| Message ID | CAGkQGiJoEPMO=a4UcOitX4JTCCBQrytB326S2SuYWTT=afvAew@mail.gmail.com |
|---|---|
| State | New |
| Headers | show |
| Series | [fortran] PRs 89843 and 90022 - C Fortran Interop fixes. | expand |
On Tue, Apr 09, 2019 at 11:18:11AM +0100, Paul Richard Thomas wrote: > > This patch is safe for trunk, even at this late stage, because its > effects are barricaded behind the tests for CFI descriptors. I believe > that it appropriately rewards the bug reporters to have this feature > work as well as possible at release. > I cannot look at this until Saturday, if no one reviews by then I'll give it a quick glance.
Thanks Paul, I tried the patch and it does fix CFI_is_contigous() There is still an issue with INTENT(IN) For example, if we declare integer :: buf(0:1025,0:1025) and then call foo(buf(1:1024,1:1024) in which foo declares buf as INTENT(IN) You previously explained buf will be copied in order to enforce INTENT(IN), and that is why base_addr is not &buf(0,0). The dim[] of the CFI_cdesc_t structure describes the subarray, CFI_is_contiguous() returns false, but (it seems) the array was copied *and packed* (base_addr was malloc'ed with size=1024*1024*4). Bottom line, the CFI_cdesc_t description does not match the contiguous layout of the (sub)array copy, and this is an issue. I previously stated that in my opinion, copy-in was not desirable for performance reasons. I also found an other potential issue with copy-in. If in Fortran, we call foo(buf(0,0)) then the C subroutine can only access buf(0,0), and other elements such as buf(1025,1025) cannot be accessed. Such elements are valid in buf, but out of bounds in the copy (that contains a single element). Strictly speaking, I cannot say whether this is a violation of the standard or not, but I can see how this will break a lot of existing apps (once again, those apps might be incorrect in the first place, but most of us got used to them working). To me, this is a second reason why copy-in is not desirable (at least as a default option). Cheers, Gilles On 4/9/2019 7:18 PM, Paul Richard Thomas wrote: > The most part of this patch is concerned with implementing calls from > C of of fortran bind c procedures with assumed shape or assumed rank > dummies to completely fix PR89843. The conversion of the descriptors > from CFI to gfc occur on entry to and reversed on exit from the > procedure. > > This patch is safe for trunk, even at this late stage, because its > effects are barricaded behind the tests for CFI descriptors. I believe > that it appropriately rewards the bug reporters to have this feature > work as well as possible at release. > > Between comments and the ChangeLogs, this patch is self explanatory. > > Bootstrapped and regtested on FC29/x86_64 - OK for trunk? > > Paul > > 2019-04-09 Paul Thomas <pault@gcc.gnu.org> > > PR fortran/89843 > * trans-decl.c (gfc_get_symbol_decl): Assumed shape and assumed > rank dummies of bind C procs require deferred initialization. > (convert_CFI_desc): New procedure to convert incoming CFI > descriptors to gfc types and back again. > (gfc_trans_deferred_vars): Call it. > * trans-expr.c (gfc_conv_gfc_desc_to_cfi_desc): Null the CFI > descriptor pointer. Free the descriptor in all cases. > > PR fortran/90022 > * trans-decl.c (gfc_get_symbol_decl): Make sure that the se > expression is a pointer type before converting it to the symbol > backend_decl type. > > 2019-04-09 Paul Thomas <pault@gcc.gnu.org> > > PR fortran/89843 > * gfortran.dg/ISO_Fortran_binding_4.f90: Modify the value of x > in ctg. Test the conversion of the descriptor types in the main > program. > * gfortran.dg/ISO_Fortran_binding_10.f90: New test. > * gfortran.dg/ISO_Fortran_binding_10.c: Called by it. > > PR fortran/90022 > * gfortran.dg/ISO_Fortran_binding_1.c: Correct the indexing for > the computation of 'ans'. Also, change the expected results for > CFI_is_contiguous to comply with standard. > * gfortran.dg/ISO_Fortran_binding_1.f90: Correct the expected > results for CFI_is_contiguous to comply with standard. > * gfortran.dg/ISO_Fortran_binding_9.f90: New test. > * gfortran.dg/ISO_Fortran_binding_9.c: Called by it. > > 2019-04-09 Paul Thomas <pault@gcc.gnu.org> > > PR fortran/89843 > * runtime/ISO_Fortran_binding.c (cfi_desc_to_gfc_desc): Only > return immediately if the source pointer is null. Bring > forward the extraction of the gfc type. Extract the kind so > that the element size can be correctly computed for sections > and components of derived type arrays. Remove the free of the > CFI descriptor since this is now done in trans-expr.c. > (gfc_desc_to_cfi_desc): Only allocate the CFI descriptor if it > is not null. > (CFI_section): Normalise the difference between the upper and > lower bounds by the stride to correctly calculate the extents > of the section. > > PR fortran/90022 > * runtime/ISO_Fortran_binding.c (CFI_is_contiguous) : Return > 1 for true and 0 otherwise to comply with the standard. Correct > the contiguity check for rank 3 and greater by using the stride > measure of the lower dimension rather than the element length.
Hi Gilles, > > I also found an other potential issue with copy-in. > > If in Fortran, we > > call foo(buf(0,0)) > > then the C subroutine can only access buf(0,0), and other elements such > as buf(1025,1025) cannot be accessed. > > Such elements are valid in buf, but out of bounds in the copy (that > contains a single element). > > Strictly speaking, I cannot say whether this is a violation of the > standard or not, but I can see how this will > > break a lot of existing apps (once again, those apps might be incorrect > in the first place, but most of us got used to them working). The above call will only be conforming if the dummy argument is declared assumed or explicit size. Otherwise, the compiler should reject it due to rank mismatch. For assumed rank, the call would be legitimate, but the rank of the dummy argument is then zero. Even if no copy-in is performed, accessing data beyond the address range of that scalar is not strictly allowed. Of more interest is the situation where the dummy argument in Fortran is declared, e.g., TYPE(*), ASYNCHRONOUS, INTENT(IN) :: BUF(..) The standard's semantics *forbids* performing copy-in/out in this case, IIRC. Otherwise ASYNCHRONOUS semantics would not work, and non-blocking MPI calls would fail due to buffers vanishing into thin air. Regards Reinhold > > To me, this is a second reason why copy-in is not desirable (at least as > a default option). > > > > Cheers, > > > Gilles > > On 4/9/2019 7:18 PM, Paul Richard Thomas wrote: > > The most part of this patch is concerned with implementing calls from > > C of of fortran bind c procedures with assumed shape or assumed rank > > dummies to completely fix PR89843. The conversion of the descriptors > > from CFI to gfc occur on entry to and reversed on exit from the > > procedure. > > > > This patch is safe for trunk, even at this late stage, because its > > effects are barricaded behind the tests for CFI descriptors. I believe > > that it appropriately rewards the bug reporters to have this feature > > work as well as possible at release. > > > > Between comments and the ChangeLogs, this patch is self explanatory. > > > > Bootstrapped and regtested on FC29/x86_64 - OK for trunk? > > > > Paul > > > > 2019-04-09 Paul Thomas <pault@gcc.gnu.org> > > > > PR fortran/89843 > > * trans-decl.c (gfc_get_symbol_decl): Assumed shape and assumed > > rank dummies of bind C procs require deferred initialization. > > (convert_CFI_desc): New procedure to convert incoming CFI > > descriptors to gfc types and back again. > > (gfc_trans_deferred_vars): Call it. > > * trans-expr.c (gfc_conv_gfc_desc_to_cfi_desc): Null the CFI > > descriptor pointer. Free the descriptor in all cases. > > > > PR fortran/90022 > > * trans-decl.c (gfc_get_symbol_decl): Make sure that the se > > expression is a pointer type before converting it to the symbol > > backend_decl type. > > > > 2019-04-09 Paul Thomas <pault@gcc.gnu.org> > > > > PR fortran/89843 > > * gfortran.dg/ISO_Fortran_binding_4.f90: Modify the value of x > > in ctg. Test the conversion of the descriptor types in the main > > program. > > * gfortran.dg/ISO_Fortran_binding_10.f90: New test. > > * gfortran.dg/ISO_Fortran_binding_10.c: Called by it. > > > > PR fortran/90022 > > * gfortran.dg/ISO_Fortran_binding_1.c: Correct the indexing for > > the computation of 'ans'. Also, change the expected results for > > CFI_is_contiguous to comply with standard. > > * gfortran.dg/ISO_Fortran_binding_1.f90: Correct the expected > > results for CFI_is_contiguous to comply with standard. > > * gfortran.dg/ISO_Fortran_binding_9.f90: New test. > > * gfortran.dg/ISO_Fortran_binding_9.c: Called by it. > > > > 2019-04-09 Paul Thomas <pault@gcc.gnu.org> > > > > PR fortran/89843 > > * runtime/ISO_Fortran_binding.c (cfi_desc_to_gfc_desc): Only > > return immediately if the source pointer is null. Bring > > forward the extraction of the gfc type. Extract the kind so > > that the element size can be correctly computed for sections > > and components of derived type arrays. Remove the free of the > > CFI descriptor since this is now done in trans-expr.c. > > (gfc_desc_to_cfi_desc): Only allocate the CFI descriptor if it > > is not null. > > (CFI_section): Normalise the difference between the upper and > > lower bounds by the stride to correctly calculate the extents > > of the section. > > > > PR fortran/90022 > > * runtime/ISO_Fortran_binding.c (CFI_is_contiguous) : Return > > 1 for true and 0 otherwise to comply with the standard. Correct > > the contiguity check for rank 3 and greater by using the stride > > measure of the lower dimension rather than the element length.
