FVP: Add support for multi-threaded CPUs
authorJeenu Viswambharan <jeenu.viswambharan@arm.com>
Tue, 15 Nov 2016 13:53:57 +0000 (13:53 +0000)
committerJeenu Viswambharan <jeenu.viswambharan@arm.com>
Tue, 1 Aug 2017 13:33:47 +0000 (14:33 +0100)
commit11ad8f208db42f7729b0ce2bd16c631c293e665c
treef4001faf329ffbaf44e8770dac44fac4b8649f46
parenteecdf19b73fb9458915fb21ea74ba8aa96e90b36
FVP: Add support for multi-threaded CPUs

ARM CPUs with multi-threading implementation has more than one
Processing Element in a single physical CPU. Such an implementation will
reflect the following changes in the MPIDR register:

  - The MT bit set;

  - Affinity levels pertaining to cluster and CPUs occupy one level
    higher than in a single-threaded implementation, and the lowest
    affinity level pertains to hardware threads. MPIDR affinity level
    fields essentially appear shifted to left than otherwise.

The FVP port henceforth assumes that both properties above to be
concomitant on a given FVP platform.

To accommodate for varied MPIDR formats at run time, this patch
re-implements the FVP platform-specific functions that translates MPIDR
values to a linear indices, along with required validation. The same
treatment is applied for GICv3 MPIDR hashing function as well.

An FVP-specific build option FVP_MAX_PE_PER_CPU is introduced which
specifies the maximum number of threads implemented per CPU. For
backwards compatibility, its value defaults to 1.

Change-Id: I729b00d3e121d16ce9a03de4f9db36dfac580e3f
Signed-off-by: Jeenu Viswambharan <jeenu.viswambharan@arm.com>
docs/user-guide.rst
plat/arm/board/fvp/aarch32/fvp_helpers.S
plat/arm/board/fvp/aarch64/fvp_helpers.S
plat/arm/board/fvp/fvp_def.h
plat/arm/board/fvp/fvp_topology.c
plat/arm/board/fvp/include/platform_def.h
plat/arm/board/fvp/platform.mk
plat/arm/common/arm_gicv3.c