[RFC] More jump threading restrictions in the presence of loops.
Aldy Hernandez
aldyh@redhat.com
Thu Oct 7 08:15:05 GMT 2021
[Andrew, ranger comment embedded below.]
On 10/7/21 1:06 AM, Jeff Law wrote:
>
>
> On 10/6/2021 7:47 AM, Aldy Hernandez via Gcc-patches wrote:
>> The pending patch I have from Richi fixes this. Even so, it's the
>> uninit code that's confused.
>>
>> Sigh...every single change to the threading code shines the light on
>> some warning bug.
>>
>> If you take the calls.ii file from the aarch64 bootstrap and break on
>> the warning, you can see that the uninitalized use is for
>> const_upper_3934 here:
>>
>> <bb 102> [local count: 315357954]:
>> # const_upper_3934 = PHI <const_upper_3937(D)(101), _6707(293)>
>> if (_881 != 0)
>> goto <bb 103>; [50.00%]
>> else
>> goto <bb 106>; [50.00%]
>>
>> <bb 103> [local count: 157678977]:
>> if (const_upper_3934 > _6699)
>> goto <bb 105>; [89.00%]
>> else
>> goto <bb 294>; [11.00%]
>>
>> <bb 294> [local count: 17344687]:
>>
>> <bb 104> [local count: 157678977]:
>> goto <bb 107>; [100.00%]
>>
>> <bb 105> [local count: 140334290]:
>> stack_usage_map.481_3930 = stack_usage_map;
>> _6441 = const_upper_3934 - _6699;
>> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>> PROBLEMATIC READ HERE
>> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>> _4819 = stack_usage_map.481_3930 + _6699;
>> __builtin_memset (_4819, 1, _6441);
>> goto <bb 104>; [11.00%]
>>
>> const_upper_3934 could be undefined if it comes from BB101
>> (const_upper_3937(D)), but it only gets read for _881 != 0, so it
>> shouldn't warn.
>>
>> I suggest -Wmaybe-uninitialized be turned off for that file until the
>> warning is fixed.
>>
>> And yes, the proposed patch will also cure this, but the underlying
>> problem in the warning is still there.
> You haven't shown enough context for me to agree or disagree. Are there
> other preds to bb105? I hate these slimmed down dumps. I work better
> with the full pred/succ lists. -fdump-tree-<whatever>-blocks-details :-)
>
> It appears to me that for _881 != 0 we certainly flow into the read of
> _const_upper_3934 in bb103 and bb105. Why do you think that's safe?
My bad, there's some missing context.
The only way to get to BB101->BB102 is through:
<bb 100>
if (_6711 != 0)
goto <bb 101>; [5.50%]
else
goto <bb 293>; [94.50%]
And there's an implicit relation between _6711 and _811:
<bb 86>
...
if (_6711 != 0)
goto <bb 287>; [5.50%]
else
goto <bb 87>; [94.50%]
<bb 287> [local count: 17344687]:
goto <bb 88>; [100.00%]
<bb 87> [local count: 298013267]:
<bb 88> [local count: 315357954]:
# _881 = PHI <1(87), 0(287)>
That is, _6711 == !_881.
[Andrew, it'd be neat if we could teach ranger the relationship between
_6711 and _811 above. And also, that _881 is [0,1]. Perhaps with the
relation oracle, the uninit code could notice that a _6711 guard is also
a !_811 guard.]
Presumably the threader shuffled things sufficiently so that the above
relationship was difficult to devise.
Your preferred full context follows ;-).