Reinhold, Thanks for the insights ! That means there is currently an other issue since copy-in is performed even if the argument is declared as ASYNCHRONOUS. I gave the copy-in mechanism some more thoughts, and as a library developers, I can clearly see a need *not* to do that on a case-by-case basis, mainly for performance reasons, but also to be friendly with legacy apps that are not strictly standard compliant. At this stage, I think the best way to move forward is to add an other directive in the interface definition. for example, we could declare module foo interface subroutine bar_f08(buf) BIND(C, name="bar_c") implicit none !GCC$ ATTRIBUTES NO_COPY_IN :: buf TYPE(*), DIMENSION(..), INTENT(IN) :: buf end subroutine end interface end module Does this make sense ? Gilles On 4/10/2019 4:22 PM, Bader, Reinhold wrote: > Hi Gilles, > >> I also found an other potential issue with copy-in. >> >> If in Fortran, we >> >> call foo(buf(0,0)) >> >> then the C subroutine can only access buf(0,0), and other elements such >> as buf(1025,1025) cannot be accessed. >> >> Such elements are valid in buf, but out of bounds in the copy (that >> contains a single element). >> >> Strictly speaking, I cannot say whether this is a violation of the >> standard or not, but I can see how this will >> >> break a lot of existing apps (once again, those apps might be incorrect >> in the first place, but most of us got used to them working). > > The above call will only be conforming if the dummy argument is declared > assumed or explicit size. > Otherwise, the compiler should reject it due to rank mismatch. For assumed > rank, the call would be > legitimate, but the rank of the dummy argument is then zero. Even if no > copy-in is performed, > accessing data beyond the address range of that scalar is not strictly > allowed. > > Of more interest is the situation where the dummy argument in Fortran is > declared, e.g., > > TYPE(*), ASYNCHRONOUS, INTENT(IN) :: BUF(..) > > The standard's semantics *forbids* performing copy-in/out in this case, IIRC. > Otherwise > ASYNCHRONOUS semantics would not work, and non-blocking MPI calls would fail > due > to buffers vanishing into thin air. > > Regards > Reinhold > >> To me, this is a second reason why copy-in is not desirable (at least as >> a default option). >> >> >> >> Cheers, >> >> >> Gilles >> >> On 4/9/2019 7:18 PM, Paul Richard Thomas wrote: >>> The most part of this patch is concerned with implementing calls from >>> C of of fortran bind c procedures with assumed shape or assumed rank >>> dummies to completely fix PR89843. The conversion of the descriptors >>> from CFI to gfc occur on entry to and reversed on exit from the >>> procedure. >>> >>> This patch is safe for trunk, even at this late stage, because its >>> effects are barricaded behind the tests for CFI descriptors. I believe >>> that it appropriately rewards the bug reporters to have this feature >>> work as well as possible at release. >>> >>> Between comments and the ChangeLogs, this patch is self explanatory. >>> >>> Bootstrapped and regtested on FC29/x86_64 - OK for trunk? >>> >>> Paul >>> >>> 2019-04-09 Paul Thomas <pault@gcc.gnu.org> >>> >>> PR fortran/89843 >>> * trans-decl.c (gfc_get_symbol_decl): Assumed shape and assumed >>> rank dummies of bind C procs require deferred initialization. >>> (convert_CFI_desc): New procedure to convert incoming CFI >>> descriptors to gfc types and back again. >>> (gfc_trans_deferred_vars): Call it. >>> * trans-expr.c (gfc_conv_gfc_desc_to_cfi_desc): Null the CFI >>> descriptor pointer. Free the descriptor in all cases. >>> >>> PR fortran/90022 >>> * trans-decl.c (gfc_get_symbol_decl): Make sure that the se >>> expression is a pointer type before converting it to the symbol >>> backend_decl type. >>> >>> 2019-04-09 Paul Thomas <pault@gcc.gnu.org> >>> >>> PR fortran/89843 >>> * gfortran.dg/ISO_Fortran_binding_4.f90: Modify the value of x >>> in ctg. Test the conversion of the descriptor types in the main >>> program. >>> * gfortran.dg/ISO_Fortran_binding_10.f90: New test. >>> * gfortran.dg/ISO_Fortran_binding_10.c: Called by it. >>> >>> PR fortran/90022 >>> * gfortran.dg/ISO_Fortran_binding_1.c: Correct the indexing for >>> the computation of 'ans'. Also, change the expected results for >>> CFI_is_contiguous to comply with standard. >>> * gfortran.dg/ISO_Fortran_binding_1.f90: Correct the expected >>> results for CFI_is_contiguous to comply with standard. >>> * gfortran.dg/ISO_Fortran_binding_9.f90: New test. >>> * gfortran.dg/ISO_Fortran_binding_9.c: Called by it. >>> >>> 2019-04-09 Paul Thomas <pault@gcc.gnu.org> >>> >>> PR fortran/89843 >>> * runtime/ISO_Fortran_binding.c (cfi_desc_to_gfc_desc): Only >>> return immediately if the source pointer is null. Bring >>> forward the extraction of the gfc type. Extract the kind so >>> that the element size can be correctly computed for sections >>> and components of derived type arrays. Remove the free of the >>> CFI descriptor since this is now done in trans-expr.c. >>> (gfc_desc_to_cfi_desc): Only allocate the CFI descriptor if it >>> is not null. >>> (CFI_section): Normalise the difference between the upper and >>> lower bounds by the stride to correctly calculate the extents >>> of the section. >>> >>> PR fortran/90022 >>> * runtime/ISO_Fortran_binding.c (CFI_is_contiguous) : Return >>> 1 for true and 0 otherwise to comply with the standard. Correct >>> the contiguity check for rank 3 and greater by using the stride >>> measure of the lower dimension rather than the element length.
Dear Gilles, > -----Ursprüngliche Nachricht----- > Von: Gilles Gouaillardet <gilles@rist.or.jp> > Gesendet: Freitag, 12. April 2019 02:25 > An: Bader, Reinhold <Reinhold.Bader@lrz.de>; Paul Richard Thomas > <paul.richard.thomas@gmail.com>; fortran@gcc.gnu.org; gcc-patches <gcc- > patches@gcc.gnu.org> > Betreff: Re: AW: [Patch, fortran] PRs 89843 and 90022 - C Fortran Interop > fixes. > > Reinhold, > > > Thanks for the insights ! > > > That means there is currently an other issue since copy-in is performed even > if the argument is declared as ASYNCHRONOUS. > > > I gave the copy-in mechanism some more thoughts, and as a library > developers, I can clearly see a need *not* to do that > > on a case-by-case basis, mainly for performance reasons, but also to be > friendly with legacy apps that are not strictly standard compliant. > > At this stage, I think the best way to move forward is to add an other > directive in the interface definition. > > > for example, we could declare > > > module foo > > interface > > subroutine bar_f08(buf) BIND(C, name="bar_c") > > implicit none > > !GCC$ ATTRIBUTES NO_COPY_IN :: buf maybe so, but some care must be taken to perform proper checking in case copy-in is necessary - these situations exist as well. For example, if the declaration for buf is TYPE(*), DIMENSION(..), INTENT(IN), CONTIGUOUS :: buf and the actual argument is a discontiguous array section. My preference is for the compiler to automatically avoid copy-in whenever possible. Cheers Reinhold > > TYPE(*), DIMENSION(..), INTENT(IN) :: buf > > end subroutine > > end interface > > end module > > > Does this make sense ? > > > Gilles > > On 4/10/2019 4:22 PM, Bader, Reinhold wrote: > > Hi Gilles, > > > >> I also found an other potential issue with copy-in. > >> > >> If in Fortran, we > >> > >> call foo(buf(0,0)) > >> > >> then the C subroutine can only access buf(0,0), and other elements such > >> as buf(1025,1025) cannot be accessed. > >> > >> Such elements are valid in buf, but out of bounds in the copy (that > >> contains a single element). > >> > >> Strictly speaking, I cannot say whether this is a violation of the > >> standard or not, but I can see how this will > >> > >> break a lot of existing apps (once again, those apps might be incorrect > >> in the first place, but most of us got used to them working). > > > > The above call will only be conforming if the dummy argument is declared > > assumed or explicit size. > > Otherwise, the compiler should reject it due to rank mismatch. For > assumed > > rank, the call would be > > legitimate, but the rank of the dummy argument is then zero. Even if no > > copy-in is performed, > > accessing data beyond the address range of that scalar is not strictly > > allowed. > > > > Of more interest is the situation where the dummy argument in Fortran is > > declared, e.g., > > > > TYPE(*), ASYNCHRONOUS, INTENT(IN) :: BUF(..) > > > > The standard's semantics *forbids* performing copy-in/out in this case, > IIRC. > > Otherwise > > ASYNCHRONOUS semantics would not work, and non-blocking MPI calls > would fail > > due > > to buffers vanishing into thin air. > > > > Regards > > Reinhold > > > >> To me, this is a second reason why copy-in is not desirable (at least as > >> a default option). > >> > >> > >> > >> Cheers, > >> > >> > >> Gilles > >> > >> On 4/9/2019 7:18 PM, Paul Richard Thomas wrote: > >>> The most part of this patch is concerned with implementing calls from > >>> C of of fortran bind c procedures with assumed shape or assumed rank > >>> dummies to completely fix PR89843. The conversion of the descriptors > >>> from CFI to gfc occur on entry to and reversed on exit from the > >>> procedure. > >>> > >>> This patch is safe for trunk, even at this late stage, because its > >>> effects are barricaded behind the tests for CFI descriptors. I believe > >>> that it appropriately rewards the bug reporters to have this feature > >>> work as well as possible at release. > >>> > >>> Between comments and the ChangeLogs, this patch is self explanatory. > >>> > >>> Bootstrapped and regtested on FC29/x86_64 - OK for trunk? > >>> > >>> Paul > >>> > >>> 2019-04-09 Paul Thomas <pault@gcc.gnu.org> > >>> > >>> PR fortran/89843 > >>> * trans-decl.c (gfc_get_symbol_decl): Assumed shape and assumed > >>> rank dummies of bind C procs require deferred initialization. > >>> (convert_CFI_desc): New procedure to convert incoming CFI > >>> descriptors to gfc types and back again. > >>> (gfc_trans_deferred_vars): Call it. > >>> * trans-expr.c (gfc_conv_gfc_desc_to_cfi_desc): Null the CFI > >>> descriptor pointer. Free the descriptor in all cases. > >>> > >>> PR fortran/90022 > >>> * trans-decl.c (gfc_get_symbol_decl): Make sure that the se > >>> expression is a pointer type before converting it to the symbol > >>> backend_decl type. > >>> > >>> 2019-04-09 Paul Thomas <pault@gcc.gnu.org> > >>> > >>> PR fortran/89843 > >>> * gfortran.dg/ISO_Fortran_binding_4.f90: Modify the value of x > >>> in ctg. Test the conversion of the descriptor types in the main > >>> program. > >>> * gfortran.dg/ISO_Fortran_binding_10.f90: New test. > >>> * gfortran.dg/ISO_Fortran_binding_10.c: Called by it. > >>> > >>> PR fortran/90022 > >>> * gfortran.dg/ISO_Fortran_binding_1.c: Correct the indexing for > >>> the computation of 'ans'. Also, change the expected results for > >>> CFI_is_contiguous to comply with standard. > >>> * gfortran.dg/ISO_Fortran_binding_1.f90: Correct the expected > >>> results for CFI_is_contiguous to comply with standard. > >>> * gfortran.dg/ISO_Fortran_binding_9.f90: New test. > >>> * gfortran.dg/ISO_Fortran_binding_9.c: Called by it. > >>> > >>> 2019-04-09 Paul Thomas <pault@gcc.gnu.org> > >>> > >>> PR fortran/89843 > >>> * runtime/ISO_Fortran_binding.c (cfi_desc_to_gfc_desc): Only > >>> return immediately if the source pointer is null. Bring > >>> forward the extraction of the gfc type. Extract the kind so > >>> that the element size can be correctly computed for sections > >>> and components of derived type arrays. Remove the free of the > >>> CFI descriptor since this is now done in trans-expr.c. > >>> (gfc_desc_to_cfi_desc): Only allocate the CFI descriptor if it > >>> is not null. > >>> (CFI_section): Normalise the difference between the upper and > >>> lower bounds by the stride to correctly calculate the extents > >>> of the section. > >>> > >>> PR fortran/90022 > >>> * runtime/ISO_Fortran_binding.c (CFI_is_contiguous) : Return > >>> 1 for true and 0 otherwise to comply with the standard. Correct > >>> the contiguity check for rank 3 and greater by using the stride > >>> measure of the lower dimension rather than the element length.
