With IBM's XL compiler family it is possible to supply two options (-qreport and -qlist) to generate reports for each source file that include information on which optimizations were applied, or which parts of the code could not be optimized (and why).
Is it possible to get a similar reporting for GNU's g++ - and if yes, how to do it?
Have a look at the -fdump-tree-[switch] flags. You can use -fdump-tree-all to get loads of information.
Also in trunk gcc -fopt-info-[options] will give you access higher level optimization information e.g. when particular optimizations were applied, missed etc e.g.
-fopt-info-inline-optimized-missed
Prints all successful and missed inlining optimizations (to stderr in this case). This is obviously pretty new functionality so I'm not sure how well supported it is yet.
In earlier releases they had -ftree-vectorizer-verbose=n which is now being deprecated in favor of opt-info.
All these options are listed here https://gcc.gnu.org/onlinedocs/gcc/Developer-Options.html though it can be a bit tricky to pick out the useful ones.
Use -S -fverbose-asm to list every silently applied option (including optimization ones) in assembler output header.
From https://gcc.gnu.org/onlinedocs/gcc-7.2.0/gcc/Optimize-Options.html#Optimize-Options:
You can invoke GCC with -Q --help=optimizers to find out the exact set of optimizations that are enabled at each level.
Example: (count number of optimization options enabled) A file is not necessary.
$ g++ -std=c++17 -O2 -Q --help=optimizers 2>&1 |grep enabled |wc -l
135
Note that many optimizations enabled by -O1/2/3 have no individual flags (see also: c++ - g++ O1 is not equal to O0 with all related optimization flags - Stack Overflow)
-Og adds to bisect to the ones that make inlining optiimization work (which is not -finline-functions btw, at least not alone), and then I use them explicitly, the optimization suddenly stops working 🤷♂️ So it clearly doesn't list all of the optimizations applied by -Og.-O0 or without any -Ox added. So no way to make a custom -Og build that is actually a -Og (the -Og option GCC provides is infamous for not doing what it's documented to).Took me way too long to find this but the key is the little-documented -fdump-passes option like so:
echo | gcc -O3 -fdump-passes -c -o /dev/null -x c -
The above gives a dump of the real effective optimizations and some warnings:
*warn_unused_result : ON
*diagnose_omp_blocks : OFF
*diagnose_tm_blocks : OFF
tree-omp_oacc_kernels_decompose : OFF
tree-omplower : ON
tree-lower : ON
tree-tmlower : OFF
tree-ehopt : OFF
tree-eh : ON
tree-coro-lower-builtins : OFF
tree-cfg : ON
*warn_function_return : ON
tree-coro-early-expand-ifns : OFF
tree-ompexp : ON
*build_cgraph_edges : ON
*free_lang_data : ON
ipa-visibility : ON
ipa-strubm : OFF
ipa-build_ssa_passes : ON
tree-fixup_cfg1 : ON
tree-ssa : ON
tree-walloca1 : ON
tree-warn-printf : OFF
*nonnullcmp : OFF
tree-early_uninit : OFF
tree-waccess1 : ON
tree-ubsan : OFF
tree-nothrow : ON
*rebuild_cgraph_edges : ON
ipa-opt_local_passes : ON
tree-fixup_cfg2 : ON
*rebuild_cgraph_edges : ON
tree-local-fnsummary1 : ON
tree-einline : ON
*infinite-recursion : OFF
tree-early_optimizations : ON
*remove_cgraph_callee_edges : ON
tree-early_objsz : ON
tree-ccp1 : ON
tree-forwprop1 : ON
tree-ethread : ON
