I check MIPS microMIPS manual [1], and found the major opcode might
be wrong. I add a comment to explicitly indicate what manual I am refering
to, and according that manual I remove microMIPS32 major opcodes 0x1f.
As for others, like 0x16, 0x17, 0x36 and 0x37, they are for higher-order
MIPS ISA level or new revision of this microMIPS architecture. Quote
from Johnson, they are belong MIPS64 [2].
[1] http://www.mips.com/products/architectures/micromips/#specifications
MIPS Architecture for Programmers Volume II-B:
The microMIPS32 Instruction Set (Revision 3.05)
MD00582-2B-microMIPS-AFP-03.05.pdf
[2] http://www.mips.com/products/architectures/mips64/
MIPS Architecture For Programmers
Volume II-A: The MIPS64 Instruction Set
MD00087-2B-MIPS64BIS-AFP-03.51.pdf
Signed-off-by: Chen Wei-Ren <chenwj@iis.sinica.edu.tw>
Reviewed-by: Eric Johnson <ericj@mips.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Current QEMU MIPS POOL32AXF encoding comes from microMIPS32
and microMIPS32 DSP. Add comment here to help reading.
Signed-off-by: Chen Wei-Ren <chenwj@iis.sinica.edu.tw>
Reviewed-by: Eric Johnson <ericj@mips.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
While reading microMIPS decoding, I found a possible wrong opcode
encoding. According to [1] page 166, the bits 13..12 for MULTU is
0x01 rather than 0x00. Please review, thanks.
[1] MIPS Architecture for Programmers VolumeIV-e: The MIPS DSP
Application-Specific Extension to the microMIPS32 Architecture
Signed-off-by: Chen Wei-Ren <chenwj@iis.sinica.edu.tw>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
The call to gen_logic_imm for OPC_LUI passes -1 for rs. This
causes the MIPS_DEBUG statement to seg fault due to the deference
of regnames[rs]. This patch fixes that.
Signed-off-by: Eric Johnson <ericj@mips.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
(aurel32: replaced static string formating by a static string)
Pass around CPUArchState instead of using global cpu_single_env.
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Acked-by: Richard Henderson <rth@twiddle.net>
Acked-by: Aurelien Jarno <aurelien@aurel32.net>
Acked-by: Guan Xuetao <gxt@mprc.pku.edu.cn>
Use the deposit op instead of and hardcoded bit field insertion. It
allows the host to emit the corresponding instruction if available.
At the same time remove the (lsb > msb) test. The MIPS64R2 instruction
set manual says "Because of the instruction format, lsb can never be
greater than msb, so there is no UNPREDICATABLE case for this
instruction."
(Bug reported as LP:1071149.)
Cc: Никита Канунников <n.kanunnikov@sbtcom.ru>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
The result of a division by 0, or a division of INT_MIN by -1 in the
signed case, is unpredictable. Just replace 0 by 1 in that case so that
it doesn't trigger a floating point exception on the host.
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Avoid the branches in movn/movz implementation and replace them with
movcond. Also update a wrong command.
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Store conditional operations only need local temps in user mode. Fix
the code to use temp local only in user mode, this spares two memory
stores in system mode.
At the same time remove a wrong a wrong copied & pasted comment,
store operations don't have a register destination.
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Load/store from helpers should be avoided as they are quite
inefficient. Rewrite unaligned loads instructions using TCG and
aligned loads. The number of actual loads operations to implement
an unaligned load instruction is reduced from up to 8 to 1.
Note: As we can't rely on shift by 32 or 64 undefined behaviour,
the code loads already shift by one constants.
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Load/store operations use macros for historical reasons. Now that there
is no point in keeping them, replace them by direct calls to qemu_ld/st.
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Use the new softfloat floatXX_muladd() functions to implement the madd,
msub, nmadd and nmsub instructions. At the same time replace the name of
the helpers by the name of the instruction, as the only reason for the
previous names was to keep the macros simple.
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
When the CPU state after a possible retranslation is going to be handled
through code retranslation, we don't need to save the CPU state before.
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
When the CPU state is restored through retranslation after an exception,
btarget should also be restored.
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
All switch() decoding instruction have a default entry, so it is possible
to have unused enum entries. Remove conditional definitions of MIPS64
opcode enums, as it only makes the code less readable.
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
With normal FP, this doesn't have much affect on the generated code,
because most of the FP operations are not CONST/PURE, and so we spill
registers in about the same frequency as the explicit load/stores.
But with Loongson multimedia instructions, which are all integral and
whose helpers are in fact CONST+PURE, this greatly improves the code.
Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
For all targets that currently call tcg_gen_debug_insn_start,
add CPU_LOG_TB_OP_OPT to the condition that gates it.
