On Wed, May 11, 2011 at 6:14 PM, saeed bishara saeed.bishara@gmail.com wrote:
On Sat, May 7, 2011 at 12:39 AM, Nicolas Pitre nicolas.pitre@linaro.org wrote:
On Fri, 6 May 2011, Ramana Radhakrishnan wrote:
On 6 May 2011 16:06, Ken Werner ken.werner@linaro.org wrote:
Currently the GCC ARM backend doesn't provide a pattern to inline 64bit __sync_* functions but the compiler emits __sync_*_8 function calls [1]. The libgcc does not provide these symbols via the usual thin wrapper around the kernel helper [2] because the ARM Linux __kernel_cmpxchg supports 32bit only. My understanding is that for ARMv7 the GCC backend could be enhanced to inline the __sync_* functions by using the LDREXD and STREXD instructions. But for ARMv5 we would still rely on a new kernel helper.
It's a bit tricky with when you want to use the kernel helper for v5t, so we've got to find a way of turning this on only with special knobs and not by default and that's a bit tricky.
What is the problem with v5t?
Hi, when I run the following on my jaunty (for armv5), I get the following
echo "void f(){__sync_synchronize();}" | gcc -O2 -x c -Wall -dA -S - -o - .arch armv6 .eabi_attribute 27, 3 .fpu vfp .eabi_attribute 20, 1 .eabi_attribute 21, 1 .eabi_attribute 23, 3 .eabi_attribute 24, 1 .eabi_attribute 25, 1 .eabi_attribute 26, 2 .eabi_attribute 30, 2 .eabi_attribute 18, 4 .file "" .text .align 2 .global f .type f, %function f: @ args = 0, pretend = 0, frame = 0 @ frame_needed = 0, uses_anonymous_args = 0 @ link register save eliminated. @ basic block 2 bx lr .size f, .-f .ident "GCC: (Ubuntu 4.4.1-4ubuntu9) 4.4.1" .section .note.GNU-stack,"",%progbits
but I can see (with debugger) that the kernel helper functions are reached. any idea how?
There was a compiler bug some time ago which mistakenly generated no code for __sync_synchronize(): https://bugs.launchpad.net/ubuntu/+source/gcc-4.4/+bug/491872
I suspect your Jaunty-era compiler may be affected by this -- you probably need to upgrade it.
If you're observing calls to the helpers, they may be coming from glibc etc. instead of your function.
You _should_ be seeing code something like this:
$ echo 'void f(void) { __sync_synchronize(); }' | arm-linux-gnueabi-gcc -march=armv7-a -O2 -S -o - -x c - [...] f: @ args = 0, pretend = 0, frame = 0 @ frame_needed = 0, uses_anonymous_args = 0 @ link register save eliminated. dmb sy bx lr [...]
(If -march is set high enough, the dmb is explicit and inline.)
$ echo 'void f(void) { __sync_synchronize(); }' | arm-linux-gnueabi-gcc -march=armv5t -O2 -S -o - -x c - [...] f: push {r3, lr} bl __sync_synchronize @ sp needed for prologue pop {r3, pc}
(-march=armv5t indicates an architecture which doesn't have the barrier instructions, so the kuser helper is needed instead)
Cheers ---Dave