Vmcompress.vm requires vstart==0, so writing vstart with 0 is redundant.
To do this, spin off VI_LOOP_END_BASE from VI_LOOP_END. VI_LOOP_END
will contain VI_LOOP_END_BASE as well as a write of 0 to vstart.
See #1623 for full discussion.
RISC-V ISA states (21.1):
"A standard-compatible global encoding can also use standard prefixes
for non-standard extensions if the associated standard extensions are
not included in the global encoding."
Currently all the instructions (either from standard or custom
extensions) are all being inserted into a single std::vector which is
then being sorted. An instruction matching process performs linear
search on that vector. The problem is that when a custom extension uses
the same opcode as standard one (i.e. match and mask are equal to the
standard counterparts) it is undefined which instruction will be picked.
That is because in std::sort "The order of equal elements is not
guaranteed to be preserved". That being said it is impossible to define
custom extension (via customext) that would use the prefix of a disabled
standard extension.
In this change I separate custom and standard extensions in two separate
std::vector's. By default we report an error if they have common
elements (There're an additional processor_t constructor's argument that
skips this check). If this error is disabled during instruction matching
we first trying to find it among custom instructions. If it has been
found the search is stopped and custom instruction is executed,
otherwise we look for it among standard instructions. Overall this
change does not completely fix the problem but at least makes it
possible to use the feature of RISC-V ISA.
Use -iquote instead. This prevents our include paths from messing up
the system headers depended upon by libstdc++. (The specific problem
was syscall.h in fesvr/, which was interfering with libstdc++'s
dependence on the system's syscall.h for SYS_futex.)
Subproject headers can now be included in the following ways:
#include "foo.h" // for a header local to this subproject
#include <bar/baz.h>" // for a header in another subproject
But no longer:
#include <baz.h> // for a header in any subproject
As a special case, libfdt needs itself to be added to the -I path,
because their coding style is to use angle brackets for local headers.
The previous implementation raises virtual instruction on WFI when TW=1 in VU-mode. According to the recent discussion, we expect an illegal instruction exception in this case.
Reference: https://github.com/riscv/riscv-isa-manual/issues/1234
..rather than unbounded, as it used to be. This led to the rather
surprising issue #1600, where a part of the address space assumed to be
vacant would allow a subset of accesses.
The H extension defines that bits VSEIP, VSTIP, and VSSIP of hvip are
writable. (The other bits of hvip are read-only 0.) Only hip.VSSIP
(mip.VSSIP) is an alias of hvip.VSSIP. The hip.VSEIP is the logical-OR
of hvip.VSEIP, selected bit of hgeip by hstatus.VGEIN, and
platform-specific external interrupt signals to VS-level, e.g., from
AIA. The hip.VSTIP is the logical-OR of hvip.VSTIP and platform-specific
timer interrupt signals to VS-level, e.g., from Sstc. Thus, the read
values of hvip.VSEIP and hvip.VSTIP differ from the ones of hip.VSEIP
and hip.VSTIP (mip.VSEIP and mip.VSTIP). In other words, the hvip isn't
an alias (proxy) of mip.
The current aliasing (proxy) implementation does not provide the desired
behavior for hvip.VSEIP and hvip.VSTIP. An ISA-level behavior difference
is that any platform-specific external and timer interrupt signals
directed to VS-level should not be observable through the hvip. For
instance, the hvip should not observe the virtual timer interrupt signal
from the vstimecmp CSR (Sstc extension), which isn't true in the current
implementation. Additionally, the hvip should not observe the virtual
external interrupt signal from the IMSIC device (AIA extension).
Another ISA-level behavior difference is that the hgeip and
hstatus.VGEIN also should not affect hvip.VSEIP, which isn't true in the
current implementation.
This commit fixes the issue by giving the hvip a specialized class,
hvip_csr_t. The hvip_csr_t aliases the hvip.VSSIP to the mip.VSSIP but
decouples the hvip.VSEIP and hvip.VSTIP from mip.VSEIP and mip.VSTIP.
Additionally, the commit updates the read value of mip to be the
logical-OR of hvip.VSEIP, hvip.VSTIP, and other sources.
When menvcfg.STCE=0, mip.STIP reverts to its defined behavior as if
unsupporting Sstc extension. When henvcfg.STCE=0, mip.VSTIP reverts
to its defined behavior as if unsupporting Sstc extension. [https://github.com/riscv/riscv-time-compare/issues/5]
The previous Sstc implementation does not respect the xenvcfg.STCE.
In other words, the Sstc may assert mip.STIP (mip.VSTIP) when
menvcfg.STCE=0 (henvcfg.STCE=0), which is a misbehaving.
The number of triggers is configurable since https://github.com/riscv-software-src/riscv-isa-sim/pull/1219.
The trigger description was for the limited implementation when developing the extension. All trigger types are supported now. The information isn't very useful anymore.
Signed-off-by: YenHaoChen <39526191+YenHaoChen@users.noreply.github.com>
Ming-Yi Lai
Andrew Waterman
rbuchner
Alexander Romanov
Jerry Zhao
LIU Yu
Katharina
YenHaoChen
YenHaoChen
Hong Cheng
Scott Johnson