Commit Graph

31 Commits

Author SHA1 Message Date
Cameron Gutman
014f507c40 Use specific acquire and release variants of InterlockedExchange on ARM
_InterlockedExchange_rel() is required for correctness on ARM because
the _ReadWriteBarrier() macro is only a compiler memory barrier, not a
hardware memory barrier. Due to ARM's relaxed memory model, this means
the '*lock = 0' write may be observed before the operations inside the
lock, causing possible corruption of data protected by the lock.

_InterlockedExchange_acq() is more efficient on ARM because it avoids an
expensive full memory barrier that _InterlockedExchange() does.
2021-01-03 12:13:40 -06:00
Cameron Gutman
59594a7891 Implement PAUSE_INSTRUCTION() for Windows ARM platforms 2021-01-02 13:43:04 -06:00
Sam Lantinga
9130f7c377 Updated copyright for 2021 2021-01-02 10:25:38 -08:00
Sam Lantinga
d25eff6505 Fixed bug 5429 - spinlock implements PAUSE_INSTRUCTION for PPC platforms
David Carlier

This form of 'or' provides a hint that performance
will probably be improved if shared resources dedicated
to the executing processor are released for use by other
processors
2021-01-02 10:06:22 -08:00
Sam Lantinga
6bd4c717a1 Fixed bug 5402 - ARM support little update proposal
David Carlier

No fix but mostly an update for ARM architecture.
2020-12-17 21:41:23 -08:00
Cameron Cawley
8f1a916ac5 Add basic support for compiling on RISC OS 2020-02-13 20:50:47 +00:00
Sam Lantinga
a8780c6a28 Updated copyright date for 2020 2020-01-16 20:49:25 -08:00
Cameron Cawley
85aabec27e atomic: Support compiling on ARMv3 2019-10-11 22:08:53 +01:00
Sam Lantinga
959cfc428e Fixed memory barrier macro check so it isn't quite so fragile 2019-06-30 23:58:31 -07:00
Sam Lantinga
a8bea85810 Limit the compile error to the case where we actually define the memory barrier macro as the function 2019-06-30 23:55:28 -07:00
Sam Lantinga
cc47810d36 Fixed bug 4683 - SDL_atomic infinite recursion on armv6/armv5 w/ thumb
The real problem is that SDL_atomic.c was built in thumb mode instead of ARM mode, which is required to use the mcr instruction on ARM platforms. Added a compiler error to catch this case so we don't generate code that does infinite recursion.

I also added a potentially better way to handle things on Linux ARM platforms, based on comments in the Chromium headers, which we can try out after 2.0.10 ships.
2019-06-30 23:26:16 -07:00
Sam Lantinga
5e13087b0f Updated copyright for 2019 2019-01-04 22:01:14 -08:00
Ozkan Sezer
d9fb77a3c1 SDL_atomic.c, SDL_spinlock.c: use lock prefix with xchg in Watcom asm. 2018-09-27 11:55:02 +03:00
Ozkan Sezer
949b8bd8c5 correct the comment correction.. 2018-06-30 20:10:40 +03:00
Ozkan Sezer
67f18a1973 comment correction 2018-06-30 20:04:01 +03:00
Ryan C. Gordon
52857de251 Corrected a comment. 2018-06-29 16:55:55 -04:00
Ryan C. Gordon
3b173f818d Patched to compile on Visual Studio (typo). 2018-06-25 23:00:38 -04:00
Ryan C. Gordon
58168a8c3c atomic: Fight with all the assemblers that don't like REP NOP. :/ 2018-06-25 16:34:16 -04:00
Ryan C. Gordon
5f123e31a3 atomic: Spin locks now try to use the x86 PAUSE instruction for short waits.
Fixes Bugzilla #4151.
2018-06-25 15:58:35 -04:00
Sam Lantinga
e3cc5b2c6b Updated copyright for 2018 2018-01-03 10:03:25 -08:00
Ozkan Sezer
500378eb52 Add atomics for Watcom/x86 as inline asm
Partially fixes Bugzilla #3758.
2017-08-18 16:35:55 -04:00
Ryan C. Gordon
1d1a0d7f03 atomic: Patched to compile on Android. 2017-04-13 15:28:56 -04:00
Ryan C. Gordon
1c9c7633d9 atomic: favor compiler intrinsics for compare-and-swap over macOS APIs.
The OSAtomicCompareAndSwap* APIs are deprecated as of macOS 10.12.
2017-04-13 13:28:52 -04:00
Ryan C. Gordon
8c00de57f6 atomic: let Clang always use atomic_load_n if available.
(Apple's Clang reports itself as GCC 4.2.1 in preprocessor macros--the final
GNU C compiler Apple shipped--as of the macOS 10.12 SDK.)
2017-04-13 13:22:23 -04:00
Sam Lantinga
7891e72dca __atomic_load_n() appears to be available in GCC 5 but not GCC 4 2017-03-30 06:52:34 -07:00
Ryan C. Gordon
f2179944cc Patched to compile on some platforms. 2017-03-29 12:04:17 -04:00
James Legg
1dc9ae5c1e Use GCC's atomic loads in SDL_AtomicGet and SDL_AtomicGetPtr
This fixes errors reported by address sanitizer, and generates simpler
code on x86 architectures.
2017-03-29 15:48:22 +01:00
Sam Lantinga
06ccb71bcd Make sure the memory barrier functions are always available, and now they are implemented on Android __ARM_ARCH_5TE__ 2017-02-10 11:21:15 -08:00
Sam Lantinga
45b774e3f7 Updated copyright for 2017 2017-01-01 18:33:28 -08:00
Sam Lantinga
42065e785d Updated copyright to 2016 2016-01-02 10:10:34 -08:00
Philipp Wiesemann
0e45984fa0 Fixed crash if initialization of EGL failed but was tried again later.
The internal function SDL_EGL_LoadLibrary() did not delete and remove a mostly
uninitialized data structure if loading the library first failed. A later try to
use EGL then skipped initialization and assumed it was previously successful
because the data structure now already existed. This led to at least one crash
in the internal function SDL_EGL_ChooseConfig() because a NULL pointer was
dereferenced to make a call to eglBindAPI().
2015-06-21 17:33:46 +02:00