[Mesa-dev] [PATCH] anv: generate different VkSemaphore handles

[Jacek Konieczny]

On 2016-02-27 19:32, Kenneth Graunke wrote:
> On Saturday, February 27, 2016 2:47:22 PM PST Jacek Konieczny wrote:
>> The same handle returned by all vkCreateSemaphore() calls confuses
>> validation layers:
>>
>> SEMAPHORE(ERROR): object: 0x1 type: 5 location: 1443 msgCode: 0: 
> vkQueueSubmit: Semaphore must not be currently signaled or in a wait state
>>
>> Signed-off-by: Jacek Konieczny <jajcus at jajcus.net>
>> ---
>>  src/intel/vulkan/anv_device.c | 4 +++-
>>  1 file changed, 3 insertions(+), 1 deletion(-)
>>
>> diff --git a/src/intel/vulkan/anv_device.c b/src/intel/vulkan/anv_device.c
>> index 5993055..96aef00 100644
>> --- a/src/intel/vulkan/anv_device.c
>> +++ b/src/intel/vulkan/anv_device.c
>> @@ -1546,7 +1546,9 @@ VkResult anv_CreateSemaphore(
>>      * rings. As such, there's nothing to do for the user space semaphore.
>>      */
>>  
>> -   *pSemaphore = (VkSemaphore)1;
>> +   static uint64_t value = 0;
>> +
>> +   *pSemaphore = (VkSemaphore) ++value;
>>  
>>     return VK_SUCCESS;
>>  }
>>
> 
> I believe this is okay on x86_64, but it would need locking (or an
> atomic increment operation) on 32-bit systems.  I'll gladly defer
> to the maintainers as to whether they care, however.

Indeed, I completely forgot about multiple threads here. Extra locking
code would probably be an overkill here and atomic increment should do.

Jacek