Errata application is painful for performance. For a start, it's done
when the core has just come out of reset, which means branch predictors
and caches will be empty so a branch to a workaround function must be
fetched from memory and that round trip is very slow. Then it also runs
with the I-cache off, which means that the loop to iterate over the
workarounds must also be fetched from memory on each iteration.
We can remove both branches. First, we can simply apply every erratum
directly instead of defining a workaround function and jumping to it.
Currently, no errata that need to be applied at both reset and runtime,
with the same workaround function, exist. If the need arose in future,
this should be achievable with a reset + runtime wrapper combo.
Then, we can construct a function that applies each erratum linearly
instead of looping over the list. If this function is part of the reset
function, then the only "far" branches at reset will be for the checker
functions. Importantly, this mitigates the slowdown even when an erratum
is disabled.
The result is ~50% speedup on N1SDP and ~20% on AArch64 Juno on wakeup
from PSCI calls that end in powerdown. This is roughly back to the
baseline of v2.9, before the errata framework regressed on performance
(or a little better). It is important to note that there are other
slowdowns since then that remain unknown.
Change-Id: Ie4d5288a331b11fd648e5c4a0b652b74160b07b9
Signed-off-by: Boyan Karatotev <boyan.karatotev@arm.com>
Errata printing is done directly via generic_errata_report.
This commit removes the unused \_cpu\()_errata_report
functions for all cores, and removes errata_func from cpu_ops.
Change-Id: I04fefbde5f0ff63b1f1cd17c864557a14070d68c
Signed-off-by: Ryan Everett <ryan.everett@arm.com>
Testing:
- Manual comparison of disassembly with and without conversion.
- Using the test script in gerrit - 19136
- Building with errata and stepping through from ArmDS and running tftf.
Signed-off-by: Govindraj Raja <govindraj.raja@arm.com>
Change-Id: I1936ab6aaef803f653e79f5c6b590a59b34a8ed1
Implements the loop workaround for Cortex-A76AE, Cortex-A78AE and
Cortex-A78C.
Signed-off-by: Bipin Ravi <bipin.ravi@arm.com>
Signed-off-by: Madhukar Pappireddy <madhukar.pappireddy@arm.com>
Change-Id: I5c838f5b9d595ed3c461a7452bd465bd54acc548
Some cores support only AArch64 mode. In those cores, only a limited
subset of the AArch32 system registers are implemented. Hence, if TF-A
is supposed to run on AArch64-only cores, it must be compiled with
CTX_INCLUDE_AARCH32_REGS=0.
Currently, the default settings for compiling TF-A are with the AArch32
system registers included. So, if we compile TF-A the default way and
attempt to run it on an AArch64-only core, we only get a runtime panic.
Now a compile-time check has been added to ensure that this flag has the
appropriate value when AArch64-only cores are included in the build.
Change-Id: I298ec550037fafc9347baafb056926d149197d4c
Signed-off-by: John Tsichritzis <john.tsichritzis@arm.com>
This patch fixes this issue:
https://github.com/ARM-software/tf-issues/issues/660
The introduced changes are the following:
1) Some cores implement cache coherency maintenance operation on the
hardware level. For those cores, such as - but not only - the DynamIQ
cores, it is mandatory that TF-A is compiled with the
HW_ASSISTED_COHERENCY flag. If not, the core behaviour at runtime is
unpredictable. To prevent this, compile time checks have been added and
compilation errors are generated, if needed.
2) To enable this change for FVP, a logical separation has been done for
the core libraries. A system cannot contain cores of both groups, i.e.
cores that manage coherency on hardware and cores that don't do it. As
such, depending on the HW_ASSISTED_COHERENCY flag, FVP includes the
libraries only of the relevant cores.
3) The neoverse_e1.S file has been added to the FVP sources.
Change-Id: I787d15819b2add4ec0d238249e04bf0497dc12f3
Signed-off-by: John Tsichritzis <john.tsichritzis@arm.com>