tree-esra : ON
tree-ealias : ON
tree-phiprop1 : ON
tree-fre1 : ON
tree-evrp : ON
tree-mergephi1 : ON
tree-dse1 : ON
tree-cddce1 : ON
tree-phiopt1 : ON
tree-tailr1 : ON
tree-iftoswitch : ON
tree-switchconv : ON
tree-ehcleanup1 : OFF
tree-sccopy1 : ON
tree-profile_estimate : ON
tree-local-pure-const1 : ON
tree-modref1 : ON
tree-fnsplit : ON
*strip_predict_hints : ON
tree-release_ssa : ON
*rebuild_cgraph_edges : ON
tree-local-fnsummary2 : ON
ipa-remove_symbols : ON
ipa-strub : OFF
ipa-ipa_oacc : OFF
ipa-pta1 : OFF
ipa-ipa_oacc_kernels : ON
tree-oacc_kernels : OFF
tree-ch1 : ON
tree-fre2 : ON
tree-lim1 : ON
tree-dom1 : ON
tree-dce1 : ON
tree-parloops1 : OFF
tree-ompexpssa1 : ON
*rebuild_cgraph_edges : ON
ipa-targetclone : ON
ipa-afdo : OFF
ipa-profile : OFF
tree-feedback_fnsplit : OFF
ipa-free-fnsummary1 : ON
ipa-increase_alignment : OFF
ipa-tmipa : OFF
ipa-emutls : OFF
ipa-analyzer : OFF
ipa-odr : OFF
ipa-whole-program : ON
ipa-profile_estimate : ON
ipa-icf : ON
ipa-devirt : ON
ipa-cdtor : OFF
ipa-cp : ON
ipa-sra : ON
ipa-fnsummary : ON
ipa-inline : ON
ipa-locality-clone : OFF
ipa-pure-const : ON
ipa-modref : ON
ipa-free-fnsummary2 : ON
ipa-static-var : ON
ipa-single-use : ON
ipa-comdats : ON
ipa-pta2 : OFF
ipa-simdclone : ON
tree-fixup_cfg3 : ON
tree-ehdisp : OFF
tree-oaccloops : OFF
tree-omp_oacc_neuter_broadcast : OFF
tree-oaccdevlow : OFF
tree-ompdevlow : ON
tree-omptargetlink : OFF
tree-adjust_alignment : ON
tree-hardcfr : OFF
*all_optimizations : ON
*remove_cgraph_callee_edges : ON
*strip_predict_hints : ON
tree-ccp2 : ON
tree-objsz1 : ON
tree-post_ipa_warn1 : OFF
tree-waccess2 : ON
tree-rebuild_frequencies1 : ON
tree-cunrolli : ON
tree-backprop : ON
tree-phiprop2 : ON
tree-forwprop2 : ON
tree-alias : ON
tree-retslot : ON
tree-fre3 : ON
tree-mergephi2 : ON
tree-threadfull1 : ON
tree-vrp1 : ON
tree-bounds : OFF
tree-dse2 : ON
tree-dce2 : ON
tree-stdarg : ON
tree-cdce : ON
tree-cselim : ON
tree-copyprop1 : ON
tree-ifcombine : ON
tree-mergephi3 : ON
tree-phiopt2 : ON
tree-tailr2 : ON
tree-ch2 : ON
tree-cplxlower1 : ON
tree-bitintlower1 : ON
tree-sra : ON
tree-thread1 : ON
tree-dom2 : ON
tree-copyprop2 : ON
tree-isolate-paths : ON
tree-reassoc1 : ON
tree-dce3 : ON
tree-forwprop3 : ON
tree-phiopt3 : ON
tree-ccp3 : ON
tree-pow : ON
tree-bswap : ON
tree-laddress : ON
tree-lim2 : ON
tree-walloca2 : ON
tree-pre : ON
tree-sink1 : ON
tree-sancov1 : OFF
tree-asan1 : OFF
tree-tsan1 : OFF
tree-dse3 : ON
tree-dce4 : ON
tree-fix_loops : ON
tree-loop : ON
tree-loopinit : ON
tree-unswitch : ON
tree-lsplit : ON
tree-sccp : ON
tree-lversion : ON
tree-unrolljam : ON
tree-cddce2 : ON
tree-ivcanon : ON
tree-ldist : ON
tree-crc : ON
tree-linterchange : ON
tree-copyprop3 : ON
tree-graphite0 : OFF
tree-graphite : OFF
tree-lim3 : ON
tree-copyprop4 : ON
tree-dce5 : ON
tree-parloops2 : OFF
tree-ompexpssa2 : ON
tree-ch_vect : ON
tree-ifcvt : ON
tree-vect : ON
tree-dce6 : ON
tree-pcom : ON
tree-cunroll : ON
*pre_slp_scalar_cleanup : OFF
tree-fre4 : ON
tree-dse4 : ON
tree-slp1 : ON
tree-aprefetch : OFF
tree-ivopts : ON
tree-lim4 : ON
tree-loopdone : ON
tree-no_loop : OFF
tree-slp2 : ON
tree-simduid1 : OFF
tree-veclower21 : ON
tree-switchlower1 : ON
tree-sincos : ON
tree-recip : OFF
tree-reassoc2 : ON
tree-slsr : ON
tree-split-paths : ON
tree-tracer : OFF
tree-fre5 : ON
