This reverts commit cd0372c515.
The patch is incorrect in that it retains copies between globals and
non-local temps, and non-local temps still die at the end of the BB.
Failing test case for hppa:
.globl _start
_start:
cmpiclr,= 0x24,%r19,%r0
cmpiclr,<> 0x2f,%r19,%r19
---- 00010057 0001005b
movi_i32 tmp0,$0x24
sub_i32 tmp1,tmp0,r19
mov_i32 tmp2,tmp0
mov_i32 tmp3,r19
movi_i32 tmp1,$0x0
---- 0001005b 0001005f
brcond_i32 tmp2,tmp3,eq,$L1
movi_i32 tmp0,$0x2f
sub_i32 tmp1,tmp0,r19
mov_i32 tmp2,tmp0
mov_i32 tmp3,r19
movi_i32 tmp1,$0x0
mov_i32 r19,tmp1
setcond_i32 psw_n,tmp2,tmp3,ne
set_label $L1
In this case, both copies of "mov_i32 tmp3,r19" are removed. The
second because opt thought it was redundant. The first is removed
later by liveness because tmp3 is known to be dead. This leaves
the setcond_i32 with an uninitialized input.
Revert the entire patch for 5.2, and a proper optimization across
the branch may be considered for the next development cycle.
Reported-by: qemu@igor2.repo.hu
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Since 6e6c4efed9, there has been a more appropriate range check
done later at the end of tcg_gen_code. There, a failing range
check results in a returned error code, which causes the TB to
be restarted at half the size.
Reported-by: Sai Pavan Boddu <saipava@xilinx.com>
Tested-by: Sai Pavan Boddu <sai.pavan.boddu@xilinx.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
We can easily propagate temp values through the entire extended
basic block (in this case, the set of blocks connected by fallthru),
simply by not discarding the register state at the branch.
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
We can easily register allocate the entire extended basic block
(in this case, the set of blocks connected by fallthru), simply
by not discarding the register state at the branch.
This does not help blocks starting with a label, as they are
reached via a taken branch, and that would require saving the
complete register state at the branch.
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
The cmp_vec opcode is mandatory; this symbol is unused.
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
When the two arguments are identical, this can be reduced to
dup_vec or to mov_vec from a tcg_constant_vec.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
The definition of INDEX_op_dupi_vec is that it operates on
units of tcg_target_ulong -- in this case 32 bits. It does
not work to use this for a uint64_t value that happens to be
small enough to fit in tcg_target_ulong.
Fixes: d2fd745fe8
Fixes: db432672dc
Cc: qemu-stable@nongnu.org
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
The previous change wrongly stated that 32-bit avx2 should have
used VPBROADCASTW. But that's a 16-bit broadcast and we want a
32-bit broadcast.
Fixes: 7b60ef3264
Cc: qemu-stable@nongnu.org
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
These are easier to set and test when they have their own fields.
Reduce the size of alias_index and sort_index to 4 bits, which is
sufficient for TCG_MAX_OP_ARGS. This leaves only the bits indicating
constants within the ct field.
Move all initialization to allocation time, rather than init
individual fields in process_op_defs.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
This wasn't actually used for anything, really. All variable
operands must accept registers, and which are indicated by the
set in TCGArgConstraint.regs.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
This uses an existing hole in the TCGArgConstraint structure
and will be convenient for keeping the data in one place.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
The union is unused; let "regs" appear in the main structure
without the "u.regs" wrapping.
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
With larger vector sizes, it turns out oprsz == maxsz, and we only
need to represent mismatch for oprsz <= 32. We do, however, need
to represent larger oprsz and do so without reducing SIMD_DATA_BITS.
Reduce the size of the oprsz field and increase the maxsz field.
Steal the oprsz value of 24 to indicate equality with maxsz.
Tested-by: Frank Chang <frank.chang@sifive.com>
Reviewed-by: Frank Chang <frank.chang@sifive.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Instead of creating GStrings and passing them into log_disas,
just print the annotations directly in tb_gen_code.
Fix the annotations for the slow paths of the TB, after the
part implementing the final guest instruction.
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
clang's C11 atomic_fetch_*() functions only take a C11 atomic type
pointer argument. QEMU uses direct types (int, etc) and this causes a
compiler error when a QEMU code calls these functions in a source file
that also included <stdatomic.h> via a system header file:
$ CC=clang CXX=clang++ ./configure ... && make
../util/async.c:79:17: error: address argument to atomic operation must be a pointer to _Atomic type ('unsigned int *' invalid)
Avoid using atomic_*() names in QEMU's atomic.h since that namespace is
used by <stdatomic.h>. Prefix QEMU's APIs with 'q' so that atomic.h
and <stdatomic.h> can co-exist. I checked /usr/include on my machine and
searched GitHub for existing "qatomic_" users but there seem to be none.
