casaroli commented on code in PR #20723:
URL: https://github.com/apache/nuttx/pull/20723#discussion_r4587643896


##########
include/nuttx/atomic.h:
##########
@@ -277,8 +333,10 @@ typedef __Atomic(int64_t) atomic64_t;
 
 #if UINTPTR_MAX > UINT32_MAX
 typedef atomic64_t atomic_ptr_t;
-#else
+#elif UINTPTR_MAX > UINT16_MAX
 typedef atomic_t atomic_ptr_t;
+#else
+typedef __Atomic(int16_t) atomic_ptr_t;

Review Comment:
   I don't think so. `NXSEM_MHOLDER()` casts the `uint32_t mholder` of every 
mutex to `atomic_t *`, and bit 31 of `mholder` is `NXSEM_MBLOCKING_BIT`. With a 
16-bit `atomic_t`, AVR reads and writes only the low half of `mholder`. 
`NXSEM_COUNT()` (`int32_t semcount`) and seqlock.h (`uint32_t sequence`) use 
the same cast.
   



##########
libs/libc/tls/Kconfig:
##########
@@ -5,6 +5,26 @@
 
 menu "Thread Local Storage (TLS)"
 
+config TLS_EMUTLS
+       bool "Per-thread copies of emulated TLS variables"
+       default y
+       depends on BUILTIN_TOOLCHAIN
+       depends on (ARCH_TOOLCHAIN_GNU && !ARCH_TOOLCHAIN_ARMCLANG && 
!ARCH_TOOLCHAIN_GHS && !ARCH_SIM) || (ARCH_SIM && SIM_TOOLCHAIN_GCC && 
!HOST_MACOS && !WINDOWS_NATIVE)

Review Comment:
   I don't think so. The toolchain selects the TLS model, not 
`SCHED_THREAD_LOCAL`. That option adds only arm_tls.c or riscv_tls.c, and no 
compiler flag. For `__thread int x`, arm-none-eabi-gcc 13.2, 
aarch64-none-elf-gcc 13.2 and riscv-none-elf-gcc 14.3 call 
`__emutls_get_address()` with or without it. `SCHED_THREAD_LOCAL` is default n, 
so with this dependency the threads of most boards share one copy again.
   



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to