Dear Reinhold, Thanks again for the insight ! My preference is also to avoid copy-in whenever possible. But since my primary concern is performance (vs enforcing correctness of the user apps), I am not willing to use existing bugs to (fallaciously) conclude copy-in should be simply dropped (unless mandatory). Cheers, Gilles On 4/12/2019 4:53 PM, Bader, Reinhold wrote: > Dear Gilles, > >> -----Ursprüngliche Nachricht----- >> Von: Gilles Gouaillardet <gilles@rist.or.jp> >> Gesendet: Freitag, 12. April 2019 02:25 >> An: Bader, Reinhold <Reinhold.Bader@lrz.de>; Paul Richard Thomas >> <paul.richard.thomas@gmail.com>; fortran@gcc.gnu.org; gcc-patches <gcc- >> patches@gcc.gnu.org> >> Betreff: Re: AW: [Patch, fortran] PRs 89843 and 90022 - C Fortran Interop >> fixes. >> >> Reinhold, >> >> >> Thanks for the insights ! >> >> >> That means there is currently an other issue since copy-in is performed even >> if the argument is declared as ASYNCHRONOUS. >> >> >> I gave the copy-in mechanism some more thoughts, and as a library >> developers, I can clearly see a need *not* to do that >> >> on a case-by-case basis, mainly for performance reasons, but also to be >> friendly with legacy apps that are not strictly standard compliant. >> >> At this stage, I think the best way to move forward is to add an other >> directive in the interface definition. >> >> >> for example, we could declare >> >> >> module foo >> >> interface >> >> subroutine bar_f08(buf) BIND(C, name="bar_c") >> >> implicit none >> >> !GCC$ ATTRIBUTES NO_COPY_IN :: buf > maybe so, but some care must be taken to perform proper checking in case copy-in is necessary - these situations exist as well. > For example, if the declaration for buf is > > TYPE(*), DIMENSION(..), INTENT(IN), CONTIGUOUS :: buf > > and the actual argument is a discontiguous array section. > > My preference is for the compiler to automatically avoid copy-in whenever possible. > > Cheers > Reinhold > > >> TYPE(*), DIMENSION(..), INTENT(IN) :: buf >> >> end subroutine >> >> end interface >> >> end module >> >> >> Does this make sense ? >> >> >> Gilles >> >> On 4/10/2019 4:22 PM, Bader, Reinhold wrote: >>> Hi Gilles, >>> >>>> I also found an other potential issue with copy-in. >>>> >>>> If in Fortran, we >>>> >>>> call foo(buf(0,0)) >>>> >>>> then the C subroutine can only access buf(0,0), and other elements such >>>> as buf(1025,1025) cannot be accessed. >>>> >>>> Such elements are valid in buf, but out of bounds in the copy (that >>>> contains a single element). >>>> >>>> Strictly speaking, I cannot say whether this is a violation of the >>>> standard or not, but I can see how this will >>>> >>>> break a lot of existing apps (once again, those apps might be incorrect >>>> in the first place, but most of us got used to them working). >>> The above call will only be conforming if the dummy argument is declared >>> assumed or explicit size. >>> Otherwise, the compiler should reject it due to rank mismatch. For >> assumed >>> rank, the call would be >>> legitimate, but the rank of the dummy argument is then zero. Even if no >>> copy-in is performed, >>> accessing data beyond the address range of that scalar is not strictly >>> allowed. >>> >>> Of more interest is the situation where the dummy argument in Fortran is >>> declared, e.g., >>> >>> TYPE(*), ASYNCHRONOUS, INTENT(IN) :: BUF(..) >>> >>> The standard's semantics *forbids* performing copy-in/out in this case, >> IIRC. >>> Otherwise >>> ASYNCHRONOUS semantics would not work, and non-blocking MPI calls >> would fail >>> due >>> to buffers vanishing into thin air. >>> >>> Regards >>> Reinhold >>> >>>> To me, this is a second reason why copy-in is not desirable (at least as >>>> a default option). >>>> >>>> >>>> >>>> Cheers, >>>> >>>> >>>> Gilles >>>> >>>> On 4/9/2019 7:18 PM, Paul Richard Thomas wrote: >>>>> The most part of this patch is concerned with implementing calls from >>>>> C of of fortran bind c procedures with assumed shape or assumed rank >>>>> dummies to completely fix PR89843. The conversion of the descriptors >>>>> from CFI to gfc occur on entry to and reversed on exit from the >>>>> procedure. >>>>> >>>>> This patch is safe for trunk, even at this late stage, because its >>>>> effects are barricaded behind the tests for CFI descriptors. I believe >>>>> that it appropriately rewards the bug reporters to have this feature >>>>> work as well as possible at release. >>>>> >>>>> Between comments and the ChangeLogs, this patch is self explanatory. >>>>> >>>>> Bootstrapped and regtested on FC29/x86_64 - OK for trunk? >>>>> >>>>> Paul >>>>> >>>>> 2019-04-09 Paul Thomas <pault@gcc.gnu.org> >>>>> >>>>> PR fortran/89843 >>>>> * trans-decl.c (gfc_get_symbol_decl): Assumed shape and assumed >>>>> rank dummies of bind C procs require deferred initialization. >>>>> (convert_CFI_desc): New procedure to convert incoming CFI >>>>> descriptors to gfc types and back again. >>>>> (gfc_trans_deferred_vars): Call it. >>>>> * trans-expr.c (gfc_conv_gfc_desc_to_cfi_desc): Null the CFI >>>>> descriptor pointer. Free the descriptor in all cases. >>>>> >>>>> PR fortran/90022 >>>>> * trans-decl.c (gfc_get_symbol_decl): Make sure that the se >>>>> expression is a pointer type before converting it to the symbol >>>>> backend_decl type. >>>>> >>>>> 2019-04-09 Paul Thomas <pault@gcc.gnu.org> >>>>> >>>>> PR fortran/89843 >>>>> * gfortran.dg/ISO_Fortran_binding_4.f90: Modify the value of x >>>>> in ctg. Test the conversion of the descriptor types in the main >>>>> program. >>>>> * gfortran.dg/ISO_Fortran_binding_10.f90: New test. >>>>> * gfortran.dg/ISO_Fortran_binding_10.c: Called by it. >>>>> >>>>> PR fortran/90022 >>>>> * gfortran.dg/ISO_Fortran_binding_1.c: Correct the indexing for >>>>> the computation of 'ans'. Also, change the expected results for >>>>> CFI_is_contiguous to comply with standard. >>>>> * gfortran.dg/ISO_Fortran_binding_1.f90: Correct the expected >>>>> results for CFI_is_contiguous to comply with standard. >>>>> * gfortran.dg/ISO_Fortran_binding_9.f90: New test. >>>>> * gfortran.dg/ISO_Fortran_binding_9.c: Called by it. >>>>> >>>>> 2019-04-09 Paul Thomas <pault@gcc.gnu.org> >>>>> >>>>> PR fortran/89843 >>>>> * runtime/ISO_Fortran_binding.c (cfi_desc_to_gfc_desc): Only >>>>> return immediately if the source pointer is null. Bring >>>>> forward the extraction of the gfc type. Extract the kind so >>>>> that the element size can be correctly computed for sections >>>>> and components of derived type arrays. Remove the free of the >>>>> CFI descriptor since this is now done in trans-expr.c. >>>>> (gfc_desc_to_cfi_desc): Only allocate the CFI descriptor if it >>>>> is not null. >>>>> (CFI_section): Normalise the difference between the upper and >>>>> lower bounds by the stride to correctly calculate the extents >>>>> of the section. >>>>> >>>>> PR fortran/90022 >>>>> * runtime/ISO_Fortran_binding.c (CFI_is_contiguous) : Return >>>>> 1 for true and 0 otherwise to comply with the standard. Correct >>>>> the contiguity check for rank 3 and greater by using the stride >>>>> measure of the lower dimension rather than the element length.
