According to Dobb’s, http://www.drdobbs.com/cpp/type-based-alias-analysis/184404273 <http://www.drdobbs.com/cpp/type-based-alias-analysis/184404273>, many compilers do this today to some degree.
> On Nov 30, 2018, at 2:51 PM, robert engels <reng...@ix.netcom.com> wrote: > > I’m not so sure about this case, as a smart compiler might decide the loop > warrants alias checking, and then might only perform a single load and store > and perform all of the computations in registers (or unroll them completely). > > I am not aware of compilers doing this today, but they certainly could. > >> On Nov 30, 2018, at 12:13 PM, Burak Serdar <bser...@ieee.org >> <mailto:bser...@ieee.org>> wrote: >> >> On Fri, Nov 30, 2018 at 11:09 AM Jakob Borg <ja...@kastelo.net >> <mailto:ja...@kastelo.net>> wrote: >>> >>> This should only be valid for a loop that doesn’t include any function >>> calls, or if the compiler can prove that said function calls don’t involve >>> any synchronization primitives anywhere in the possible call stack. >> >> ...or any potential pointer aliases: >> >> p:=&value >> q:=&value >> >> for ... { >> total+=*p >> (*q)++ >> } >> >>> >>> So maybe it’s optimized in >>> >>> for … { >>> total += *p >>> } >>> >>> but I wouldn’t bet on it in >>> >>> for … { >>> total += *p >>> foo() >>> } >>> >>> //jb >>> >>> On 30 Nov 2018, at 19:01, robert engels <reng...@ix.netcom.com >>> <mailto:reng...@ix.netcom.com>> wrote: >>> >>> I would think the compiler would be free to optimize this as a single load >>> outside the loop - as Go has no volatile declaration (to inform the >>> compiler that some other thread might of changed the value of *p) >>> >>> On Nov 30, 2018, at 11:58 AM, robert engels <reng...@ix.netcom.com >>> <mailto:reng...@ix.netcom.com>> wrote: >>> >>> It’s as simple as: >>> >>> var count = 1 >>> var p = &count >>> >>> func X() { >>> total :=0 >>> >>> for i:=0; i <1000000;i++ { >>> total = total + *p >>> } >>> } >>> >>> no ? >>> >>> On Nov 30, 2018, at 11:48 AM, Mark Volkmann <r.mark.volkm...@gmail.com >>> <mailto:r.mark.volkm...@gmail.com>> wrote: >>> >>> I think I failed to come up with a good code example. But suppose I need to >>> do something with a pointer inside a loop that really does require >>> dereferencing it and the pointer never changes inside the loop. In that >>> case will the Go compiler optimize that so the dereference doesn't happen >>> in each loop iteration? >>> >>> On Fri, Nov 30, 2018 at 11:36 AM Jan Mercl <0xj...@gmail.com >>> <mailto:0xj...@gmail.com>> wrote: >>>> >>>> >>>> On Fri, Nov 30, 2018 at 6:16 PM Mark Volkmann <r.mark.volkm...@gmail.com >>>> <mailto:r.mark.volkm...@gmail.com>> wrote: >>>> >>>>> Will the Go compiler optimize the pointer dereference so it doesn't >>>>> happen in every loop iteration? If not, is it common practice to do that >>>>> outside the loop like this? >>>>> >>>>> myValue := *myPtr >>>>> for _, v := range values { >>>>> fmt.Printf("%v %v\n", myValue, v) >>>>> } >>>> >>>> There's no pointer dereference inside the loop. But there's an allocation >>>> inside the loop, on every iteration, that puts a pointer to a copy of >>>> myValue into an interface{} that's passed to fmt.Printf. The allocation >>>> can be probably avoided: >>>> >>>> myValue := interface{}(*myPtr) >>>> for _, v := range values { >>>> fmt.Printf("%v %v\n", myValue, v) >>>> } >>>> >>>> But the %v verb handle pointers in many cases well. If that holds for the >>>> type of *myPtr then simply: >>>> >>>> for _, v := range values { >>>> fmt.Printf("%v %v\n", myPtr, v) >>>> } >>>> >>>> is what could be good enough. >>>> >>>> -- >>>> >>>> -j >>> >>> >>> >>> -- >>> R. Mark Volkmann >>> Object Computing, Inc. >>> >>> -- >>> You received this message because you are subscribed to the Google Groups >>> "golang-nuts" group. >>> To unsubscribe from this group and stop receiving emails from it, send an >>> email to golang-nuts+unsubscr...@googlegroups.com >>> <mailto:golang-nuts+unsubscr...@googlegroups.com>. >>> For more options, visit https://groups.google.com/d/optout >>> <https://groups.google.com/d/optout>. >>> >>> >>> >>> -- >>> You received this message because you are subscribed to the Google Groups >>> "golang-nuts" group. >>> To unsubscribe from this group and stop receiving emails from it, send an >>> email to golang-nuts+unsubscr...@googlegroups.com >>> <mailto:golang-nuts+unsubscr...@googlegroups.com>. >>> For more options, visit https://groups.google.com/d/optout >>> <https://groups.google.com/d/optout>. >>> >>> >>> >>> -- >>> You received this message because you are subscribed to the Google Groups >>> "golang-nuts" group. >>> To unsubscribe from this group and stop receiving emails from it, send an >>> email to golang-nuts+unsubscr...@googlegroups.com >>> <mailto:golang-nuts+unsubscr...@googlegroups.com>. >>> For more options, visit https://groups.google.com/d/optout >>> <https://groups.google.com/d/optout>. >>> >>> >>> -- >>> You received this message because you are subscribed to the Google Groups >>> "golang-nuts" group. >>> To unsubscribe from this group and stop receiving emails from it, send an >>> email to golang-nuts+unsubscr...@googlegroups.com >>> <mailto:golang-nuts+unsubscr...@googlegroups.com>. >>> For more options, visit https://groups.google.com/d/optout >>> <https://groups.google.com/d/optout>. >> >> -- >> You received this message because you are subscribed to the Google Groups >> "golang-nuts" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to golang-nuts+unsubscr...@googlegroups.com >> <mailto:golang-nuts+unsubscr...@googlegroups.com>. >> For more options, visit https://groups.google.com/d/optout >> <https://groups.google.com/d/optout>. > > > -- > You received this message because you are subscribed to the Google Groups > "golang-nuts" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to golang-nuts+unsubscr...@googlegroups.com > <mailto:golang-nuts+unsubscr...@googlegroups.com>. > For more options, visit https://groups.google.com/d/optout > <https://groups.google.com/d/optout>. -- You received this message because you are subscribed to the Google Groups "golang-nuts" group. To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
