On Fri, 13 Dec 2024 02:18:35 GMT, Vladimir Ivanov wrote:
>> The double check needs less time vs enter to 'synchronize' block when it not
>> needed. This patch for the xml.validation workload for runs with 112 threads
>> reports ~2.5% improvement on the server with Xeon 8480:
>> orig: 1T - 433.6
On Fri, 13 Dec 2024 02:18:35 GMT, Vladimir Ivanov wrote:
>> The double check needs less time vs enter to 'synchronize' block when it not
>> needed. This patch for the xml.validation workload for runs with 112 threads
>> reports ~2.5% improvement on the server with Xeon 8480:
>> orig: 1T - 433.6
On Fri, 13 Dec 2024 02:18:35 GMT, Vladimir Ivanov wrote:
>> The double check needs less time vs enter to 'synchronize' block when it not
>> needed. This patch for the xml.validation workload for runs with 112 threads
>> reports ~2.5% improvement on the server with Xeon 8480:
>> orig: 1T - 433.6
> The double check needs less time vs enter to 'synchronize' block when it not
> needed. This patch for the xml.validation workload for runs with 112 threads
> reports ~2.5% improvement on the server with Xeon 8480:
> orig: 1T - 433.64 ops/m, 112T - 17344.25 ops/m
> patched: 1T - 437.01 ops/m, 11
The double check needs less time vs enter to 'synchronize' block when it not
needed. This patch for the xml.validation workload for runs with 112 threads
reports ~2.5% improvement on the server with Xeon 8480:
orig: 1T - 433.64 ops/m, 112T - 17344.25 ops/m
patched: 1T - 437.01 ops/m, 112T - 17771
