem5-dev mailing list
gem5-dev@gem5.org
http://m5sim.org/mailman/listinfo/gem5-dev

Re: [gem5-dev] Review Request 3090: cpu: Support virtual addr in elastic traces

[Curtis Dunham]

---
This is an automatically generated e-mail. To reply, visit:
http://reviews.gem5.org/r/3090/
---

(Updated Dec. 3, 2015, 1:31 a.m.)


Review request for Default.


Repository: gem5


Description
---

This patch adds support to optionally capture the virtual address and asid
for load/store instructions in the elastic traces. If they are present in
the traces, Trace CPU will set those fields of the request during replay.

This patch adds to the "Elastic Traces" patch series of RB3026-3031,3089.


Diffs (updated)
-

  src/cpu/o3/probe/elastic_trace.hh PRE-CREATION 
  src/cpu/o3/probe/elastic_trace.cc PRE-CREATION 
  src/cpu/trace/trace_cpu.hh PRE-CREATION 
  src/cpu/trace/trace_cpu.cc PRE-CREATION 
  src/mem/request.hh 135c16fa409d814dfbda979e59583534601dc1c7 
  src/proto/inst_dep_record.proto PRE-CREATION 
  util/decode_inst_dep_trace.py PRE-CREATION 
  util/encode_inst_dep_trace.py PRE-CREATION 
  src/cpu/o3/probe/ElasticTrace.py PRE-CREATION 

Diff: http://reviews.gem5.org/r/3090/diff/


Testing
---


Thanks,

Curtis Dunham

___
gem5-dev mailing list
gem5-dev@gem5.org
http://m5sim.org/mailman/listinfo/gem5-dev

Re: [gem5-dev] Review Request 3090: cpu: Support virtual addr in elastic traces

[Andreas Sandberg]

> On Dec. 1, 2015, 5:58 p.m., Nilay Vaish wrote:
> > src/cpu/o3/probe/elastic_trace.hh, lines 357-359
> > 
> >
> > Do we really need this?  I would rather always trace the virtual 
> > address as well.  I don't think there are any code correctness or clarity 
> > or performance gains to be made from this check.

I haven't done the comparison, but my guess is that it'd grow the trace by 
about 10%. AFAIK, the protobuf format should be pretty good at handling unused 
fields. Besides, there are cases where the virtual address isn't meaningful 
(e.g., when tracing device accesses).

Also, making virtual addresses mandatory would only really affect the probe. We 
would still have to handle the case where old traces don't contain them.


- Andreas


---
This is an automatically generated e-mail. To reply, visit:
http://reviews.gem5.org/r/3090/#review7665
---


On Sept. 3, 2015, 5:17 p.m., Curtis Dunham wrote:
> 
> ---
> This is an automatically generated e-mail. To reply, visit:
> http://reviews.gem5.org/r/3090/
> ---
> 
> (Updated Sept. 3, 2015, 5:17 p.m.)
> 
> 
> Review request for Default.
> 
> 
> Repository: gem5
> 
> 
> Description
> ---
> 
> This patch adds support to optionally capture the virtual address and asid
> for load/store instructions in the elastic traces. If they are present in
> the traces, Trace CPU will set those fields of the request during replay.
> 
> This patch adds to the "Elastic Traces" patch series of RB3026-3031,3089.
> 
> 
> Diffs
> -
> 
>   src/cpu/o3/probe/ElasticTrace.py PRE-CREATION 
>   src/cpu/o3/probe/elastic_trace.hh PRE-CREATION 
>   src/cpu/o3/probe/elastic_trace.cc PRE-CREATION 
>   src/cpu/trace/trace_cpu.hh PRE-CREATION 
>   src/cpu/trace/trace_cpu.cc PRE-CREATION 
>   src/proto/inst_dep_record.proto PRE-CREATION 
>   util/decode_inst_dep_trace.py PRE-CREATION 
>   util/encode_inst_dep_trace.py PRE-CREATION 
> 
> Diff: http://reviews.gem5.org/r/3090/diff/
> 
> 
> Testing
> ---
> 
> 
> Thanks,
> 
> Curtis Dunham
> 
>

