The mcontrol trigger can select either address or data for checking. The
The selection decides the priority of the trigger. For instance, the
address trigger has a higher priority over the page fault, and the page
fault has a higher priority over the data trigger.

The previous implementation only has the checking functions for data
trigger, which results in incorrect priority of address trigger.

This commit adds a has_data argument to indicate address trigger and the
priority of the trigger.

Fix build using Xcode 14

Don't include .a files in other .a files.  Doing so trips a sanity
check in the Xcode 14 toolchain, because the included .a files are
not themselves mach-o format, even though their contents are.

Fix vmv.x.s for RV32

The Spike internals require that, when XLEN is narrower than reg_t,
values be sign-extended to the width of reg_t.

Build and install the lib and header dependencies required by Hammer

Actually run a program in CI

To avoid long CI times, keep the program in binary form.

To avoid storing binary files in this repository, store the binaries as
an asset on a github release.

https://github.com/rivosinc/hammer



Signed-off-by: Jerin Joy <joy@rivosinc.com>

In CI, check that help message prints without error

Run CI with -Werror

Detect loading isa-incompatible code

add support for sscofpmf extension v0.5.2

detects the loading of isa-incompatible (i.e. 32 bit code to 64bit HART) code and emits an error message to help avoid unintentionally loading wrong elf.

The trigger matching only checks on TLB hit (after refill_tlb()).
However, instructions on plugin devices will never be filled into
the TLB; thus, triggers cannot fire when executing instructions on
the plugin devices.

The PR removes the if-condition of TLB hit for trigger checking.

Co-authored-by: Howard Yen-Hao Chen <yhchen@andestech.com>

Initialize triggers using default member initializers

Instead of constructor member initializer lists, which require
repeating the list of members, and cause ugly merge conflicts when the
list of members changes.

Remove obsolete enum trigger_operation_t

Non-functional changes for RVV

Moved to triggers.h and renamed in
a2a25874, but the old enum was
mistakenly left behind.

and VI_XI_LOOP_WITH_CARRY: require_vm is also included in
VI_CHECK_SSS

require_vector_novtype

require_vector is also included in VI_GENERAL_LOOP_BASE

vmv<nf>r.v depends on vtype, and therefore should check vill

It's nearly free to do so, because it's just virtual address space.

@davidbiancolin recently fell into this pit when using context_t with VCS.

Fix tval on illegal instruction faults with long illegal instruction

The autoconf @libdir@ and @includedir@ variables include the
prefix part so we were generating things like "/usr//usr/lib".

The current spike implementation does not include the ILEN (maximum
instruction length supported by the implementation), which is required
to constrain the value of tval on an illegal instruction exception.

Consider an implementation supporting only an RV64I base instruction
set. The ILEN is 32 bits (spec sec. 1.5), and the MXLEN is 64 bits (spec
sec. 5.1). Under an illegal instruction exception with the instruction
longer than the ILEN, the mtval should contain the first ILEN (32 bits)
of the faulting instruction. However, the current spike implementation
lets the mtval be the instruction's first MXLEN (64 bits).

To fix this bug, this PR masks out the upper bits of the tval and leaves
the first ILEN bits of the faulting instruction. When this PR is being
made, all official instructions are either 16 or 32 bits. So, We hard-
code the ILEN to 32 bits.

Rename misleading variable

Because it's always better to do so where possible.

No reason to use a variable misleadingly named 'paddr' to hold the
virtual address.

These are just HINTs, so Spike only needs ISA-parser support.