Each slot in a packet can be assigned to at most one instruction.
Although the assembler generally ought to enforce this rule, we better
be safe than sorry and also do some check to properly throw an "invalid
packet" exception on wrong slot assignments.
This should also make it easier to debug possible future errors caused
by missing updates to `find_iclass_slots()` rules in
target/hexagon/iclass.c.
Co-authored-by: Taylor Simpson <tsimpson@quicinc.com>
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Signed-off-by: Matheus Tavares Bernardino <quic_mathbern@quicinc.com>
Reviewed-by: Taylor Simpson <tsimpson@quicinc.com>
Tested-by: Taylor Simpson <tsimpson@quicinc.com>
Message-Id: <f8b829443523568823d062adf8bf6659bc6d4a3f.1683552984.git.quic_mathbern@quicinc.com>
The Hexagon PRM says that "The assembler automatically encodes
instructions in the packet in the proper order. In the binary encoding
of a packet, the instructions must be ordered from Slot 3 down to
Slot 0."
Prior to the architecture version v73, the slot constraints from
instruction "hintjr" only allowed it to be executed at slot 2.
With that in mind, consider the packet:
{
hintjr(r0)
nop
nop
if (!p0) memd(r1+#0) = r1:0
}
To satisfy the ordering rule quoted from the PRM, the assembler would,
thus, move one of the nops to the first position, so that it can be
assigned to slot 3 and the subsequent hintjr to slot 2.
However, since v73, hintjr can be executed at either slot 2 or 3. So
there is no need to reorder that packet and the assembler will encode it
as is. When QEMU tries to execute it, however, we end up hitting a
"misaliged store" exception because both the store and the hintjr will
be assigned to store 0, and some functions like `slot_is_predicated()`
expect the decode machinery to assign only one instruction per slot. In
particular, the mentioned function will traverse the packet until it
finds the first instruction at the desired slot which, for slot 0, will
be hintjr. Since hintjr is not predicated, the result is that we try to
execute the store regardless of the predicate. And because the predicate
is false, we had not previously loaded hex_store_addr[0] or
hex_store_width[0]. As a result, the store will decide de width based on
trash memory, causing it to be misaligned.
Update the slot constraints for hintjr so that QEMU can properly handle
such encodings.
Note: to avoid similar-but-not-identical issues in the future, we should
look for multiple instructions at the same slot during decoding time and
throw an invalid packet exception. That will be done in the subsequent
commit.
Signed-off-by: Matheus Tavares Bernardino <quic_mathbern@quicinc.com>
Reviewed-by: Taylor Simpson <tsimpson@quicinc.com>
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Message-Id: <0fcd8293642c6324119fbbab44741164bcbd04fb.1673616964.git.quic_mathbern@quicinc.com>
Running qemu-hexagon with a binary that was compiled for an arch version
unknown by qemu can produce a somewhat confusing message:
qemu-hexagon: unable to find CPU model 'unknown'
Let's give a bit more info by appending the eflags so that the message
becomes:
qemu-hexagon: unable to find CPU model 'unknown (0x69)'
Signed-off-by: Matheus Tavares Bernardino <quic_mathbern@quicinc.com>
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Tested-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Taylor Simpson <tsimpson@quicinc.com>
Message-Id: <8a8d013cc619b94fd4fb577ae6a8df26cedb972b.1683225804.git.quic_mathbern@quicinc.com>
Currently, qemu-hexagon only models the v67 cpu. Nonetheless if we try
to get this information with `-cpu help`, qemu just exists with an error
code and no output. Let's correct that.
The code is basically a copy from target/alpha/cpu.h, but we strip the
"-hexagon-cpu" suffix before printing. This is to avoid confusing
situations like the following:
$ qemu-hexagon -cpu help
Available CPUs:
v67-hexagon-cpu
$ qemu-hexagon -cpu v67-hexagon-cpu ./prog
qemu-hexagon: unable to find CPU model 'v67-hexagon-cpu'
Signed-off-by: Matheus Tavares Bernardino <quic_mathbern@quicinc.com>
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Tested-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Taylor Simpson <tsimpson@quicinc.com>
Message-Id: <b946e17c7e17eed9095700b54c5ead36e5d55dfa.1683225804.git.quic_mathbern@quicinc.com>
Currently, the python scripts used for the hexagon building will not
abort the compilation when there is an error parsing a register. Let's
make the compilation properly fail in such cases by rasing an exception
instead of just printing a warning message, which might get lost in the
output.
