Simplify cpu_loop by doing all of the decode in translate.
This fixes a bug in that cpu_loop was not handling the
different layout of the R6 version of break16. This fixes
a bug in that cpu_loop extracted the wrong bits for the
mips16e break16 instruction.
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20220107213243.212806-17-richard.henderson@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
When using the Loongson-3A4000 CPU, the MSAIR is returned with a
zero value (because unimplemented). Checking on real hardware,
this value appears incorrect:
$ cat /proc/cpuinfo
system type : generic-loongson-machine
machine : loongson,generic
cpu model : Loongson-3 V0.4 FPU V0.1
model name : Loongson-3A R4 (Loongson-3A4000) @ 1800MHz
isa : mips1 mips2 mips3 mips4 mips5 mips32r1 mips32r2 mips64r1 mips64r2
ASEs implemented : vz msa loongson-mmi loongson-cam loongson-ext loongson-ext2
...
Checking the CFCMSA opcode result with gdb we get 0x60140:
Breakpoint 1, 0x00000001200037c4 in main ()
1: x/i $pc
=> 0x1200037c4 <main+52>: cfcmsa v0,msa_ir
(gdb) si
0x00000001200037c8 in main ()
(gdb) i r v0
v0: 0x60140
MSAIR bits 17 and 18 are "reserved" per the spec revision 1.12,
so mask them out, and set MSAIR=0x0140 for the Loongson-3A4000
CPU model added in commit af868995e1.
Cc: Huacai Chen <chenhuacai@kernel.org>
Reviewed-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20211026180920.1085516-1-f4bug@amsat.org>
Only the MSA generic opcode was overlapping with the other
instructions. Since the previous commit removed it, we can
now remove the overlap group. The decodetree script forces
us to re-indent the opcodes.
Diff trivial to review using `git-diff --ignore-all-space`.
Reviewed-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20211028210843.2120802-32-f4bug@amsat.org>
Convert 3-register operations to decodetree.
Per the Encoding of Operation Field for 3R Instruction Format'
(Table 3.25), these instructions are not defined for the BYTE
format. Therefore the TRANS_DF_iii_b() macro returns 'false'
in that case, because no such instruction is decoded.
Reviewed-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20211028210843.2120802-22-f4bug@amsat.org>
Convert 3-register operations to decodetree.
Since the 'data format' field is a constant value, use
tcg_constant_i32() instead of a TCG temporary.
Note, the format definition could be named @3rf_b (for
3R with a df field BYTE-based) but since the instruction
class is named '3R', we simply call the format @3r to
ease reviewing the msa.decode file.
However we directly call the trans_msa_3rf() function,
which handles the BYTE-based df field.
Reviewed-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20211028210843.2120802-21-f4bug@amsat.org>
Convert the LDI opcode (Immediate Load) to decodetree. Since it
overlaps with the generic MSA handler, use a decodetree overlap
group.
Since the 'data format' field is a constant value, use
tcg_constant_i32() instead of a TCG temporary.
Reviewed-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20211028210843.2120802-9-f4bug@amsat.org>
This 'shift amount' format is not always 16-bit, so name it
generically as 'sa'. This will help to unify the various
arg_msa decodetree generated structures.
Rename the @bz format -> @bz_v (specific @bz with df=3) and
@bz_df -> @bz (generic @bz).
Since we modify &msa_bz, re-align its arguments, so the other
structures added in the following commits stay visually aligned.
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20211028210843.2120802-8-f4bug@amsat.org>
While the first 'off' variable assignment is unused, it helps
to better understand the code logic. Move the assignation where
it would have been used so it is easier to compare the MSA
registers based on FPU ones versus the MSA specific registers.
Reviewed-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20211023214803.522078-34-f4bug@amsat.org>
The fallback code in cpu_loop_exit_sigsegv is sufficient
for mips linux-user.
This means we can remove tcg/user/tlb_helper.c entirely.
Remove the code from cpu_loop that raised SIGSEGV.
Reviewed-by: Warner Losh <imp@bsdimp.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
While for the DEXTR_S.H opcode:
"The shift argument is provided in the instruction."
For the DEXTRV_S.H opcode we have:
"The five least-significant bits of register rs provide the
shift argument, interpreted as a five-bit unsigned integer;
the remaining bits in rs are ignored."
While 't1' contains the 'rs' register content (the shift value
for DEXTR_S.H), we need to load the value of 'rs' for DEXTRV_S.H.
We can directly use the v1_t TCG register which already contains
this shift value.
Fixes: b53371ed5d ("target-mips: Add ASE DSP accumulator instructions")
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20211013215652.1764551-1-f4bug@amsat.org>
Commits cbe50b9a8e ("target-mips: add MSA VEC/2R format instructions")
and 3bdeb68866 ("target-mips: add MSA 2RF format instructions") added
the MSA 2R/2RF instructions. However these instructions don't use any
target vector register, so remove the unused TCG temporaries.
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20211003175743.3738710-2-f4bug@amsat.org>
Per the "MIPS Architecture Extension: nanoMIPS32 DSP TRM" rev 0.04,
MULT and MULTU opcodes:
The value of ac selects an accumulator numbered from 0 to 3.
When ac=0, this refers to the original HI/LO register pair of the
MIPS32 architecture.
In Release 6 of the MIPS Architecture, accumulators are eliminated
from MIPS32.
Ensure pre-Release 6 is restricted to HI/LO registers pair.
Fixes: 8b3698b294 ("target/mips: Add emulation of DSP ASE for nanoMIPS - part 4")
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
As per an ancient comment in mips_tr_translate_insn about the
expectations of gdb, when restarting the insn in a delay slot
we also re-execute the branch. Which means that we are
expected to execute two insns in this case.
This has been broken since 8b86d6d258, where we forced max_insns
to 1 while single-stepping. This resulted in an exit from the
translator loop after the branch but before the delay slot is
translated.
Increase the max_insns to 2 for this case. In addition, bypass
the end-of-page check, for when the branch itself ends the page.
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Rather than use 4-16 separate operations, use 2 operations
plus some byte reordering as necessary.
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
