[Bug bootstrap/111021] [14 Regression] Serial build broken for CRIS, ARM, and others
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111021 --- Comment #11 from Hans-Peter Nilsson --- (In reply to Hans-Peter Nilsson from comment #8) > I might have misunderstood things of course, ...like still having to include tree.h to get the code_helper class definition. Doh!
[Bug bootstrap/111021] [14 Regression] Serial build broken for CRIS, ARM, and others
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111021 --- Comment #10 from Kewen Lin --- (In reply to Hans-Peter Nilsson from comment #7) > Exactly; I'm happy that we seem to be on the same page here. > > I'm testing a patch for CRIS (making the hook function just a wrapper, > reverting the cris-protos.h change) that may be re-usable for the other > targets similarly affected. > https://gcc.gnu.org/pipermail/gcc-patches/2023-August/627444.html Thanks for the suggestion and doing the patch! If Richi & Richard prefer this, I'll follow up with a patch for the other affected ports. > I believe you could add a code_helper constructor for another type, maybe > even "int"? Then the "default" value for the code_helper argument for the > legitimate address hook function would just look like "code_helper = 0". > > Maybe that's too error-prone and a pointer-type or some entirely different > type would be better. I might have misunderstood things of course, but > that's what I mean by "or something". OK, thanks for further clarifying, I agree it's doable, I didn't consider this before because it's a "code" helper, as its comments says it's for tree codes and builtin function codes, shouldn't make it support things that isn't "code". Even for a int which can be equivalent to a tree_code (with an appropriate assertion check for unexpected range), it's still easy to be questioned like why not just use tree_code directly (having good readability and without any expected range checking etc.).
[Bug bootstrap/111021] [14 Regression] Serial build broken for CRIS, ARM, and others
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111021 --- Comment #9 from Hans-Peter Nilsson --- https://gcc.gnu.org/pipermail/gcc-patches/2023-August/627444.html
[Bug fortran/101602] [F2018] local and local_init are not supported in DO CONCURRENT
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101602 --- Comment #3 from Michael Klemm --- The locality specifiers cannot directly map to the OpenMP data-sharing clauses. While SHARED and LOCAL can be mapped, LOCAL_INIT cannot. The latter needs to initialize the variable anew for each iteration of the DO CONCURRENT loop, while FIRSTPRIVATE will initialize the variable only once per thread that executes chunks of said loop. So, the code transformation for that case will have to be more involved. There has been discussions in the OpenMP language committee if a LINEAR(x:0) clause can substitute LOCAL_INIT(x). That might be one solution to reduce the implementation burden and map everything to OpenMP constructs.
[Bug bootstrap/111021] [14 Regression] Serial build broken for CRIS, ARM, and others
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111021 --- Comment #8 from Hans-Peter Nilsson --- (In reply to Kewen Lin from comment #6) > > btw, thanks for the suggestion on "defaulting to NULL or something" in ML, > but I guess a default value like zero rather than ERROR_MARK has to base on > the assertion that the default value is equal to ERROR_MARK, IMHO it isn't > quite maintainable. I believe you could add a code_helper constructor for another type, maybe even "int"? Then the "default" value for the code_helper argument for the legitimate address hook function would just look like "code_helper = 0". Maybe that's too error-prone and a pointer-type or some entirely different type would be better. I might have misunderstood things of course, but that's what I mean by "or something".
[Bug bootstrap/111021] [14 Regression] Serial build broken for CRIS, ARM, and others
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111021
--- Comment #7 from Hans-Peter Nilsson ---
(In reply to Kewen Lin from comment #6)
> (In reply to Hans-Peter Nilsson from comment #5)
> [...]
> but the concern is that one day some file just includes tm_p.h but
> not recog.h, the issue will show up again. I'm thinking how to deal with the
> problems on ${port}-protos.h, adding $(TREE_H) in TM_P is an alternative,
> but not sure if people would think it's an overkill since the current
> affected ports are:
> [...]
Exactly; I'm happy that we seem to be on the same page here.
I'm testing a patch for CRIS (making the hook function just a wrapper,
reverting the cris-protos.h change) that may be re-usable for the other targets
similarly affected.
[Bug fortran/111022] New: ES0.0E0 format gave ES0.dE0 output with d too high.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111022 Bug ID: 111022 Summary: ES0.0E0 format gave ES0.dE0 output with d too high. Product: gcc Version: 12.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: fortran Assignee: unassigned at gcc dot gnu.org Reporter: john.harper at vuw dot ac.nz Target Milestone: --- This 3-line Fortran program: print "(ES0.0E0)", -666e0 print "(ES0.0E0)", -666d0 end program printed -6.66000E+2 -6.66000E+2 but I think it should have printed -7.E+2 -7.E+2 which is what ifort printed with the same program.
[Bug bootstrap/111021] [14 Regression] Serial build broken for CRIS, ARM, and others
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111021
Kewen Lin changed:
What|Removed |Added
CC||rguenth at gcc dot gnu.org,
||rsandifo at gcc dot gnu.org
--- Comment #6 from Kewen Lin ---
(In reply to Hans-Peter Nilsson from comment #5)
> (In reply to Hans-Peter Nilsson from comment #4)
> > (In reply to Andrew Pinski from comment #1)
> > > https://gcc.gnu.org/pipermail/gcc-patches/2023-August/627435.html
> >
> > Nope, that just fixes recog.h. See the quoted error; it's the patched
> > *-protos.h.
Thanks for clarifying, yeah, some ${port}-protos.h includes tree.h in that
commit.
>
> ...but I'm "guessing" that it will appear to work even for the
> tm-protos.h-on-tree.h dependency, just incidentally.
It can work for this build failure on build/gencondmd.o since the same required
dependency on tree.h is supported by checking the dependence of recog.h, but
the concern is that one day some file just includes tm_p.h but not recog.h, the
issue will show up again. I'm thinking how to deal with the problems on
${port}-protos.h, adding $(TREE_H) in TM_P is an alternative, but not sure if
people would think it's an overkill since the current affected ports are:
gcc/config/arm/arm-protos.h:#include "tree.h" /* For ERROR_MARK. */
gcc/config/cris/cris-protos.h:#include "tree.h" /* For ERROR_MARK. */
gcc/config/microblaze/microblaze-protos.h:#include "tree.h" /* For ERROR_MARK.
*/
gcc/config/rl78/rl78-protos.h:#include "tree.h" /* For ERROR_MARK. */
gcc/config/stormy16/stormy16-protos.h:#include "tree.h" /* For ERROR_MARK. */
I checked the existing tm_p_file_list and tm_p_include_list which are specific
for port, but tm_p_include_list is generated based on tm_p_file_list, it seems
not for this purpose.
Hi Richi & Richard, what do you think of this?
btw, thanks for the suggestion on "defaulting to NULL or something" in ML, but
I guess a default value like zero rather than ERROR_MARK has to base on the
assertion that the default value is equal to ERROR_MARK, IMHO it isn't quite
maintainable.
[Bug bootstrap/111021] [14 Regression] Serial build broken for CRIS, ARM, and others
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111021 --- Comment #5 from Hans-Peter Nilsson --- (In reply to Hans-Peter Nilsson from comment #4) > (In reply to Andrew Pinski from comment #1) > > https://gcc.gnu.org/pipermail/gcc-patches/2023-August/627435.html > > Nope, that just fixes recog.h. See the quoted error; it's the patched > *-protos.h. ...but I'm "guessing" that it will appear to work even for the tm-protos.h-on-tree.h dependency, just incidentally.
[no subject]
ทุนหมุนเวียนธุรกิจระยะสั้นสำหรับ ผู้ประกอบการ โรงงานฯ หจก. บริษัท ธุรกิจ SMEs ง่ายกว่าธนาคาร | ไม่เช็คบูโร | ลดต้น ลดดอกเบี้ย | ไม่ต้องค้ำ | คุยง่าย อนุมัติไวเร็ว 📞โทร 082 5928519 คุณเอก 📞โทร 063 2543219 ตะวัน 💬ไลน์ไอดี esc.credit ✅ดอกเบี้ยต่ำสุด 1.25%* ✅ปิดยอดได้ตลอดเวลา ไม่ต้องรอให้ครบสัญญา ✅ฟรีค่าธรรมเนียม ไม่เรียกเก็บเงินก่อนทำสัญญาทุกกรณี
[Bug bootstrap/111021] [14 Regression] Serial build broken for CRIS, ARM, and others
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111021 --- Comment #4 from Hans-Peter Nilsson --- (In reply to Andrew Pinski from comment #1) > https://gcc.gnu.org/pipermail/gcc-patches/2023-August/627435.html Nope, that just fixes recog.h. See the quoted error; it's the patched *-protos.h.
[Bug bootstrap/111021] [14 Regression] Serial build broken for CRIS, ARM, and others
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111021 --- Comment #3 from Hans-Peter Nilsson --- (In reply to Kewen Lin from comment #2) > Thanks for reporting, I think the culprit is r14-3093 instead of r14-3092? Not for this PR!
[Bug bootstrap/111021] [14 Regression] Serial build broken for CRIS, ARM, and others
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111021 Kewen Lin changed: What|Removed |Added Status|UNCONFIRMED |ASSIGNED URL||https://gcc.gnu.org/piperma ||il/gcc-patches/2023-August/ ||627435.html Ever confirmed|0 |1 Last reconfirmed||2023-08-15 Assignee|unassigned at gcc dot gnu.org |linkw at gcc dot gnu.org --- Comment #2 from Kewen Lin --- Thanks for reporting, I think the culprit is r14-3093 instead of r14-3092? I think the other build/gen*.cc building don't have this issue, since none of them includes recog.h themselves, but emit the source files which includ recog.h (those insn-*.cc, I had a double check, they are fine). build/gencondmd.cc is special, unlike the other build/gen*.cc, it's generated by genconditions and it include "recog.h". Only build/gencondmd.o depends on RECOG_H.
[Bug target/111020] RFE: RISC-V: ability to cherry-pick additional instructions
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111020 --- Comment #5 from H. Peter Anvin --- I don't think source code modifications are a huge problem, but at this point they require tracking down each individual bit. As far as trapping implementations are concerned: 1. In deeply embedded implementations, it is entirely possible that firmware/microcode might be *more* expensive than logic. Although memory arrays are, of course, very dense, they are still extremely general and RISC-V isn't a very sparse instruction set. 2. It seems like it almost would require an implementation-specific performance model. Now, one can validly argue that by setting the cost of unimplemented instructions to a (near-)infinite value such instructions should never be generated even if they are "enabled". That might also be a possible avenue for achieving this. As far as an explosion of subsets, yes, this is really what this means. Bloating a tiny on-chip control processor both in area and timing to implement instructions that never actually appears in the code is at best painful. That being said, I do intend to submit a proposal to the RISC-V ISA folks to subset the Zbb subset. It is worth noting that there are overlaps between the Zb* and Zbk* subsets, but the individual intersection sets do not have their own names. The Zbb instruction set is particularly noxious (and this is indeed an ISA definition problem), because it implements multiple things that are, from an implementation point of view, completely separate and require separate code paths in the ALU: § 1.2.1 Logical with negate - minimal cost; in fact in some implementations it might have zero or even negative cost due to decoder simplification. - Extremely common in embedded operations. § 1.2.2 Count leading/trailing zero bits - Requires dedicated logic. - ctz and clz have very different uses. - Typically clz and ctz will not be able to share logic, either, requiring *two* dedicated units. § 1.2.3 Count population - Requires dedicated logic. - May be useless depending on what the processor needs. § 1.2.4 Integer minimum/maximum - May be cheap or expensive, depending on if an existing comparator can be leveraged. - Quite possibly free or almost free if the AMO instruction set is already supported in its entirety, as that requires max/min already. § 1.2.5 Sign- and zero-extension § 1.2.6 Bitwise rotation - May be very cheap or quite expensive, depending on the implementation of the shift instructions. § 1.2.7 OR combine - Requires dedicated logic. - Virtually useless in control processors that do not process text. § 1.2.8 Byte-reverse - Requires dedicated logic. - These, and some other instructions, are special cases of a bit swap extension proposed in the original bitmanip proposal, but was not included even as a separate set. - Virtually useless in control processors that does not need to interface with cross-endian data. These 8 groups really ought to be given separate names. Is this going to happen again? Quite likely. It seems, as you say, that chopping the public ISA to pieces to support every single use case would seem unlikely. It really comes down to: out of multiple suboptimal cases (forced hardware bloat, custom subsets, extremely fine grained public subsets, vendor-hacked trees that lag behind and/or diverge from upstream), what option is the least amount of badness?
[Bug bootstrap/111021] [14 Regression] Serial build broken for CRIS, ARM, and others
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111021 --- Comment #1 from Andrew Pinski --- https://gcc.gnu.org/pipermail/gcc-patches/2023-August/627435.html
[Bug libfortran/105456] Child I/O does not propage iostat
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105456 Jerry DeLisle changed: What|Removed |Added Ever confirmed|0 |1 Status|UNCONFIRMED |NEW Assignee|unassigned at gcc dot gnu.org |jvdelisle at gcc dot gnu.org CC||jvdelisle at gcc dot gnu.org Last reconfirmed||2023-08-15 --- Comment #3 from Jerry DeLisle --- On my list.
