Andrew

Currently AMX co-processor is undocumented and uninstrumented. It is quite legal to tinker, though nobody prevents OEM to microcode your toys away. Wrapper can be written at ILP64->LP64 level -...

Actually dubiously named co-procesor is not mentioned in that abstraction API anywhere. Or by Apple. Though you can try yourself if 4x4 * 4x1 generates any non-disassemblable functions or calls...

Each race pair has "failed to restore stack" on one side. Could you compile with gcc -O0 -g or something that gives proper backtraces?

It is miscompiled if it uses threads for such a small sample!!!! It does not relieve thread safety issue if any BUT.

Simple explanation: tsan breaks wildly if mixing instrumented and non-instrumented objects in same file I fail to repeat claim gcc4(helgrind...) gcc10 .. gcc12 ``` git clone https://github.com/xianyi/openblas cd openblas make...

Just that work is not divided N.N.N.N.M but around and about like N,N+1,N-1,M but correctly in every case, including marginal or 1.1.0.0.0.0 case. No problem at all, not at least...

Regarding optimizing threshold away - initializing variable to 1 at declaration will not be blown away.

@aborzunov you must use -fsanitize=thread in your project too, and check with `ldd` that target binary does not import libpthread or other version of sanitizersomewhere besides thread sanitizer emulation. What...

@dw-loewe @dataCobra would be nice to post device identifiers like from auc output so that @aparcar can look into them.

CPE does not support any flow control. It is on CPU port, switch port 0 and beams to port on their own wlans unrelated to vlan 1/2 and isolated by...