Gilles & Reinhold, Following the discussion, I have decided to remove the copy-in for intent(in). Dominique and I have located the problem with -m32 for PR90022 and I am working on PR89846 right now. I will be resubmitting a bit later on. Cheers Paul On Fri, 12 Apr 2019 at 01:25, Gilles Gouaillardet <gilles@rist.or.jp> wrote: > > Reinhold, > > > Thanks for the insights ! > > > That means there is currently an other issue since copy-in is performed > even if the argument is declared as ASYNCHRONOUS. > > > I gave the copy-in mechanism some more thoughts, and as a library > developers, I can clearly see a need *not* to do that > > on a case-by-case basis, mainly for performance reasons, but also to be > friendly with legacy apps that are not strictly standard compliant. > > At this stage, I think the best way to move forward is to add an other > directive in the interface definition. > > > for example, we could declare > > > module foo > > interface > > subroutine bar_f08(buf) BIND(C, name="bar_c") > > implicit none > > !GCC$ ATTRIBUTES NO_COPY_IN :: buf > > TYPE(*), DIMENSION(..), INTENT(IN) :: buf > > end subroutine > > end interface > > end module > > > Does this make sense ? > > > Gilles > > On 4/10/2019 4:22 PM, Bader, Reinhold wrote: > > Hi Gilles, > > > >> I also found an other potential issue with copy-in. > >> > >> If in Fortran, we > >> > >> call foo(buf(0,0)) > >> > >> then the C subroutine can only access buf(0,0), and other elements such > >> as buf(1025,1025) cannot be accessed. > >> > >> Such elements are valid in buf, but out of bounds in the copy (that > >> contains a single element). > >> > >> Strictly speaking, I cannot say whether this is a violation of the > >> standard or not, but I can see how this will > >> > >> break a lot of existing apps (once again, those apps might be incorrect > >> in the first place, but most of us got used to them working). > > > > The above call will only be conforming if the dummy argument is declared > > assumed or explicit size. > > Otherwise, the compiler should reject it due to rank mismatch. For assumed > > rank, the call would be > > legitimate, but the rank of the dummy argument is then zero. Even if no > > copy-in is performed, > > accessing data beyond the address range of that scalar is not strictly > > allowed. > > > > Of more interest is the situation where the dummy argument in Fortran is > > declared, e.g., > > > > TYPE(*), ASYNCHRONOUS, INTENT(IN) :: BUF(..) > > > > The standard's semantics *forbids* performing copy-in/out in this case, IIRC. > > Otherwise > > ASYNCHRONOUS semantics would not work, and non-blocking MPI calls would fail > > due > > to buffers vanishing into thin air. > > > > Regards > > Reinhold > > > >> To me, this is a second reason why copy-in is not desirable (at least as > >> a default option). > >> > >> > >> > >> Cheers, > >> > >> > >> Gilles > >> > >> On 4/9/2019 7:18 PM, Paul Richard Thomas wrote: > >>> The most part of this patch is concerned with implementing calls from > >>> C of of fortran bind c procedures with assumed shape or assumed rank > >>> dummies to completely fix PR89843. The conversion of the descriptors > >>> from CFI to gfc occur on entry to and reversed on exit from the > >>> procedure. > >>> > >>> This patch is safe for trunk, even at this late stage, because its > >>> effects are barricaded behind the tests for CFI descriptors. I believe > >>> that it appropriately rewards the bug reporters to have this feature > >>> work as well as possible at release. > >>> > >>> Between comments and the ChangeLogs, this patch is self explanatory. > >>> > >>> Bootstrapped and regtested on FC29/x86_64 - OK for trunk? > >>> > >>> Paul > >>> > >>> 2019-04-09 Paul Thomas <pault@gcc.gnu.org> > >>> > >>> PR fortran/89843 > >>> * trans-decl.c (gfc_get_symbol_decl): Assumed shape and assumed > >>> rank dummies of bind C procs require deferred initialization. > >>> (convert_CFI_desc): New procedure to convert incoming CFI > >>> descriptors to gfc types and back again. > >>> (gfc_trans_deferred_vars): Call it. > >>> * trans-expr.c (gfc_conv_gfc_desc_to_cfi_desc): Null the CFI > >>> descriptor pointer. Free the descriptor in all cases. > >>> > >>> PR fortran/90022 > >>> * trans-decl.c (gfc_get_symbol_decl): Make sure that the se > >>> expression is a pointer type before converting it to the symbol > >>> backend_decl type. > >>> > >>> 2019-04-09 Paul Thomas <pault@gcc.gnu.org> > >>> > >>> PR fortran/89843 > >>> * gfortran.dg/ISO_Fortran_binding_4.f90: Modify the value of x > >>> in ctg. Test the conversion of the descriptor types in the main > >>> program. > >>> * gfortran.dg/ISO_Fortran_binding_10.f90: New test. > >>> * gfortran.dg/ISO_Fortran_binding_10.c: Called by it. > >>> > >>> PR fortran/90022 > >>> * gfortran.dg/ISO_Fortran_binding_1.c: Correct the indexing for > >>> the computation of 'ans'. Also, change the expected results for > >>> CFI_is_contiguous to comply with standard. > >>> * gfortran.dg/ISO_Fortran_binding_1.f90: Correct the expected > >>> results for CFI_is_contiguous to comply with standard. > >>> * gfortran.dg/ISO_Fortran_binding_9.f90: New test. > >>> * gfortran.dg/ISO_Fortran_binding_9.c: Called by it. > >>> > >>> 2019-04-09 Paul Thomas <pault@gcc.gnu.org> > >>> > >>> PR fortran/89843 > >>> * runtime/ISO_Fortran_binding.c (cfi_desc_to_gfc_desc): Only > >>> return immediately if the source pointer is null. Bring > >>> forward the extraction of the gfc type. Extract the kind so > >>> that the element size can be correctly computed for sections > >>> and components of derived type arrays. Remove the free of the > >>> CFI descriptor since this is now done in trans-expr.c. > >>> (gfc_desc_to_cfi_desc): Only allocate the CFI descriptor if it > >>> is not null. > >>> (CFI_section): Normalise the difference between the upper and > >>> lower bounds by the stride to correctly calculate the extents > >>> of the section. > >>> > >>> PR fortran/90022 > >>> * runtime/ISO_Fortran_binding.c (CFI_is_contiguous) : Return > >>> 1 for true and 0 otherwise to comply with the standard. Correct > >>> the contiguity check for rank 3 and greater by using the stride > >>> measure of the lower dimension rather than the element length. -- "If you can't explain it simply, you don't understand it well enough" - Albert Einstein
Dear All,
Please find attached the updated patch, which fixes the problem with
-m32 in PR90022, eliminates the temporary creation for INTENT(IN)
dummies and fixes PR89846.
While it looks like it should be intrusive because of its size, I
believe that the patch is still safe for trunk since it is hidden
behind tests for CFI descriptors.
Bootstraps and regtests on FC29/x86_64 - OK for trunk?
Cheers
Paul
2019-04-14 Paul Thomas <pault@gcc.gnu.org>
PR fortran/89843
* trans-decl.c (gfc_get_symbol_decl): Assumed shape and assumed
rank dummies of bind C procs require deferred initialization.
(convert_CFI_desc): New procedure to convert incoming CFI
descriptors to gfc types and back again.
(gfc_trans_deferred_vars): Call it.
* trans-expr.c (gfc_conv_gfc_desc_to_cfi_desc): Null the CFI
descriptor pointer. Free the descriptor in all cases.
PR fortran/89846
* expr.c (is_CFI_desc): New function.
(is_subref_array): Tidy up by referencing the symbol directly.
* gfortran.h : Prototype for is_CFI_desc.
* trans_array.c (get_CFI_desc): New function.
(gfc_get_array_span, gfc_conv_scalarized_array_ref,
gfc_conv_array_ref): Use it.
* trans.c (get_array_span): Extract the span from descriptors
that are indirect references.
PR fortran/90022
* trans-decl.c (gfc_get_symbol_decl): Make sure that the se
expression is a pointer type before converting it to the symbol
backend_decl type.
* trans-expr.c (gfc_conv_gfc_desc_to_cfi_desc): Eliminate
temporary creation for intent(in).
2019-04-14 Paul Thomas <pault@gcc.gnu.org>
PR fortran/89843
* gfortran.dg/ISO_Fortran_binding_4.f90: Modify the value of x
in ctg. Test the conversion of the descriptor types in the main
program.
* gfortran.dg/ISO_Fortran_binding_10.f90: New test.
* gfortran.dg/ISO_Fortran_binding_10.c: Called by it.
PR fortran/89846
* gfortran.dg/ISO_Fortran_binding_11.f90: New test.
* gfortran.dg/ISO_Fortran_binding_11.c: Called by it.
PR fortran/90022
* gfortran.dg/ISO_Fortran_binding_1.c: Correct the indexing for
the computation of 'ans'. Also, change the expected results for
CFI_is_contiguous to comply with standard.
* gfortran.dg/ISO_Fortran_binding_1.f90: Correct the expected
results for CFI_is_contiguous to comply with standard.
* gfortran.dg/ISO_Fortran_binding_9.f90: New test.
* gfortran.dg/ISO_Fortran_binding_9.c: Called by it.
2019-04-14 Paul Thomas <pault@gcc.gnu.org>
PR fortran/89843
* runtime/ISO_Fortran_binding.c (cfi_desc_to_gfc_desc): Only
return immediately if the source pointer is null. Bring
forward the extraction of the gfc type. Extract the kind so
that the element size can be correctly computed for sections
and components of derived type arrays. Remove the free of the
CFI descriptor since this is now done in trans-expr.c.
(gfc_desc_to_cfi_desc): Only allocate the CFI descriptor if it
is not null.
(CFI_section): Normalise the difference between the upper and
lower bounds by the stride to correctly calculate the extents
of the section.
PR fortran/89846
* runtime/ISO_Fortran_binding.c (cfi_desc_to_gfc_desc): Use
the stride measure for the gfc span if it is not a multiple
of the element length. Otherwise use the element length.
PR fortran/90022
* runtime/ISO_Fortran_binding.c (CFI_is_contiguous) : Return
1 for true and 0 otherwise to comply with the standard. Correct
the contiguity check for rank 3 and greater by using the stride
measure of the lower dimension rather than the element length.
On Fri, 12 Apr 2019 at 10:08, Paul Richard Thomas
<paul.richard.thomas@gmail.com> wrote:
>
> Gilles & Reinhold,
>
> Following the discussion, I have decided to remove the copy-in for
> intent(in). Dominique and I have located the problem with -m32 for
> PR90022 and I am working on PR89846 right now.
>
> I will be resubmitting a bit later on.
>
> Cheers
>
> Paul
>
>
> On Fri, 12 Apr 2019 at 01:25, Gilles Gouaillardet <gilles@rist.or.jp> wrote:
> >
> > Reinhold,
> >
> >
> > Thanks for the insights !
> >
> >
> > That means there is currently an other issue since copy-in is performed
> > even if the argument is declared as ASYNCHRONOUS.
> >
> >
> > I gave the copy-in mechanism some more thoughts, and as a library
> > developers, I can clearly see a need *not* to do that
> >
> > on a case-by-case basis, mainly for performance reasons, but also to be
> > friendly with legacy apps that are not strictly standard compliant.
> >
> > At this stage, I think the best way to move forward is to add an other
> > directive in the interface definition.
> >
> >
> > for example, we could declare
> >
> >
> > module foo
> >
> > interface
> >
> > subroutine bar_f08(buf) BIND(C, name="bar_c")
> >
> > implicit none
> >
> > !GCC$ ATTRIBUTES NO_COPY_IN :: buf
> >
> > TYPE(*), DIMENSION(..), INTENT(IN) :: buf
> >
> > end subroutine
> >
> > end interface
> >
> > end module
> >
> >
> > Does this make sense ?
> >
> >
> > Gilles
> >
> > On 4/10/2019 4:22 PM, Bader, Reinhold wrote:
> > > Hi Gilles,
> > >
> > >> I also found an other potential issue with copy-in.
> > >>
> > >> If in Fortran, we
> > >>
> > >> call foo(buf(0,0))
> > >>
> > >> then the C subroutine can only access buf(0,0), and other elements such
> > >> as buf(1025,1025) cannot be accessed.
> > >>
> > >> Such elements are valid in buf, but out of bounds in the copy (that
> > >> contains a single element).
> > >>
> > >> Strictly speaking, I cannot say whether this is a violation of the
> > >> standard or not, but I can see how this will
> > >>
> > >> break a lot of existing apps (once again, those apps might be incorrect
> > >> in the first place, but most of us got used to them working).
> > >
> > > The above call will only be conforming if the dummy argument is declared
> > > assumed or explicit size.
> > > Otherwise, the compiler should reject it due to rank mismatch. For assumed
> > > rank, the call would be
> > > legitimate, but the rank of the dummy argument is then zero. Even if no
> > > copy-in is performed,
> > > accessing data beyond the address range of that scalar is not strictly
> > > allowed.
> > >
> > > Of more interest is the situation where the dummy argument in Fortran is
> > > declared, e.g.,
> > >
> > > TYPE(*), ASYNCHRONOUS, INTENT(IN) :: BUF(..)
> > >
> > > The standard's semantics *forbids* performing copy-in/out in this case, IIRC.
> > > Otherwise
> > > ASYNCHRONOUS semantics would not work, and non-blocking MPI calls would fail
> > > due
> > > to buffers vanishing into thin air.
> > >
> > > Regards
> > > Reinhold
> > >
> > >> To me, this is a second reason why copy-in is not desirable (at least as
> > >> a default option).
> > >>
> > >>
> > >>
> > >> Cheers,
> > >>
> > >>
> > >> Gilles
> > >>
> > >> On 4/9/2019 7:18 PM, Paul Richard Thomas wrote:
> > >>> The most part of this patch is concerned with implementing calls from
> > >>> C of of fortran bind c procedures with assumed shape or assumed rank
> > >>> dummies to completely fix PR89843. The conversion of the descriptors
> > >>> from CFI to gfc occur on entry to and reversed on exit from the
> > >>> procedure.
