Closed cyring closed 1 year ago
@wjc404 Hello
Can you try to read the following MSR on your Skylake/X ?
## MSR_SKL_CORE_PERF_LIMIT_REASONS
rdmsr -ax 0x0000064f
## MSR_GRAPHICS_PERF_LIMIT_REASONS
rdmsr -ax 0x000006b0
## MSR_RING_PERF_LIMIT_REASONS
rdmsr -ax 0x000006b1
Thank you
request | result | note |
---|---|---|
0x0000064f | 0xe1848080 | same for all cores |
0x000006b0 | cannot read | for all cores |
0x000006b1 | 0x0 | same for all cores |
@wjc404 Thank you for the MSR tests.
Can you first pull and test branch hotfix_Skylake_X
?
If branch does not crash, can you then allow MSR_RING_PERF_LIMIT_REASONS
by replacing comments in the following lines:
https://github.com/cyring/CoreFreq/blob/c761e4657bd29d37fcb73ef07eb20c6a458b05d2/corefreqk.c#L11818
/*TODO(Unsolved)
GraphicsPerfLimitReasons(Core);
*/
RingPerfLimitReasons(Core);
https://github.com/cyring/CoreFreq/blob/c761e4657bd29d37fcb73ef07eb20c6a458b05d2/corefreqk.c#L16726
/*TODO(Unsolved)
Monitor_GraphicsPerfLimitReasons(PUBLIC(RO(Proc)));
*/
Monitor_RingPerfLimitReasons(PUBLIC(RO(Proc)));
- Can you first pull and test branch
hotfix_Skylake_X
?
Succeeded, nothing wrong with corefreq-cli.
- If branch does not crash, can you then allow
MSR_RING_PERF_LIMIT_REASONS
by replacing comments in the following lines:https://github.com/cyring/CoreFreq/blob/c761e4657bd29d37fcb73ef07eb20c6a458b05d2/corefreqk.c#L11818
/*TODO(Unsolved) GraphicsPerfLimitReasons(Core); */ RingPerfLimitReasons(Core);
https://github.com/cyring/CoreFreq/blob/c761e4657bd29d37fcb73ef07eb20c6a458b05d2/corefreqk.c#L16726
/*TODO(Unsolved) Monitor_GraphicsPerfLimitReasons(PUBLIC(RO(Proc))); */ Monitor_RingPerfLimitReasons(PUBLIC(RO(Proc)));
Also succeeded.
@wjc404 Thank you for your contribution.
Do you mind to run the current development branch develop
for non regression tests ?
Any CLI outputs and screenshots are greatly appreciated. Such as CPU-support
Well, maybe there's no need to call Monitor_GraphicsPerfLimitReasons
and GraphicsPerfLimitReasons
for skylake-X processors, since they don't have built-in graphics. I will test after this fix is taken into the target branch.
Well, maybe there's no need to call
Monitor_GraphicsPerfLimitReasons
andGraphicsPerfLimitReasons
for skylake-X processors, since they don't have built-in graphics.
Exactly.
And I'm looking for an iGPU discriminant to jump around or not GraphicsPerfLimitReasons
in code.
CSR registers for capabilities may tell if a Graphic is present or not:
https://github.com/cyring/CoreFreq/blob/35b504f3a26c6e2a1d18fd1a6f197e3881a3027b/intel_reg.h#L4394
But this required to test various Skylake-{X}
and successors for the validity of such bit.
I will test after this fix is taken into the target branch.
Are we closing the issue ?
Master branch is released with new version: can you tell if it works OK ?
Hello,
I need your processor for ACPI testings!
Based on develop
branch, can you change the code of the function Query_Skylake_X
as below
https://github.com/cyring/CoreFreq/blob/e41ec49b5f3ff9085f5c4396ce236182fc8f2711/corefreqk.c#L6876
static void Query_Skylake_X(unsigned int cpu)
{
if (PUBLIC(RO(Proc))->Features.Power.EAX.TurboIDA)
{
PUBLIC(RO(Proc))->Features.Uncore_Unlock = 1;
}
Skylake_X_Platform_Info(cpu);
HyperThreading_Technology();
Haswell_Uncore_Ratio(NULL);
Intel_PowerInterface();
Intel_Hardware_Performance();
if (cpu == PUBLIC(RO(Proc))->Service.Core) {
Read_ACPI_PCT_Registers(cpu);
Read_ACPI_PSS_Registers(cpu);
Read_ACPI_PPC_Registers(cpu);
Read_ACPI_CST_Registers(cpu);
}
}
Then rebuild, start CoreFreq, and post the output of corefreq-cli -s
Thank you
Originally posted by @wjc404 in https://github.com/cyring/CoreFreq/issues/373#issuecomment-1312450134