This patch was generated with:
git grep -l "Bad register" *hexagon* | \
xargs sed -i "" -e 's/print("Bad register parse: "[, ]*\([^)]*\))/hex_common.bad_register(\1)/g'
Plus the bad_register() helper added to hex_common.py.
Signed-off-by: Matheus Tavares Bernardino <quic_mathbern@quicinc.com>
Reviewed-by: Anton Johansson <anjo@rev.ng>
Tested-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Taylor Simpson <tsimpson@quicinc.com>
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Message-Id: <1f5dbd92f68fdd89e2647e4ba527a2c32cf0f070.1683217043.git.quic_mathbern@quicinc.com>
**** Changes in v2 ****
Fix yyassert's for sign and zero extends
Coverity reports a parameter that is "set but never used". This is caused
by an assignment operator being used instead of equality.
Co-authored-by: Taylor Simpson <tsimpson@quicinc.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Anton Johansson <anjo@rev.ng>
Tested-by: Anton Johansson <anjo@rev.ng>
Message-Id: <20230428204411.1400931-1-tsimpson@quicinc.com>
**** Changes in v3 ****
Fix bugs exposed by dpmpyss_rnd_s0 instruction
Set correct size/signedness for constants
Test cases added to tests/tcg/hexagon/misc.c
**** Changes in v2 ****
Fix bug in imm_print identified in clang build
Currently, idef-parser skips all floating point instructions. However,
there are some floating point instructions that can be handled.
The following instructions are now parsed
F2_sfimm_p
F2_sfimm_n
F2_dfimm_p
F2_dfimm_n
F2_dfmpyll
F2_dfmpylh
To make these instructions work, we fix some bugs in parser-helpers.c
gen_rvalue_extend
gen_cast_op
imm_print
lexer properly sets size/signedness of constants
Test cases added to tests/tcg/hexagon/fpstuff.c
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Tested-by: Anton Johansson <anjo@rev.ng>
Reviewed-by: Anton Johansson <anjo@rev.ng>
Message-Id: <20230501203125.4025991-1-tsimpson@quicinc.com>
The following items in the CPUHexagonState are only used for bookkeeping
within the translation of a packet. With recent changes that eliminate
the need to free TCGv variables, these make more sense to be transient
and kept in DisasContext.
The following items are moved
dczero_addr
branch_taken
this_PC
Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230427230012.3800327-22-tsimpson@quicinc.com>
The pkt_has_store_s1 field is only used for bookkeeping helpers with
a load. With recent changes that eliminate the need to free TCGv
variables, it makes more sense to make this transient.
These helpers already take the instruction slot as an argument. We
combine the slot and pkt_has_store_s1 into a single argument called
slotval.
Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230427230012.3800327-21-tsimpson@quicinc.com>
The pred_written variable in the CPUHexagonState is only used for
bookkeeping within the translation of a packet. With recent changes
that eliminate the need to free TCGv variables, these make more sense
to be transient and kept in DisasContext.
Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230427230012.3800327-20-tsimpson@quicinc.com>
The new_pred_value array in the CPUHexagonState is only used for
bookkeeping within the translation of a packet. With recent changes
that eliminate the need to free TCGv variables, these make more sense
to be transient and kept in DisasContext.
Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230427230012.3800327-19-tsimpson@quicinc.com>
The new_value array in the CPUHexagonState is only used for bookkeeping
within the translation of a packet. With recent changes that eliminate
the need to free TCGv variables, these make more sense to be transient
and kept in DisasContext.
Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230427230012.3800327-18-tsimpson@quicinc.com>
Precursor to moving new_value from the global state to DisasContext
USR will need to stay in the global state because some helpers will
set it's value
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230427230012.3800327-17-tsimpson@quicinc.com>
The following have overrides
S2_insert
S2_insert_rp
S2_asr_r_svw_trun
A2_swiz
These instructions have semantics that write to the destination
before all the operand reads have been completed. Therefore,
the idef-parser versions were disabled with the short-circuit patch.
