Split module interface rule into interface and code outputs

[crtschin]

I initially wanted to optimize interaction times and ran benchmarks. This led me to the high rulesBuilt number for the simple edit benchmark. Using #5055, shows the built rules being dominated by (re)loading module interfaces. This despite the edit only modifying a string in the benchmark, which doesn't touch the module's interface.

Investigating this further leads to the fingerprint used to determine when to reload a modules interface. The fingerprint is defined as:

hiFileFingerPrint :: HiFileResult -> ByteString
hiFileFingerPrint HiFileResult{..} = hirIfaceFp hirIface <> fromMaybe "" hirCoreFp

What I think's wrong here is that the core file is more susceptible to being invalidated despite no interface changes occurring. A module's interface being invalidated is infectious, and dirties a module's transitive reverse closure.

This PR fixes it by being explicit on which of a module's artefacts a rule depends on. GetLinkable can reference the on-disk core file, while GetModIface can stick to strictly a module's ABI, primarily so GetModIface can ignore non-ABI changes.

---

Running the edit bench experiment on both MultiLayerModules and MultiLayerModulesNoTH using cabal bench and the rule build histogram counts from #5055.

before with TH

+--------------------------------+------+
|              rule              | edit |
+--------------------------------+------+
| GhcSessionDeps                 | 601  |
| GetModIfaceFromDisk            | 599  |
| GenerateCore                   | 3    |
| GetHieAst                      | 3    |
| TypeCheck                      | 3    |
| GetModIface                    | 2    |
| GetParsedModule                | 2    |
| NonLocalCompletions            | 2    |
| GetFileContents                | 1    |
| GetModSummary                  | 1    |
| GetModSummaryWithoutTimestamps | 1    |
| GetModificationTime            | 1    |
| TOTAL                          | 1219 |
+--------------------------------+------+
+------+---------+---------+---------+-------+-------+----------+-----------+-------------+--------+------------+--------------+--------------+------------+-----------+-------------+
| name | success | samples | startup | setup | userT | delayedT | 1stBuildT | avgPerRespT | totalT | rulesBuilt | rulesChanged | rulesVisited | rulesTotal | ruleEdges | ghcRebuilds |
+------+---------+---------+---------+-------+-------+----------+-----------+-------------+--------+------------+--------------+--------------+------------+-----------+-------------+
| edit | True    | 50      | 0.24s   | 0.00s | 1m05s | 2.07s    | 2.06s     | 0.44s       | 1m07s  | 1219       | 619          | 5141         | 11556      | 118622    | 141         |
+------+---------+---------+---------+-------+-------+----------+-----------+-------------+--------+------------+--------------+--------------+------------+-----------+-------------+

after with TH

+--------------------------------+------+
|              rule              | edit |
+--------------------------------+------+
| GenerateCore                   | 3    |
| GetHieAst                      | 3    |
| GetParsedModule                | 3    |
| TypeCheck                      | 3    |
| GetModArtefacts                | 2    |
| GetModIface                    | 2    |
| GetFileContents                | 1    |
| GetModSummary                  | 1    |
| GetModSummaryWithoutTimestamps | 1    |
| GetModificationTime            | 1    |
| TOTAL                          | 20   |
+--------------------------------+------+
+------+---------+---------+---------+-------+--------+----------+-----------+-------------+--------+------------+--------------+--------------+------------+-----------+-------------+
| name | success | samples | startup | setup | userT  | delayedT | 1stBuildT | avgPerRespT | totalT | rulesBuilt | rulesChanged | rulesVisited | rulesTotal | ruleEdges | ghcRebuilds |
+------+---------+---------+---------+-------+--------+----------+-----------+-------------+--------+------------+--------------+--------------+------------+-----------+-------------+
| edit | True    | 50      | 0.26s   | 0.00s | 14.34s | 2.16s    | 2.17s     | 0.10s       | 16.51s | 20         | 17           | 5782         | 12197      | 143303    | 101         |
+------+---------+---------+---------+-------+--------+----------+-----------+-------------+--------+------------+--------------+--------------+------------+-----------+-------------+