Aldy
----
Here we see that _881 == 0 when _6711 != 0:
;; basic block 286, loop depth 1, count 640272214 (estimated locally),
maybe hot
;; prev block 85, next block 86, flags: (NEW)
;; pred: 85 [67.0% (guessed)] count:640272214 (estimated
locally) (FALSE_VALUE,EXECUTABLE)
goto <bb 112>; [100.00%]
;; succ: 112 [always] count:640272214 (estimated locally)
(FALLTHRU)
;; basic block 86, loop depth 1, count 315357954 (estimated locally),
maybe hot
;; prev block 286, next block 287, flags: (NEW, REACHABLE, VISITED)
;; pred: 85 [33.0% (guessed)] count:315357953 (estimated
locally) (TRUE_VALUE,EXECUTABLE)
_4293 = MEM[(long int *)_5338 + 80B];
_6727 = MEM[(long int *)_5338 + 88B];
_6715 = MEM[(long int *)_5338 + 32B];
_4305 = _4293 + _6715;
_4299 = MEM[(long int *)_5338 + 40B];
_4300 = _4299 + _6727;
_6707 = (long unsigned int) _4305;
_6711 = (long unsigned int) _4300;
_6699 = (long unsigned int) _4293;
if (_6711 != 0)
goto <bb 287>; [5.50%]
else
goto <bb 87>; [94.50%]
;; succ: 287 [5.5% (guessed)] count:17344687 (estimated
locally) (TRUE_VALUE,EXECUTABLE)
;; 87 [94.5% (guessed)] count:298013267 (estimated
locally) (FALSE_VALUE,EXECUTABLE)
;; basic block 287, loop depth 1, count 17344687 (estimated locally),
maybe hot
;; prev block 86, next block 87, flags: (NEW)
;; pred: 86 [5.5% (guessed)] count:17344687 (estimated
locally) (TRUE_VALUE,EXECUTABLE)
goto <bb 88>; [100.00%]
;; succ: 88 [always] count:17344687 (estimated locally) (FALLTHRU)
;; basic block 87, loop depth 1, count 298013267 (estimated locally),
maybe hot
;; prev block 287, next block 88, flags: (NEW, VISITED)
;; pred: 86 [94.5% (guessed)] count:298013267 (estimated
locally) (FALSE_VALUE,EXECUTABLE)
;; succ: 88 [always] count:298013267 (estimated locally)
(FALLTHRU,EXECUTABLE)
;; basic block 88, loop depth 1, count 315357954 (estimated locally),
maybe hot
;; prev block 87, next block 89, flags: (NEW, REACHABLE, VISITED)
;; pred: 87 [always] count:298013267 (estimated locally)
(FALLTHRU,EXECUTABLE)
;; 287 [always] count:17344687 (estimated locally)
(FALLTHRU)
# const_upper_3863 = PHI <_6707(87), 18446744073709551615(287)>
# _881 = PHI <1(87), 0(287)>
[snip]
[snip]
[snip]
And here we see that the "undefined" read is guarded by _6711:
;; basic block 100, loop depth 1, count 315357955 (estimated locally),
maybe hot
;; prev block 99, next block 293, flags: (NEW, REACHABLE, VISITED)
;; pred: 98 [always] count:94607385 (estimated locally)
(FALLTHRU,EXECUTABLE)
;; 99 [always] count:220750569 (estimated locally)
(FALLTHRU,EXECUTABLE)
# iftmp.581_254 = PHI <_669(98), _3922(99)>
MEM[(struct poly_int *)&D.150826] ={v} {CLOBBER};
_4417 = MEM[(long int *)_5338 + 56B];
MEM[(struct poly_int *)&D.150826].D.21135.coeffs[0] = _4417;
_4410 = MEM[(long int *)_5338 + 64B];
MEM[(struct poly_int *)&D.150826].D.21135.coeffs[1] = _4410;
_673 = gen_int_mode (D.150826, 16);
MEM[(struct poly_int *)&D.150832] ={v} {CLOBBER};
MEM[(struct poly_int *)&D.150832].D.21135.coeffs[0] = 0;
MEM[(struct poly_int *)&D.150832].D.21135.coeffs[1] = 0;
emit_push_insn (val_598, mode_597, 0B, 0B, parm_align_601,
partial_600, reg_599, D.150832, argblock_88, _673, 0, iftmp.581_254, 0);
D.150832 ={v} {CLOBBER};
D.150826 ={v} {CLOBBER};
D.150830 ={v} {CLOBBER};