This is useful for comparing optimization dumps, when the
pre-optimization dump is merely noise.
Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Implements all of the COP2 instructions except for the S<cond>
family of comparisons. The documentation is unclear for those.
Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
this will prevent some of the compilation errors with debugging
enabled from creeping back in.
Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
The macro uses the DisasContext. Pass it around as needed.
Signed-off-by: Richard Henderson <rth@twiddle.net>
Acked-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Used by MIPS_DEBUG, when enabled.
Signed-off-by: Richard Henderson <rth@twiddle.net>
Acked-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Add an explicit CPUState parameter instead of relying on AREG0
and switch to AREG0 free mode.
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Acked-by: Aurelien Jarno <aurelien@aurel32.net>
This change updates the CPU reset sequence to use a common piece of code
that figures out CPU state flags, fixing the problem with MIPS_HFLAG_COP1X
not being set where applicable that causes floating-point MADD family
instructions (and other instructions from the MIPS IV FP subset) to trap.
As compute_hflags is now shared between op_helper.c and translate.c, the
function is now moved to a common header. There are no changes to this
function.
The problem was seen with the 24Kf MIPS32r2 processor in user emulation.
The new approach prevents system and user emulation from diverging -- all
the hflags state is initialized in one place now.
Signed-off-by: Maciej W. Rozycki <macro@codesourcery.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
The microMIPS SWP and SDP instructions do not modify GPRs. So their
behavior is well defined when RD equals BASE. The MIPS Architecture
Verification Programs (AVPs) check that they work as expected. This
is required for AVPs to pass.
Signed-off-by: Eric Johnson <ericj@mips.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
The MIPS Architecture Verification Programs (AVPs) check privileged
instructions for the required privilege level. These changes are needed
to pass the AVP suite.
Signed-off-by: Eric Johnson <ericj@mips.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
The kernel will emulate this instruction if it's not supported
natively. This insn is used for TLS, among other things, and
so is required by modern glibc.
Signed-off-by: Richard Henderson <rth@twiddle.net>
Cc: Riku Voipio <riku.voipio@iki.fi>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
The FS input to CVT.PS.S is the high half and FT is the low half.
tcg_gen_concat_i32_i64 takes the low half first, so the operands
were in the wrong order.
Signed-off-by: Richard Sandiford <rdsandiford@googlemail.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Read the second input operand of RECIP2.S and RECIP2.PS from FT rather
than FD. RECIP2.D is already correct.
Signed-off-by: Richard Sandiford <rdsandiford@googlemail.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
While running in the usermode emulator all of the required*
MIPS32r2 RDHWR hardware registers should be accessible (the
Linux kernel enables access to these same registers). Note
that these registers are still enabled when the MIPS ISA is
not release 2. This is OK since the Linux kernel emulates
access to them when they are not available in hardware.
* There is also the ULR register which is only recommended
for full release 2 compliance. Incidentally, accessing
this register in the current implementation works fine
without flipping its access bit.
Signed-off-by: Meador Inge <meadori@codesourcery.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
This change addresses a problem where QEMU incorrectly traps on
floating-point MADD group instructions with SIGILL, at least while
emulating MIPS32r2 processors. These instructions use the COP1X major
opcode and include ones like:
madd.d $f2,$f4,$f2,$f6
Here's Nathan's original analysis of the problem:
"QEMU essentially does:
d = find_cpu (cpu_string) // get CPU definition
fpu_init (env, d) // initialize fpu state (init FCR0, basically)
cpu_reset (env)
...and the cpu_reset call clears all interesting state that fpu_init
setup, then proceeds to reinitialize all the CP0 registers...but not
FCR0."
I have verified this change with system emulation running the GDB test
suite for the mips-sde-elf target (o32, big endian, 24Kf CPU emulated),
there were 55 progressions and no regressions.
Signed-off-by: Maciej W. Rozycki <macro@codesourcery.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Commit 0f71a7095d (target-mips: QOM'ify
CPU) hooked up cpu_state_reset() to CPUClass::reset(). Dropping the
introduction of subclasses for 1.1, due to mips_def_t the reset code
could not be QOM'ified yet, i.e. cpu_state_reset() will not forward to
CPUClass::reset().
Update cpu_mips_init() with cpu_reset() nontheless, indirectly calling
cpu_state_reset(), so that generic code can be converted to call
cpu_reset().
Signed-off-by: Andreas Färber <afaerber@suse.de>
Evgeny Voevodin
Aurelien Jarno
陳韋任 (Wei-Ren Chen)
Eric Johnson
Blue Swirl
Jia Liu
Richard Henderson
Maciej W. Rozycki
Richard Sandiford
Meador Inge
Nathan Froyd
Andreas Färber