perf(x86_64): skip APIC probe on uhyve

[n0toose]

• Skip memory probes, use default address directly on Uhyve.
• Checking whether the processor supports x2apic in init_x2apic goes against the component's own conventions.
• We need a local APIC, but IO-APIC is not supported.
• Show more relevant info/warning messages.
• Fix minor formatting issue that shows "0x0x" prefix.

Fixes #1352

[n0toose]

As pointed out in #1352 (comment), a manual skip of x2apic causes an error (adding that it appears both under QEMU and Uhyve). I don't currently believe it is relevant to my changes.

[github-actions]

Benchmark Results

Benchmark	Current: 08867bb	Previous: 8e08755	Performance Ratio
startup_benchmark Build Time	115.54 s	118.64 s	0.97
startup_benchmark File Size	0.90 MB	0.90 MB	1.00
Startup Time - 1 core	0.88 s (±0.03 s)	0.92 s (±0.03 s)	0.96
Startup Time - 2 cores	0.90 s (±0.03 s)	0.93 s (±0.03 s)	0.97
Startup Time - 4 cores	0.90 s (±0.04 s)	0.90 s (±0.03 s)	0.99
multithreaded_benchmark Build Time	114.31 s	115.52 s	0.99
multithreaded_benchmark File Size	1.01 MB	1.01 MB	1.00
Multithreaded Pi Efficiency - 2 Threads	87.84 % (±9.47 %)	89.14 % (±8.69 %)	0.99
Multithreaded Pi Efficiency - 4 Threads	43.64 % (±3.30 %)	43.33 % (±4.80 %)	1.01
Multithreaded Pi Efficiency - 8 Threads	25.03 % (±1.86 %)	24.89 % (±3.25 %)	1.01
micro_benchmarks Build Time	110.99 s	121.35 s	0.91
micro_benchmarks File Size	1.01 MB	1.01 MB	1.00
Scheduling time - 1 thread	65.85 ticks (±5.14 ticks)	64.78 ticks (±3.48 ticks)	1.02
Scheduling time - 2 threads	37.95 ticks (±5.22 ticks)	39.58 ticks (±5.81 ticks)	0.96
Micro - Time for syscall (getpid)	3.16 ticks (±0.50 ticks)	3.11 ticks (±0.35 ticks)	1.01
Memcpy speed - (built_in) block size 4096	77124.23 MByte/s (±53565.81 MByte/s)	76892.26 MByte/s (±53208.68 MByte/s)	1.00
Memcpy speed - (built_in) block size 1048576	42717.77 MByte/s (±29575.57 MByte/s)	42437.70 MByte/s (±29401.77 MByte/s)	1.01
Memcpy speed - (built_in) block size 16777216	24106.08 MByte/s (±20044.63 MByte/s)	24216.95 MByte/s (±20131.45 MByte/s)	1.00
Memset speed - (built_in) block size 4096	76675.88 MByte/s (±53248.29 MByte/s)	76746.14 MByte/s (±53110.30 MByte/s)	1.00
Memset speed - (built_in) block size 1048576	42897.69 MByte/s (±29697.86 MByte/s)	42677.86 MByte/s (±29565.07 MByte/s)	1.01
Memset speed - (built_in) block size 16777216	24505.42 MByte/s (±20221.68 MByte/s)	24944.26 MByte/s (±20616.34 MByte/s)	0.98
Memcpy speed - (rust) block size 4096	69331.51 MByte/s (±48574.36 MByte/s)	72842.39 MByte/s (±51003.78 MByte/s)	0.95
Memcpy speed - (rust) block size 1048576	42474.29 MByte/s (±29435.83 MByte/s)	42614.04 MByte/s (±29546.77 MByte/s)	1.00
Memcpy speed - (rust) block size 16777216	24490.51 MByte/s (±20300.02 MByte/s)	25125.16 MByte/s (±20845.63 MByte/s)	0.97
Memset speed - (rust) block size 4096	69543.27 MByte/s (±48704.40 MByte/s)	73212.71 MByte/s (±51231.34 MByte/s)	0.95
Memset speed - (rust) block size 1048576	42680.82 MByte/s (±29575.26 MByte/s)	42865.24 MByte/s (±29720.96 MByte/s)	1.00
Memset speed - (rust) block size 16777216	24750.78 MByte/s (±20393.15 MByte/s)	25840.21 MByte/s (±21299.14 MByte/s)	0.96
alloc_benchmarks Build Time	106.69 s	106.38 s	1.00
alloc_benchmarks File Size	0.97 MB	0.97 MB	1.00
Allocations - Allocation success	100.00 %	100.00 %	1
Allocations - Deallocation success	69.99 % (±0.30 %)	70.03 % (±0.28 %)	1.00
Allocations - Pre-fail Allocations	100.00 %	100.00 %	1
Allocations - Average Allocation time	12712.40 Ticks (±157.18 Ticks)	11224.94 Ticks (±674.74 Ticks)	1.13
Allocations - Average Allocation time (no fail)	12712.40 Ticks (±157.18 Ticks)	11224.94 Ticks (±674.74 Ticks)	1.13
Allocations - Average Deallocation time	760.79 Ticks (±128.50 Ticks)	733.24 Ticks (±64.51 Ticks)	1.04
mutex_benchmark Build Time	104.79 s	111.63 s	0.94
mutex_benchmark File Size	1.01 MB	1.01 MB	1.00
Mutex Stress Test Average Time per Iteration - 1 Threads	12.50 ns (±0.94 ns)	13.10 ns (±0.85 ns)	0.95
Mutex Stress Test Average Time per Iteration - 2 Threads	13.96 ns (±1.23 ns)	14.56 ns (±0.90 ns)	0.96

This comment was automatically generated by workflow using github-action-benchmark.

[mkroening]

Could you split this up into several commits to ease review of separate changes? :)

I am not too convinced about the log changes. Not being able to find APIC via ACPI or MP is expected for certain machines (such as Uhyve). I don't think this warrants a warning. In what kind of cases are the logs useful in practice?

[mkroening]

Do I understand correctly that this is improving the performance and not fixing a problem?

[n0toose]

Do I understand correctly that this is improving the performance and not fixing a problem?

I considered performing checks "that will not go anywhere" in the Uhyve environment to be a problem/bug, but I understand why the fix: might be misleading. Sorry.

To answer your question: Yes. To the best extend that I am aware of, I don't find that the previous checks are necessary within Uhyve (as far as its main branch is concerned) for Uhyve to work. This change removes this "small freeze effect"-feeling when launching Uhyve.

[mkroening]

All right, I have further separated the refactoring from the first commit. I think you missed adding a condition in application_processor_init. I wonder why this caused no problems, though. It should have, as far as I understand.

Are you okay with my changes? :) c2159c8..9d15052

[n0toose]

Yes, thanks for catching this and you have my apologies for the errors. :)

[mkroening]

Thanks! :)

[n0toose]

CI reports:

thread 'main' panicked at /home/runner/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/uhyve-0.5.0/src/linux/x86_64/kvm_cpu.rs:145:14:
Unable to disable exists due pause instructions: Error(22)
note: run with RUST_BACKTRACE=1 environment variable to display a backtrace

[mkroening]

Yeah, that's #1992. I thought we might be able to avoid rebasing.