-
Notifications
You must be signed in to change notification settings - Fork 506
-
hello,
on Sapphire Rapids server with Intel(R) Xeon(R) Silver 4416+ running compression only workload we're getting such readings:
./pcm-accel -qat
Intel(r) Performance Counter Monitor ($Format:%ci ID=%h$)
This utility measures Sapphire Rapids-SP accelerators information.
===== Processor information =====
Linux arch_perfmon flag : yes
Hybrid processor : no
IBRS and IBPB supported : yes
STIBP supported : yes
Spec arch caps supported : yes
Max CPUID level : 32
CPU model number : 143
Number of logical cores: 80
Number of online logical cores: 76
Threads (logical cores) per physical core: 2 (maybe imprecise due to core offlining/hybrid CPU)
Offlined cores: 45 46 60 61
Num sockets: 2
Physical cores per socket: 20 (maybe imprecise due to core offlining/hybrid CPU)
Last level cache slices per socket: 20
Core PMU (perfmon) version: 5
Number of core PMU generic (programmable) counters: 8
Width of generic (programmable) counters: 48 bits
Number of core PMU fixed counters: 4
Width of fixed counters: 48 bits
Nominal core frequency: 2000000000 Hz
IBRS enabled in the kernel : yes
STIBP enabled in the kernel : no
The processor is not susceptible to Rogue Data Cache Load: yes
The processor supports enhanced IBRS : yes
Package thermal spec power: 165 Watt; Package minimum power: 106 Watt; Package maximum power: 357 Watt;
INFO: Linux perf interface to program uncore PMUs is NOT present
Socket 0: 4 memory controllers detected with total number of 12 channels. 2 UPI ports detected. 4 M2M (mesh to memory) blocks detected. 0 HBM M2M blocks detected. 0 EDC/HBM channels detected. 0 Home Agents detected. 3 M3UPI blocks detected.
Socket 1: 4 memory controllers detected with total number of 12 channels. 2 UPI ports detected. 4 M2M (mesh to memory) blocks detected. 0 HBM M2M blocks detected. 0 EDC/HBM channels detected. 0 Home Agents detected. 3 M3UPI blocks detected.
ERROR: MDF PMU not found
ERROR: MDF PMU not found
Info: IDX - Detected 2 IAA devices, 2 DSA devices, 2 QAT devices.
Warning: IDX - BAR0 of B:0x75,D:0x2,F:0x0 is invalid(memBar=0x39bffff5000c, pciCmd=0x100140), skipped.
Warning: IDX - BAR0 of B:0xf2,D:0x2,F:0x0 is invalid(memBar=0x3b7ffff5000c, pciCmd=0x100140), skipped.
Warning: IDX - BAR0 of B:0x75,D:0x1,F:0x0 is invalid(memBar=0x39bffff4000c, pciCmd=0x100144), skipped.
Warning: IDX - BAR0 of B:0xf2,D:0x1,F:0x0 is invalid(memBar=0x3b7ffff4000c, pciCmd=0x100144), skipped.
Initializing RMIDs
Detected Intel(R) Xeon(R) Silver 4416+ "Intel(r) microarchitecture codename Sapphire Rapids-SP" stepping 8 microcode level 0x2b0001b0
Info: IDX - Please ensure the required driver(e.g idxd driver for iaa/dsa, qat driver and etc) correct enabled with this system, else the tool may fail to run.
|Accelerator|Socket|Inbound_BW(Bps)|Outbound_BW(Bps)|util_comp0(%)|util_decomp0(%)|util_decomp1(%)|util_decomp2(%)|util_xlt0(%)|
|___________|______|_______________|________________|_____________|_______________|_______________|_______________|____________|
| QAT#0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| QAT#1 | 1 | 27 G |4158 M | 98 | 45 | 0 | 0 | 4 |
|___________|______|_______________|________________|_____________|_______________|_______________|_______________|____________|
Notice that util_decomp0 is 45%. Is it ok if we only compress data? How to interpret that?
Cheers,
Filip
Beta Was this translation helpful? Give feedback.
All reactions
Replies: 1 comment 3 replies
-
It's OK. While QAT is compressing data, it is also decompressing it. The purpose is to verify the compressed data is able to decompress and ensuring the data integrity.
Beta Was this translation helpful? Give feedback.
All reactions
-
Yes, agree with Pan.
Beta Was this translation helpful? Give feedback.
All reactions
-
thank you guys. Do you happen to know if this verification could be disabled? If so, would that release some resources so that more are available for parallel compression requests? And what would be the cost, I mean probability of a failure? Like a probability of some external event like a soft-error or is this internal cause?
Beta Was this translation helpful? Give feedback.
All reactions
-
It's related to QAT HW implementation. So far as I know, the verification can not be disabled. And QAT use different HW resources for compression and decompression, so disable verification can not boost compression performance. I'm not certain the answer of question "And what would be the cost, I mean probability of a failure? Like a probability of some external event like a soft-error or is this internal cause?", I guess maybe both.
Beta Was this translation helpful? Give feedback.