tree-thread2 : ON
tree-dom3 : ON
tree-strlen1 : ON
tree-threadfull2 : ON
tree-vrp2 : ON
tree-ccp4 : ON
tree-wrestrict : ON
tree-dse5 : ON
tree-dce7 : ON
tree-forwprop4 : ON
tree-sink2 : ON
tree-phiopt4 : ON
tree-fab1 : ON
tree-widening_mul : ON
tree-store-merging : ON
tree-cddce3 : ON
tree-sccopy2 : ON
tree-tailc : ON
tree-crited1 : ON
tree-uninit1 : OFF
tree-local-pure-const2 : ON
tree-modref2 : ON
tree-uncprop1 : ON
*all_optimizations_g : OFF
*remove_cgraph_callee_edges : ON
*strip_predict_hints : ON
tree-cplxlower2 : ON
tree-bitintlower2 : ON
tree-veclower22 : ON
tree-switchlower2 : ON
tree-ccp5 : ON
tree-post_ipa_warn2 : OFF
tree-objsz2 : ON
tree-fab2 : ON
tree-strlen2 : ON
tree-copyprop5 : ON
tree-dce8 : ON
tree-rebuild_frequencies2 : ON
tree-sancov2 : OFF
tree-asan2 : OFF
tree-tsan2 : OFF
tree-crited2 : ON
tree-uninit2 : OFF
tree-uncprop2 : ON
tree-assumptions : OFF
*tminit : OFF
tree-tmmark : ON
tree-tmmemopt : OFF
tree-tmedge : ON
tree-simduid2 : OFF
tree-vtable-verify : OFF
tree-lower_vaarg : ON
tree-veclower : ON
tree-cplxlower0 : ON
tree-bitintlower0 : ON
tree-sancov_O0 : OFF
tree-switchlower_O0 : OFF
tree-asan0 : OFF
tree-tsan0 : OFF
tree-sanopt : OFF
tree-ehcleanup2 : OFF
tree-musttail : OFF
tree-resx : OFF
tree-nrv : ON
tree-isel : ON
tree-hardcbr : OFF
tree-hardcmp : OFF
tree-waccess3 : ON
tree-optimized : ON
*warn_function_noreturn : OFF
rtl-expand : ON
*rest_of_compilation : ON
rtl-vregs : ON
rtl-into_cfglayout : ON
rtl-jump : ON
rtl-subreg1 : ON
rtl-dfinit : ON
rtl-cse1 : ON
rtl-fwprop1 : ON
rtl-cprop1 : ON
rtl-rtl pre : ON
rtl-hoist : OFF
rtl-hardreg_pre : OFF
rtl-cprop2 : ON
rtl-store_motion : OFF
rtl-cse_local : OFF
rtl-ce1 : ON
rtl-apx_nfcvt : OFF
rtl-reginfo : ON
rtl-loop2 : ON
rtl-loop2_init : ON
rtl-loop2_invariant : ON
rtl-loop2_unroll : ON
rtl-loop2_doloop : OFF
rtl-loop2_done : ON
rtl-subreg2 : OFF
rtl-web : OFF
rtl-cprop3 : ON
rtl-stv1 : ON
rtl-cse2 : ON
rtl-dse1 : ON
rtl-fwprop2 : ON
rtl-auto_inc_dec : OFF
rtl-init-regs : ON
rtl-ud_dce : ON
rtl-ext_dce : ON
rtl-combine : ON
rtl-late_combine1 : ON
rtl-rpad : OFF
rtl-stv2 : ON
rtl-ce2 : ON
rtl-jump_after_combine : ON
rtl-bbpart : ON
rtl-outof_cfglayout : ON
rtl-split1 : ON
rtl-subreg3 : ON
rtl-no-opt dfinit : OFF
*stack_ptr_mod : ON
rtl-mode_sw : ON
rtl-asmcons : ON
rtl-sms : OFF
rtl-lr_shrinkage : OFF
rtl-sched1 : OFF
rtl-avoid_store_forwarding : OFF
rtl-early_remat : OFF
rtl-ira : ON
rtl-reload : ON
*all-postreload : OFF
rtl-postreload : OFF
rtl-vzeroupper : OFF
rtl-late_combine2 : ON
rtl-gcse2 : ON
rtl-split2 : ON
rtl-ree : ON
rtl-cmpelim : ON
rtl-pro_and_epilogue : ON
rtl-dse2 : ON
rtl-csa : ON
rtl-jump2 : ON
rtl-compgotos : ON
rtl-sched_fusion : OFF
rtl-peephole2 : ON
rtl-ce3 : ON
rtl-rnreg : OFF
rtl-fold_mem_offsets : ON
rtl-cprop_hardreg : ON
rtl-rtl_dce : ON
rtl-bbro : ON
*leaf_regs : ON
rtl-split3 : ON
rtl-sched2 : ON
*stack_regs : ON
rtl-split4 : OFF
rtl-stack : ON
rtl-late_pro_and_epilogue : OFF
*all-late_compilation : OFF
rtl-zero_call_used_regs : ON
rtl-alignments : ON
rtl-vartrack : OFF
*free_cfg : ON
rtl-mach : ON
rtl-barriers : ON
rtl-dbr : OFF
rtl-split5 : OFF
rtl-eh_ranges : OFF
rtl-endbr_and_patchable_area : OFF
rtl-align_tight_loops : ON
rtl-shorten : ON
rtl-nothrow : ON
rtl-dwarf2 : ON
rtl-final : ON
rtl-dfinish : ON
*clean_state : ON
Compiler returned: 0
