Hi,
On Fri, Sep 17 2021, Kewen.Lin wrote:
>
[...]
>
> Against v2 [2], this v3 addressed Martin's review comments:
> - Replace HWI auto_vec with unsigned int for target_info
> to avoid overkill (also Segher's comments), adjust some
> places need to be updated for this change.
> - Annotate target_info won't be streamed for offloading
> target compilers.
> - Scan for all gimple statements instead of those with
> non-zero size/time weights.
>
> Against v1 [1], the v2 addressed Richi's and Segher's review
> comments, mainly consists of:
> - Extend it to cover non always_inline.
> - Exclude the case for offload streaming.
> - Some function naming and formatting issues.
> - Adjust rs6000_can_inline_p.
> - Add new cases.
>
> Bootstrapped and regress-tested on powerpc64le-linux-gnu Power9.
>
> Any comments are highly appreciated!
>
> [1] https://gcc.gnu.org/pipermail/gcc-patches/2021-September/578555.html
> [2] https://gcc.gnu.org/pipermail/gcc-patches/2021-September/579045.html
> ------
> gcc/ChangeLog:
>
> PR ipa/102059
> * config/rs6000/rs6000-call.c (rs6000_fn_has_any_of_these_mask_bits):
> New function.
> * config/rs6000/rs6000-internal.h
> (rs6000_fn_has_any_of_these_mask_bits): New declare.
> * config/rs6000/rs6000.c (TARGET_NEED_IPA_FN_TARGET_INFO): New macro.
> (TARGET_UPDATE_IPA_FN_TARGET_INFO): Likewise.
> (rs6000_need_ipa_fn_target_info): New function.
> (rs6000_update_ipa_fn_target_info): Likewise.
> (rs6000_can_inline_p): Adjust for ipa function summary target info.
> * config/rs6000/rs6000.h (RS6000_FN_TARGET_INFO_HTM): New macro.
> * ipa-fnsummary.c (ipa_dump_fn_summary): Adjust for ipa function
> summary target info.
> (analyze_function_body): Adjust for ipa function summary target
> info and call hook rs6000_need_ipa_fn_target_info and
> rs6000_update_ipa_fn_target_info.
> (ipa_merge_fn_summary_after_inlining): Adjust for ipa function
> summary target info.
> (inline_read_section): Likewise.
> (ipa_fn_summary_write): Likewise.
> * ipa-fnsummary.h (ipa_fn_summary::target_info): New member.
> * doc/tm.texi: Regenerate.
> * doc/tm.texi.in (TARGET_UPDATE_IPA_FN_TARGET_INFO): Document new
> hook.
> (TARGET_NEED_IPA_FN_TARGET_INFO): Likewise.
> * target.def (update_ipa_fn_target_info): New hook.
> (need_ipa_fn_target_info): Likewise.
> * targhooks.c (default_need_ipa_fn_target_info): New function.
> (default_update_ipa_fn_target_info): Likewise.
> * targhooks.h (default_update_ipa_fn_target_info): New declare.
> (default_need_ipa_fn_target_info): Likewise.
>
[...]
> diff --git a/gcc/target.def b/gcc/target.def
> index 28a34f1d51b..28ff639684b 100644
> --- a/gcc/target.def
> +++ b/gcc/target.def
> @@ -6600,6 +6600,41 @@ specific target options and the caller does not use
> the same options.",
> bool, (tree caller, tree callee),
> default_target_can_inline_p)
>
> +DEFHOOK
> +(update_ipa_fn_target_info,
> + "Allow target to analyze all gimple statements for the given function to\n\
> +record and update some target specific information for inlining. A
> typical\n\
> +example is that a caller with one isa feature disabled is normally not\n\
> +allowed to inline a callee with that same isa feature enabled even which
> is\n\
> +attributed by always_inline, but with the conservative analysis on all\n\
> +statements of the callee if we are able to guarantee the callee does not\n\
> +exploit any instructions from the mismatch isa feature, it would be safe
> to\n\
> +allow the caller to inline the callee.\n\
> +@var{info} is one @code{unsigned int} value to record information in which\n\
> +one set bit indicates one corresponding feature is detected in the
> analysis,\n\
> +@var{stmt} is the statement being analyzed. Return true if target still\n\
> +need to analyze the subsequent statements, otherwise return false to stop\n\
> +subsequent analysis.\n\
> +The default version of this hook returns false.",
> + bool, (unsigned int& info, const gimple* stmt),
> + default_update_ipa_fn_target_info)
> +
> +DEFHOOK
> +(need_ipa_fn_target_info,
> + "Allow target to check early whether it is necessary to analyze all
> gimple\n\
> +statements in the given function to update target specific information for\n\
> +inlining. See hook @code{update_ipa_fn_target_info} for usage example of\n\
> +target specific information. This hook is expected to be invoked ahead of\n\
> +the iterating with hook @code{update_ipa_fn_target_info}.\n\
> +@var{decl} is the function being analyzed, @var{info} is the same as what\n\
> +in hook @code{update_ipa_fn_target_info}, target can do one time update\n\
> +into @var{info} without iterating for some case. Return true if target\n\
> +decides to analyze all gimple statements to collect information, otherwise\n\
> +return false.\n\
> +The default version of this hook returns false.",
> + bool, (const_tree decl, unsigned int& info),
> + default_need_ipa_fn_target_info)
> +
> DEFHOOK
> (relayout_function,
> "This target hook fixes function @var{fndecl} after attributes are
> processed.\n\
> diff --git a/gcc/targhooks.c b/gcc/targhooks.c
> index c9b5208853d..5ba539cef1a 100644
> --- a/gcc/targhooks.c
> +++ b/gcc/targhooks.c
> @@ -1765,6 +1765,22 @@ default_target_can_inline_p (tree caller, tree callee)
> return callee_opts == caller_opts;
> }
>
> +/* By default, return false to not need to collect any target information
> + for inlining. Target maintainer should re-define the hook if the
> + target want to take advantage of it. */
> +
> +bool
> +default_need_ipa_fn_target_info (const_tree, uint16_t &)
> +{
> + return false;
> +}
> +
> +bool
> +default_update_ipa_fn_target_info (uint16_t &, const gimple *)
> +{
> + return false;
> +}
The parameters have uint16_t type here but you apparently decided to use
unsigned int everywhere else, you probably forgot to change them here
too.
Martin