> > >>>
> > >>> This patch is safe for trunk, even at this late stage, because its
> > >>> effects are barricaded behind the tests for CFI descriptors. I believe
> > >>> that it appropriately rewards the bug reporters to have this feature
> > >>> work as well as possible at release.
> > >>>
> > >>> Between comments and the ChangeLogs, this patch is self explanatory.
> > >>>
> > >>> Bootstrapped and regtested on FC29/x86_64 - OK for trunk?
> > >>>
> > >>> Paul
> > >>>
> > >>> 2019-04-09 Paul Thomas <pault@gcc.gnu.org>
> > >>>
> > >>> PR fortran/89843
> > >>> * trans-decl.c (gfc_get_symbol_decl): Assumed shape and assumed
> > >>> rank dummies of bind C procs require deferred initialization.
> > >>> (convert_CFI_desc): New procedure to convert incoming CFI
> > >>> descriptors to gfc types and back again.
> > >>> (gfc_trans_deferred_vars): Call it.
> > >>> * trans-expr.c (gfc_conv_gfc_desc_to_cfi_desc): Null the CFI
> > >>> descriptor pointer. Free the descriptor in all cases.
> > >>>
> > >>> PR fortran/90022
> > >>> * trans-decl.c (gfc_get_symbol_decl): Make sure that the se
> > >>> expression is a pointer type before converting it to the symbol
> > >>> backend_decl type.
> > >>>
> > >>> 2019-04-09 Paul Thomas <pault@gcc.gnu.org>
> > >>>
> > >>> PR fortran/89843
> > >>> * gfortran.dg/ISO_Fortran_binding_4.f90: Modify the value of x
> > >>> in ctg. Test the conversion of the descriptor types in the main
> > >>> program.
> > >>> * gfortran.dg/ISO_Fortran_binding_10.f90: New test.
> > >>> * gfortran.dg/ISO_Fortran_binding_10.c: Called by it.
> > >>>
> > >>> PR fortran/90022
> > >>> * gfortran.dg/ISO_Fortran_binding_1.c: Correct the indexing for
> > >>> the computation of 'ans'. Also, change the expected results for
> > >>> CFI_is_contiguous to comply with standard.
> > >>> * gfortran.dg/ISO_Fortran_binding_1.f90: Correct the expected
> > >>> results for CFI_is_contiguous to comply with standard.
> > >>> * gfortran.dg/ISO_Fortran_binding_9.f90: New test.
> > >>> * gfortran.dg/ISO_Fortran_binding_9.c: Called by it.
> > >>>
> > >>> 2019-04-09 Paul Thomas <pault@gcc.gnu.org>
> > >>>
> > >>> PR fortran/89843
> > >>> * runtime/ISO_Fortran_binding.c (cfi_desc_to_gfc_desc): Only
> > >>> return immediately if the source pointer is null. Bring
> > >>> forward the extraction of the gfc type. Extract the kind so
> > >>> that the element size can be correctly computed for sections
> > >>> and components of derived type arrays. Remove the free of the
> > >>> CFI descriptor since this is now done in trans-expr.c.
> > >>> (gfc_desc_to_cfi_desc): Only allocate the CFI descriptor if it
> > >>> is not null.
> > >>> (CFI_section): Normalise the difference between the upper and
> > >>> lower bounds by the stride to correctly calculate the extents
> > >>> of the section.
> > >>>
> > >>> PR fortran/90022
> > >>> * runtime/ISO_Fortran_binding.c (CFI_is_contiguous) : Return
> > >>> 1 for true and 0 otherwise to comply with the standard. Correct
> > >>> the contiguity check for rank 3 and greater by using the stride
> > >>> measure of the lower dimension rather than the element length.
>
>
>
> --
> "If you can't explain it simply, you don't understand it well enough"
> - Albert Einstein
Hi Paul, > Please find attached the updated patch, which fixes the problem with > -m32 in PR90022, eliminates the temporary creation for INTENT(IN) > dummies and fixes PR89846. > > While it looks like it should be intrusive because of its size, I > believe that the patch is still safe for trunk since it is hidden > behind tests for CFI descriptors. > > Bootstraps and regtests on FC29/x86_64 - OK for trunk? OK. I we're going into the gcc 9 release with an implementation of the C interop features, it will be better with fewer bugs :-) Thanks a lot for working on it! Regards Thomas
Hi Thomas, Thanks a lot. Committed as revision 270353. I was determined not to repeat the PDT experience, which is still nagging at me. That has to be the next major gfc project, I guess. Regards Paul On Sun, 14 Apr 2019 at 18:08, Thomas Koenig <tkoenig@netcologne.de> wrote: > > Hi Paul, > > > > Please find attached the updated patch, which fixes the problem with > > -m32 in PR90022, eliminates the temporary creation for INTENT(IN) > > dummies and fixes PR89846. > > > > While it looks like it should be intrusive because of its size, I > > believe that the patch is still safe for trunk since it is hidden > > behind tests for CFI descriptors. > > > > Bootstraps and regtests on FC29/x86_64 - OK for trunk? > > OK. > > I we're going into the gcc 9 release with an implementation of the > C interop features, it will be better with fewer bugs :-) > > Thanks a lot for working on it! > > Regards > > Thomas
Dear Paul, mostly looks good. Apart from a regression with optional arguments reported as https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90093 all other test cases I have now execute correctly. Cheers Reinhold > -----Ursprüngliche Nachricht----- > Von: Paul Richard Thomas <paul.richard.thomas@gmail.com> > Gesendet: Sonntag, 14. April 2019 20:16 > An: Thomas Koenig <tkoenig@netcologne.de> > Cc: Gilles Gouaillardet <gilles@rist.or.jp>; Bader, Reinhold > <Reinhold.Bader@lrz.de>; fortran@gcc.gnu.org; gcc-patches <gcc- > patches@gcc.gnu.org> > Betreff: Re: [Patch, fortran] PRs 89843 and 90022 - C Fortran Interop fixes. > > Hi Thomas, > > Thanks a lot. Committed as revision 270353. > > I was determined not to repeat the PDT experience, which is still nagging at > me. That has to be the next major gfc project, I guess. > > Regards > > Paul > > On Sun, 14 Apr 2019 at 18:08, Thomas Koenig <tkoenig@netcologne.de> > wrote: > > > > Hi Paul, > > > > > > > Please find attached the updated patch, which fixes the problem with > > > -m32 in PR90022, eliminates the temporary creation for INTENT(IN) > > > dummies and fixes PR89846. > > > > > > While it looks like it should be intrusive because of its size, I > > > believe that the patch is still safe for trunk since it is hidden > > > behind tests for CFI descriptors. > > > > > > Bootstraps and regtests on FC29/x86_64 - OK for trunk? > > > > OK. > > > > I we're going into the gcc 9 release with an implementation of the C > > interop features, it will be better with fewer bugs :-) > > > > Thanks a lot for working on it! > > > > Regards > > > > Thomas > > > > -- > "If you can't explain it simply, you don't understand it well enough" > - Albert Einstein
Dear Dominique, Gilles and Reinhold, Thank you for your rapid feedback. We might even get a reasonably functional ISO Fortran binding in place for 9-branch release :-) On your remaining nits: (i) ISO_Fortran_binding_4.f90 -m32 -O1/Os looks awful. I will take a look, though. (ii) pr89844 being fixed by an earlier patch led me to give it lower priority. I will look to see whether another testcase is required to nail it down. (iii) I will take a look at 90093 - it should be straight forward. I do not regard it as being a regression, however, since the arguments were not being correctly handled until now - ie. were not converted from cfi to gfc descriptors. Cheers Paul On Mon, 15 Apr 2019 at 10:27, Bader, Reinhold <Reinhold.Bader@lrz.de> wrote: > > Dear Paul, > > mostly looks good. Apart from a regression with optional arguments reported as > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90093 > all other test cases I have now execute correctly. > > Cheers > Reinhold > > > -----Ursprüngliche Nachricht----- > > Von: Paul Richard Thomas <paul.richard.thomas@gmail.com> > > Gesendet: Sonntag, 14. April 2019 20:16 > > An: Thomas Koenig <tkoenig@netcologne.de> > > Cc: Gilles Gouaillardet <gilles@rist.or.jp>; Bader, Reinhold > > <Reinhold.Bader@lrz.de>; fortran@gcc.gnu.org; gcc-patches <gcc- > > patches@gcc.gnu.org> > > Betreff: Re: [Patch, fortran] PRs 89843 and 90022 - C Fortran Interop fixes. > > > > Hi Thomas, > > > > Thanks a lot. Committed as revision 270353. > > > > I was determined not to repeat the PDT experience, which is still nagging at > > me. That has to be the next major gfc project, I guess. > > > > Regards > > > > Paul > > > > On Sun, 14 Apr 2019 at 18:08, Thomas Koenig <tkoenig@netcologne.de> > > wrote: > > > > > > Hi Paul, > > > > > > > > > > Please find attached the updated patch, which fixes the problem with > > > > -m32 in PR90022, eliminates the temporary creation for INTENT(IN) > > > > dummies and fixes PR89846. > > > > > > > > While it looks like it should be intrusive because of its size, I > > > > believe that the patch is still safe for trunk since it is hidden > > > > behind tests for CFI descriptors. > > > > > > > > Bootstraps and regtests on FC29/x86_64 - OK for trunk? > > > > > > OK. > > > > > > I we're going into the gcc 9 release with an implementation of the C > > > interop features, it will be better with fewer bugs :-) > > > > > > Thanks a lot for working on it! > > > > > > Regards > > > > > > Thomas > > > > > > > > -- > > "If you can't explain it simply, you don't understand it well enough" > > - Albert Einstein