Test cases added to tests/tcg/hexagon/read_write_overlap.c
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230427230012.3800327-16-tsimpson@quicinc.com>
The generated helpers for HVX use pass-by-reference, so they can't
short-circuit when the reads/writes overlap. The instructions with
overrides are OK because they use tcg_gen_gvec_*.
We add a flag has_hvx_helper to DisasContext and extend gen_analyze_funcs
to set the flag when the instruction is an HVX instruction with a
generated helper.
We add an override for V6_vcombine so that it can be short-circuited
along with a test case in tests/tcg/hexagon/hvx_misc.c
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230427230012.3800327-15-tsimpson@quicinc.com>
In certain cases, we can avoid the overhead of writing to future_VRegs
and write directly to VRegs. We consider HVX reads/writes when computing
ctx->need_commit. Then, we can early-exit from gen_commit_hvx.
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230427230012.3800327-14-tsimpson@quicinc.com>
In certain cases, we can avoid the overhead of writing to hex_new_pred_value
and write directly to hex_pred. We consider predicate reads/writes when
computing ctx->need_commit. The get_result_pred() function uses this
field to decide between hex_new_pred_value and hex_pred. Then, we can
early-exit from gen_pred_writes.
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230427230012.3800327-13-tsimpson@quicinc.com>
In certain cases, we can avoid the overhead of writing to hex_new_value
and write directly to hex_gpr. We add need_commit field to DisasContext
indicating if the end-of-packet commit is needed. If it is not needed,
get_result_gpr() and get_result_gpr_pair() can return hex_gpr.
We pass the ctx->need_commit to helpers when needed.
Finally, we can early-exit from gen_reg_writes during packet commit.
There are a few instructions whose semantics write to the result before
reading all the inputs. Therefore, the idef-parser generated code is
incompatible with short-circuit. We tell idef-parser to skip them.
For debugging purposes, we add a cpu property to turn off short-circuit.
When the short-circuit property is false, we skip the analysis and force
the end-of-packet commit.
Here's a simple example of the TCG generated for
0x004000b4: 0x7800c020 { R0 = #0x1 }
BEFORE:
---- 004000b4
movi_i32 new_r0,$0x1
mov_i32 r0,new_r0
AFTER:
---- 004000b4
movi_i32 r0,$0x1
This patch reintroduces a use of check_for_attrib, so we remove the
G_GNUC_UNUSED added earlier in this series.
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Brian Cain <bcain@quicinc.com>
Message-Id: <20230427230012.3800327-12-tsimpson@quicinc.com>
Have gen_analyze_funcs mark the registers that are read by the
instruction. We also mark the implicit reads using instruction
attributes.
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230427230012.3800327-11-tsimpson@quicinc.com>
When generating TCG, make sure we have read all the operand registers
before writing to the destination registers.
This is a prerequesite for short-circuiting where the source and dest
operands could be the same.
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230427230012.3800327-10-tsimpson@quicinc.com>
Only endloop instructions will conditionally write to a predicate.
When there is an endloop instruction, we preload the values into
new_pred_value.
The only place pred_written is needed is when HEX_DEBUG is on.
We remove the last use of check_for_attrib. However, new uses will be
introduced later in this series, so we mark it with G_GNUC_UNUSED.
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230427230012.3800327-9-tsimpson@quicinc.com>
These instructions have implicit writes to registers, so we don't
want them to be helpers when idef-parser is off.
The following instructions are overriden
S2_cabacdecbin
SA1_cmpeqi
Remove the log_pred_write function from op_helper.c
Remove references in macros.h
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Acked-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230427230012.3800327-8-tsimpson@quicinc.com>
With the overrides added in prior commits, this function is not used
Remove references in macros.h
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230427230012.3800327-7-tsimpson@quicinc.com>
These instructions have implicit reads from p0, so we don't want
them in helpers when idef-parser is off.
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230427230012.3800327-6-tsimpson@quicinc.com>
These instructions have implicit writes to registers, so we don't
want them to be helpers when idef-parser is off.
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230427230012.3800327-5-tsimpson@quicinc.com>
These instructions have implicit writes to registers, so we don't
want them to be helpers when idef-parser is off.
