New keyword arguments MPS_KEY_ARENA_COMMIT_LIMIT and MPS_KEY_SPARE_ARENA_COMMIT_LIMIT can be passed to mps_arena_create_k or mps_arena_configure.
Deprecate mps_arena_commit_limit_set and mps_arena_spare_commit_limit_set.
Copied from Perforce
Change: 188085
ServerID: perforce.ravenbrook.com
The CONTROL_EXTEND_BY value for the extend by on the control pool was set really small (4096) on the assumption that the control pool was not heavily used. However, since we allocate Seg's in the control pool and recycle them frequently, this assumption is no longer true.
Copied from Perforce
Change: 186832
ServerID: perforce.ravenbrook.com
Remove hack from rnd_grain -- this was covering up a bug.
Base the AMC "large" segment determination on the requested object size, not the (rounded up) segment size.
Assert in the case where largeSize < extendBy: this has bad space performance.
Copied from Perforce
Change: 186755
ServerID: perforce.ravenbrook.com
AMC also uses MPS_KEY_LARGE_SIZE keyword argument to specify what size objects need to be to live on their own segment.
Copied from Perforce
Change: 186480
ServerID: perforce.ravenbrook.com
* Move Windows-specific modules like vmw3.c out of commpre.nmk and into the platform-specific Nmake files.
* Set StackProbeDEPTH to 0 on the ANSI platform.
* New Nmake file ananmv.nmk builds the MPS for the ANSI platform using Microsoft Visual C/C++.
Copied from Perforce
Change: 186063
ServerID: perforce.ravenbrook.com
MVFF and MVT support alignments down to FreelistMinimumAlignment.
For MVFF and MVT, alignments that are too small are rounded up automatically, to make it easier to write portable programs.
Copied from Perforce
Change: 185369
ServerID: perforce.ravenbrook.com
* CONFIG_POLL_NONE (because the user-visible consequence is that polling is no longer supported; was CONFIG_PROTECTION_NONE).
* DISABLE_LOCKS (was THREAD_SINGLE).
* DISABLE_SHIELD (was THREAD_SINGLE && PROTECTION_NONE)
* DISABLE_REMEMBERED_SET (was PROTECTION_NONE)
When the shield is disabled, ArenaLeave asserts that there are no busy traces, and ArenaPoll is a no-op.
By having functions implemented using the corresponding macro, we can avoid duplicated code, and avoid testing DISABLE_SHIELD in global.c.
Remove all remaining references to MPS_PROD_EPCORE.
Copied from Perforce
Change: 185176
ServerID: perforce.ravenbrook.com
* Don't test __has_feature unless we know we are on MPS_BUILD_LL.
* Fix type-punned pointer aliasing in expt825.c.
Copied from Perforce
Change: 185104
ServerID: perforce.ravenbrook.com
Add __attribute__((__format__(printf))) to functions that take a printf-compatible format string (when building using GCC or Clang), so that format string mistakes can be detected statically.
Copied from Perforce
Change: 185021
ServerID: perforce.ravenbrook.com
* New nmake file w3i6pc.nmk.
* New stack scanner ssw3i6pc.c supplies missing jump buffer definition.
* Some platform tests change from defined(MPS_PF_W3I6MV) to defined(MPS_OS_w3) && defined(MPS_ARCH_I6).
* Make reasonable changes to the source code to avoid warnings from Pelles C:
** Ensure that printf formats are consistent with arguments by using PRIuLONGEST and casting to ulongest_t.
** Use size_t for loop variables that index into arrays.
** Suppress "Consider changing type to 'size_t' for loop variable" warnings for the MPS core.
Copied from Perforce
Change: 184996
ServerID: perforce.ravenbrook.com