before without TH

+--------------------------------+------+
|              rule              | edit |
+--------------------------------+------+
| GenerateCore                   | 3    |
| GetHieAst                      | 3    |
| GetParsedModule                | 3    |
| TypeCheck                      | 3    |
| GetModIface                    | 2    |
| GetFileContents                | 1    |
| GetModSummary                  | 1    |
| GetModSummaryWithoutTimestamps | 1    |
| GetModificationTime            | 1    |
| TOTAL                          | 18   |
+--------------------------------+------+
+------+---------+---------+---------+-------+--------+----------+-----------+-------------+--------+------------+--------------+--------------+------------+-----------+-------------+
| name | success | samples | startup | setup | userT  | delayedT | 1stBuildT | avgPerRespT | totalT | rulesBuilt | rulesChanged | rulesVisited | rulesTotal | ruleEdges | ghcRebuilds |
+------+---------+---------+---------+-------+--------+----------+-----------+-------------+--------+------------+--------------+--------------+------------+-----------+-------------+
| edit | True    | 50      | 0.30s   | 0.00s | 17.02s | 2.37s    | 2.39s     | 0.12s       | 19.39s | 18         | 15           | 5142         | 11556      | 118620    | 101         |
+------+---------+---------+---------+-------+--------+----------+-----------+-------------+--------+------------+--------------+--------------+------------+-----------+-------------+

after without TH

+--------------------------------+------+
|              rule              | edit |
+--------------------------------+------+
| GenerateCore                   | 3    |
| GetHieAst                      | 3    |
| GetParsedModule                | 3    |
| TypeCheck                      | 3    |
| GetModArtefacts                | 2    |
| GetFileContents                | 1    |
| GetModSummary                  | 1    |
| GetModSummaryWithoutTimestamps | 1    |
| GetModificationTime            | 1    |
| TOTAL                          | 18   |
+--------------------------------+------+
+------+---------+---------+---------+-------+--------+----------+-----------+-------------+--------+------------+--------------+--------------+------------+-----------+-------------+
| name | success | samples | startup | setup | userT  | delayedT | 1stBuildT | avgPerRespT | totalT | rulesBuilt | rulesChanged | rulesVisited | rulesTotal | ruleEdges | ghcRebuilds |
+------+---------+---------+---------+-------+--------+----------+-----------+-------------+--------+------------+--------------+--------------+------------+-----------+-------------+
| edit | True    | 50      | 0.26s   | 0.00s | 14.58s | 2.14s    | 2.13s     | 0.10s       | 16.72s | 18         | 15           | 5142         | 12197      | 143301    | 101         |
+------+---------+---------+---------+-------+--------+----------+-----------+-------------+--------+------------+--------------+--------------+------------+-----------+-------------+

csv comparison results

Example              , Version, Configuration, name, success, samples, startup, setup, userT  , delayedT, 1stBuildT, avgPerRespT, totalT , rulesBuilt, rulesChanged, rulesVisited, rulesTotal, ruleEdges, ghcRebuilds, maxResidency, allocatedBytes
MultiLayerModules    , after  , All          , edit, True   , 0.00%  , 8.33%  , NA   , -78.04%, 4.35%   , 5.34%    , -77.27%    , -75.50%, -98.36%   , -97.25%     , 12.47%      , 5.55%     , 20.81%   , -28.37%    , -47.18%     , -84.20%       
MultiLayerModulesNoTH, after  , All          , edit, True   , 0.00%  , -13.33%, NA   , -14.34%, -9.70%  , -10.88%  , -16.67%    , -13.77%, 0.00%     , 0.00%       , 0.00%       , 5.55%     , 20.81%   , 0.00%      , -1.28%      , 0.74%         
                     

[wz1000]

Very nice results, this makes a lot of sense.

[fendor]

LGTM, thank you!