[Bug bootstrap/111021] New: [14 Regression] Serial build broken for CRIS, ARM, and others
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111021 Bug ID: 111021 Summary: [14 Regression] Serial build broken for CRIS, ARM, and others Product: gcc Version: 14.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: bootstrap Assignee: unassigned at gcc dot gnu.org Reporter: hp at gcc dot gnu.org CC: linkw at gcc dot gnu.org Target Milestone: --- Since r14-3092-g165b1f6ad1d396 "targhooks: Extend legitimate_address_p with code_helper [PR110248]", build has been broken (at least with "GNU Make 4.3") for *serial* builds (no "-j" option) for those target where that commit added an include of tree.h was added, exposed by building for example cris-elf and arm-eabi (wrapped lines copy-pasted): g++ -c -g -O2 -DIN_GCC -DCROSS_DIRECTORY_STRUCTURE -fno-exceptions -fno-rtti -fasynchronous-unwind-tables -W -Wall -Wno-narrowing -Wwrite-strings -Wcast-qual -Wmissing-format-attribute -Wconditionally-supported -Woverloaded-virtual -pedantic -Wno-long-long -Wno-variadic-macros -Wno-overlength-strings -fno-common -DHAVE_CONFIG_H -DGENERATOR_FILE -I. -Ibuild -I/srctop/gcc -I/srctop/gcc/build -I/srctop/gcc/../include -I/srctop/gcc/../libcpp/include \ -o build/gencondmd.o build/gencondmd.cc In file included from /srctop/gcc/tree.h:23, from /srctop/gcc/config/arm/arm-protos.h:26, from ./tm_p.h:5, from build/gencondmd.cc:29: /srctop/gcc/tree-core.h:145:10: fatal error: all-tree.def: No such file or directory 145 | #include "all-tree.def" | ^~ compilation terminated. make[2]: *** [Makefile:2918: build/gencondmd.o] Error 1 At a glance, (at least) build/gencondmd.o lacks a dependency on the generated all-tree.def. But adding only that dependency leads to a similar error for a missing of tree-check.h; looks like TREE_H contains the right bits. Maybe it's as "simple" as including $(TREE_H) in TM_P (or a new intermediary macro) or add $(TREE_H) to all build/gen*.o rules where the gen*.cc includes "tm_p.h" (but within the filter-out for build/gencondmd.o). I *did* try the latter; simply adding $(TREE_H) (within the filter-out call) for build/gencondmd.o and that seemed to work, but it doesn't seem *right* and may work just by happenstance, as other build/gen*.o seems like they could fail given unfortunate timing or Makefile tweak. It seems proper to leave that headache to the author of that the offending commit. (I'm going to bail out for CRIS with a patch to cris-protos.h, wrapping the function that needs to be prototyped so it doesn't need to refer to ERROR_MARK.)
[Bug target/111020] RFE: RISC-V: ability to cherry-pick additional instructions
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111020
palmer at gcc dot gnu.org changed:
What|Removed |Added
CC||palmer at gcc dot gnu.org
--- Comment #4 from palmer at gcc dot gnu.org ---
(In reply to Andrew Pinski from comment #3)
> (In reply to H. Peter Anvin from comment #2)
> > Named subsets are, inherently, designed to make sense toward mass-produced
> > products where the hardware and software are designed (mostly)
> > independently. However, what I mean with "very deep embedded use" is
> > hardware and software being co-designed.
> >
> > The RISC-V ISA policy is that those are considered vendor-specific subsets
> > and are to be given an X* name; however, gcc obviously needs to be able to
> > understand the meaning of this X* name. At this point there is no way to do
> > without changing the source code in nontrivial ways.
> >
> > Regardless of if it is done in source code or at runtime, by implementing a
> > fine-grained, preferably table-driven, approach to subsets in gcc then it
> > would be very simple for a hardware implementor to define their custom
> > X-subsets without a lot of surgery to the code, *and* it makes it possible
> > to take it one step further and allowing custom (or newly defined! - there
> > have been multiple instances already of new subsets of existing instructions
> > defined a posteori) instruction subsets to be defined in a configuration
> > file.
>
> I am 100% disagree here. Because if you do this there would be a huge
> explosion of what is and is not considered a subset. THIS is why it should
> be defined at the ISA level instead. Why just CTZ for ZBB what next just
> bseti or bexti of ZBS?
>
> defining the specific set during your development is different from a
> production compiler really. GCC should aim for production compiler quality
> even for highly embedded targets.
IMO adding some config file for custom subsets is going to make more headaches
than it fixes. For a while we had args like "-mno-div", but that's kind of
hacky and we eventually ended up with Zmmul to handle it -- having an external
config file controlling this would expose a lot of interface surface we don't
have a sane way to test.
If vendors want a custom subset then they can make one, it'll just be called
"X${vendor}${subset}". We've already got a few forks/subsets floating around,
look at the T-Head and Ventana stuff. For a few instructions it's pretty
mechanical, aside from fixing whatever fallout comes from splitting off the
subset.
We do currently require (IIRC we still didn't write this down) some amount of
public commitment to hardware availability to take that code, but if that's the
problem we should try and figure something out. It's certainly a pain for
vendors to keep in-development trees around, but we're trading that off with
upstream pain -- I've found these sorts of subsets drift around until the HW
actually ships, so we don't want to end up stuck keeping around subsets that
didn't ship.
Vendors also have the option of just implementing all the instructions (via
some trap or microcode or whatever), thus turning this into a performance
problem. That sort of just trades one problem for another, but we've got some
examples of this as well (SiFive traps on a bunch of stuff, for example).
[Bug target/111020] RFE: RISC-V: ability to cherry-pick additional instructions
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111020 --- Comment #3 from Andrew Pinski --- (In reply to H. Peter Anvin from comment #2) > Named subsets are, inherently, designed to make sense toward mass-produced > products where the hardware and software are designed (mostly) > independently. However, what I mean with "very deep embedded use" is > hardware and software being co-designed. > > The RISC-V ISA policy is that those are considered vendor-specific subsets > and are to be given an X* name; however, gcc obviously needs to be able to > understand the meaning of this X* name. At this point there is no way to do > without changing the source code in nontrivial ways. > > Regardless of if it is done in source code or at runtime, by implementing a > fine-grained, preferably table-driven, approach to subsets in gcc then it > would be very simple for a hardware implementor to define their custom > X-subsets without a lot of surgery to the code, *and* it makes it possible > to take it one step further and allowing custom (or newly defined! - there > have been multiple instances already of new subsets of existing instructions > defined a posteori) instruction subsets to be defined in a configuration > file. I am 100% disagree here. Because if you do this there would be a huge explosion of what is and is not considered a subset. THIS is why it should be defined at the ISA level instead. Why just CTZ for ZBB what next just bseti or bexti of ZBS? defining the specific set during your development is different from a production compiler really. GCC should aim for production compiler quality even for highly embedded targets.
[Bug target/111020] RFE: RISC-V: ability to cherry-pick additional instructions
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111020 --- Comment #2 from H. Peter Anvin --- Named subsets are, inherently, designed to make sense toward mass-produced products where the hardware and software are designed (mostly) independently. However, what I mean with "very deep embedded use" is hardware and software being co-designed. The RISC-V ISA policy is that those are considered vendor-specific subsets and are to be given an X* name; however, gcc obviously needs to be able to understand the meaning of this X* name. At this point there is no way to do without changing the source code in nontrivial ways. Regardless of if it is done in source code or at runtime, by implementing a fine-grained, preferably table-driven, approach to subsets in gcc then it would be very simple for a hardware implementor to define their custom X-subsets without a lot of surgery to the code, *and* it makes it possible to take it one step further and allowing custom (or newly defined! - there have been multiple instances already of new subsets of existing instructions defined a posteori) instruction subsets to be defined in a configuration file.
[Bug fortran/110996] RISC-V vector Fortran: SEGV ICE during parsing
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110996 --- Comment #4 from JuzheZhong --- (In reply to Jeremy Bennett from comment #3) > @JuzheZhong I believe this is in someway related to RVV. If I remove `v' > from the march: > > riscv64-unknown-linux-gnu-gfortran -march=rv64gc -mabi=lp64d -c -Ofast > testcase.f90 > > The output I get is correct: > > testcase.f90:6:20: > > 6 |SUBROUTINE c(d) e > |1 > Error: Syntax error in SUBROUTINE statement at (1) > > Why does adding `v' to the -march string cause a SEGV? I didn't reproduce the issue. What I see is GCC has ICE even without 'v' in -march. And I have no idea for it.
[Bug target/110985] RISC-V: Incorrect code gen for RVV VLS
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110985 Li Pan changed: What|Removed |Added Status|NEW |RESOLVED Resolution|--- |FIXED --- Comment #4 from Li Pan --- Close this bug as committed already.
[Bug c/96952] __builtin_thread_pointer support cannot be probed
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96952 H. Peter Anvin changed: What|Removed |Added CC||hpa at zytor dot com --- Comment #10 from H. Peter Anvin --- Is this bug still relevant? RISC-V doesn't even seem to support disabling tls support, and __builtin_thread_pointer() appears to be properly supported. So it would presumably be up to any remaining target that doesn't have __builtin_thread_pointer() (or not in all configurations) to verify that __has_builtin(__builtin_thread_pointer) evaluates to false?
[Bug target/111020] RFE: RISC-V: ability to cherry-pick additional instructions
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111020 --- Comment #1 from Andrew Pinski --- This sounds more like something which should be designed on at ISA level and since RISC-V is an open source ISA, it should be discussed at that level ... There are already extensions which are designed this way too. E.g. Zmmul which is a subset of the M extension.
[Bug target/111020] New: RFE: RISC-V: ability to cherry-pick additional instructions
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111020 Bug ID: 111020 Summary: RFE: RISC-V: ability to cherry-pick additional instructions Product: gcc Version: unknown Status: UNCONFIRMED Severity: normal Priority: P3 Component: target Assignee: unassigned at gcc dot gnu.org Reporter: hpa at zytor dot com Target Milestone: --- For very deeply embedded use, it is sometimes highly desirable to control the instruction set on a very fine grained basis. For example, the Zbb extension contains a mixture of things that most likely requires separate functional units. However, as an example, the ctz instruction is highly useful to speed up interrupt latency in designs that do not have vectorized interrupt handling (which is, in its most basic form, a dedicated ctz unit.) It would be massive hardware bloat to require the full Zbb set to add this one instruction. Once the instruction is added, though, one would like to be able to use it as fully as possible. This, obviously, creates binaries that are specifically tuned toward a single processor implementation, but that is pretty much the essence of deeply embedded, where in the normal case the entire software stack from the OS to application is linked together in a single binary, or at the very least compiled together, often from a single source tree. As far as object code compatibility is concerned, this is very much a "programmer beware" situation. There is no need for heroics in terms of tagging objects with the exact instruction set, for example.
[Bug middle-end/110986] [14 Regression] aarch64 has support for conditional not (and vectorized conditional not ) after r14-3110-g7fb65f10285
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110986 --- Comment #17 from Andrew Pinski --- For the scalar version we could match: (set (reg:SI 106) (xor:SI (neg:SI (ne:SI (reg:CC 66 cc) (const_int 0 [0]))) (reg:SI 107 [ MEM[(int *)a_14(D) + ivtmp.14_9 * 1] ]))) In the backend to do the csinv but it might not catch all of them because combine does not match everything.
[Bug tree-optimization/103035] [meta-bug] YARPGen bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103035 Bug 103035 depends on bug 110954, which changed state. Bug 110954 Summary: [14 Regression] Wrong code with -O0 https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110954 What|Removed |Added Status|ASSIGNED|RESOLVED Resolution|--- |FIXED
[Bug middle-end/110954] [14 Regression] Wrong code with -O0
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110954 Andrew Pinski changed: What|Removed |Added Resolution|--- |FIXED Status|ASSIGNED|RESOLVED --- Comment #8 from Andrew Pinski --- Fixed.
[Bug c++/111019] New: Optimizer incorrectly assumes variable is not changed while change happens through another pointer
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111019
Bug ID: 111019
Summary: Optimizer incorrectly assumes variable is not changed
while change happens through another pointer
Product: gcc
Version: 12.3.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: boskidialer at gmail dot com
Target Milestone: ---
Created attachment 55737
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=55737&action=edit
Smallest reproduction i managed to create
Hello,
I was investigating one of the tests failures in the product, test failure that
only happens while compiling with -O3 or -O2, but one that does not happen with
-O1 or when not using any optimization.
GCC Version:
dashboard@dashboard-desktop:~$ /usr/bin/g++ -v
Using built-in specs.
