chore(ci): set regression targets for cpu

[soonum]

This change is 

[soonum]

I need to add support for shortint layer in the data extractor.
That being said, execution time for this new regression profile is around 2 hours 15 mins.

[IceTDrinker]

A few questions/comments, also, why is it taking 2h+ to get those benchmarks in that case ?

@IceTDrinker reviewed 1 of 1 files at r1, 1 of 2 files at r2, 1 of 1 files at r3, all commit messages.
Reviewable status: all files reviewed, 3 unresolved discussions (waiting on @soonum)

---

ci/regression.toml line 84 at r3 (raw file):

    "swap_claim::no_cmux"
]
target.hlapi-erc20 = ["transfer::whitepaper", "transfer::no_cmux"]

why two variants here ?

---

ci/perf_regression/perf_regression.py line 634 at r3 (raw file):

        try:
            op_perf = OperationPerformance(op_name, baseline_data, head_branch_value)
        except statistics.StatisticsError as err:

when did you encounter this issue ?

---

ci/perf_regression/perf_regression.py line 740 at r3 (raw file):

        "<summary><strong>View All Benchmarks</strong></summary>",
        "",
        f"* Minor change threshold: +/- {MINOR_CHANGE_SCALE_FACTOR}x StdDev",

nice idea, do we print by how many std dev the latest value diverges ?

[soonum]

Reviewable status: all files reviewed, 3 unresolved discussions (waiting on @antoniupop and @IceTDrinker)

---

ci/regression.toml line 84 at r3 (raw file):

Previously, IceTDrinker wrote…

why two variants here ?

Because @antoniupop told me they use both of them in the coprocessor code.

---

ci/perf_regression/perf_regression.py line 634 at r3 (raw file):

Previously, IceTDrinker wrote…

when did you encounter this issue ?

When there is only one data point in main. Which is unlikely to happen except when we introduce a brand new benchmark (either new operation or a new params set)

---

ci/perf_regression/perf_regression.py line 740 at r3 (raw file):

Previously, IceTDrinker wrote…

nice idea, do we print by how many std dev the latest value diverges ?

Nope but I can add it easily. Should I do it ?

[IceTDrinker]

ci/perf_regression/perf_regression.py line 634 at r3 (raw file):

Previously, soonum (David Testé) wrote…

When there is only one data point in main. Which is unlikely to happen except when we introduce a brand new benchmark (either new operation or a new params set)

ok yeah

[IceTDrinker]

Thanks a lot ! :)

Reviewable status: all files reviewed, 1 unresolved discussion (waiting on @soonum)

---

ci/regression.toml line 84 at r3 (raw file):

Previously, soonum (David Testé) wrote…

Because @antoniupop told me they use both of them in the coprocessor code.

ok surprising which variant is set to be standardized in an ERC ?

---

ci/perf_regression/perf_regression.py line 740 at r3 (raw file):

Previously, soonum (David Testé) wrote…

Nope but I can add it easily. Should I do it ?

not necessarilyin this PR to keep the focus and avoid having this PR being delayed

[soonum]

Reviewable status: all files reviewed, 1 unresolved discussion (waiting on @soonum)

---

ci/regression.toml line 84 at r3 (raw file):

Previously, IceTDrinker wrote…

ok surprising which variant is set to be standardized in an ERC ?

No idea. Let's keep the two variants for now and remove the unused if they finally use only one.

[IceTDrinker]

Reviewable status: all files reviewed, 1 unresolved discussion

---

ci/regression.toml line 84 at r3 (raw file):

Previously, soonum (David Testé) wrote…

No idea. Let's keep the two variants for now and remove the unused if they finally use only one.

let's have a backlog issue maybe to avoid forgetting about it ?