This patch was generated using:
$ git grep -h -o '\<atomic\(64\)\?_[a-z0-9_]\+' include/qemu/atomic.h | \
sort -u >/tmp/changed_identifiers
$ for identifier in $(</tmp/changed_identifiers); do
sed -i "s%\<$identifier\>%q$identifier%g" \
$(git grep -I -l "\<$identifier\>")
done
I manually fixed line-wrap issues and misaligned rST tables.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20200923105646.47864-1-stefanha@redhat.com>
We already support duplication of 128-bit blocks. This extends
that support to 256-bit blocks. This will be needed by SVE2.
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Do not store back to the exact memory from which we just loaded.
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
The fallback inline expansion for vectorized absolute value,
when the host doesn't support such an insn was flawed.
E.g. when a vector of bytes has all elements negative, mask
will be 0xffff_ffff_ffff_ffff. Subtracting mask only adds 1
to the low element instead of all elements becase -mask is 1
and not 0x0101_0101_0101_0101.
Signed-off-by: Stephen Long <steplong@quicinc.com>
Message-Id: <20200813161818.190-1-steplong@quicinc.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
With Makefiles that have automatically generated dependencies, you
generated includes are set as dependencies of the Makefile, so that they
are built before everything else and they are available when first
building the .c files.
Alternatively you can use a fine-grained dependency, e.g.
target/arm/translate.o: target/arm/decode-neon-shared.inc.c
With Meson you have only one choice and it is a third option, namely
"build at the beginning of the corresponding target"; the way you
express it is to list the includes in the sources of that target.
The problem is that Meson decides if something is a source vs. a
generated include by looking at the extension: '.c', '.cc', '.m', '.C'
are sources, while everything else is considered an include---including
'.inc.c'.
Use '.c.inc' to avoid this, as it is consistent with our other convention
of using '.rst.inc' for included reStructuredText files. The editorconfig
file is adjusted.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Forgetting this asserts when tcg_gen_cmp_vec is called from
within tcg_gen_cmpsel_vec.
Fixes: 72b4c792c7
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
If the output of the move is dead, then the last use is in
the store. If we propagate the input to the store, then we
can remove the move opcode entirely.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
We already had support for rotlv, using a target-specific opcode;
convert to use the generic opcode. Handle rotrv via simple negation.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
For immediate rotate , we can implement this in two instructions,
using SLI. For variable rotate, the oddness of aarch64 right-shift-
as-negative-left-shift means a backend-specific expansion works best.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
For immediates, we must continue the special casing of 8-bit
elements. The other element sizes and shift types are trivially
implemented with shifts.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
No host backend support yet, but the interfaces for rotls
are in place. Only implement left-rotate for now, as the
only known use of vector rotate by scalar is s390x, so any
right-rotate would be unused and untestable.
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
We do not reflect this expansion in tcg_can_emit_vecop_list,
so it is unused and unusable. However, we actually perform
the same expansion in do_gvec_shifts, so it is also unneeded.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
No host backend support yet, but the interfaces for rotlv
and rotrv are in place.
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
v3: Drop the generic expansion from rot to shift; we can do better
for each backend, and then this code becomes unused.
No host backend support yet, but the interfaces for rotli
are in place. Canonicalize immediate rotate to the left,
based on a survey of architectures, but provide both left
and right shift interfaces to the translators.
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
For the benefit of compatibility of function pointer types,
we have standardized on int32_t and int64_t as the integral
argument to tcg expanders.
We converted most of them in 474b2e8f0f, but missed the rotates.
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
We have this same parameter for GVecGen2i, GVecGen3,
and GVecGen3i. This will make some SVE2 insns easier
to parameterize.
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Add a version of tcg_gen_dup_* that takes both immediate and
a vector element size operand. This will replace the set of
tcg_gen_gvec_dup{8,16,32,64}i functions that encode the element
size within the function name.
Reviewed-by: LIU Zhiwei <zhiwei_liu@c-sky.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
We were only constructing the 64-bit element, and not
replicating the 64-bit element across the rest of the vector.
Cc: qemu-stable@nongnu.org
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
A given RISU testcase for SVE can produce
tcg-op-vec.c:511: do_shifti: Assertion `i >= 0 && i < (8 << vece)' failed.
because expand_vec_sari gave a shift count of 32 to a MO_32
vector shift.
In 44f1441dbe, we changed from direct expansion of vector opcodes
to re-use of the tcg expanders. So while the comment correctly notes
that the hw will handle such a shift count, we now have to take our
own sanity checks into account. Which is easy in this particular case.
Fixes: 44f1441dbe
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
It is, after all, just two instructions.
Profiling on a cortex-a15, using -d nochain to increase the number
of exit_tb that are executed, shows a minor improvement of 0.5%.
Signed-off-by: Richard Henderson <rth@twiddle.net>
We will shortly use this function from tcg_out_op as well.
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>
