Re: divide 64-bit by constant for 32-bit target machines

[Dinar Temirbulatov <dtemirbulatov at gmail dot com>]

Hi,
Here is new version of patch based up on Paolo review, again tested on
arm-7l, mips-32r2 (74k), i686 without new regressions.
                          thanks, Dinar.

On Sat, May 26, 2012 at 4:45 PM, Paolo Bonzini <bonzini@gnu.org> wrote:
> Il 26/05/2012 14:35, Paolo Bonzini ha scritto:
>> ? ? ? ?/* We have to return
>>
>> ? ? ? ? ? ? z2 + ((u0 + u1) >> GET_MODE_BITSIZE (word_mode)).
>>
>> ? ? ? ? ? u0 + u1 are the upper two words of the three-word
>> ? ? ? ? ? intermediate result and they could have up to
>> ? ? ? ? ? 2 * GET_MODE_BITSIZE (word_mode) + 1 bits of precision.
>> ? ? ? ? ? We compute the extra bit by checking for carry, and add
>> ? ? ? ? ? 1 << GET_MODE_BITSIZE (word_mode) to z2 if there is carry. ?*/
>
> Oops, GET_MODE_BITSIZE (word_mode) is more concisely BITS_PER_WORD.
>
>>> > + ? ? ?tmp = expand_binop (mode, add_optab, u0, u1, tmp, 1, OPTAB_LIB_WIDEN);
>>> > + ? ? ?if (!tmp)
>>> > + ? ? ? ? ? ? return 0;
>> ? ? ? ? ?/* We have to return z2 + (tmp >> 32). ?We need
>>> > + ? ? ?/* Checking for overflow. ?*/
>> This is not overflow, it's carry (see above).
>>
>>> > + ? ? ?c = gen_reg_rtx (mode);
>>> > + ? ? ?c1 = gen_reg_rtx (mode);
>>> > + ? ? ?cres = gen_reg_rtx (mode);
>>> > +
>>> > + ? ? ?emit_store_flag_force (c, GT, u0, tmp, mode, 1, 1);
>>> > + ? ? ?emit_store_flag_force (c1, GT, u1, tmp, mode, 1, 1);
>>> > + ? ? ?result = expand_binop (mode, ior_optab, c, c1, cres, 1, OPTAB_LIB_WIDEN);
>>> > + ? ? ?if (!result)
>>> > + ? ? ? ? ? return 0;
>>> > +
>>> > + ? ? ?ccst = gen_reg_rtx (mode);
>>> > + ? ? ?ccst = expand_shift (LSHIFT_EXPR, mode, cres, 32, ccst, 1);
>> This 32 should be GET_MODE_BITSIZE (word_mode).
>
> Here, too.
>
> Paolo
>

Attachment: ChangeLog.txt
Description: Text document

Attachment: 28.patch
Description: Binary data