COLLECT_GCC=/usr/bin/g++
COLLECT_LTO_WRAPPER=/usr/lib/gcc/x86_64-linux-gnu/12/lto-wrapper
OFFLOAD_TARGET_NAMES=nvptx-none:amdgcn-amdhsa
OFFLOAD_TARGET_DEFAULT=1
Target: x86_64-linux-gnu
Configured with: ../src/configure -v --with-pkgversion='Ubuntu
12.3.0-1ubuntu1~23.04' --with-bugurl=file:///usr/share/doc/gcc-12/README.Bugs
--enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++,m2 --prefix=/usr
--with-gcc-major-version-only --program-suffix=-12
--program-prefix=x86_64-linux-gnu- --enable-shared --enable-linker-build-id
--libexecdir=/usr/lib --without-included-gettext --enable-threads=posix
--libdir=/usr/lib --enable-nls --enable-clocale=gnu --enable-libstdcxx-debug
--enable-libstdcxx-time=yes --with-default-libstdcxx-abi=new
--enable-gnu-unique-object --disable-vtable-verify --enable-plugin
--enable-default-pie --with-system-zlib --enable-libphobos-checking=release
--with-target-system-zlib=auto --enable-objc-gc=auto --enable-multiarch
--disable-werror --enable-cet --with-arch-32=i686 --with-abi=m64
--with-multilib-list=m32,m64,mx32 --enable-multilib --with-tune=generic
--enable-offload-targets=nvptx-none=/build/gcc-12-DAPbBt/gcc-12-12.3.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-12-DAPbBt/gcc-12-12.3.0/debian/tmp-gcn/usr
--enable-offload-defaulted --without-cuda-driver --enable-checking=release
--build=x86_64-linux-gnu --host=x86_64-linux-gnu --target=x86_64-linux-gnu
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 12.3.0 (Ubuntu 12.3.0-1ubuntu1~23.04)
Reproduction:
/usr/bin/g++ gcc-err.cpp -O3 -o gcc-err.out && ./gcc-err.out
(gcc-err.cpp is provided as the attachment to the bug report).
Issue is that generated output freezes when compiled it under -O3 or -O2 but
not when compiling under -O1 or without any optimizations.
Just in case to verify the issue is not on my end, i pasted the reproduction
code and required compiler flags onto a godbolt:
https://godbolt.org/z/Ez7vrz77W - and it shows that the compiled program times
out. This is a confirmation that the generated output is stuck. After changing
the compiler options on the right side on the godbolt site to -O1, the code
compiles as well but the executable now correctly finishes within time limit
and outputs a single line "test".
Based on the debugging i did on this code, it looks to be related to the
Target::~Target code where there is the `whlie (this->next)` loop where i
suspect compiler or optimizer incorrectly assumes that value of `this->next` is
unchanged between iterations however that is not true because in this case
there is `n` variable set to `this->next` which points to a second item in the
double linked list, which means `n->previous == this` and as such
`n->previous->next = ...` line is effectively changing value of the
`this->next`, but indirectly.
When generating the assembly from the given reproduction using `/usr/bin/g++
-masm=intel gcc-err.cpp -O3 -S -o gcc-err.S`, instructions produced seem to be
incorrect as they are missing the repeated checks if the value of `this->next`
was changed in the next iteration:
.L21:
mov rcx, QWORD PTR [rax]
mov rdx, QWORD PTR 8[rax]
testrcx, rcx
je .L19 // if (n->previous)
mov QWORD PTR 8[rcx], rdx
mov rdx, QWORD PTR 8[rax] // n->previous->next = n->next;
.L19:
testrdx, rdx
je .L20 // if (n->next)
mov QWORD PTR [rdx], rcx // n->next->previous = n->previous;
.L20:
xor edx, edx
movups XMMWORD PTR [rax], xmm0
mov QWORD PTR 16[rax], rdx
jmp .L21
When any external function calls, barrier instructions (like 'asm
volatile("":::"memory")') or more complex code is added, the loop produces the
correct code:
.L18:
mov rax, QWORD PTR 8[rbx]
testrax, rax
je .L74 // quits the loop if `this->next ==
nullptr`
mov rcx, QWORD PTR [rax]
mov rdx, QWORD PTR 8[rax]
testrcx, rcx
[Bug c/102989] Implement C2x's n2763 (_BitInt)
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102989 Trevor Gross changed: What|Removed |Added CC||tmgross at umich dot edu --- Comment #95 from Trevor Gross --- Just as a heads up: there is an ongoing conversation at the x86 psABI about adjusting `_BitInt(128)` to have the same alignment as `__int128`, which would help address some of the issues mentioned here. Please join in the discussion if you have any comments: https://groups.google.com/g/x86-64-abi/c/-JeR9HgUU20
[Bug c++/111018] lexical interpretation of friendship rules depends on whether the friend function has a definition
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111018 Andrew Pinski changed: What|Removed |Added See Also||https://gcc.gnu.org/bugzill ||a/show_bug.cgi?id=58993 --- Comment #4 from Andrew Pinski --- I am going to leave this one open only because of what is mentioned in the commit that changed behavior here: r13-465-g4df735e01e3199978
[Bug c++/111018] lexical interpretation of friendship rules depends on whether the friend function has a definition
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111018 --- Comment #3 from Andrew Pinski --- https://wg21.link/cwg1699
[Bug c++/111018] lexical interpretation of friendship rules depends on whether the friend function has a definition
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111018 Andrew Pinski changed: What|Removed |Added Keywords|rejects-valid | --- Comment #2 from Andrew Pinski --- MSVC rejects both cases
[Bug c++/111018] lexical interpretation of friendship rules depends on whether the friend function has a definition
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111018
--- Comment #1 from Andrew Pinski ---
So if we make a definition outside of the class:
```
template
auto bar(Self s) -> decltype(::T::foo(s))
{
}
```
Then clang rejects the code too ...
So
[Bug target/106671] aarch64: BTI instruction are not inserted for cross-section direct calls
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106671 --- Comment #11 from Jiangning Liu --- Hi Wilco, > "it means we will need a linker optimization to remove those redundant BTIs > (eg. by changing them into NOPs)" It will be only for performance optimization, right? If we don't care about performance, the linker doesn't need to optimize it to be NOP, right? It could still be useful if we only do this operation for a specific module. Thanks, -Jiangning
[Bug c++/111018] New: lexical interpretation of friendship rules depends on whether the friend function has a definition
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111018
Bug ID: 111018
Summary: lexical interpretation of friendship rules depends on
whether the friend function has a definition
Product: gcc
Version: 14.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: eric.niebler at gmail dot com
Target Milestone: ---
Starting in gcc 12, transitive friendship is extended to friend functions that
are defined lexically within the body of the friend class. E.g.:
struct S;
struct T {
friend struct S;
private:
template
static void foo(T) {}
};
struct S {
template
friend auto bar(Self s) -> decltype(::T::foo(s)) { // (1)
return ::T::foo(s);
}
};
Prior to gcc-12, the commented line would have been rejected, but now it is
accepted. Great, it brings gcc in line with clang and is arguably more
sensible.
HOWEVER, it does NOT work if the friend function is merely declared but not
defined. For instance, this is still an error:
struct S;
struct T {
friend struct S;
private:
template
static void foo(T) {}
};
struct S {
template
friend auto bar(Self s) -> decltype(::T::foo(s)); // NO FN DEFINITION
};
int main() {
S s;
using T = decltype(bar(s)); // ERROR: T::foo is private
}
This is very confusing and inconsistent behavior.
See: https://godbolt.org/z/WT9P37Wba
[Bug fortran/101602] [F2018] local and local_init are not supported in DO CONCURRENT
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101602 Marshall Ward changed: What|Removed |Added CC||marshall.ward at gmail dot com --- Comment #2 from Marshall Ward --- I've tested this in 13.0.0 and it appears that `local`, `local_init`, `shared` are still not supported in `do concurrent` and produce syntax errors. Has there been any activity on this issue recently? If not, could anyone comment on the proposal by Jeff Hammond about utilizing the analogous OpenMP constructs? Would this be a feasible option?
[Bug fortran/110360] ABI issue with character,value dummy argument
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110360 --- Comment #39 from CVS Commits --- The master branch has been updated by Mikael Morin : https://gcc.gnu.org/g:564b637f4a32883cbf3c3019d3cfcf0b0aec9b82 commit r14-3207-g564b637f4a32883cbf3c3019d3cfcf0b0aec9b82 Author: Mikael Morin Date: Mon Aug 14 21:51:54 2023 +0200 fortran: Fix length one character dummy arg type [PR110419] Revision r14-2171-g8736d6b14a4dfdfb58c80ccd398981b0fb5d00aa changed the argument passing convention for length 1 value dummy arguments to pass just the single character by value. However, the procedure declarations weren't updated to reflect the change in the argument types. This change does the missing argument type update. The change of argument types generated an internal error in gfc_conv_string_parameter with value_9.f90. Indeed, that function is not prepared for bare character type, so it is updated as well. The condition guarding the single character argument passing code is loosened to not exclude non-interoperable kind (this fixes a regression with c_char_tests_2.f03). Finally, the constant string argument passing code is updated as well to extract the single char and pass it instead of passing it as a length one string. As the code taking care of non-constant arguments was already doing this, the condition guarding it is just removed. With these changes, value_9.f90 passes on 32 bits big-endian powerpc. PR fortran/110360 PR fortran/110419 gcc/fortran/ChangeLog: * trans-types.cc (gfc_sym_type): Use a bare character type for length one value character dummy arguments. * trans-expr.cc (gfc_conv_string_parameter): Handle single character case. (gfc_conv_procedure_call): Don't exclude interoperable kinds from single character handling. For single character dummy arguments, extend the existing handling of non-constant expressions to constant expressions. gcc/testsuite/ChangeLog: * gfortran.dg/bind_c_usage_13.f03: Update tree dump patterns.
[Bug testsuite/110419] [14 regression] new test case gfortran.dg/value_9.f90 in r14-2050-gd130ae8499e0c6 fails
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110419 --- Comment #20 from CVS Commits --- The master branch has been updated by Mikael Morin : https://gcc.gnu.org/g:564b637f4a32883cbf3c3019d3cfcf0b0aec9b82 commit r14-3207-g564b637f4a32883cbf3c3019d3cfcf0b0aec9b82 Author: Mikael Morin Date: Mon Aug 14 21:51:54 2023 +0200 fortran: Fix length one character dummy arg type [PR110419] Revision r14-2171-g8736d6b14a4dfdfb58c80ccd398981b0fb5d00aa changed the argument passing convention for length 1 value dummy arguments to pass just the single character by value. However, the procedure declarations weren't updated to reflect the change in the argument types. This change does the missing argument type update. The change of argument types generated an internal error in gfc_conv_string_parameter with value_9.f90. Indeed, that function is not prepared for bare character type, so it is updated as well. The condition guarding the single character argument passing code is loosened to not exclude non-interoperable kind (this fixes a regression with c_char_tests_2.f03). Finally, the constant string argument passing code is updated as well to extract the single char and pass it instead of passing it as a length one string. As the code taking care of non-constant arguments was already doing this, the condition guarding it is just removed. With these changes, value_9.f90 passes on 32 bits big-endian powerpc. PR fortran/110360 PR fortran/110419 gcc/fortran/ChangeLog: * trans-types.cc (gfc_sym_type): Use a bare character type for length one value character dummy arguments. * trans-expr.cc (gfc_conv_string_parameter): Handle single character case. (gfc_conv_procedure_call): Don't exclude interoperable kinds from single character handling. For single character dummy arguments, extend the existing handling of non-constant expressions to constant expressions. gcc/testsuite/ChangeLog: * gfortran.dg/bind_c_usage_13.f03: Update tree dump patterns.
[Bug fortran/110995] segfault for function in declaration of module function
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110995 anlauf at gcc dot gnu.org changed: What|Removed |Added Last reconfirmed||2023-08-14 Status|UNCONFIRMED |NEW Ever confirmed|0 |1 --- Comment #1 from anlauf at gcc dot gnu.org --- Confirmed. A reduced testcase giving the same failure may give some insight: module strings implicit none contains pure function charArrToString(data) character, intent(in) :: data(:) ! character(size(data)) :: charArrToString ! OK character(ubound(data,1)) :: charArrToString ! ICE if used in main charArrToString = "" end function charArrToString end module strings program test_str use strings implicit none character :: c(2) = (/ 'T', 'T' /) write(*,*) "write module-", charArrToString(c), "-" ! ICE write(*,*) "write module-", charArrToString2(c), "-" ! OK contains pure function charArrToString2(data) character, intent(in) :: data(:) ! character(size(data)) :: charArrToString2 ! OK character(ubound(data,1)) :: charArrToString2 ! OK charArrToString2 = "" end function charArrToString2 end program test_str So (apparently): - if the function definition is in the same compilation unit as where it is used, there is no problem. - if the function interface is passed via a module file, some declarations work (e.g. using the SIZE intrinsic), but some don't (e.g. using a pure user-defined function, or some other intrinsics).