D.150828 ={v} {CLOBBER};
if (_6711 != 0)
goto <bb 101>; [5.50%]
else
goto <bb 293>; [94.50%]
;; succ: 101 [5.5% (guessed)] count:17344687 (estimated
locally) (TRUE_VALUE,EXECUTABLE)
;; 293 [94.5% (guessed)] count:298013268 (estimated
locally) (FALSE_VALUE,EXECUTABLE)
;; basic block 293, loop depth 1, count 298013268 (estimated locally),
maybe hot
;; prev block 100, next block 101, flags: (NEW)
;; pred: 100 [94.5% (guessed)] count:298013268 (estimated
locally) (FALSE_VALUE,EXECUTABLE)
goto <bb 102>; [100.00%]
;; succ: 102 [always] count:298013268 (estimated locally)
(FALLTHRU)
;; basic block 101, loop depth 1, count 17344687 (estimated locally),
maybe hot
;; prev block 293, next block 102, flags: (NEW, VISITED)
;; pred: 100 [5.5% (guessed)] count:17344687 (estimated
locally) (TRUE_VALUE,EXECUTABLE)
;; succ: 102 [always] count:17344687 (estimated locally)
(FALLTHRU,EXECUTABLE)
;; basic block 102, loop depth 1, count 315357954 (estimated locally),
maybe hot
;; prev block 101, next block 103, flags: (NEW, REACHABLE, VISITED)
;; pred: 101 [always] count:17344687 (estimated locally)
(FALLTHRU,EXECUTABLE)
;; 293 [always] count:298013268 (estimated locally)
(FALLTHRU)
# const_upper_3934 = PHI <const_upper_3937(D)(101), _6707(293)>
if (_881 != 0)
goto <bb 103>; [50.00%]
else
goto <bb 106>; [50.00%]
;; succ: 103 [50.0% (guessed)] count:157678977 (estimated
locally) (TRUE_VALUE,EXECUTABLE)
;; 106 [50.0% (guessed)] count:157678977 (estimated
locally) (FALSE_VALUE,EXECUTABLE)
;; basic block 103, loop depth 1, count 157678977 (estimated locally),
maybe hot
;; prev block 102, next block 294, flags: (NEW, REACHABLE, VISITED)
;; pred: 102 [50.0% (guessed)] count:157678977 (estimated
locally) (TRUE_VALUE,EXECUTABLE)
if (const_upper_3934 > _6699)
goto <bb 105>; [89.00%]
else
goto <bb 294>; [11.00%]
;; succ: 105 [89.0% (guessed)] count:140334290 (estimated
locally) (TRUE_VALUE,EXECUTABLE)
;; 294 [11.0% (guessed)] count:17344687 (estimated
locally) (FALSE_VALUE,EXECUTABLE)
;; basic block 294, loop depth 1, count 17344687 (estimated locally),
maybe hot
;; prev block 103, next block 104, flags: (NEW)
;; pred: 103 [11.0% (guessed)] count:17344687 (estimated
locally) (FALSE_VALUE,EXECUTABLE)
;; succ: 104 [always] count:17344687 (estimated locally)
(FALLTHRU)
;; basic block 104, loop depth 1, count 157678977 (estimated locally),
maybe hot
;; Invalid sum of incoming counts 32781459 (estimated locally), should
be 157678977 (estimated locally)
;; prev block 294, next block 105, flags: (NEW, VISITED)
;; pred: 294 [always] count:17344687 (estimated locally)
(FALLTHRU)
;; 105 [11.0% (guessed)] count:15436772 (estimated
locally) (FALLTHRU,EXECUTABLE)
goto <bb 107>; [100.00%]
;; succ: 107 [always] count:157678977 (estimated locally)
(FALLTHRU,EXECUTABLE)
;; basic block 105, loop depth 1, count 140334290 (estimated locally),
maybe hot
;; prev block 104, next block 106, flags: (NEW, REACHABLE, VISITED)
;; pred: 103 [89.0% (guessed)] count:140334290 (estimated
locally) (TRUE_VALUE,EXECUTABLE)
stack_usage_map.481_3930 = stack_usage_map;
_6441 = const_upper_3934 - _6699;
_4819 = stack_usage_map.481_3930 + _6699;
__builtin_memset (_4819, 1, _6441);
goto <bb 104>; [11.00%]
More information about the Gcc-patches
mailing list