Index: gcc/fortran/trans-decl.c =================================================================== *** gcc/fortran/trans-decl.c (revision 270149) --- gcc/fortran/trans-decl.c (working copy) *************** gfc_get_symbol_decl (gfc_symbol * sym) *** 1494,1499 **** --- 1494,1506 ---- && sym->attr.dummy) gfc_defer_symbol_init (sym); + if (sym->attr.dummy + && sym->ns->proc_name->attr.is_bind_c + && sym->attr.dimension + && (sym->as->type == AS_ASSUMED_SHAPE + || sym->as->type == AS_ASSUMED_RANK)) + gfc_defer_symbol_init (sym); + /* All deferred character length procedures need to retain the backend decl, which is a pointer to the character length in the caller's namespace and to declare a local character length. */ *************** gfc_null_and_pass_deferred_len (gfc_symb *** 4268,4273 **** --- 4275,4353 ---- } + /* Convert CFI descriptor dummies into gfc types and back again. */ + static void + convert_CFI_desc (gfc_wrapped_block * block, gfc_symbol *sym) + { + tree gfc_desc; + tree gfc_desc_ptr; + tree CFI_desc; + tree CFI_desc_ptr; + tree dummy_ptr; + tree tmp; + tree incoming; + tree outgoing; + stmtblock_t tmpblock; + + /* dummy_ptr will be the pointer to the passed array descriptor, + while CFI_desc is the descriptor itself. */ + dummy_ptr = sym->backend_decl; + CFI_desc = sym->backend_decl; + if (POINTER_TYPE_P (TREE_TYPE (CFI_desc))) + CFI_desc = build_fold_indirect_ref_loc (input_location, CFI_desc); + if (!GFC_DESCRIPTOR_TYPE_P (TREE_TYPE (CFI_desc))) + { + if (DECL_LANG_SPECIFIC (sym->backend_decl)) + CFI_desc = GFC_DECL_SAVED_DESCRIPTOR (sym->backend_decl); + else + CFI_desc = NULL; + dummy_ptr = CFI_desc; + } + + if (CFI_desc) + { + if (POINTER_TYPE_P (TREE_TYPE (CFI_desc))) + CFI_desc = build_fold_indirect_ref_loc (input_location, CFI_desc); + + /* The compiler will have given CFI_desc the correct gfortran + type. Use this new variable to store the converted + descriptor. */ + gfc_desc = gfc_create_var (TREE_TYPE (CFI_desc), "gfc_desc"); + tmp = build_pointer_type (TREE_TYPE (gfc_desc)); + gfc_desc_ptr = gfc_create_var (tmp, "gfc_desc_ptr"); + CFI_desc_ptr = gfc_create_var (pvoid_type_node, "CFI_desc_ptr"); + + gfc_init_block (&tmpblock); + /* Pointer to the gfc descriptor. */ + gfc_add_modify (&tmpblock, gfc_desc_ptr, + gfc_build_addr_expr (NULL, gfc_desc)); + /* Store the pointer to the CFI descriptor. */ + gfc_add_modify (&tmpblock, CFI_desc_ptr, + fold_convert (pvoid_type_node, dummy_ptr)); + tmp = gfc_build_addr_expr (ppvoid_type_node, CFI_desc_ptr); + /* Convert the CFI descriptor. */ + incoming = build_call_expr_loc (input_location, + gfor_fndecl_cfi_to_gfc, 2, gfc_desc_ptr, tmp); + gfc_add_expr_to_block (&tmpblock, incoming); + /* Set the dummy pointer to point to the gfc_descriptor. */ + gfc_add_modify (&tmpblock, dummy_ptr, + fold_convert (TREE_TYPE (dummy_ptr), gfc_desc_ptr)); + incoming = gfc_finish_block (&tmpblock); + + gfc_init_block (&tmpblock); + /* Convert the gfc descriptor back to the CFI type before going + out of scope. */ + tmp = gfc_build_addr_expr (ppvoid_type_node, CFI_desc_ptr); + outgoing = build_call_expr_loc (input_location, + gfor_fndecl_gfc_to_cfi, 2, tmp, gfc_desc_ptr); + gfc_add_expr_to_block (&tmpblock, outgoing); + outgoing = gfc_finish_block (&tmpblock); + + /* Add the lot to the procedure init and finally blocks. */ + gfc_add_init_cleanup (block, incoming, outgoing); + } + } + /* Get the result expression for a procedure. */ static tree *************** gfc_trans_deferred_vars (gfc_symbol * pr *** 4844,4849 **** --- 4924,4940 ---- } else if (!(UNLIMITED_POLY(sym)) && !is_pdt_type) gcc_unreachable (); + + /* Assumed shape and assumed rank arrays are passed to BIND(C) procedures + as ISO Fortran Interop descriptors. These have to be converted to + gfortran descriptors and back again. This has to be done here so that + the conversion occurs at the start of the init block. */ + if (sym->attr.dummy + && sym->ns->proc_name->attr.is_bind_c + && sym->attr.dimension + && (sym->as->type == AS_ASSUMED_SHAPE + || sym->as->type == AS_ASSUMED_RANK)) + convert_CFI_desc (block, sym); } gfc_init_block (&tmpblock); Index: gcc/fortran/trans-expr.c =================================================================== *** gcc/fortran/trans-expr.c (revision 270149) --- gcc/fortran/trans-expr.c (working copy) *************** gfc_conv_gfc_desc_to_cfi_desc (gfc_se *p *** 4990,4997 **** --- 4990,4999 ---- tree ptr = NULL_TREE; tree size; tree type; + tree cond; int attribute; symbol_attribute attr = gfc_expr_attr (e); + stmtblock_t block; /* If this is a full array or a scalar, the allocatable and pointer attributes can be passed. Otherwise it is 'CFI_attribute_other'*/ *************** gfc_conv_gfc_desc_to_cfi_desc (gfc_se *p *** 5135,5140 **** --- 5137,5144 ---- /* Variables to point to the gfc and CFI descriptors. */ gfc_desc_ptr = parmse->expr; cfi_desc_ptr = gfc_create_var (pvoid_type_node, "cfi"); + gfc_add_modify (&parmse->pre, cfi_desc_ptr, + build_int_cst (pvoid_type_node, 0)); /* Allocate the CFI descriptor and fill the fields. */ tmp = gfc_build_addr_expr (NULL_TREE, cfi_desc_ptr); *************** gfc_conv_gfc_desc_to_cfi_desc (gfc_se *p *** 5147,5154 **** if (ptr) { ! /* Free both the temporary data and the CFI descriptor for ! INTENT(IN) arrays. */ tmp = gfc_call_free (ptr); gfc_prepend_expr_to_block (&parmse->post, tmp); tmp = gfc_call_free (cfi_desc_ptr); --- 5151,5157 ---- if (ptr) { ! /* Free the temporary data for INTENT(IN) arrays. */ tmp = gfc_call_free (ptr); gfc_prepend_expr_to_block (&parmse->post, tmp); tmp = gfc_call_free (cfi_desc_ptr); *************** gfc_conv_gfc_desc_to_cfi_desc (gfc_se *p *** 5156,5162 **** return; } ! /* Transfer values back to gfc descriptor and free the CFI descriptor. */ tmp = gfc_build_addr_expr (NULL_TREE, parmse->expr); tmp = build_call_expr_loc (input_location, gfor_fndecl_cfi_to_gfc, 2, gfc_desc_ptr, tmp); --- 5159,5177 ---- return; } ! /* Free the CFI descriptor. */ ! gfc_init_block (&block); ! cond = fold_build2_loc (input_location, NE_EXPR, ! logical_type_node, cfi_desc_ptr, ! build_int_cst (TREE_TYPE (cfi_desc_ptr), 0)); ! tmp = gfc_call_free (cfi_desc_ptr); ! gfc_add_expr_to_block (&block, tmp); ! tmp = build3_v (COND_EXPR, cond, ! gfc_finish_block (&block), ! build_empty_stmt (input_location)); ! gfc_prepend_expr_to_block (&parmse->post, tmp); ! ! /* Transfer values back to gfc descriptor. */ tmp = gfc_build_addr_expr (NULL_TREE, parmse->expr); tmp = build_call_expr_loc (input_location, gfor_fndecl_cfi_to_gfc, 2, gfc_desc_ptr, tmp); Index: gcc/testsuite/gfortran.dg/ISO_Fortran_binding_1.c =================================================================== *** gcc/testsuite/gfortran.dg/ISO_Fortran_binding_1.c (revision 270149) --- gcc/testsuite/gfortran.dg/ISO_Fortran_binding_1.c (working copy) *************** float section_c(int *std_case, CFI_cdesc *** 105,111 **** CFI_index_t idx[CFI_MAX_RANK], lower[CFI_MAX_RANK], strides[CFI_MAX_RANK], upper[CFI_MAX_RANK]; CFI_CDESC_T(1) section; ! int ind, size; float *ret_addr; float ans = 0.0; --- 105,111 ---- CFI_index_t idx[CFI_MAX_RANK], lower[CFI_MAX_RANK], strides[CFI_MAX_RANK], upper[CFI_MAX_RANK]; CFI_CDESC_T(1) section; ! int ind; float *ret_addr; float ans = 0.0; *************** float section_c(int *std_case, CFI_cdesc *** 121,129 **** if (ind) return -2.0; /* Sum over the section */ ! size = (section.dim[0].extent - 1) ! * section.elem_len/section.dim[0].sm + 1; ! for (idx[0] = 0; idx[0] < size; idx[0]++) ans += *(float*)CFI_address ((CFI_cdesc_t*)§ion, idx); return ans; } --- 121,127 ---- if (ind) return -2.0; /* Sum over the section */ ! for (idx[0] = 0; idx[0] < section.dim[0].extent; idx[0]++) ans += *(float*)CFI_address ((CFI_cdesc_t*)§ion, idx); return ans; } *************** float section_c(int *std_case, CFI_cdesc *** 143,151 **** if (ind) return -2.0; /* Sum over the section */ ! size = (section.dim[0].extent - 1) ! * section.elem_len/section.dim[0].sm + 1; ! for (idx[0] = 0; idx[0] < size; idx[0]++) ans += *(float*)CFI_address ((CFI_cdesc_t*)§ion, idx); return ans; } --- 141,147 ---- if (ind) return -2.0; /* Sum over the section */ ! for (idx[0] = 0; idx[0] < section.dim[0].extent; idx[0]++) ans += *(float*)CFI_address ((CFI_cdesc_t*)§ion, idx); return ans; } *************** int setpointer_c(CFI_cdesc_t * ptr, int *** 191,205 **** int assumed_size_c(CFI_cdesc_t * desc) { ! int ierr; ! ierr = CFI_is_contiguous(desc); ! if (ierr) return 1; if (desc->rank) ! ierr = 2 * (desc->dim[desc->rank-1].extent != (CFI_index_t)(long long)(-1)); else ! ierr = 3; ! return ierr; } --- 187,201 ---- int assumed_size_c(CFI_cdesc_t * desc) { ! int res; ! res = CFI_is_contiguous(desc); ! if (!res) return 1; if (desc->rank) ! res = 2 * (desc->dim[desc->rank-1].extent != (CFI_index_t)(long long)(-1)); else ! res = 3; ! return res; } Index: gcc/testsuite/gfortran.dg/ISO_Fortran_binding_1.f90 =================================================================== *** gcc/testsuite/gfortran.dg/ISO_Fortran_binding_1.f90 (revision 270149) --- gcc/testsuite/gfortran.dg/ISO_Fortran_binding_1.f90 (working copy) *************** end subroutine test_CFI_address *** 170,185 **** integer, dimension (2,*) :: arg character(4), dimension(2) :: chr ! These are contiguous ! if (c_contiguous (arg) .ne. 0) stop 20 if (.not.allocated (x)) allocate (x(2, 2)) ! if (c_contiguous (x) .ne. 0) stop 22 deallocate (x) ! if (c_contiguous (chr) .ne. 0) stop 23 ! These are not contiguous ! if (c_contiguous (der%i) .eq. 0) stop 24 ! if (c_contiguous (arg(1:1,1:2)) .eq. 0) stop 25 ! if (c_contiguous (d(4:2:-2, 1:3:2)) .eq. 0) stop 26 ! if (c_contiguous (chr(:)(2:3)) .eq. 0) stop 27 end subroutine test_CFI_contiguous subroutine test_CFI_section (arg) --- 170,185 ---- integer, dimension (2,*) :: arg character(4), dimension(2) :: chr ! These are contiguous ! if (c_contiguous (arg) .ne. 1) stop 20 if (.not.allocated (x)) allocate (x(2, 2)) ! if (c_contiguous (x) .ne. 1) stop 22 deallocate (x) ! if (c_contiguous (chr) .ne. 1) stop 23 ! These are not contiguous ! if (c_contiguous (der%i) .eq. 1) stop 24 ! if (c_contiguous (arg(1:1,1:2)) .eq. 1) stop 25 ! if (c_contiguous (d(4:2:-2, 1:3:2)) .eq. 1) stop 26 ! if (c_contiguous (chr(:)(2:3)) .eq. 1) stop 27 end subroutine test_CFI_contiguous subroutine test_CFI_section (arg) Index: gcc/testsuite/gfortran.dg/ISO_Fortran_binding_10.c =================================================================== *** gcc/testsuite/gfortran.dg/ISO_Fortran_binding_10.c (nonexistent) --- gcc/testsuite/gfortran.dg/ISO_Fortran_binding_10.c (working copy) *************** *** 0 **** --- 1,73 ---- + /* Test the fix of PR89843. */ + + /* Contributed by Reinhold Bader <Bader@lrz.de> */ + + #include "../../../libgfortran/ISO_Fortran_binding.h" + #include <stdlib.h> + #include <stdio.h> + #include <stdbool.h> + + void sa(CFI_cdesc_t *, int, int *); + + void si(CFI_cdesc_t *this, int flag, int *status) + { + int value, sum; + bool err; + CFI_CDESC_T(1) that; + CFI_index_t lb[] = { 0, 0 }; + CFI_index_t ub[] = { 4, 1 }; + CFI_index_t st[] = { 2, 0 }; + int chksum[] = { 9, 36, 38 }; + + if (flag == 1) + { + lb[0] = 0; lb[1] = 2; + ub[0] = 2; ub[1] = 2; + st[0] = 1; st[1] = 0; + } + else if (flag == 2) + { + lb[0] = 1; lb[1] = 0; + ub[0] = 1; ub[1] = 3; + st[0] = 0; st[1] = 1; + } + + CFI_establish((CFI_cdesc_t *) &that, NULL, CFI_attribute_other, + CFI_type_float, 0, 1, NULL); + + *status = CFI_section((CFI_cdesc_t *) &that, this, lb, ub, st); + + if (*status != CFI_SUCCESS) + { + printf("FAIL C: status is %i\n",status); + return; + } + + value = CFI_is_contiguous((CFI_cdesc_t *) &that); + err = ((flag == 0 && value != 0) + || (flag == 1 && value != 1) + || (flag == 2 && value != 0)); + + if (err) + { + printf("FAIL C: contiguity for flag value %i - is %i\n",flag, value); + *status = 10; + return; + } + + sum = 0; + for (int i = 0; i < that.dim[0].extent; i++) + { + CFI_index_t idx[] = {i}; + sum += (int)(*(float *)CFI_address ((CFI_cdesc_t *)&that, idx)); + } + + if (sum != chksum[flag]) + { + printf ("FAIL C: check sum = %d(%d)\n", sum, chksum[flag]); + *status = 11; + return; + } + + sa((CFI_cdesc_t *) &that, flag, status); + } Index: gcc/testsuite/gfortran.dg/ISO_Fortran_binding_10.f90 =================================================================== *** gcc/testsuite/gfortran.dg/ISO_Fortran_binding_10.f90 (nonexistent) --- gcc/testsuite/gfortran.dg/ISO_Fortran_binding_10.f90 (working copy) *************** *** 0 **** --- 1,99 ---- + ! { dg-do run { target c99_runtime } } + ! { dg-additional-sources ISO_Fortran_binding_10.c } + ! + ! Test the fix of PR89843. + ! + ! Contributed by Reinhold Bader <Bader@lrz.de> + ! + module mod_section_01 + use, intrinsic :: iso_c_binding + implicit none + interface + subroutine si(this, flag, status) bind(c) + import :: c_float, c_int + real(c_float) :: this(:,:) + integer(c_int), value :: flag + integer(c_int) :: status + end subroutine si + end interface + contains + subroutine sa(this, flag, status) bind(c) + real(c_float) :: this(:) + integer(c_int), value :: flag + integer(c_int) :: status + + status = 0 + + select case (flag) + case (0) + if (is_contiguous(this)) then + write(*,*) 'FAIL 1:' + status = status + 1 + end if + if (size(this,1) /= 3) then + write(*,*) 'FAIL 2:',size(this) + status = status + 1 + goto 10 + end if + if (maxval(abs(this - [ 1.0, 3.0, 5.0 ])) > 1.0e-6) then + write(*,*) 'FAIL 3:',this + status = status + 1 + end if + 10 continue + case (1) + if (size(this,1) /= 3) then + write(*,*) 'FAIL 4:',size(this) + status = status + 1 + goto 20 + end if + if (maxval(abs(this - [ 11.0, 12.0, 13.0 ])) > 1.0e-6) then + write(*,*) 'FAIL 5:',this + status = status + 1 + end if + 20 continue + case (2) + if (size(this,1) /= 4) then + write(*,*) 'FAIL 6:',size(this) + status = status + 1 + goto 30 + end if + if (maxval(abs(this - [ 2.0, 7.0, 12.0, 17.0 ])) > 1.0e-6) then + write(*,*) 'FAIL 7:',this + status = status + 1 + end if + 30 continue + end select + + ! if (status == 0) then + ! write(*,*) 'OK' + ! end if + end subroutine sa + end module mod_section_01 + + program section_01 + use mod_section_01 + implicit none + real(c_float) :: v(5,4) + integer :: i + integer :: status + + v = reshape( [ (real(i), i = 1, 20) ], [ 5, 4 ] ) + call si(v, 0, status) + if (status .ne. 0) stop 1 + + call sa(v(1:5:2, 1), 0, status) + if (status .ne. 0) stop 2 + + call si(v, 1, status) + if (status .ne. 0) stop 3 + + call sa(v(1:3, 3), 1, status) + if (status .ne. 0) stop 4 + + call si(v, 2, status) + if (status .ne. 0) stop 5 + + call sa(v(2,1:4), 2, status) + if (status .ne. 0) stop 6 + + end program section_01 Index: gcc/testsuite/gfortran.dg/ISO_Fortran_binding_4.f90 =================================================================== *** gcc/testsuite/gfortran.dg/ISO_Fortran_binding_4.f90 (revision 270149) --- gcc/testsuite/gfortran.dg/ISO_Fortran_binding_4.f90 (working copy) *************** contains *** 10,18 **** --- 10,20 ---- if (any(abs(x - [2.,4.,6.]) > 1.e-6)) then write(*,*) 'FAIL' + stop 1 else write(*,*) 'OK' end if + x = [2.,4.,6.]*10.0 end subroutine end module program p *************** program p *** 23,27 **** x = [ (real(i), i=1, size(x)) ] call ctg(x(2::2)) ! end program --- 25,29 ---- x = [ (real(i), i=1, size(x)) ] call ctg(x(2::2)) ! if (any (abs (x - [1.,20.,3.,40.,5.,60.]) > 1.e-6)) stop 2 end program Index: gcc/testsuite/gfortran.dg/ISO_Fortran_binding_9.c =================================================================== *** gcc/testsuite/gfortran.dg/ISO_Fortran_binding_9.c (nonexistent) --- gcc/testsuite/gfortran.dg/ISO_Fortran_binding_9.c (working copy) *************** *** 0 **** --- 1,14 ---- + /* Test fix of a problem with CFI_is_contiguous. */ + + /* Contributed by Gilles Gouaillardet <gilles@rist.or.jp> */ + + #include "../../../libgfortran/ISO_Fortran_binding.h" + #include <stdlib.h> + + int cdesc_c(CFI_cdesc_t* x, long *expected) + { + int res; + res = CFI_is_contiguous (x); + if (x->base_addr != (void *)*expected) res = 0; + return res; + } \ No newline at end of file Index: gcc/testsuite/gfortran.dg/ISO_Fortran_binding_9.f90 =================================================================== *** gcc/testsuite/gfortran.dg/ISO_Fortran_binding_9.f90 (nonexistent) --- gcc/testsuite/gfortran.dg/ISO_Fortran_binding_9.f90 (working copy) *************** *** 0 **** --- 1,28 ---- + ! { dg-do run { target c99_runtime } } + ! { dg-additional-sources ISO_Fortran_binding_9.c } + ! + ! Fix a problem with CFI_is_contiguous + ! + ! Contributed by Gilles Gouaillardet <gilles@rist.or.jp> + ! + module cdesc + interface + function cdesc_f08(buf, expected) result (res) BIND(C, name="cdesc_c") + USE, INTRINSIC :: ISO_C_BINDING + implicit none + INTEGER(C_INT) :: res + type(*), dimension(..), INTENT(INOUT) :: buf + integer(kind=8),INTENT(IN) :: expected + end function cdesc_f08 + end interface + end module + + program cdesc_test + use cdesc + implicit none + integer :: a0, a1(10), a2(10,10), a3(10,10,10) + if (cdesc_f08(a0, LOC(a0)) .ne. 1) stop 1 + if (cdesc_f08(a1, LOC(a1(1))) .ne. 1) stop 2 + if (cdesc_f08(a2, LOC(a2(1,1))) .ne. 1) stop 3 + if (cdesc_f08(a3, LOC(a3(1,1,1))) .ne. 1) stop 4 + end program Index: libgfortran/runtime/ISO_Fortran_binding.c =================================================================== *** libgfortran/runtime/ISO_Fortran_binding.c (revision 270149) --- libgfortran/runtime/ISO_Fortran_binding.c (working copy) *************** void *** 37,59 **** cfi_desc_to_gfc_desc (gfc_array_void *d, CFI_cdesc_t **s_ptr) { int n; CFI_cdesc_t *s = *s_ptr; ! /* If not a full pointer or allocatable array free the descriptor ! and return. */ ! if (!s || s->attribute == CFI_attribute_other) ! goto finish; GFC_DESCRIPTOR_DATA (d) = s->base_addr; - - if (!s->rank || s->dim[0].sm == (CFI_index_t)s->elem_len) - GFC_DESCRIPTOR_SIZE (d) = s->elem_len; - else - GFC_DESCRIPTOR_SIZE (d) = (index_type)s->dim[0].sm; - - d->dtype.version = s->version; - GFC_DESCRIPTOR_RANK (d) = (signed char)s->rank; GFC_DESCRIPTOR_TYPE (d) = (signed char)(s->type & CFI_type_mask); /* Correct the unfortunate difference in order with types. */ if (GFC_DESCRIPTOR_TYPE (d) == BT_CHARACTER) --- 37,51 ---- cfi_desc_to_gfc_desc (gfc_array_void *d, CFI_cdesc_t **s_ptr) { int n; + index_type kind; CFI_cdesc_t *s = *s_ptr; ! if (!s) ! return; GFC_DESCRIPTOR_DATA (d) = s->base_addr; GFC_DESCRIPTOR_TYPE (d) = (signed