Quoting Joern Rennecke :
Quoting Richard Sandiford :
1) As Richard B says, having "locked lengths" with the comment "care must
be taken to avoid cycles" doesn't sound like good design. So the
question was: without this, why would the length be going up and down
"arbitrarily", even thou
Quoting Richard Sandiford :
[responding because you kept me on cc:]
Joern Rennecke writes:
This uses the same interface as my previous patch:
http://gcc.gnu.org/ml/gcc-patches/2012-11/msg00473.html ,
but I refined the algorithm for the get_insn_variants
mechanism to work properly with the rew
[responding because you kept me on cc:]
Joern Rennecke writes:
> This uses the same interface as my previous patch:
> http://gcc.gnu.org/ml/gcc-patches/2012-11/msg00473.html ,
> but I refined the algorithm for the get_insn_variants
> mechanism to work properly with the reworked ARC port -
> http:
This uses the same interface as my previous patch:
http://gcc.gnu.org/ml/gcc-patches/2012-11/msg00473.html ,
but I refined the algorithm for the get_insn_variants
mechanism to work properly with the reworked ARC port -
http://gcc.gnu.org/ml/gcc-patches/2012-11/msg01891.html -
the only user so far,
Quoting Joern Rennecke :
+ varying_length[uid] = (varying_length[inner_uid] & 1);
Typo; I meant:
+ varying_length[uid] |= (varying_length[inner_uid] & 1);
Quoting Joern Rennecke :
The return value is a pointer to a function, to be used for adjusting
instruction length in the place of ADJUST_INSN_LENGTHS - and where it
is missing, or zero if no such functionis required.
(I first thought of having separate target hooks for everything, but then
I re
Quoting Richard Sandiford :
Joern Rennecke writes:
So, if we want to go with a declarative approach, each instruction can have
a plurality of variants, each described by a size, an instruction fetch
alignment base, an instruction fetch alignment offset set, a cache
line base,
a cache lin
Quoting Richard Sandiford :
> But I think extending the FSM from just being an on-the-fly
transformatino during final (as for ARM) to something that happens
earlier is a pretty major step in terms of the interface between the
backend and the rest of GCC. There are then undocumented restriction
Joern Rennecke writes:
> Quoting Richard Sandiford :
>
>> But what I'm trying to get at is: why can't the backend tell
>> shorten_branches about the amount of alignment/misalignment
>> that the target wants, and where? Via an attribute or a hook,
>> I don't mind which. But it should be declarati
Quoting Richard Sandiford :
But what I'm trying to get at is: why can't the backend tell
shorten_branches about the amount of alignment/misalignment
that the target wants, and where? Via an attribute or a hook,
I don't mind which. But it should be declarative, rather than
a part of the shorten
Joern Rennecke writes:
> Quoting Richard Sandiford :
>>> The length variation for the ARC are not alike: there are branches that
>>> are subject to branch shortening in the usual way, but they might
>>> shrink when other things shrink. When we are iterating starting at
>>> minimal length and incr
Quoting Richard Biener :
On Wed, Oct 31, 2012 at 2:06 PM, Joern Rennecke
wrote:
Quoting Richard Biener :
Maybe we should split the thing then into a adjust_insn_length attribute
without the iteration parameter
Attributes don't get any parameter but the instruction, and don't apply
to dela
On Wed, Oct 31, 2012 at 2:06 PM, Joern Rennecke
wrote:
> Quoting Richard Biener :
>
>> Maybe we should split the thing then into a adjust_insn_length attribute
>> without the iteration parameter
>
>
> Attributes don't get any parameter but the instruction, and don't apply
> to delay slot SEQUENCEs
Quoting Richard Biener :
Maybe we should split the thing then into a adjust_insn_length attribute
without the iteration parameter
Attributes don't get any parameter but the instruction, and don't apply
to delay slot SEQUENCEs.
On Wed, Oct 31, 2012 at 1:33 PM, Joern Rennecke
wrote:
> Quoting Richard Sandiford :
>
>>> It's about describing complex interactions of length adjustments that
>>> derive from branch shortening and length added for (un)alignment for
>>> scheduling purposes. Expressed naively, these can lead to c
Quoting Richard Sandiford :
It's about describing complex interactions of length adjustments that
derive from branch shortening and length added for (un)alignment for
scheduling purposes. Expressed naively, these can lead to cycles.
But shorten_branches should be written to avoid cycles, and
Joern Rennecke writes:
> Quoting Richard Sandiford :
>> I should probably have piped up earlier, but I'm really not sure about it.
>> ADJUST_INSN_LENGTH as defined now is telling us a property of the target.
>> The patch instead ties the new hook directly into the shorten_branches
>> algorithm, wh
Quoting Richard Sandiford :
I should probably have piped up earlier, but I'm really not sure about it.
ADJUST_INSN_LENGTH as defined now is telling us a property of the target.
The patch instead ties the new hook directly into the shorten_branches
algorithm, which I think is a bad idea.
IMO, th
On Wed, Oct 31, 2012 at 11:22 AM, Richard Sandiford
wrote:
> Richard Biener writes:
>> On Tue, Oct 30, 2012 at 5:14 PM, Joern Rennecke
>> wrote:
>>> Quoting Richard Biener :
>>>
Apart from the iteration_threshold the hookization would be
straight-forward.
Now I cannot decipher fro
Richard Biener writes:
> On Tue, Oct 30, 2012 at 5:14 PM, Joern Rennecke
> wrote:
>> Quoting Richard Biener :
>>
>>> Apart from the iteration_threshold the hookization would be
>>> straight-forward.
>>> Now I cannot decipher from the patch what functional change it introduces
>>> ;)
>>
>>
>> The
On Tue, Oct 30, 2012 at 5:14 PM, Joern Rennecke
wrote:
> Quoting Richard Biener :
>
>> Apart from the iteration_threshold the hookization would be
>> straight-forward.
>> Now I cannot decipher from the patch what functional change it introduces
>> ;)
>
>
> The only change occurs if we reach an ite
Quoting Richard Biener :
Apart from the iteration_threshold the hookization would be straight-forward.
Now I cannot decipher from the patch what functional change it introduces ;)
The only change occurs if we reach an iteration count of MAX_INT iterations -
which should already be indicative o
On Sun, Oct 28, 2012 at 6:13 PM, Joern Rennecke
wrote:
> Quoting Richard Biener :
>
>>> Thus, you can allow the length to vary downwards as well as upwards
>>> across iterations with suitable definitions of the @code{length}
>>> attribute
>>> and/or @code{ADJUST_INSN_LENGTH}. Care has to be taken
Quoting Richard Biener :
Thus, you can allow the length to vary downwards as well as upwards
across iterations with suitable definitions of the @code{length} attribute
and/or @code{ADJUST_INSN_LENGTH}. Care has to be taken that this does not
lead to infinite loops.
I don't see that you can sh
24 matches
Mail list logo