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Acked-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230427230012.3800327-4-tsimpson@quicinc.com>
Add DisasContext arg to gen_log_reg_write_pair also
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230427230012.3800327-3-tsimpson@quicinc.com>
Enable conditional compilation depending on whether idef-parser
is configured
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230427230012.3800327-2-tsimpson@quicinc.com>
Tests added for the following instructions
J2_callrh
J2_jumprh
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Anton Johansson <anjo@rev.ng>
Message-Id: <20230427224057.3766963-10-tsimpson@quicinc.com>
The following instructions are added
J2_callrh
J2_junprh
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Anton Johansson <anjo@rev.ng>
Message-Id: <20230427224057.3766963-9-tsimpson@quicinc.com>
The following instructions are tested
V6_vasrvuhubrndsat
V6_vasrvuhubsat
V6_vasrvwuhrndsat
V6_vasrvwuhsat
V6_vassign_tmp
V6_vcombine_tmp
V6_vmpyuhvs
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Anton Johansson <anjo@rev.ng>
Message-Id: <20230427224057.3766963-8-tsimpson@quicinc.com>
The following instructions are added
V6_vasrvuhubrndsat
V6_vasrvuhubsat
V6_vasrvwuhrndsat
V6_vasrvwuhsat
V6_vassign_tmp
V6_vcombine_tmp
V6_vmpyuhvs
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Anton Johansson <anjo@rev.ng>
Message-Id: <20230427224057.3766963-7-tsimpson@quicinc.com>
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Anton Johansson <anjo@rev.ng>
Message-Id: <20230427224057.3766963-6-tsimpson@quicinc.com>
The following instructions are added
V6_v6mpyvubs10_vxx
V6_v6mpyhubs10_vxx
V6_v6mpyvubs10
V6_v6mpyhubs10
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Anton Johansson <anjo@rev.ng>
Message-Id: <20230427224057.3766963-5-tsimpson@quicinc.com>
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Anton Johansson <anjo@rev.ng>
Message-Id: <20230427224057.3766963-4-tsimpson@quicinc.com>
The following instructions are added
L2_loadw_aq
L4_loadd_aq
R6_release_at_vi
R6_release_st_vi
S2_storew_rl_at_vi
S4_stored_rl_at_vi
S2_storew_rl_st_vi
S4_stored_rl_st_vi
The release instructions are nop's in qemu. The others behave as
loads/stores.
The encodings for these instructions changed some "don't care" bits
L2_loadw_locked
L4_loadd_locked
S2_storew_locked
S4_stored_locked
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Anton Johansson <anjo@rev.ng>
Message-Id: <20230427224057.3766963-3-tsimpson@quicinc.com>
Add support for the ELF flags
Move target/hexagon/cpu.[ch] to be v73
Change the compiler flag used by "make check-tcg"
The decbin instruction is removed in Hexagon v73, so check the
version before trying to compile the instruction.
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Anton Johansson <anjo@rev.ng>
Message-Id: <20230427224057.3766963-2-tsimpson@quicinc.com>
This header is supposed to be private to tcg and in fact
does not need to be included here at all.
Reviewed-by: Song Gao <gaosong@loongson.cn>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Instead of trying to unify all operations on uint64_t, use
mmu_lookup() to perform the basic tlb hit and resolution.
Create individual functions to handle access by size.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Instead of trying to unify all operations on uint64_t, pull out
mmu_lookup() to perform the basic tlb hit and resolution.
Create individual functions to handle access by size.
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Instead of playing with offsetof in various places, use
MMUAccessType to index an array. This is easily defined
instead of the previous dummy padding array in the union.
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Like cpu_in_exclusive_context, but also true if
there is no other cpu against which we could race.
Use it in tb_flush as a direct replacement.
Use it in cpu_loop_exit_atomic to ensure that there
is no loop against cpu_exec_step_atomic.
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
All uses have now been expunged.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Mark all memory operations that are not already marked with UNALIGN.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
In gen_ldx/gen_stx, the only two locations for memory operations,
mark the operation as either aligned (softmmu) or unaligned
(user-only, as if emulated by the kernel).
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
The opposite of MO_UNALN is MO_ALIGN.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Memory operations that are not already aligned, or otherwise
marked up, require addition of ctx->default_tcg_memop_mask.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