char)(s->type & CFI_type_mask); + kind = (index_type)((s->type - (s->type & CFI_type_mask)) >> CFI_type_kind_shift); /* Correct the unfortunate difference in order with types. */ if (GFC_DESCRIPTOR_TYPE (d) == BT_CHARACTER) *************** cfi_desc_to_gfc_desc (gfc_array_void *d, *** 61,72 **** else if (GFC_DESCRIPTOR_TYPE (d) == BT_DERIVED) GFC_DESCRIPTOR_TYPE (d) = BT_CHARACTER; d->dtype.attribute = (signed short)s->attribute; if (s->rank) d->span = (index_type)s->dim[0].sm; - /* On the other hand, CFI_establish can change the bounds. */ d->offset = 0; for (n = 0; n < GFC_DESCRIPTOR_RANK (d); n++) { --- 53,73 ---- else if (GFC_DESCRIPTOR_TYPE (d) == BT_DERIVED) GFC_DESCRIPTOR_TYPE (d) = BT_CHARACTER; + if (!s->rank || s->dim[0].sm == (CFI_index_t)s->elem_len) + GFC_DESCRIPTOR_SIZE (d) = s->elem_len; + else if (GFC_DESCRIPTOR_TYPE (d) != BT_DERIVED) + GFC_DESCRIPTOR_SIZE (d) = kind; + else + GFC_DESCRIPTOR_SIZE (d) = s->elem_len; + + d->dtype.version = s->version; + GFC_DESCRIPTOR_RANK (d) = (signed char)s->rank; + d->dtype.attribute = (signed short)s->attribute; if (s->rank) d->span = (index_type)s->dim[0].sm; d->offset = 0; for (n = 0; n < GFC_DESCRIPTOR_RANK (d); n++) { *************** cfi_desc_to_gfc_desc (gfc_array_void *d, *** 76,86 **** GFC_DESCRIPTOR_STRIDE(d, n) = (index_type)(s->dim[n].sm / s->elem_len); d->offset -= GFC_DESCRIPTOR_STRIDE(d, n) * GFC_DESCRIPTOR_LBOUND(d, n); } - - finish: - if (s) - free (s); - s = NULL; } extern void gfc_desc_to_cfi_desc (CFI_cdesc_t **, const gfc_array_void *); --- 77,82 ---- *************** gfc_desc_to_cfi_desc (CFI_cdesc_t **d_pt *** 95,102 **** /* Play it safe with allocation of the flexible array member 'dim' by setting the length to CFI_MAX_RANK. This should not be necessary but valgrind complains accesses after the allocated block. */ ! d = malloc (sizeof (CFI_cdesc_t) + (CFI_type_t)(CFI_MAX_RANK * sizeof (CFI_dim_t))); d->base_addr = GFC_DESCRIPTOR_DATA (s); d->elem_len = GFC_DESCRIPTOR_SIZE (s); --- 91,101 ---- /* Play it safe with allocation of the flexible array member 'dim' by setting the length to CFI_MAX_RANK. This should not be necessary but valgrind complains accesses after the allocated block. */ ! if (*d_ptr == NULL) ! d = malloc (sizeof (CFI_cdesc_t) + (CFI_type_t)(CFI_MAX_RANK * sizeof (CFI_dim_t))); + else + d = *d_ptr; d->base_addr = GFC_DESCRIPTOR_DATA (s); d->elem_len = GFC_DESCRIPTOR_SIZE (s); *************** gfc_desc_to_cfi_desc (CFI_cdesc_t **d_pt *** 115,121 **** d->type = (CFI_type_t)(d->type + ((CFI_type_t)d->elem_len << CFI_type_kind_shift)); ! /* Full pointer or allocatable arrays have zero lower_bound. */ for (n = 0; n < GFC_DESCRIPTOR_RANK (s); n++) { if (d->attribute != CFI_attribute_other) --- 114,120 ---- d->type = (CFI_type_t)(d->type + ((CFI_type_t)d->elem_len << CFI_type_kind_shift)); ! /* Full pointer or allocatable arrays retain their lower_bounds. */ for (n = 0; n < GFC_DESCRIPTOR_RANK (s); n++) { if (d->attribute != CFI_attribute_other) *************** gfc_desc_to_cfi_desc (CFI_cdesc_t **d_pt *** 134,140 **** d->dim[n].sm = (CFI_index_t)(GFC_DESCRIPTOR_STRIDE(s, n) * s->span); } ! *d_ptr = d; } void *CFI_address (const CFI_cdesc_t *dv, const CFI_index_t subscripts[]) --- 133,140 ---- d->dim[n].sm = (CFI_index_t)(GFC_DESCRIPTOR_STRIDE(s, n) * s->span); } ! if (*d_ptr == NULL) ! *d_ptr = d; } void *CFI_address (const CFI_cdesc_t *dv, const CFI_index_t subscripts[]) *************** int CFI_is_contiguous (const CFI_cdesc_t *** 416,422 **** if (dv == NULL) { fprintf (stderr, "CFI_is_contiguous: C descriptor is NULL.\n"); ! return CFI_INVALID_DESCRIPTOR; } /* Base address must not be NULL. */ --- 416,422 ---- if (dv == NULL) { fprintf (stderr, "CFI_is_contiguous: C descriptor is NULL.\n"); ! return 0; } /* Base address must not be NULL. */ *************** int CFI_is_contiguous (const CFI_cdesc_t *** 424,430 **** { fprintf (stderr, "CFI_is_contiguous: Base address of C Descriptor " "is already NULL.\n"); ! return CFI_ERROR_BASE_ADDR_NULL; } /* Must be an array. */ --- 424,430 ---- { fprintf (stderr, "CFI_is_contiguous: Base address of C Descriptor " "is already NULL.\n"); ! return 0; } /* Must be an array. */ *************** int CFI_is_contiguous (const CFI_cdesc_t *** 432,444 **** { fprintf (stderr, "CFI_is_contiguous: C Descriptor must describe an " "array (0 < dv->rank = %d).\n", dv->rank); ! return CFI_INVALID_RANK; } } /* Assumed size arrays are always contiguous. */ if (dv->rank > 0 && dv->dim[dv->rank - 1].extent == -1) ! return CFI_SUCCESS; /* If an array is not contiguous the memory stride is different to the element * length. */ --- 432,444 ---- { fprintf (stderr, "CFI_is_contiguous: C Descriptor must describe an " "array (0 < dv->rank = %d).\n", dv->rank); ! return 0; } } /* Assumed size arrays are always contiguous. */ if (dv->rank > 0 && dv->dim[dv->rank - 1].extent == -1) ! return 1; /* If an array is not contiguous the memory stride is different to the element * length. */ *************** int CFI_is_contiguous (const CFI_cdesc_t *** 447,461 **** if (i == 0 && dv->dim[i].sm == (CFI_index_t)dv->elem_len) continue; else if (i > 0 ! && dv->dim[i].sm == (CFI_index_t)(dv->elem_len * dv->dim[i - 1].extent)) continue; ! return CFI_FAILURE; } /* Array sections are guaranteed to be contiguous by the previous test. */ ! return CFI_SUCCESS; } --- 447,461 ---- if (i == 0 && dv->dim[i].sm == (CFI_index_t)dv->elem_len) continue; else if (i > 0 ! && dv->dim[i].sm == (CFI_index_t)(dv->dim[i - 1].sm * dv->dim[i - 1].extent)) continue; ! return 0; } /* Array sections are guaranteed to be contiguous by the previous test. */ ! return 1; } *************** int CFI_section (CFI_cdesc_t *result, co *** 670,676 **** } int idx = i - aux; result->dim[idx].lower_bound = lower[i]; ! result->dim[idx].extent = upper[i] - lower[i] + 1; result->dim[idx].sm = stride[i] * source->dim[i].sm; /* Adjust 'lower' for the base address offset. */ lower[idx] = lower[idx] - source->dim[i].lower_bound; --- 670,676 ---- } int idx = i - aux; result->dim[idx].lower_bound = lower[i]; ! result->dim[idx].extent = 1 + (upper[i] - lower[i])/stride[i]; result->dim[idx].sm = stride[i] * source->dim[i].sm; /* Adjust 'lower' for the base address offset. */ lower[idx] = lower[idx] - source->dim[i].lower_bound;
The most part of this patch is concerned with implementing calls from C of of fortran bind c procedures with assumed shape or assumed rank dummies to completely fix PR89843. The conversion of the descriptors from CFI to gfc occur on entry to and reversed on exit from the procedure. This patch is safe for trunk, even at this late stage, because its effects are barricaded behind the tests for CFI descriptors. I believe that it appropriately rewards the bug reporters to have this feature work as well as possible at release. Between comments and the ChangeLogs, this patch is self explanatory. Bootstrapped and regtested on FC29/x86_64 - OK for trunk? Paul 2019-04-09 Paul Thomas <pault@gcc.gnu.org> PR fortran/89843 * trans-decl.c (gfc_get_symbol_decl): Assumed shape and assumed rank dummies of bind C procs require deferred initialization. (convert_CFI_desc): New procedure to convert incoming CFI descriptors to gfc types and back again. (gfc_trans_deferred_vars): Call it. * trans-expr.c (gfc_conv_gfc_desc_to_cfi_desc): Null the CFI descriptor pointer. Free the descriptor in all cases. PR fortran/90022 * trans-decl.c (gfc_get_symbol_decl): Make sure that the se expression is a pointer type before converting it to the symbol backend_decl type. 2019-04-09 Paul Thomas <pault@gcc.gnu.org> PR fortran/89843 * gfortran.dg/ISO_Fortran_binding_4.f90: Modify the value of x in ctg. Test the conversion of the descriptor types in the main program. * gfortran.dg/ISO_Fortran_binding_10.f90: New test. * gfortran.dg/ISO_Fortran_binding_10.c: Called by it. PR fortran/90022 * gfortran.dg/ISO_Fortran_binding_1.c: Correct the indexing for the computation of 'ans'. Also, change the expected results for CFI_is_contiguous to comply with standard. * gfortran.dg/ISO_Fortran_binding_1.f90: Correct the expected results for CFI_is_contiguous to comply with standard. * gfortran.dg/ISO_Fortran_binding_9.f90: New test. * gfortran.dg/ISO_Fortran_binding_9.c: Called by it. 2019-04-09 Paul Thomas <pault@gcc.gnu.org> PR fortran/89843 * runtime/ISO_Fortran_binding.c (cfi_desc_to_gfc_desc): Only return immediately if the source pointer is null. Bring forward the extraction of the gfc type. Extract the kind so that the element size can be correctly computed for sections and components of derived type arrays. Remove the free of the CFI descriptor since this is now done in trans-expr.c. (gfc_desc_to_cfi_desc): Only allocate the CFI descriptor if it is not null. (CFI_section): Normalise the difference between the upper and lower bounds by the stride to correctly calculate the extents of the section. PR fortran/90022 * runtime/ISO_Fortran_binding.c (CFI_is_contiguous) : Return 1 for true and 0 otherwise to comply with the standard. Correct the contiguity check for rank 3 and greater by using the stride measure of the lower dimension rather than the element length.