[Bug fortran/87326] [F18] Support the NEW_INDEX= specifier in the FORM TEAM statement
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87326 --- Comment #8 from anlauf at gcc dot gnu.org --- (In reply to Nathan Weeks from comment #7) > (In reply to anlauf from comment #6) > > (In reply to Nathan Weeks from comment #5) > > > (In reply to Brad Richardson from comment #3) > > > > Was there any more progress on this? I've just run into it. > > > > > > > > FYI I'm trying implement a polymorphic send/receive: > > > > https://gitlab.com/everythingfunctional/communicator > > > > > > The FSF copyright assignment ended up being an unexpectedly-difficult > > > hurdle > > > at the time. I could try again if there is interest, though it would also > > > require some effort to rework the original patch for GCC 14. If another > > > contributor were willing to submit a clean-room implementation, that may > > > be > > > more expedient. > > > > Besides the copyright assignment there is also the possibility to use the > > Developer's Certificate of Origin sign-off: > > > > https://gcc.gnu.org/dco.html > > Is that an option in this case? I was originally advised to pursue a > copyright assignment: > > https://gcc.gnu.org/pipermail/fortran/2019-January/051674.html That requirement was changed in 2021: https://gcc.gnu.org/pipermail/gcc/2021-June/236182.html See also: https://gcc.gnu.org/contribute.html
[Bug c++/110513] Invalid use of incomplete type std::bool_constant inside requires expression
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110513 Patrick Palka changed: What|Removed |Added CC||ed at catmur dot uk --- Comment #4 from Patrick Palka --- *** Bug 111016 has been marked as a duplicate of this bug. ***
[Bug c++/111016] Confusing "used in its own initializer" for non-dependent ad-hoc constraint
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111016 Patrick Palka changed: What|Removed |Added Resolution|--- |DUPLICATE Status|UNCONFIRMED |RESOLVED CC||ppalka at gcc dot gnu.org --- Comment #1 from Patrick Palka --- dup of PR110513 I think *** This bug has been marked as a duplicate of bug 110513 ***
[Bug analyzer/110543] RFE: Add optional trim of the analyzer diagnostics through system headers.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110543 Benjamin Priour changed: What|Removed |Added Resolution|--- |FIXED Status|UNCONFIRMED |RESOLVED --- Comment #2 from Benjamin Priour --- Fixed by the above patch
[Bug c++/110127] -fimplicit-constexpr leads to extremely slow and memory intensive compilation
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110127 Patrick Palka changed: What|Removed |Added Resolution|--- |FIXED Target Milestone|--- |13.3 Status|NEW |RESOLVED --- Comment #3 from Patrick Palka --- Fixed for GCC 13.3 after the backport r13-7713-gd3088f0ed25fe7.
[Bug c++/55004] [meta-bug] constexpr issues
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=55004 Bug 55004 depends on bug 110127, which changed state. Bug 110127 Summary: -fimplicit-constexpr leads to extremely slow and memory intensive compilation https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110127 What|Removed |Added Status|NEW |RESOLVED Resolution|--- |FIXED
[Bug c++/111008] '>' in a lambda as a template argument causes a syntax error
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111008
--- Comment #6 from Andrew Pinski ---
Here is another valid C++20 which is also causing issues, this time with `>>`
rather than `>`:
```
typedef int b;
templatestruct F1{};
templatestruct F2{};
F10)}> a;
constexpr int t = 3;
F2>0}>{}> a0;
```
[Bug analyzer/110543] RFE: Add optional trim of the analyzer diagnostics through system headers.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110543 --- Comment #1 from CVS Commits --- The trunk branch has been updated by Benjamin Priour : https://gcc.gnu.org/g:ce8cdf5bcf96a2db6d7b9f656fc9ba58d7942a83 commit r14-3205-gce8cdf5bcf96a2db6d7b9f656fc9ba58d7942a83 Author: benjamin priour Date: Mon Aug 14 17:36:21 2023 +0200 analyzer: New option fanalyzer-show-events-in-system-headers [PR110543] This patch introduces -fanalyzer-show-events-in-system-headers, disabled by default. This option reduces the noise of the analyzer emitted diagnostics when dealing with system headers. The new option only affects the display of the diagnostics, but doesn't hinder the actual analysis. Given a diagnostics path diving into a system header in the form [ prefix events..., system header call, system header entry, events within system headers..., system header return, suffix events... ] then disabling the option (either by default or explicitly) will shorten the path into: [ prefix events..., system header call, system header return, suffix events... ] Signed-off-by: benjamin priour gcc/analyzer/ChangeLog: PR analyzer/110543 * analyzer.opt: Add new option. * diagnostic-manager.cc (diagnostic_manager::prune_path): Call prune_system_headers. (prune_frame): New function that deletes all events in a frame. (diagnostic_manager::prune_system_headers): New function. * diagnostic-manager.h: Add prune_system_headers declaration. gcc/ChangeLog: PR analyzer/110543 * doc/invoke.texi: Add documentation of fanalyzer-show-events-in-system-headers gcc/testsuite/ChangeLog: PR analyzer/110543 * g++.dg/analyzer/fanalyzer-show-events-in-system-headers-default.C: New test. * g++.dg/analyzer/fanalyzer-show-events-in-system-headers-no.C: New test. * g++.dg/analyzer/fanalyzer-show-events-in-system-headers.C: New test.
[Bug libstdc++/110990] `format_to_n` returns wrong value
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110990 Jonathan Wakely changed: What|Removed |Added Resolution|--- |FIXED Status|ASSIGNED|RESOLVED --- Comment #6 from Jonathan Wakely --- Fixed for 13.3, thanks for the report.
[Bug libstdc++/110860] std::format("{:f}",2e304) invokes undefined behaviour
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110860 Jonathan Wakely changed: What|Removed |Added Status|REOPENED|RESOLVED Resolution|--- |FIXED --- Comment #18 from Jonathan Wakely --- Should be fixed for real now. Thanks for the report and the patch. Looks like I managed to mess up the authorship of the gcc-13 backport though (because I squashed together my incorrect commit and your correct one for the backport). I'll fix it in the ChangeLog file after that is regenerated overnight, but it will stay wrong in Git. Sorry.
[Bug libstdc++/110860] std::format("{:f}",2e304) invokes undefined behaviour
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110860 --- Comment #17 from CVS Commits --- The releases/gcc-13 branch has been updated by Jonathan Wakely : https://gcc.gnu.org/g:559341b5b5a30448362f3f205cd1bf043a919945 commit r13-7722-g559341b5b5a30448362f3f205cd1bf043a919945 Author: Jonathan Wakely Date: Fri Aug 11 18:10:29 2023 +0100 libstdc++: Avoid problematic use of log10 in std::format [PR110860] If abs(__v) is smaller than one, the result will be of the form 0.x. It is only if the magnitude is large that more digits are needed before the decimal dot. This uses frexp instead of log10 which should be less expensive and have sufficient precision for the desired purpose. It removes the problematic cases where log10 will be negative or not fit in an int. Signed-off-by: Paul Dreik libstdc++-v3/ChangeLog: PR libstdc++/110860 * include/std/format (__formatter_fp::format): Use frexp instead of log10. (cherry picked from commit 2d2b05f0691799f03062bf5c436462f14cad3e7c)
[Bug libstdc++/110990] `format_to_n` returns wrong value
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110990 --- Comment #5 from CVS Commits --- The releases/gcc-13 branch has been updated by Jonathan Wakely : https://gcc.gnu.org/g:8f82863df33c63693a355d9244c315ef5cb2158e commit r13-7721-g8f82863df33c63693a355d9244c315ef5cb2158e Author: Jonathan Wakely Date: Fri Aug 11 23:02:44 2023 +0100 libstdc++: Fix std::format_to_n return value [PR110990] When writing to a contiguous iterator, std::format_to_n(out, n, ...) always returns out + n, even if it wrote fewer than n characters to the iterator. The problem is in the _M_finish() member function of the _Iter_sink specialization for contiguous iterators. _M_finish() calls _M_overflow() to update its count of characters written, so it can return the count of characters that would be written if there was room. But _M_overflow() assumes it's only called when the buffer is full, and so switches to the internal buffer. _M_finish() then thinks that if the internal buffer is in use, we already wrote at least n characters and so returns out+n as the output position. We can fix the problem by adding a check in _M_overflow() so that we don't update the count and switch to the internal buffer unless we've run out of room, i.e. _M_unused().size() is zero. The caller then needs to be prepared for _M_count not being the final total, and so add _M_used.size() to it. However, there's not actually any need for _M_finish() to call _M_overflow() to get the count. We now need to use _M_count and _M_used.size() to get the total anyway so _M_overflow() doesn't help with that. And we don't need to use _M_overflow() to flush unwritten characters to the output, because the specialization for contiguous iterators always writes directly to the output without buffering (except when we've exceeded the maximum number of characters, in which case we want to discard the buffered characters anyway). So _M_finish() can be simplified and can avoid calling _M_overflow(). This change also fixes some member functions of other sink classes to only call _M_overflow() when there are characters in the buffer, which is needed to meet _M_overflow's precondition that _M_used().size()!=0. libstdc++-v3/ChangeLog: PR libstdc++/110990 * include/std/format (_Seq_sink::get): Only call _M_overflow if its precondition is met. (_Iter_sink::_M_finish): Likewise. (_Iter_sink::_M_overflow): Only switch to the internal buffer after running out of space. (_Iter_sink::_M_finish): Do not use _M_overflow. (_Counting_sink::count): Likewise. * testsuite/std/format/functions/format_to_n.cc: Check cases where the output fits into the buffer. (cherry picked from commit 003016a40844701c48851020df672b70f3446bdb)
[Bug fortran/110677] UBSAN error: load of value 1818451807, which is not a valid value for type 'expr_t' when compiling pr49213.f90
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110677 --- Comment #2 from Martin Jambor --- I have proposed a fix on the mailing list: https://gcc.gnu.org/pipermail/gcc-patches/2023-August/627379.html ...and also posted it to the Fortran mailing list: https://gcc.gnu.org/pipermail/fortran/2023-August/059687.html
[Bug c++/110216] tuple_size requirements for structured binding has not been updated after DR 2386
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110216 --- Comment #4 from CVS Commits --- The trunk branch has been updated by Jason Merrill : https://gcc.gnu.org/g:1a43af04dd62b80f45700f94ed241347263ed773 commit r14-3204-g1a43af04dd62b80f45700f94ed241347263ed773 Author: gnaggnoyil Date: Sat Aug 12 16:16:52 2023 +0800 c++: follow DR 2386 and update implementation of get_tuple_size [PR110216] DR 2386 updated the tuple_size requirements for structured binding and it now requires tuple_size to be considered only if std::tuple_size names a complete class type with member value. GCC before this patch does not follow the updated requrements, and this patch is intended to implement it. (jason) Accepting pseudonym sign-off because a change this small is not legally significant for copyright. DR 2386 PR c++/110216 gcc/cp/ChangeLog: * decl.cc (get_tuple_size): Update implementation for DR 2386. gcc/testsuite/ChangeLog: * g++.dg/cpp1z/decomp10.C: Update expected error for DR 2386. * g++.dg/cpp1z/pr110216.C: New test. Signed-off-by: gnaggnoyil Reviewed-by: Jason Merrill
[Bug libstdc++/110860] std::format("{:f}",2e304) invokes undefined behaviour
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110860 --- Comment #16 from CVS Commits --- The master branch has been updated by Jonathan Wakely : https://gcc.gnu.org/g:2d2b05f0691799f03062bf5c436462f14cad3e7c commit r14-3202-g2d2b05f0691799f03062bf5c436462f14cad3e7c Author: Paul Dreik Date: Mon Aug 14 15:42:33 2023 +0100 libstdc++: Avoid problematic use of log10 in std::format [PR110860] If abs(__v) is smaller than one, the result will be of the form 0.x. It is only if the magnitude is large that more digits are needed before the decimal dot. This uses frexp instead of log10 which should be less expensive and have sufficient precision for the desired purpose. It removes the problematic cases where log10 will be negative or not fit in an int. Signed-off-by: Paul Dreik libstdc++-v3/ChangeLog: PR libstdc++/110860 * include/std/format (__formatter_fp::format): Use frexp instead of log10.
[Bug middle-end/111017] New: [OpenMP] Wrong code with non-rectangular loop nest
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111017
Bug ID: 111017
Summary: [OpenMP] Wrong code with non-rectangular loop nest
Product: gcc
Version: 14.0
Status: UNCONFIRMED
Keywords: openmp, wrong-code
Severity: normal
Priority: P3
Component: middle-end
Assignee: unassigned at gcc dot gnu.org
Reporter: burnus at gcc dot gnu.org
CC: jakub at gcc dot gnu.org
Target Milestone: ---
At a glance, the following looks fine to me - but it fails with UBSAN as
follows:
test.c:10:15: runtime error: index -42 out of bounds for type 'int [1024]'
test.c:10:8: runtime error: index -42 out of bounds for type 'int [1024]'
=
==619433==ERROR: AddressSanitizer: stack-buffer-overflow on address
0x7f2942e01060 at pc 0x00401856 bp 0x7f29415fecb0 sp 0x7f29415feca8
READ of size 4 at 0x7f2942e01060 thread T1
test.c:10:22: runtime error: index -4210096 out of bounds for type 'int [1024]'
#define DIM 32
#define N (DIM*DIM)
int
main() {
int a[N] = {}, b[N] = {}, c[N];
#pragma omp parallel for collapse(2)
for (int i = 0; i < DIM; i++) {
for (int j = (i*DIM); j < (i*DIM + DIM); j++) {
c[j] = a[j] + b[j];
}
}
}
Longer version using 'target teams loop', which segfault here (w/o offloading
configured + w/o UBSAN):
https://github.com/SOLLVE/sollve_vv/blob/master/tests/5.0/teams_loop/test_target_teams_loop_collapse.c
[Bug c++/111016] New: Confusing "used in its own initializer" for non-dependent ad-hoc constraint
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111016
Bug ID: 111016
Summary: Confusing "used in its own initializer" for
non-dependent ad-hoc constraint
Product: gcc
Version: 14.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: ed at catmur dot uk
Target Milestone: ---
#include
struct S { int i; };
static_assert(requires(S s) { requires std::destructible; });
In file included from :1:
include/c++/14.0.0/concepts:148:13: required for the satisfaction of
'destructible'
include/c++/14.0.0/concepts:148:38: error: the value of
'std::__detail::__destructible' is not usable in a constant
expression
148 | concept destructible = __detail::__destructible<_Tp>;
|~~^~~
include/c++/14.0.0/concepts:127:22: note:
'std::__detail::__destructible' used in its own initializer
127 | constexpr bool __destructible = __destructible_impl<_Tp>;
| ^~
Obviously this is IFNDR, but it would be nice to emit a diagnostic which gives
some better clue to what is going on (e.g. "warning: constraint is
non-dependent"). Unfortunately both clang and MSVC accept with no diagnostic,
making this look like a gcc bug.
