| Safe Haskell | None |
|---|---|
| Language | Haskell2010 |
GHC.Driver.GenerateCgIPEStub
Synopsis
- generateCgIPEStub :: HscEnv -> Module -> InfoTableProvMap -> (NonCaffySet, ModuleLFInfos, Map CmmInfoTable (Maybe IpeSourceLocation), IPEStats) -> Stream IO CmmGroupSRTs CmmCgInfos
- lookupEstimatedTicks :: HscEnv -> Map CmmInfoTable (Maybe IpeSourceLocation) -> IPEStats -> CmmGroupSRTs -> IO (Map CmmInfoTable (Maybe IpeSourceLocation), IPEStats)
Documentation
generateCgIPEStub :: HscEnv -> Module -> InfoTableProvMap -> (NonCaffySet, ModuleLFInfos, Map CmmInfoTable (Maybe IpeSourceLocation), IPEStats) -> Stream IO CmmGroupSRTs CmmCgInfos Source #
lookupEstimatedTicks :: HscEnv -> Map CmmInfoTable (Maybe IpeSourceLocation) -> IPEStats -> CmmGroupSRTs -> IO (Map CmmInfoTable (Maybe IpeSourceLocation), IPEStats) Source #
Given:
* an initial mapping from info tables to possible source locations,
* initial IPEStats,
* a CmmGroupSRTs,
map every info table listed in the CmmProcs of the group to their possible
source locations and update IPEStats for skipped stack info tables (in case
both -finfo-table-map and -fno-info-table-map-with-stack were given). See:
Note [Stacktraces from Info Table Provenance Entries (IPE based stack unwinding)]
Note: While it would be cleaner if we could keep the recursion and
accumulation internal to this function, this cannot be done without
separately traversing stream of CmmGroupSRTs in Main. The
initial implementation of this logic did such a thing, and code generation
performance suffered considerably as a result (see #23103).