Re: [TCWG CI] Regression caused by gcc: Add new constant data structure.

[Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>]

Hi Michael,

This was sent out due to a bug in our CI.  Sorry for the noise.

> On 20 Dec 2021, at 09:46, Michael Meissner via Gcc-regression <gcc-regression@gcc.gnu.org> wrote:
> 
> On Sat, Dec 18, 2021 at 09:12:06PM +0000, ci_notify@linaro.org wrote:
>> [TCWG CI] Regression caused by gcc: Add new constant data structure.:
>> commit c6756b3bc1d2af1c8e86f0ad1711e9b9134520ba
>> Author: Michael Meissner <meissner@linux.ibm.com>
>> 
>>    Add new constant data structure.
>> 
...
> 
> Given the files that I changed were in the config/rs6000 directory, and you are
> building an arm toolchain, I don't see how this particular change broke your
> build.  I believe that it must have been one of the changes after my particular
> commit that broke your build.  You should really list all of the commits that
> occurred between your last sucessful build and the failure.

This CI bisects performance problems down to a single commit.  Due to a bug in script that compares two sets of results it thought that your patch increased code-size of CoreMark.

> 
> Also, it was nearly impossible to trace through the build to find what really
> caused the error.

Because this report was due to a bug it didn’t have the actual details on code-size regression (no surprise, since there was no regression :-) ).  Here’s an example of a proper recent report: https://gcc.gnu.org/pipermail/gcc-regression/2021-December/076034.html .

Regards,

--
Maxim Kuvyrkov
https://www.linaro.org


> 
>> Full commit (up to 1000 lines):
>> <cut>
>> commit c6756b3bc1d2af1c8e86f0ad1711e9b9134520ba
>> Author: Michael Meissner <meissner@linux.ibm.com>
>> Date:   Wed Dec 15 00:56:25 2021 -0500
>> 
>>    Add new constant data structure.
>> 
>>    This patch provides the data structure and function to convert a
>>    CONST_INT, CONST_DOUBLE, CONST_VECTOR, or VEC_DUPLICATE of a constant) to
>>    an array of bytes, half-words, words, and  double words that can be loaded
>>    into a 128-bit vector register.
>> 
>>    The next patches will use this data structure to generate code that
>>    generates load of the vector/floating point registers using the XXSPLTIDP,
>>    XXSPLTIW, and LXVKQ instructions that were added in power10.
>> 
>>    2021-12-15  Michael Meissner  <meissner@the-meissners.org>
>> 
>>    gcc/
>> 
>>            * config/rs6000/rs6000-protos.h (VECTOR_128BIT_BITS): New macro.
>>            (VECTOR_128BIT_BYTES): Likewise.
>>            (VECTOR_128BIT_HALF_WORDS): Likewise.
>>            (VECTOR_128BIT_WORDS): Likewise.
>>            (VECTOR_128BIT_DOUBLE_WORDS): Likewise.
>>            (vec_const_128bit_type): New structure type.
>>            (vec_const_128bit_to_bytes): New declaration.
>>            * config/rs6000/rs6000.c (constant_int_to_128bit_vector): New
>>            helper function.
>>            (constant_fp_to_128bit_vector): New helper function.
>>            (vec_const_128bit_to_bytes): New function.
> 
> -- 
> Michael Meissner, IBM
> PO Box 98, Ayer, Massachusetts, USA, 01432
> email: meissner@linux.ibm.com