[Bug tree-optimization/110988] [14 regression] ICE when building 523.xalancbmk_r with pgo and lto
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110988 --- Comment #1 from CVS Commits --- The master branch has been updated by Jan Hubicka : https://gcc.gnu.org/g:882af290c137dfab5d99b88e6dbecc5e75d85a0b commit r14-3201-g882af290c137dfab5d99b88e6dbecc5e75d85a0b Author: Jan Hubicka Date: Mon Aug 14 17:55:33 2023 +0200 Avoid division by zero in fold_loop_internal_call My patch to fix profile after folding internal call is missing check for the case profile was already zero before if-conversion. gcc/ChangeLog: PR gcov-profile/110988 * tree-cfg.cc (fold_loop_internal_call): Avoid division by zero.
[Bug tree-optimization/111012] [14 Regression] Dead Code Elimination Regression at -O3 since r14-573-g69f1a8af45d
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111012
Andrew Pinski changed:
What|Removed |Added
Last reconfirmed||2023-08-14
Status|UNCONFIRMED |NEW
Ever confirmed|0 |1
--- Comment #1 from Andrew Pinski ---
So the difference here is basically doing the following manually:
```
#if 0
static char i(int k) {
if (k)
bar173_();
c <= 0 >= b;
if (k)
return c;
return c;
}
#else
static char i(int k) {
if (k)
bar173_();
c <= 0 >= b;
int t;
if (k)
t = c;
else
t = c;
return t;
}
#endif
```
and we used to do some jump threading previously such that the load of the c
would be in the `if (k)` branch after the call to `bar173_()`. The only pass
which moves it like that is pre which happens maybe later but too late for jump
threading to happen for the optimizations.
I have not looked into why this makes a difference either. maybe because there
is another jump threading where we know c is 0 ...
Confirmed.
[Bug tree-optimization/111013] [14 Regression] Dead Code Elimination Regression at -O2 since r14-338-g1dd154f6407
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111013 Andrew Pinski changed: What|Removed |Added Ever confirmed|0 |1 Last reconfirmed||2023-08-14 Status|UNCONFIRMED |NEW --- Comment #2 from Andrew Pinski --- Confirmed.
[Bug tree-optimization/111013] [14 Regression] Dead Code Elimination Regression at -O2 since r14-338-g1dd154f6407
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111013 --- Comment #1 from Andrew Pinski --- THe only difference phiopt does is: GCC 13: ``` _19 = d.8_18 & 1; _6 = (bool) _19; _21 = (int) _6; ``` vs trunk: ``` _19 = d.8_18 & 1; _33 = _19 != 0; _21 = (int) _33; ```
[Bug fortran/110996] RISC-V vector Fortran: SEGV ICE during parsing
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110996 --- Comment #3 from Jeremy Bennett --- @JuzheZhong I believe this is in someway related to RVV. If I remove `v' from the march: riscv64-unknown-linux-gnu-gfortran -march=rv64gc -mabi=lp64d -c -Ofast testcase.f90 The output I get is correct: testcase.f90:6:20: 6 |SUBROUTINE c(d) e |1 Error: Syntax error in SUBROUTINE statement at (1) Why does adding `v' to the -march string cause a SEGV?
[Bug target/111010] [13 regression] error: unable to find a register to spill compiling GCDAProfiling.c
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111010 --- Comment #4 from Rainer Orth --- Created attachment 55736 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=55736&action=edit original testcase I just noticed that the original testcase still fails on trunk.
[Bug middle-end/110994] RISC-V Fortran: Illegal instruction ICE with scalable autovec
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110994 Jeremy Bennett changed: What|Removed |Added Resolution|--- |FIXED Status|UNCONFIRMED |RESOLVED --- Comment #3 from Jeremy Bennett --- Confirmed the patch resolves this issue, and the code correctly produces a syntax error.
[Bug middle-end/110989] RISC-V vector ICE due to invalid tree code in GIMPLE vect pass
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110989 Jeremy Bennett changed: What|Removed |Added Resolution|--- |FIXED Status|UNCONFIRMED |RESOLVED --- Comment #4 from Jeremy Bennett --- Confirmed the patch resolves this issue.
[Bug target/110964] RISC-V vector ICE in expand_cond_len_ternop
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110964 Jeremy Bennett changed: What|Removed |Added Resolution|--- |FIXED Status|UNCONFIRMED |RESOLVED --- Comment #2 from Jeremy Bennett --- Confirmed the patch resolves this issue.
[Bug middle-end/110962] RISC-V vector Fortran ICE in expand_expr_real_2
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110962 Jeremy Bennett changed: What|Removed |Added Status|UNCONFIRMED |RESOLVED Resolution|--- |FIXED --- Comment #2 from Jeremy Bennett --- Confirmed patch resolves the issue.
[Bug target/110950] RISC-V vector ICE in expand_const_vector
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110950 Jeremy Bennett changed: What|Removed |Added Status|UNCONFIRMED |RESOLVED Resolution|--- |FIXED --- Comment #5 from Jeremy Bennett --- Patch resolves the issue.
[Bug target/111010] [13 regression] error: unable to find a register to spill compiling GCDAProfiling.c
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111010 --- Comment #3 from ro at CeBiTec dot Uni-Bielefeld.DE --- I've just completed a reghunt which identified commit 4e0b504f26f78ff02e80ad98ebbf8ded3aa6ffa1 Author: Richard Biener Date: Tue Jan 10 13:48:51 2023 +0100 tree-optimization/106293 - missed DSE with virtual LC PHI as the culprit.
[Bug tree-optimization/111013] [14 Regression] Dead Code Elimination Regression at -O2 since r14-338-g1dd154f6407
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111013 Richard Biener changed: What|Removed |Added CC||pinskia at gcc dot gnu.org Target Milestone|--- |14.0
[Bug tree-optimization/111015] [11/12/13/14 Regression] __int128 bitfields optimized incorrectly to the 64 bit operations
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111015 Richard Biener changed: What|Removed |Added Status|UNCONFIRMED |NEW Summary|__int128 bitfields |[11/12/13/14 Regression] |optimized incorrectly to|__int128 bitfields |the 64 bit operations |optimized incorrectly to ||the 64 bit operations Known to work||7.5.0 Last reconfirmed||2023-08-14 Target Milestone|--- |11.5 Priority|P3 |P2 Component|rtl-optimization|tree-optimization Ever confirmed|0 |1 Keywords||needs-bisection Known to fail||11.4.0, 13.2.0 --- Comment #2 from Richard Biener --- Confirmed.
[Bug rtl-optimization/111015] __int128 bitfields optimized incorrectly to the 64 bit operations
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111015 --- Comment #1 from Richard Biener --- Created attachment 55735 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=55735&action=edit testcase from godbolt
[Bug rtl-optimization/111015] New: __int128 bitfields optimized incorrectly to the 64 bit operations
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111015
Bug ID: 111015
Summary: __int128 bitfields optimized incorrectly to the 64
bit operations
Product: gcc
Version: 13.2.1
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: rtl-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: pshevchuk at pshevchuk dot com
Target Milestone: ---
godbolt: https://godbolt.org/z/r5d6ToY1z
Basically, a store of one half of a 70-bit bitfield gets completely optimized
away.
i.e. for
struct Entry {
unsigned left : 4;
unsigned right : 4;
uint128 key : KEY_BITS;
} data;
the code:
data.left = left;
data.right = right;
data.key = key & KEY_BITS_MASK;
produces the following (amd64):
andl$15, %ecx
salq$4, %rcx
andl$15, %edx
orq %rdx, %rcx
movq%rdi, %rax
salq$8, %rax
orq %rax, %rcx
movq%rcx, data(%rip)
andw$-16384, data+8(%rip)
critically, at no point is there any attempt to actually initialize data+8
The problem does not disappear if the bitfields gets moved around; it is,
however, very finicky with respect to the size of the bitfields.
-O1 -fstore-merging appears to be close to the smallest set of compilation
options at which it fails.
If you replace -O1 with the list of -O1 optimization from here:
https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html , it will start
working correctly, so probably we have a documentation issue
[Bug driver/111014] New: do_spec_1 terminates arguments too eagerly when processing spec function
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111014
Bug ID: 111014
Summary: do_spec_1 terminates arguments too eagerly when
processing spec function
Product: gcc
Version: 14.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: driver
Assignee: unassigned at gcc dot gnu.org
Reporter: p...@gcc-bugzilla.mail.kapsi.fi
Target Milestone: ---
Hi.
do_spec_1 always terminates the current argument (via end_going_arg) if
processing a spec function.
6881 if (processing_spec_function)
6882end_going_arg ();
If a spec function returns a non-null string it is expanded using do_spec_1.
7043 funcval = eval_spec_function (func, args, soft_matched_part);
7044 if (funcval != NULL && do_spec_1 (funcval, 0, NULL) < 0)
However, because processing_spec_function is in this case non-zero, there is
always a termination after each nested %-sequence has been processed.
Thus, e.g. "%:version-compare(>= 1.0 -version-compare= %%(do_spec))" and
"%(do_spec)" behave differently if the expanded %-sequences have any trailing
parts:
# cat version_compare_string.spec
*self_spec:
+ --version-compare=1.0 %<-version-compare=*
*do_spec:
-W%{!Werror:no-}error
*cc1_options:
+ %:version-compare(>= 1.0 -version-compare= %%(do_spec)) %(do_spec)
# gcc -c -xc -specs=version_compare_string.spec - < /dev/null
cc1: error: unrecognized command-line option ‘-Wno-’
[...]
In the first case, do_spec is not expanded as "-Wno-error" but as "-Wno-" and
"error". I wonder if this termination is really necessary; the same effect can
be archieved by simply having a whitespace character to follow the spec
function, if concatenation is not to be desired.
[Bug tree-optimization/111013] New: [14 Regression] Dead Code Elimination Regression at -O2 since r14-338-g1dd154f6407
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111013
Bug ID: 111013
Summary: [14 Regression] Dead Code Elimination Regression at
-O2 since r14-338-g1dd154f6407
Product: gcc
Version: 14.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: scherrer.sv at gmail dot com
Target Milestone: ---
static int b, c, e;
static unsigned d;
static short f;
void foo(void);
void(a)();
static char g(int h) {
int i = 2749857453;
j:
a();
k:
if (b)
goto j;
f = 1;
if ((unsigned)(7 ^ e | (h & d && f)) >= 2)
i = 0;
e = 0;
if (c)
goto k;
if (!(i <= 6))
foo();
return h;
}
int main() {
d--;
b = c = e && g(1);
}
gcc-9ec5d6de735 (trunk) -O2 cannot eliminate the call to foo but
gcc-releases/gcc-13.1.0 -O2 can.