___
gem5-dev mailing list
gem5-dev@gem5.org
http://m5sim.org/mailman/listinfo/gem5-dev

Re: [gem5-dev] Review Request 3090: cpu: Support virtual addr in elastic traces

[Nilay Vaish]

---
This is an automatically generated e-mail. To reply, visit:
http://reviews.gem5.org/r/3090/#review7665
---



src/cpu/o3/probe/elastic_trace.hh (lines 357 - 359)


Do we really need this?  I would rather always trace the virtual address as 
well.  I don't think there are any code correctness or clarity or performance 
gains to be made from this check.


- Nilay Vaish


On Sept. 3, 2015, 4:17 p.m., Curtis Dunham wrote:
> 
> ---
> This is an automatically generated e-mail. To reply, visit:
> http://reviews.gem5.org/r/3090/
> ---
> 
> (Updated Sept. 3, 2015, 4:17 p.m.)
> 
> 
> Review request for Default.
> 
> 
> Repository: gem5
> 
> 
> Description
> ---
> 
> This patch adds support to optionally capture the virtual address and asid
> for load/store instructions in the elastic traces. If they are present in
> the traces, Trace CPU will set those fields of the request during replay.
> 
> This patch adds to the "Elastic Traces" patch series of RB3026-3031,3089.
> 
> 
> Diffs
> -
> 
>   src/cpu/o3/probe/ElasticTrace.py PRE-CREATION 
>   src/cpu/o3/probe/elastic_trace.hh PRE-CREATION 
>   src/cpu/o3/probe/elastic_trace.cc PRE-CREATION 
>   src/cpu/trace/trace_cpu.hh PRE-CREATION 
>   src/cpu/trace/trace_cpu.cc PRE-CREATION 
>   src/proto/inst_dep_record.proto PRE-CREATION 
>   util/decode_inst_dep_trace.py PRE-CREATION 
>   util/encode_inst_dep_trace.py PRE-CREATION 
> 
> Diff: http://reviews.gem5.org/r/3090/diff/
> 
> 
> Testing
> ---
> 
> 
> Thanks,
> 
> Curtis Dunham
> 
>

___
gem5-dev mailing list
gem5-dev@gem5.org
http://m5sim.org/mailman/listinfo/gem5-dev

[gem5-dev] Review Request 3090: cpu: Support virtual addr in elastic traces

[Curtis Dunham]

---
This is an automatically generated e-mail. To reply, visit:
http://reviews.gem5.org/r/3090/
---

Review request for Default.


Repository: gem5


Description
---

This patch adds support to optionally capture the virtual address and asid
for load/store instructions in the elastic traces. If they are present in
the traces, Trace CPU will set those fields of the request during replay.

This patch adds to the "Elastic Traces" patch series of RB3026-3031,3089.


Diffs
-

  src/cpu/o3/probe/ElasticTrace.py PRE-CREATION 
  src/cpu/o3/probe/elastic_trace.hh PRE-CREATION 
  src/cpu/o3/probe/elastic_trace.cc PRE-CREATION 
  src/cpu/trace/trace_cpu.hh PRE-CREATION 
  src/cpu/trace/trace_cpu.cc PRE-CREATION 
  src/proto/inst_dep_record.proto PRE-CREATION 
  util/decode_inst_dep_trace.py PRE-CREATION 
  util/encode_inst_dep_trace.py PRE-CREATION 

Diff: http://reviews.gem5.org/r/3090/diff/


Testing
---


Thanks,

Curtis Dunham

___
gem5-dev mailing list
gem5-dev@gem5.org
http://m5sim.org/mailman/listinfo/gem5-dev