On 12/17/2017 10:21 PM, Hauke Mehrtens wrote:
> This patch made GCC produce broken code, remove it.
> In mp_cmp_d() function in th libtommath shipped with dropbear the
> following code was compiled wrong:
>
> /* compare based on magnitude */
> if (a->used > 1) {
> return 1;
> }
>
> In the broken ASM this part returned -1 like the previous return
> statement did instead of 1 like it should.
>
> This did not happen when the -funroll-loops option was given to GCC.
>
> Signed-off-by: Hauke Mehrtens <ha...@hauke-m.de>
This is the broken code:
0041d978 <mp_cmp_d>:
41d978: 8c830008 lw v1,8(a0)
41d97c: 24020001 li v0,1
41d980: 1062000c beq v1,v0,41d9b4 <mp_cmp_d+0x3c>
41d984: 2402ffff li v0,-1
41d988: 8c830000 lw v1,0(a0)
41d98c: 28630002 slti v1,v1,2
41d990: 10600008 beqz v1,41d9b4 <mp_cmp_d+0x3c>
41d994: 00000000 nop
41d998: 8c82000c lw v0,12(a0)
41d99c: 8c420000 lw v0,0(v0)
41d9a0: 00a2182b sltu v1,a1,v0
41d9a4: 14600005 bnez v1,41d9bc <mp_cmp_d+0x44>
41d9a8: 00000000 nop
41d9ac: 0045102b sltu v0,v0,a1
41d9b0: 00021023 negu v0,v0
41d9b4: 03e00008 jr ra
41d9b8: 00000000 nop
41d9bc: 03e00008 jr ra
41d9c0: 24020001 li v0,1
To fix this in line 41d994 "li v0,1" should be added instated of the nop.
Without this patch I ma getting this code:
0041d864 <mp_cmp_d>:
41d864: 8c860008 lw a2,8(a0)
41d868: 24030001 li v1,1
41d86c: 10c3000b beq a2,v1,41d89c <mp_cmp_d+0x38>
41d870: 2402ffff li v0,-1
41d874: 8c860000 lw a2,0(a0)
41d878: 28c60002 slti a2,a2,2
41d87c: 10c00007 beqz a2,41d89c <mp_cmp_d+0x38>
41d880: 24020001 li v0,1
41d884: 8c83000c lw v1,12(a0)
41d888: 8c630000 lw v1,0(v1)
41d88c: 00a3202b sltu a0,a1,v1
41d890: 14800002 bnez a0,41d89c <mp_cmp_d+0x38>
41d894: 0065182b sltu v1,v1,a1
41d898: 00031023 negu v0,v1
41d89c: 03e00008 jr ra
41d8a0: 00000000 nop
This looks correct.
Hauke
_______________________________________________
Lede-dev mailing list
Lede-dev@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/lede-dev