---
gcc-9ec5d6de7355c15b3811150d1581dab5bd489966 -O2 case.c -S -o case.s
- OUTPUT -
main:
.LFB1:
.cfi_startproc
subq$8, %rsp
.cfi_def_cfa_offset 16
movle(%rip), %eax
subl$1, d(%rip)
testl %eax, %eax
jne .L3
.L2:
movl%eax, c(%rip)
movl%eax, b(%rip)
xorl%eax, %eax
addq$8, %rsp
.cfi_remember_state
.cfi_def_cfa_offset 8
ret
.p2align 4,,10
.p2align 3
.L3:
.cfi_restore_state
xorl%eax, %eax
calla
movlb(%rip), %edx
testl %edx, %edx
jne .L3
movld(%rip), %ecx
movle(%rip), %eax
movl$-1545109843, %edx
xorl%esi, %esi
movlc(%rip), %edi
andl$1, %ecx
jmp .L5
.p2align 4,,10
.p2align 3
.L8:
xorl%eax, %eax
.L5:
xorl$7, %eax
orl %ecx, %eax
cmpl$2, %eax
cmovnb %esi, %edx
testl %edi, %edi
jne .L8
xorl%eax, %eax
movl%eax, e(%rip)
cmpl$6, %edx
jg .L15
.L6:
movl$1, %eax
jmp .L2
.L15:
callfoo
jmp .L6
-- END OUTPUT -
---
gcc-2b98cc24d6af0432a74f6dad1c722ce21c1f7458 -O2 case.c -S -o case.s
- OUTPUT -
main:
.LFB1:
.cfi_startproc
subq$8, %rsp
.cfi_def_cfa_offset 16
movle(%rip), %eax
subl$1, d(%rip)
testl %eax, %eax
jne .L3
.L2:
movl%eax, c(%rip)
movl%eax, b(%rip)
xorl%eax, %eax
addq$8, %rsp
.cfi_remember_state
.cfi_def_cfa_offset 8
ret
.p2align 4,,10
.p2align 3
.L3:
.cfi_restore_state
xorl%eax, %eax
calla
movlb(%rip), %edx
movlc(%rip), %eax
testl %edx, %edx
jne .L3
testl %eax, %eax
je .L4
.L5:
jmp .L5
.L4:
xorl%eax, %eax
movl%eax, e(%rip)
movl$1, %eax
jmp .L2
-- END OUTPUT -
---
Bisects to r14-338-g1dd154f6407
[Bug tree-optimization/111012] [14 Regression] Dead Code Elimination Regression at -O3 since r14-573-g69f1a8af45d
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111012 Richard Biener changed: What|Removed |Added Target Milestone|--- |14.0 CC||pinskia at gcc dot gnu.org
[Bug tree-optimization/111012] New: [14 Regression] Dead Code Elimination Regression at -O3 since r14-573-g69f1a8af45d
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111012
Bug ID: 111012
Summary: [14 Regression] Dead Code Elimination Regression at
-O3 since r14-573-g69f1a8af45d
Product: gcc
Version: 14.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: scherrer.sv at gmail dot com
Target Milestone: ---
static int b, c;
static char d;
static short e = -1L;
static int *j = &c;
void foo(void);
void bar150_(void);
void bar173_(void);
static char(a)(char k, char l) { return k + l; }
static void g(unsigned k, int l) {
if (l)
if (!k)
foo();
if (k)
bar150_();
}
static const unsigned char h();
static char i(int k) {
if (k)
bar173_();
c <= 0 >= b;
if (k)
return c;
return c;
}
static void f(char k, unsigned) {
char m = h(8 != c);
g(m && 8, k);
}
static const unsigned char h(int k) {
d = i(c);
*j = a(e, d < k < k && c) ^ k;
b = 0;
return c;
}
int main() { f(c, b); }
gcc-9ec5d6de735 (trunk) -O3 cannot eliminate the call to foo but
gcc-releases/gcc-13.1.0 -O3 can.
---
gcc-9ec5d6de7355c15b3811150d1581dab5bd489966 -O3 case.c -S -o case.s
- OUTPUT -
main:
.LFB5:
.cfi_startproc
pushq %rbp
.cfi_def_cfa_offset 16
.cfi_offset 6, -16
pushq %rbx
.cfi_def_cfa_offset 24
.cfi_offset 3, -24
subq$8, %rsp
.cfi_def_cfa_offset 32
movlc(%rip), %ebx
cmpl$8, %ebx
setne %bpl
testl %ebx, %ebx
jne .L9
movl$-2, c(%rip)
xorl%eax, %eax
movl%eax, b(%rip)
.L4:
callbar150_
.L6:
addq$8, %rsp
.cfi_remember_state
.cfi_def_cfa_offset 24
xorl%eax, %eax
popq%rbx
.cfi_def_cfa_offset 16
popq%rbp
.cfi_def_cfa_offset 8
ret
.p2align 4,,10
.p2align 3
.L9:
.cfi_restore_state
callbar173_
movlc(%rip), %edx
movzbl %bpl, %ebp
movl$0, b(%rip)
movsbl %dl, %eax
cmpl%eax, %ebp
setg%al
movzbl %al, %eax
cmpl%eax, %ebp
setg%al
testl %edx, %edx
setne %dl
andl%edx, %eax
subl$1, %eax
movsbl %al, %eax
xorl%ebp, %eax
movl%eax, c(%rip)
testb %bl, %bl
je .L3
testb %al, %al
jne .L4
callfoo
jmp .L6
.L3:
testb %al, %al
je .L6
jmp .L4
-- END OUTPUT -
---
gcc-2b98cc24d6af0432a74f6dad1c722ce21c1f7458 -O3 case.c -S -o case.s
- OUTPUT -
main:
.LFB5:
.cfi_startproc
movlc(%rip), %eax
pushq %rbx
.cfi_def_cfa_offset 16
.cfi_offset 3, -16
xorl%ebx, %ebx
cmpl$8, %eax
setne %bl
testl %eax, %eax
jne .L10
.L2:
cmpl%eax, %ebx
setg%al
movzbl %al, %eax
cmpl%eax, %ebx
jg .L3
notl%ebx
.L4:
movl%ebx, c(%rip)
movl$0, b(%rip)
callbar150_
xorl%eax, %eax
popq%rbx
.cfi_remember_state
.cfi_def_cfa_offset 8
ret
.L3:
.cfi_restore_state
cmpl$1, c(%rip)
sbbb%bl, %bl
xorl$1, %ebx
movsbl %bl, %ebx
jmp .L4
.L10:
callbar173_
movsbl c(%rip), %eax
jmp .L2
-- END OUTPUT -
---
Bisects to r14-573-g69f1a8af45d
[Bug c++/108080] ICE: in core_vals, at cp/module.cc:6262 with -fmodule-header
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108080 --- Comment #7 from Gregory Dushkin --- Sorry for splitting the comments, but the issue indeed appears to be tied to optimization options. When manually specifying -Og, -O1, -O2, -O3, -Os the compilation is successful. I only observe the ICE with -O0 or no optimization options.
[Bug middle-end/111009] [12/13/14 regression] -fno-strict-overflow erroneously elides null pointer checks and causes SIGSEGV on perf from linux-6.4.10
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111009
--- Comment #3 from Richard Biener ---
bool
operator_addr_expr::fold_range (irange &r, tree type,
const irange &lh,
const irange &rh,
relation_trio) const
{
if (empty_range_varying (r, type, lh, rh))
return true;
// Return a non-null pointer of the LHS type (passed in op2).
if (lh.zero_p ())
r = range_zero (type);
not sure how this is called, but we can only derive this if the offset
is zero as well, definitely if targetm.addr_space.zero_address_valid,
but I think this is true in general.
else if (!contains_zero_p (lh))
r = range_nonzero (type);
and this is only true for TYPE_OVERFLOW_UNDEFINED (type), with
-fwrapv-pointer we could wrap to zero.
That is, it's _not_ GIMPLE undefined behavior to compute &0->bar.
It looks like without -fwrapv-pointer we elide the if (!a) check,
dereferencing it when dso && dso != curr. I suppose that looks reasonable
with a = &dso->maj, when dso != 0 then a != 0 unless ->maj wraps.
[Bug middle-end/111009] [12/13/14 regression] -fno-strict-overflow erroneously elides null pointer checks and causes SIGSEGV on perf from linux-6.4.10
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111009
Richard Biener changed:
What|Removed |Added
Status|UNCONFIRMED |NEW
CC||amacleod at redhat dot com,
||rguenth at gcc dot gnu.org
Priority|P3 |P2
Target Milestone|--- |12.4
Last reconfirmed||2023-08-14
Ever confirmed|0 |1
--- Comment #2 from Richard Biener ---
There's nothing really wrong with lifting the &dso->maj computation, on GIMPLE
&dso->maj is just address arithmetic.
Interestingly we unswitch the loop but only with -fwrapv-pointer.
OK, so the bug looks like we have
if (&dso->maj != 0)
for (;;)
{
if (!dso) return 1;
if (dso == curr) return 1;
...
}
and the if (!dso) test is optimized away since &dso->maj != 0.
That's done by DOM3 here:
Optimizing statement _21 = dso_8(D) == _11;
LKUP STMT _21 = dso_8(D) eq_expr _11
2>>> STMT _21 = dso_8(D) eq_expr _11
Optimizing statement _22 = _21 | _13;
Replaced '_13' with constant '0'
Applying pattern match.pd:201, gimple-match-10.cc:6318
gimple_simplified to _22 = _21;
Folded to: _22 = _21;
I don't see where _13 = 0 is entered, this is possibly ranger related:
_13 : CACHE: BB 9 DOM query for _13, found [irange] _Bool VARYING at BB3
797 GORI recomputation attempt on edge 3->16 for _13 = dso_8(D) == 0B;
798 GORIoutgoing_edge for dso_8(D) on edge 3->16
799 GORI compute op 1 (a_9) at if (a_9 == 0B)
GORILHS =[irange] _Bool [1, 1]
GORIComputes a_9 = [irange] int * [0, 0] intersect Known range
: [irange] int * VARYING
GORI TRUE : (799) produces (a_9) [irange] int * [0, 0]
800 GORI compute op 1 (dso_8(D)) at a_9 = &dso_8(D)->maj;
GORILHS =[irange] int * [0, 0]
GORIComputes dso_8(D) = [irange] struct dso * [0, 0] intersect
Known range : [irange] struct dso * VARYING
GORI TRUE : (800) produces (dso_8(D)) [irange] struct dso * [0,
0]
GORITRUE : (798) outgoing_edge (dso_8(D)) [irange] struct dso * [0,
0]
GORI TRUE : (797) recomputation (_13) [irange] _Bool [1, 1]
I don't think we can do this. Andrew?
[Bug c++/108080] ICE: in core_vals, at cp/module.cc:6262 with -fmodule-header
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108080 --- Comment #6 from Gregory Dushkin --- Created attachment 55734 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=55734&action=edit Preprocessed source file /tmp/ccoSmHEM.out (gzip compressed)
[Bug c++/108080] ICE: in core_vals, at cp/module.cc:6262 with -fmodule-header
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108080
Gregory Dushkin changed:
What|Removed |Added
CC||yagreg7 at gmail dot com
--- Comment #5 from Gregory Dushkin ---
I have a similar issue with GCC 13.2.1. The weird part about it is that it
seems to depend on the exact value of some compiler flags rather than the
actual source file content. I'm trying to use g++ to compile {fmt} as a module
and this is what I get:
When building with CMake the {fmt} itself (go to repository, `mkdir build; cd
build; cmake .. -DFMT_MODULE=ON && make`, the command
```
/usr/bin/c++
-I/home/greg/.cpm/fmt/c85658eda638008e7e9290fcd31836ed9f7be1a4/include -O3
-DNDEBUG -std=gnu++20 -fvisibility=hidden -fvisibility-inlines-hidden
-fmodules-ts -MD -MT CMakeFiles/fmt.dir/src/fmt.cc.o -MF
CMakeFiles/fmt.dir/src/fmt.cc.o.d -o CMakeFiles/fmt.dir/src/fmt.cc.o -c
/home/greg/.cpm/fmt/c85658eda638008e7e9290fcd31836ed9f7be1a4/src/fmt.cc
```
executes successfully and the compilation is OK.
However, when compiling {fmt} as a CMake subproject, the command is a little
different:
```
cd /home/greg/projects/cpp/4seudo/build/_deps/fmt-build && /usr/bin/c++
-I/home/greg/.cpm/fmt/c85658eda638008e7e9290fcd31836ed9f7be1a4/include
-std=c++20 -fmodules-ts -MD -MT _deps/fmt-build/CMakeFiles/fmt.dir/src/fmt.cc.o
-MF CMakeFiles/fmt.dir/src/fmt.cc.o.d -o CMakeFiles/fmt.dir/src/fmt.cc.o -c
/home/greg/.cpm/fmt/c85658eda638008e7e9290fcd31836ed9f7be1a4/src/fmt.cc
```
And _this_ command fails with an ICE:
```
$ cd /home/greg/projects/cpp/4seudo/build/_deps/fmt-build && /usr/bin/c++
-I/home/greg/.cpm/fmt/c85658eda638008e7e9290fcd31836ed9f7be1a4/include
-std=c++20 -fmodules-ts -MD -MT _deps/fmt-build/CMakeFiles/fmt.dir/src/fmt.cc.o
-MF CMakeFiles/fmt.dir/src/fmt.cc.o.d -o CMakeFiles/fmt.dir/src/fmt.cc.o -c
/home/greg/.cpm/fmt/c85658eda638008e7e9290fcd31836ed9f7be1a4/src/fmt.cc
-freport-bug
/home/greg/.cpm/fmt/c85658eda638008e7e9290fcd31836ed9f7be1a4/src/fmt.cc:73:8:
internal compiler error: in core_vals, at cp/module.cc:6262
73 | export module fmt;
|^~
0x1ad33c8 internal_error(char const*, ...)
???:0
0x6b7b63 fancy_abort(char const*, int, char const*)
???:0
0x7c98b7 trees_out::tree_value(tree_node*)
???:0
0x7c763d trees_out::tree_node(tree_node*)
???:0
0x7c8806 trees_out::core_vals(tree_node*)
???:0
0x7c8d64 trees_out::tree_node_vals(tree_node*)
???:0
0x7c616b trees_out::decl_value(tree_node*, depset*)
???:0
0x7cb47f depset::hash::find_dependencies(module_state*)
???:0
0x7cc422 module_state::write_begin(elf_out*, cpp_reader*, module_state_config&,
unsigned int&)
???:0
0x7dc105 finish_module_processing(cpp_reader*)
???:0
0x7714fd c_parse_final_cleanups()
???:0
0x9444b4 c_common_parse_file()
???:0
Please submit a full bug report, with preprocessed source.
