Replace rdcycle instruction with rdtime

[marv]

The rdcycle instruction causes SIGILL with RISCV_PMU_SBI=y in the kernel[1]

According to a PR for the highway project[2] which fixed the same problem there rdtime is the proper equivalent for the x86 rdtsc instruction on RISC-V

Using rdtime instead of rdcycle makes all tests pass on my VisionFive 2 board

[1] https://lists.infradead.org/pipermail/linux-riscv/2022-September/018862.html
[2] google/highway#961

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

[derekmauro]

@bbarenblat had proposed a similar (identical?) change internally a few months ago. I pointed out that it is a little bit more complicated because ABSL_INTERNAL_UNSCALED_CYCLECLOCK_FREQUENCY_IS_CPU_FREQUENCY would also need to change.

[aurel32]

Removing the ABSL_INTERNAL_UNSCALED_CYCLECLOCK_FREQUENCY_IS_CPU_FREQUENCY define for RISC-V seems easy to do. But then it seems we have to also update UnscaledCycleClock::Frequency. The frequency of the rdtime timer varies from board to board, and TTBOMK there is no easy way to get its frequency. In such condition, is it fine changing UnscaledCycleClock::Frequency to return 1.0, like it is done for the CPU when it is not possible to determine the frequency?

[aurel32]

Removing the ABSL_INTERNAL_UNSCALED_CYCLECLOCK_FREQUENCY_IS_CPU_FREQUENCY define for RISC-V seems easy to do. But then it seems we have to also update UnscaledCycleClock::Frequency. The frequency of the rdtime timer varies from board to board, and TTBOMK there is no easy way to get its frequency. In such condition, is it fine changing UnscaledCycleClock::Frequency to return 1.0, like it is done for the CPU when it is not possible to determine the frequency?

Given I got no answer I gave it a tried, and I have found it is not necessary to provide UnscaledCycleClock::Frequency. I have therefore opened a new PR: #1631

[derekmauro]

This was addressed by #1644.