Re: [PATCH] [MIPS] Loongson EXT fixes

[Jiaxun Yang <jiaxun dot yang at flygoat dot com>]

>> +++ b/gas/doc/as.texi
>> @@ -1644,7 +1644,7 @@ This tells the assembler to accept Loongson EXT instructions.
>> @item -mloongson-ext2
>> @itemx -mno-loongson-ext2
>> Generate code for the Loongson EXTensions R2 (EXT2) instructions.
>> -This option implies @samp{-mloongson-ext}.
>> +This option implies @samp{-mloongson-ext2}.
>> This tells the assembler to accept Loongson EXT2 instructions.
>> @samp{-mno-loongson-ext2} turns off this option.
> 
> It is ok, the -mloongson-ext2 will trun on -mloongson-ext.

Thanks for the clarification.

> 
>> diff --git a/opcodes/mips-dis.c b/opcodes/mips-dis.c
>> index 755bbe294b..18dde79294 100644
>> --- a/opcodes/mips-dis.c
>> +++ b/opcodes/mips-dis.c
>> @@ -824,6 +824,14 @@ mips_convert_abiflags_ases (unsigned long afl_ases)
>> opcode_ases |= ASE_DSPR3;
>> if (afl_ases & AFL_ASE_MIPS16E2)
>> opcode_ases |= ASE_MIPS16E2;
>> + if (afl_ases & AFL_ASE_LOONGSON_MMI)
>> + opcode_ases |= ASE_LOONGSON_MMI;
>> + if (afl_ases & AFL_ASE_LOONGSON_CAM)
>> + opcode_ases |= ASE_LOONGSON_CAM;
>> + if (afl_ases & AFL_ASE_LOONGSON_EXT)
>> + opcode_ases |= ASE_LOONGSON_EXT;
>> + if (afl_ases & AFL_ASE_LOONGSON_EXT2)
>> + opcode_ases |= ASE_LOONGSON_EXT2;
>> return opcode_ases;
>> }
> 
> Thanks for catch this.
> 
>> diff --git a/opcodes/mips-opc.c b/opcodes/mips-opc.c
>> index 5270aeefa8..625b745edc 100644
>> --- a/opcodes/mips-opc.c
>> +++ b/opcodes/mips-opc.c
>> @@ -526,6 +526,8 @@ const struct mips_opcode mips_builtin_opcodes[] =
>> {"ctz", "d,s", 0x70000022, 0xfc1f07ff, WR_1|RD_2, 0, 0, LEXT2, 0 },
>> {"dcto", "d,s", 0x700000e2, 0xfc1f07ff, WR_1|RD_2, 0, 0, LEXT2, 0 },
>> {"dctz", "d,s", 0x700000a2, 0xfc1f07ff, WR_1|RD_2, 0, 0, LEXT2, 0 },
>> +{"gsandn", "d,s,t", 0x70000483, 0xfc0007ff, WR_1|RD_2|RD_3, 0, 0, LEXT2, 0 },
>> +{"gsorn", "d,s,t", 0x70000486, 0xfc0007ff, WR_1|RD_2|RD_3, 0, 0, LEXT2, 0 },
> 
> The gsandn and gsorn belong to upcomming loongson-ext3.

According to Loongson-3A2000 Vol.2 [1], they do exist in GS464E, so I thought it should
be a part of ext2, is it true?

I just discovered that test for ext2 and cam never got run, will fix it in v2.

Btw, as we still didn't see any see action from Loongson in mainline LoongISA2 support,
I'm cooperating with Xuerui Wang <wangxuerui@qiniu.com> to archive this. I'm working
on Binutils side while he is working on GCC.
We've already got binutils patches from Loongson about MSA2, EXT3, mipsr6p,
MSA Crypto, AMO, LSX, LASX and trying to improve them to meet upstream requirements.

Thanks.

[1]: http://loongson.cn/uploadfile/cpu/3A2000/Loongson3A2000_user2.pdf

--
Jiaxun Yang