Please include the complete backtrace with any bug report.
See
[Bug target/111010] [13 regression] error: unable to find a register to spill compiling GCDAProfiling.c
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111010
Richard Biener changed:
What|Removed |Added
Status|UNCONFIRMED |NEW
Last reconfirmed||2023-08-14
Priority|P3 |P2
Ever confirmed|0 |1
Target|i386-pc-solaris2.11 |i386-pc-solaris2.11
||i?86-linux-gnu
Keywords||needs-bisection
--- Comment #2 from Richard Biener ---
I can confirm on x86_64-unknown-linux-gnu if you add -fno-omit-frame-pointer:
> ../../gcc13-g/gcc/cc1 -quiet t.c -m32 -O3 -fPIC -march=pentium4
> -mtune=generic -fpreprocessed -w -fno-omit-frame-pointer
t.c: In function 'c':
t.c:21:1: error: unable to find a register to spill
21 | }
| ^
t.c:21:1: error: this is the insn:
(insn 109 283 243 9 (set (reg:DI 295)
(ior:DI (ashift:DI (zero_extend:DI (mem:SI (plus:SI (mult:SI (reg:SI
329 [orig:229 _118 ] [229])
(const_int 4 [0x4]))
(reg/f:SI 328 [orig:83 a.0_1 ] [83])) [3
MEM[(unsigned int *)_11]+0 S4 A32]))
(const_int 32 [0x20]))
(zero_extend:DI (mem:SI (plus:SI (mult:SI (reg:SI 294 [233])
(const_int 4 [0x4]))
(reg/f:SI 328 [orig:83 a.0_1 ] [83])) [3 MEM[(unsigned
int *)_15]+0 S4 A32] "t.c":15:7 681 {*concatsidi3_3}
(expr_list:REG_DEAD (reg/f:SI 328 [orig:83 a.0_1 ] [83])
(expr_list:REG_DEAD (reg:SI 329 [orig:229 _118 ] [229])
(expr_list:REG_DEAD (reg:SI 294 [233])
(nil)
during RTL pass: reload
t.c:21:1: internal compiler error: in lra_split_hard_reg_for, at
lra-assigns.cc:1871
0x15066db _fatal_insn(char const*, rtx_def const*, char const*, int, char
const*)
/space/rguenther/src/gcc-13-branch/gcc/rtl-error.cc:108
0x131bb9a lra_split_hard_reg_for()
/space/rguenther/src/gcc-13-branch/gcc/lra-assigns.cc:1871
0x1314d74 lra(_IO_FILE*)
/space/rguenther/src/gcc-13-branch/gcc/lra.cc:2451
0x12bdb69 do_reload
/space/rguenther/src/gcc-13-branch/gcc/ira.cc:5963
0x12be060 execute
/space/rguenther/src/gcc-13-branch/gcc/ira.cc:6149
Please submit a full bug report, with preprocessed source (by using
-freport-bug).
Please include the complete backtrace with any bug report.
See
[Bug rtl-optimization/111011] gcc-13 incorrectly decrements by 2. It's twice as fast as gcc-12 and clang!
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111011
Richard Biener changed:
What|Removed |Added
Resolution|--- |INVALID
Status|UNCONFIRMED |RESOLVED
--- Comment #1 from Richard Biener ---
There's nothing wrong, we unroll the loop.
> ./cc1 -quiet t.c -O3 -fopt-info
t.c:5:15: optimized: loop unrolled 1 times
adding "# foo" to the asm text you'll see
.L2:
#APP
# 6 "t.c" 1
# foo
# 0 "" 2
# 6 "t.c" 1
# foo
# 0 "" 2
#NO_APP
subq$2, %rax
jne .L2
there's no data dependence with 'count' for the asm. You can instead use
#include
int main() {
int64_t count=270;
while (count>0) {
__asm__ __volatile__("" : "=g" (count) : "0" (count) : "memory");
--count;
}
return 0;
}
to get the desired effect.
[Bug target/88160] Error: register save offset not a multiple of 4 only with optimize
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88160 Thorsten Otto changed: What|Removed |Added CC||ad...@tho-otto.de --- Comment #5 from Thorsten Otto --- Another possible patch would be: --- a/gcc/config/m68k/m68k.cc2023-07-27 10:13:04.0 +0200 +++ b/gcc/config/m68k/m68k.cc 2023-08-13 08:59:00.959510772 +0200 @@ -712,6 +712,14 @@ m68k_option_override (void) else m68k_sched_mac = MAC_NO; } + + /* + * disable -fcombine-stack-adjustments for coldfire/mshort combination, + * which generates wrong CFI offsets. + * https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88160 + */ + if (TARGET_COLDFIRE && TARGET_SHORT && (write_symbols & DWARF2_DEBUG)) +flag_combine_stack_adjustments = 0; } This is only a workaround, but should prevent the bug.
[Bug rtl-optimization/111011] New: gcc-13 incorrectly decrements by 2. It's twice as fast as gcc-12 and clang!
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111011
Bug ID: 111011
Summary: gcc-13 incorrectly decrements by 2. It's twice as fast
as gcc-12 and clang!
Product: gcc
Version: 13.2.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: rtl-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: adam.warner.nz at gmail dot com
Target Milestone: ---
(Please fix my guess at the correct component for this bug report)
I'm amused by a ghost in the GCC virtual machine. I'm running this code on a
Debian Linux x86-64 desktop with these software versions:
gcc-12 (Debian 12.3.0-7) 12.3.0
gcc-13 (Debian 13.2.0-2) 13.2.0
gcc (Debian 20230718-1) 14.0.0 20230718 (experimental) [master
r14-2597-g6bab2772dbc]
Debian clang version 17.0.0 (++20230128060150+75153adeda1a-1~exp1)
My CPU is locked at 2.7GHz. It should take a nice round 10 seconds to decrement
2.7x10^10 to zero if each decrement takes one clock cycle.
And indeed it used to:
$ cat countdown.c
#include
int main() {
int64_t count=270;
while (count>0) {
__asm__ __volatile__("" : : : "memory");
--count;
}
return 0;
}
$ gcc-12 -O3 countdown.c && time ./a.out
real0m10.029s
user0m10.024s
sys 0m0.004s
$ clang-17 -O3 countdown.c && time ./a.out
real0m10.032s
user0m10.030s
sys 0m0.000s
But now it only takes 5 seconds:
$ gcc-13 -O3 countdown.c && time ./a.out
real0m5.022s
user0m5.021s
sys 0m0.001s
$ gcc-snapshot.sh -O3 countdown.c && time ./a.out
real0m5.023s
user0m5.022s
sys 0m0.000s
By disassembling the machine code we can clearly see why:
$ gcc-13 -O3 countdown.c && objdump -d -m i386:x86-64:intel a.out
...
1040 :
1040: 48 b8 00 4e 53 49 06movabs rax,0x649534e00
1047: 00 00 00
104a: 66 0f 1f 44 00 00 nopWORD PTR [rax+rax*1+0x0]
1050: 48 83 e8 02 subrax,0x2
1054: 75 fa jne1050
1056: 31 c0 xoreax,eax
1058: c3 ret
1059: 0f 1f 80 00 00 00 00nopDWORD PTR [rax+0x0]
...
[Bug target/111010] [13 regression] error: unable to find a register to spill compiling GCDAProfiling.c
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111010 Rainer Orth changed: What|Removed |Added Known to work||12.3.1, 14.0 Known to fail||13.2.1 Target Milestone|--- |13.3
[Bug target/111010] [13 regression] error: unable to find a register to spill compiling GCDAProfiling.c
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111010 --- Comment #1 from Rainer Orth --- Created attachment 55733 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=55733&action=edit reduced testcase
[Bug target/111010] New: [13 regression] error: unable to find a register to spill compiling GCDAProfiling.c
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111010
Bug ID: 111010
Summary: [13 regression] error: unable to find a register to
spill compiling GCDAProfiling.c
Product: gcc
Version: 13.2.1
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: target
Assignee: unassigned at gcc dot gnu.org
Reporter: ro at gcc dot gnu.org
Target Milestone: ---
Host: i386-pc-solaris2.11
Target: i386-pc-solaris2.11
Build: i386-pc-solaris2.11
Building current LLVM main with GCC 13.1.0 fails on Solaris/amd64, compiling
compiler-rt/lib/profile/GCDAProfiling.c. I could reduce that file to the
attached
testcase:
$ gcc -O3 -m32 -fPIC -c GCDAProfiling.i -w
GCDAProfiling.i: In function ‘c’:
GCDAProfiling.i:21:1: error: unable to find a register to spill
21 | }
| ^
GCDAProfiling.i:21:1: error: this is the insn:
(insn 109 333 249 9 (set (reg:DI 300)
(ior:DI (ashift:DI (zero_extend:DI (mem:SI (plus:SI (mult:SI (reg:SI
377 [orig:229 _118 ] [229])
(const_int 4 [0x4]))
(reg/f:SI 338 [orig:83 a.0_1 ] [83])) [3
MEM[(unsigned int *)_11]+0 S4 A32]))
(const_int 32 [0x20]))
(zero_extend:DI (mem:SI (plus:SI (mult:SI (reg:SI 299 [233])
(const_int 4 [0x4]))
(reg/f:SI 338 [orig:83 a.0_1 ] [83])) [3 MEM[(unsigned
int *)_15]+0 S4 A32] "GCDAProfiling.i":15:7 680 {*concatsidi3_3}
(expr_list:REG_DEAD (reg:SI 377 [orig:229 _118 ] [229])
(expr_list:REG_DEAD (reg/f:SI 338 [orig:83 a.0_1 ] [83])
(expr_list:REG_DEAD (reg:SI 299 [233])
(nil)
The corresponding cc1 invocation is
cc1 -fpreprocessed GCDAProfiling.i -quiet -m32 -mtune=generic -march=pentium4
-O3 -w -fPIC
The testcase compiles on the gcc-12 branch and trunk, so this is gcc-13 branch
regression only.
However, it does compile with a Linux/x86_64 gcc 13.1.0.
[Bug middle-end/111009] [12/13/14 regression] -fno-strict-overflow erroneously elides null pointer checks and causes SIGSEGV on perf from linux-6.4.10
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111009 Alexander Monakov changed: What|Removed |Added CC||amonakov at gcc dot gnu.org --- Comment #1 from Alexander Monakov --- Triggered by GIMPLE loop invariant motion lifting a_9 = &dso_8(D)->maj; across a (dso != NULL) test.
[Bug libstdc++/110860] std::format("{:f}",2e304) invokes undefined behaviour
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110860 --- Comment #15 from Jonathan Wakely --- Hi Paul, could you please submit the patch to the gcc-patches mailing list for formal review, CCing the libstd...@gcc.gnu.org list? Please either complete a copyright assignment to the FSF or use DCO sign-off as per https://gcc.gnu.org/dco.html so that we can use your patch in GCC. Thanks!
[Bug tree-optimization/111000] [14 Regression] Wrong code at -O3 on x86_64-linux-gnu since r14-2944-g3d48c11ad08
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111000
Richard Biener changed:
What|Removed |Added
Status|NEW |ASSIGNED
Assignee|unassigned at gcc dot gnu.org |rguenth at gcc dot
gnu.org
--- Comment #2 from Richard Biener ---
we are computing 647 >> t even when that would invoke undefined behavior (but
LIM does this already). Disabling either loop splitting or final value
replacement avoids the miscompile.
The only "interesting" replacement is the PHI defining the guard of the
second loop, thus 'd' after the first loop:
final value replacement:
_54 = PHI <_4(10)>
with expr: (d.6_22 + 1) + (d.6_22 <= _53 ? (int) ((unsigned int) _53 -
(unsigned int) d.6_22) : 0)
final stmt:
_54 = _67 + _73;
d.6_22 is 'd' at the start of the program, _53 is the loop split adjusted
exit test value, MIN (647 >> t, 7 - (d + 1)).
When disabling vectorization we instead get
0
Aborted (core dumped)
likewise when disabling unswitching. When disabling early invariant motion
splitting is disabled. Changing the test to the following, avoiding the
undefined behavior resolves the miscompile.
int printf(const char *, ...);
volatile int a = 31;
int b, d, e;
int main()
{
int t = a;
for (; d <= 6; d++) {
for (b = 0; b <= 6; b++) {
if (t >= 30)
e = d;
else if (d > (647 >> t))
e = d;
else
e = 0;
}
}
printf("%d\n", e);
if (e != 6)
__builtin_abort();
}
so in the end I think we see the effect of hoisting the shift and then
involving it in a complex computation + comparison, eventually enabling
folding that's "invalid" (taking advantage of the undefined behavior).
I will need to see to avoid hoisting shifts (we already rewrite undefined
signed overflow to unsigned arithmetic). Unfortunately there's no "safe"
way to write shifts without adding computation (but maybe we should do
that, rewrite it to 647 >> min (t, 31)).
