[Bug target/94420] [8/9/10 Regression] ICE error: insn does not satisfy its constraints
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94420 Jeffrey A. Law changed: What|Removed |Added CC||law at redhat dot com Resolution|--- |FIXED Status|ASSIGNED|RESOLVED --- Comment #11 from Jeffrey A. Law --- Commit referenced by c#10 fixed this problem.
[Bug target/94420] [8/9/10 Regression] ICE error: insn does not satisfy its constraints
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94420 --- Comment #10 from CVS Commits --- The master branch has been updated by Segher Boessenkool : https://gcc.gnu.org/g:032f2366a4cd57f781f2093d977b9cf9600c83b8 commit r10-7497-g032f2366a4cd57f781f2093d977b9cf9600c83b8 Author: Segher Boessenkool Date: Wed Apr 1 16:51:04 2020 + rs6000: Make code questionably using r2 not ICE (PR94420) The example code in the PR uses r2 (the TOC register) directly. In the RTL generated for that, r2 is copied to some pseudo, and then cprop propagates that into a "*tocref" insn, because nothing is preventing it from doing that. So, put the same condition in the insn condition for this as we will later encounter in the constraint anyway, fixing this. 2020-04-01 Segher Boessenkool PR target/94420 * config/rs6000/rs6000.md (*tocref for P): Add insn condition on operands[1].
[Bug target/94420] [8/9/10 Regression] ICE error: insn does not satisfy its constraints
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94420 Richard Biener changed: What|Removed |Added Priority|P3 |P2
[Bug target/94420] [8/9/10 Regression] ICE error: insn does not satisfy its constraints
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94420 --- Comment #9 from Segher Boessenkool --- diff --git a/gcc/config/rs6000/rs6000.md b/gcc/config/rs6000/rs6000.md index dcccb03..11ab745 100644 --- a/gcc/config/rs6000/rs6000.md +++ b/gcc/config/rs6000/rs6000.md @@ -10311,7 +10311,8 @@ (define_insn "*largetoc_low_aix" (define_insn_and_split "*tocref" [(set (match_operand:P 0 "gpc_reg_operand" "=b") (match_operand:P 1 "small_toc_ref" "R"))] - "TARGET_TOC" + "TARGET_TOC +&& legitimate_constant_pool_address_p (operands[1], QImode, false)" "la %0,%a1" "&& TARGET_CMODEL != CMODEL_SMALL && reload_completed" [(set (match_dup 0) (high:P (match_dup 1))) This works on the testcase fine; regression testing now.
[Bug target/94420] [8/9/10 Regression] ICE error: insn does not satisfy its constraints
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94420 Segher Boessenkool changed: What|Removed |Added Status|NEW |ASSIGNED Assignee|unassigned at gcc dot gnu.org |segher at gcc dot gnu.org --- Comment #8 from Segher Boessenkool --- I put it in the insn condition, testing it now.
[Bug target/94420] [8/9/10 Regression] ICE error: insn does not satisfy its constraints
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94420 Segher Boessenkool changed: What|Removed |Added Keywords|ice-on-invalid-code |ice-on-valid-code --- Comment #7 from Segher Boessenkool --- But okay, let's call it valid code, it doesn't actually *do* anything with the undefined value in this example :-)
[Bug target/94420] [8/9/10 Regression] ICE error: insn does not satisfy its constraints
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94420 --- Comment #6 from Jakub Jelinek --- So shouldn't predicate on that operand verify the same thing as the constraint does (or insn condition)?
[Bug target/94420] [8/9/10 Regression] ICE error: insn does not satisfy its constraints
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94420 --- Comment #5 from Segher Boessenkool --- It is undefined behaviour to access such a register in any way -- sure, you *can* inspect it, there just is no guarantee at all what value you will get. It still is a useful thing to do in debug code and the like, of course.
[Bug target/94420] [8/9/10 Regression] ICE error: insn does not satisfy its constraints
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94420 --- Comment #4 from Segher Boessenkool --- cprop1 has done LOCAL COPY-PROP: Replacing reg 2 in insn 7 with reg 118 which is wrong: the insn isn't valid after that. But there is nothing that expresses that, the "R" constraint "just" becomes unsatisfiable.
[Bug target/94420] [8/9/10 Regression] ICE error: insn does not satisfy its constraints
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94420 --- Comment #3 from Jakub Jelinek --- I'm not sure about the invalid, of course if one uses a global register var for a fixed reg and stores something into it, then it is clearly UB, but this case is ont about changing it, just inspecting, and I believe users use both global and local register vars for various fixed regs quite often, whether it is a stack pointer, hard frame pointer, _GLOBAL_OFFSET_TABLE_ etc.
[Bug target/94420] [8/9/10 Regression] ICE error: insn does not satisfy its constraints
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94420 --- Comment #2 from Segher Boessenkool --- Reserving a register that already *is* reserved (by the ABI) is undefined, of course. But we shouldn't ICE.
[Bug target/94420] [8/9/10 Regression] ICE error: insn does not satisfy its constraints
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94420 Jakub Jelinek changed: What|Removed |Added Last reconfirmed||2020-03-31 Ever confirmed|0 |1 CC||jakub at gcc dot gnu.org Status|UNCONFIRMED |NEW Summary|ICE error: insn does not|[8/9/10 Regression] ICE |satisfy its constraints |error: insn does not ||satisfy its constraints Target Milestone|--- |8.5 --- Comment #1 from Jakub Jelinek --- With additional -fno-common it is quite old regression.