[Bug middle-end/111009] New: [12/13/14 regression] -fno-strict-overflow erroneously elides null pointer checks and causes SIGSEGV on perf from linux-6.4.10
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111009
Bug ID: 111009
Summary: [12/13/14 regression] -fno-strict-overflow erroneously
elides null pointer checks and causes SIGSEGV on perf
from linux-6.4.10
Product: gcc
Version: 14.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: middle-end
Assignee: unassigned at gcc dot gnu.org
Reporter: slyfox at gcc dot gnu.org
Target Milestone: ---
Initially observed the failure as a perf SIGSEGV when running against
r14-3191-g614052dd4ea083.
I hope I did not break it too much when minimizing. Self-contained example:
// $ cat bug c.c
struct dso {
struct dso * next;
int maj;
};
__attribute__((noipa)) static void __dso_id__cmp_(void) {}
__attribute__((noipa))
static int bug(struct dso * d, struct dso *dso)
{
struct dso **p = &d;
struct dso *curr = 0;
while (*p) {
curr = *p;
// prevent null deref below
if (!dso) return 1;
if (dso == curr) return 1;
int *a = &dso->maj;
// null deref
if (!(a && *a)) __dso_id__cmp_();
p = &curr->next;
}
return 0;
}
__attribute__((noipa))
int main(void) {
struct dso d = { 0, 0, };
bug(&d, 0);
}
Triggering the bug:
$ gcc -fno-strict-overflow -O3 bug.c.c -o bug3 && ./bug3
Segmentation fault (core dumped)
$ gcc -fno-strict-overflow -O2 bug.c.c -o bug2 && ./bug2
$ gcc -v
Using built-in specs.
COLLECT_GCC=/<>/gcc-14.0.0/bin/gcc
COLLECT_LTO_WRAPPER=/<>/gcc-14.0.0/libexec/gcc/x86_64-unknown-linux-gnu/14.0.0/lto-wrapper
Target: x86_64-unknown-linux-gnu
Configured with: ../source/configure --prefix=/<>/gcc-14.0.0
--with-gmp-include=/<>/gmp-6.3.0-dev/include
--with-gmp-lib=/<>/gmp-6.3.0/lib
--with-mpfr-include=/<>/mpfr-4.2.0-12-dev/include
--with-mpfr-lib=/<>/mpfr-4.2.0-12/lib --with-mpc=/<>/libmpc-1.3.1
--with-native-system-header-dir=/<>/glibc-2.38-dev/include
--with-build-sysroot=/ --program-prefix= --enable-lto --disable-libstdcxx-pch
--without-included-gettext --with-system-zlib --enable-checking=release
--enable-static --enable-languages=c,c++ --disable-multilib --enable-plugin
--disable-libcc1 --with-isl=/<>/isl-0.20 --disable-bootstrap
--build=x86_64-unknown-linux-gnu --host=x86_64-unknown-linux-gnu
--target=x86_64-unknown-linux-gnu
Thread model: posix
Supported LTO compression algorithms: zlib
gcc version 14.0.0 (experimental) (GCC)
[Bug tree-optimization/110248] ivopts could under-cost for some addressing modes on len_{load,store}
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110248 Jan-Benedict Glaw changed: What|Removed |Added CC||jbg...@lug-owl.de --- Comment #12 from Jan-Benedict Glaw --- The second patch (that pulls in tree.h into recog.h) breaks building for me on an usual amd64 Linux system: configure '--with-pkgversion=basepoints/gcc-14-3093-g4a8e6fa8016, built at 1691996332'\ --prefix=/var/lib/laminar/run/gcc-local/82/toolchain-install \ --enable-werror-always \ --enable-languages=all \ --disable-multilib make V=1 all-gcc echo timestamp > s-preds-h TARGET_CPU_DEFAULT="" \ HEADERS="config/i386/i386-d.h" DEFINES="" \ /bin/bash ../../gcc/gcc/mkconfig.sh tm_d.h /var/lib/laminar/run/gcc-local/82/local-toolchain-install/bin/g++ -std=c++11 -c -g -O2 -DIN_GCC-fno-exceptions -fno-rtti -fasynchronous-unwind-tables -W -Wall -Wno-narrowing -Wwrite-strings -Wcast-qual -Wmissing-format-attribute -Wconditionally-supported -Woverloaded-virtual -pedantic -Wno-long-long -Wno-variadic-macros -Wno-overlength-strings -Werror -fno-common -DHAVE_CONFIG_H -DGENERATOR_FILE -I. -Ibuild -I../../gcc/gcc -I../../gcc/gcc/build -I../../gcc/gcc/../include -I../../gcc/gcc/../libcpp/include \ -o build/genflags.o ../../gcc/gcc/genflags.cc /var/lib/laminar/run/gcc-local/82/local-toolchain-install/bin/g++ -std=c++11 -g -O2 -DIN_GCC-fno-exceptions -fno-rtti -fasynchronous-unwind-tables -W -Wall -Wno-narrowing -Wwrite-strings -Wcast-qual -Wmissing-format-attribute -Wconditionally-supported -Woverloaded-virtual -pedantic -Wno-long-long -Wno-variadic-macros -Wno-overlength-strings -Werror -fno-common -DHAVE_CONFIG_H -DGENERATOR_FILE -static-libstdc++ -static-libgcc -o build/genflags \ build/genflags.o build/rtl.o build/read-rtl.o build/ggc-none.o build/vec.o build/min-insn-modes.o build/gensupport.o build/print-rtl.o build/hash-table.o build/sort.o build/read-md.o build/errors.o ../build-x86_64-pc-linux-gnu/libiberty/libiberty.a /var/lib/laminar/run/gcc-local/82/local-toolchain-install/bin/g++ -std=c++11 -c -g -O2 -DIN_GCC-fno-exceptions -fno-rtti -fasynchronous-unwind-tables -W -Wall -Wno-narrowing -Wwrite-strings -Wcast-qual -Wmissing-format-attribute -Wconditionally-supported -Woverloaded-virtual -pedantic -Wno-long-long -Wno-variadic-macros -Wno-overlength-strings -Werror -fno-common -DHAVE_CONFIG_H -DGENERATOR_FILE -I. -Ibuild -I../../gcc/gcc -I../../gcc/gcc/build -I../../gcc/gcc/../include -I../../gcc/gcc/../libcpp/include \ -o build/genconditions.o ../../gcc/gcc/genconditions.cc /var/lib/laminar/run/gcc-local/82/local-toolchain-install/bin/g++ -std=c++11 -g -O2 -DIN_GCC-fno-exceptions -fno-rtti -fasynchronous-unwind-tables -W -Wall -Wno-narrowing -Wwrite-strings -Wcast-qual -Wmissing-format-attribute -Wconditionally-supported -Woverloaded-virtual -pedantic -Wno-long-long -Wno-variadic-macros -Wno-overlength-strings -Werror -fno-common -DHAVE_CONFIG_H -DGENERATOR_FILE -static-libstdc++ -static-libgcc -o build/genconditions \ build/genconditions.o build/rtl.o build/read-rtl.o build/ggc-none.o build/vec.o build/min-insn-modes.o build/gensupport.o build/print-rtl.o build/hash-table.o build/sort.o build/read-md.o build/errors.o ../build-x86_64-pc-linux-gnu/libiberty/libiberty.a build/genconditions ../../gcc/gcc/common.md ../../gcc/gcc/config/i386/i386.md > tmp-condmd.cc /bin/bash ../../gcc/gcc/../move-if-change tmp-condmd.cc build/gencondmd.cc echo timestamp > s-conditions build/genpreds -c ../../gcc/gcc/common.md ../../gcc/gcc/config/i386/i386.md > tmp-constrs.h /bin/bash ../../gcc/gcc/../move-if-change tmp-constrs.h tm-constrs.h echo timestamp > s-constrs-h /var/lib/laminar/run/gcc-local/82/local-toolchain-install/bin/g++ -std=c++11 -c -g -O2 -DIN_GCC-fno-exceptions -fno-rtti -fasynchronous-unwind-tables -W -Wall -Wno-narrowing -Wwrite-strings -Wcast-qual -Wmissing-format-attribute -Wconditionally-supported -Woverloaded-virtual -pedantic -Wno-long-long -Wno-variadic-macros -Wno-overlength-strings -Werror -fno-common -DHAVE_CONFIG_H -DGENERATOR_FILE -I. -Ibuild -I../../gcc/gcc -I../../gcc/gcc/build -I../../gcc/gcc/../include -I../../gcc/gcc/../libcpp/include \ -o build/gencondmd.o build/gencondmd.cc In file included from ../../gcc/gcc/tree.h:23, from ../../gcc/gcc/recog.h:24, from build/gencondmd.cc:40: ../../gcc/gcc/tree-core.h:145:10: fatal error: all-tree.def: No such file or directory 145 | #include "all-tree.def" | ^~ compilation t
[Bug tree-optimization/111003] [14 Regression] Dead Code Elimination Regression at -O3 since r14-2161-g237e83e2158
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111003 Richard Biener changed: What|Removed |Added CC|rguenther at suse dot de |rguenth at gcc dot gnu.org --- Comment #2 from Richard Biener --- So the only additional hoisting is in LIM4 Moving statement _60 = _13 ^ iftmp.9_18; (cost 1) out of loop 1. Moving statement _62 = _60 == 10; (cost 21) out of loop 1. @@ -239,6 +279,8 @@ goto ; [100.00%] [local count: 477815112]: + _60 = _13 ^ iftmp.9_18; + _62 = _60 == 10; [local count: 4343774241]: _9 = a (); @@ -296,8 +338,6 @@ goto ; [100.00%] [local count: 14585209535]: - _60 = _13 ^ iftmp.9_18; - _62 = _60 == 10; [local count: 35145083376]: # prephitmp_63 = PHI <0(73), _62(14), 0(74)> and -fdisable-tree-lim4 restores the missed optimization. The difference is then in CCP4 which is able to remove the call to foo() when not hoisting the compare. When not hoisting the global range of _60 is stricter: - # RANGE [irange] unsigned int [0, 511] MASK 0x1c7 VALUE 0x0 - _60 = _13 ^ iftmp.9_18; + # RANGE [irange] unsigned int [0, 16383] MASK 0x3fff VALUE 0x0 + _60 = _13 ^ iftmp.9_18;
[Bug tree-optimization/110991] [14 Regression] Dead Code Elimination Regression at -O2 since r14-1135-gc53f51005de
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110991
Richard Biener changed:
What|Removed |Added
Status|NEW |ASSIGNED
Target||x86_64-*-*
Target Milestone|--- |14.0
Assignee|unassigned at gcc dot gnu.org |rguenth at gcc dot
gnu.org
--- Comment #3 from Richard Biener ---
So the difference is that GCC 14 vectorizes the loop and that vectorized loop
is not completely unrolled because
Loop 1 likely iterates at most 2 times.
Estimating sizes for loop 1
BB: 3, after_exit: 0
size: 1 _34 = vect_vec_iv_.15_33 + { 252, 252, 252, 252 };
size: 0 vect_a.16_35 = VIEW_CONVERT_EXPR(vect_vec_iv_.15_33);
size: 1 vect_iftmp.17_36 = vect_a.16_35 << 3;
size: 1 mask__23.18_38 = vect_a.16_35 < { 0, 0, 0, 0 };
size: 1 vect_iftmp.19_40 = VEC_COND_EXPR ;
size: 1 ivtmp_44 = ivtmp_43 + 1;
Induction variable computation will be folded away.
size: 2 if (ivtmp_44 < 3)
Exit condition will be eliminated in peeled copies.
Exit condition will be eliminated in last copy.
Constant conditional.
BB: 9, after_exit: 1
size: 7-3, last_iteration: 7-3
Loop size: 7
Estimated size after unrolling: 8
Not unrolling loop 1: size would grow.
when we still have a loop there's nothing that can fully elide things.
Without vectorization we have
Loop 2 likely iterates at most 11 times.
Estimating sizes for loop 2
BB: 10, after_exit: 0
size: 0 a.2_13 = (signed char) a.6_22;
Induction variable computation will be folded away.
size: 2 if (a.2_13 < 0)
Constant conditional.
BB: 13, after_exit: 1
BB: 12, after_exit: 0
size: 1 _26 = a.6_22 + 255;
Induction variable computation will be folded away.
size: 1 ivtmp_27 = ivtmp_4 - 1;
Induction variable computation will be folded away.
size: 2 if (ivtmp_27 != 0)
Exit condition will be eliminated in peeled copies.
Exit condition will be eliminated in last copy.
Constant conditional.
BB: 11, after_exit: 0
size: 1 iftmp.0_12 = a.2_13 << 3;
Induction variable computation will be folded away.
size: 7-7, last_iteration: 7-7
Loop size: 7
Estimated size after unrolling: 1
unrolling relies on constant_after_peeling which relies on SCEV which
doesn't handle vector IVs.
I have a patch improving it to
size: 7-4, last_iteration: 7-4
Loop size: 7
Estimated size after unrolling: 6
IIRC I also had a patch more appropriately "propagating" constness at some
point.
