| Message ID | 583F02B0.3030406@foss.arm.com |
|---|---|
| State | New |
| Headers | show |
Ping. https://gcc.gnu.org/ml/gcc-patches/2016-11/msg03078.html Thanks, Kyrill On 30/11/16 16:47, Kyrill Tkachov wrote: > Hi all, > > In this awkward ICE we have a *load_multiple pattern that is being transformed in reload from: > (insn 55 67 151 3 (parallel [ > (set (reg:SI 0 r0) > (mem/u/c:SI (reg/f:SI 147) [2 c+0 S4 A32])) > (set (reg:SI 158 [ c+4 ]) > (mem/u/c:SI (plus:SI (reg/f:SI 147) > (const_int 4 [0x4])) [2 c+4 S4 A32])) > ]) arm-crash.c:25 393 {*load_multiple} > (expr_list:REG_UNUSED (reg:SI 0 r0) > (nil))) > > > into the invalid: > (insn 55 67 70 3 (parallel [ > (set (reg:SI 0 r0) > (mem/u/c:SI (reg/f:SI 5 r5 [147]) [2 c+0 S4 A32])) > (set (mem/c:SI (plus:SI (reg/f:SI 102 sfp) > (const_int -4 [0xfffffffffffffffc])) [4 %sfp+-12 S4 A32]) > (mem/u/c:SI (plus:SI (reg/f:SI 5 r5 [147]) > (const_int 4 [0x4])) [2 c+4 S4 A32])) > ]) arm-crash.c:25 393 {*load_multiple} > (nil)) > > The operands of *load_multiple are not validated through constraints like LRA is used to, but rather through > a match_parallel predicate which ends up calling ldm_stm_operation_p to validate the multiple sets. > But this means that LRA cannot reason about the constraints properly. > This two-regiseter load should not have used *load_multiple anyway, it should have used *ldm2_ from ldmstm.md > and indeed it did until the loop2_invariant pass which copied the ldm2_ pattern: > (insn 27 23 28 4 (parallel [ > (set (reg:SI 0 r0) > (mem/u/c:SI (reg/f:SI 147) [2 c+0 S4 A32])) > (set (reg:SI 1 r1) > (mem/u/c:SI (plus:SI (reg/f:SI 147) > (const_int 4 [0x4])) [2 c+4 S4 A32])) > ]) "ldm.c":25 385 {*ldm2_} > (nil)) > > into: > (insn 55 19 67 3 (parallel [ > (set (reg:SI 0 r0) > (mem/u/c:SI (reg/f:SI 147) [2 c+0 S4 A32])) > (set (reg:SI 158) > (mem/u/c:SI (plus:SI (reg/f:SI 147) > (const_int 4 [0x4])) [2 c+4 S4 A32])) > ]) "ldm.c":25 404 {*load_multiple} > (expr_list:REG_UNUSED (reg:SI 0 r0) > (nil))) > > Note that it now got recognised as load_multiple because the second register is not a hard register but the pseudo 158. > In any case, the solution suggested in the PR (and I agree with it) is to restrict *load_multiple to after reload. > The similar pattern *load_multiple_with_writeback also has a similar condition and the comment above *load_multiple says that > it's used to generate epilogues, which is done after reload anyway. For pre-reload load-multiples the patterns in ldmstm.md > should do just fine. > > Bootstrapped and tested on arm-none-linux-gnueabihf. > > Ok for trunk? > > Thanks, > Kyrill > > 2016-11-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com> > > PR target/71436 > * config/arm/arm.md (*load_multiple): Add reload_completed to > matching condition. > > 2016-11-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com> > > PR target/71436 > * gcc.c-torture/compile/pr71436.c: New test.
Ping. Thanks, Kyrill On 08/12/16 11:55, Kyrill Tkachov wrote: > Ping. > https://gcc.gnu.org/ml/gcc-patches/2016-11/msg03078.html > > Thanks, > Kyrill > > On 30/11/16 16:47, Kyrill Tkachov wrote: >> Hi all, >> >> In this awkward ICE we have a *load_multiple pattern that is being transformed in reload from: >> (insn 55 67 151 3 (parallel [ >> (set (reg:SI 0 r0) >> (mem/u/c:SI (reg/f:SI 147) [2 c+0 S4 A32])) >> (set (reg:SI 158 [ c+4 ]) >> (mem/u/c:SI (plus:SI (reg/f:SI 147) >> (const_int 4 [0x4])) [2 c+4 S4 A32])) >> ]) arm-crash.c:25 393 {*load_multiple} >> (expr_list:REG_UNUSED (reg:SI 0 r0) >> (nil))) >> >> >> into the invalid: >> (insn 55 67 70 3 (parallel [ >> (set (reg:SI 0 r0) >> (mem/u/c:SI (reg/f:SI 5 r5 [147]) [2 c+0 S4 A32])) >> (set (mem/c:SI (plus:SI (reg/f:SI 102 sfp) >> (const_int -4 [0xfffffffffffffffc])) [4 %sfp+-12 S4 A32]) >> (mem/u/c:SI (plus:SI (reg/f:SI 5 r5 [147]) >> (const_int 4 [0x4])) [2 c+4 S4 A32])) >> ]) arm-crash.c:25 393 {*load_multiple} >> (nil)) >> >> The operands of *load_multiple are not validated through constraints like LRA is used to, but rather through >> a match_parallel predicate which ends up calling ldm_stm_operation_p to validate the multiple sets. >> But this means that LRA cannot reason about the constraints properly. >> This two-regiseter load should not have used *load_multiple anyway, it should have used *ldm2_ from ldmstm.md >> and indeed it did until the loop2_invariant pass which copied the ldm2_ pattern: >> (insn 27 23 28 4 (parallel [ >> (set (reg:SI 0 r0) >> (mem/u/c:SI (reg/f:SI 147) [2 c+0 S4 A32])) >> (set (reg:SI 1 r1) >> (mem/u/c:SI (plus:SI (reg/f:SI 147) >> (const_int 4 [0x4])) [2 c+4 S4 A32])) >> ]) "ldm.c":25 385 {*ldm2_} >> (nil)) >> >> into: >> (insn 55 19 67 3 (parallel [ >> (set (reg:SI 0 r0) >> (mem/u/c:SI (reg/f:SI 147) [2 c+0 S4 A32])) >> (set (reg:SI 158) >> (mem/u/c:SI (plus:SI (reg/f:SI 147) >> (const_int 4 [0x4])) [2 c+4 S4 A32])) >> ]) "ldm.c":25 404 {*load_multiple} >> (expr_list:REG_UNUSED (reg:SI 0 r0) >> (nil))) >> >> Note that it now got recognised as load_multiple because the second register is not a hard register but the pseudo 158. >> In any case, the solution suggested in the PR (and I agree with it) is to restrict *load_multiple to after reload. >> The similar pattern *load_multiple_with_writeback also has a similar condition and the comment above *load_multiple says that >> it's used to generate epilogues, which is done after reload anyway. For pre-reload load-multiples the patterns in ldmstm.md >> should do just fine. >> >> Bootstrapped and tested on arm-none-linux-gnueabihf. >> >> Ok for trunk? >> >> Thanks, >> Kyrill >> >> 2016-11-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com> >> >> PR target/71436 >> * config/arm/arm.md (*load_multiple): Add reload_completed to >> matching condition. >> >> 2016-11-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com> >> >> PR target/71436 >> * gcc.c-torture/compile/pr71436.c: New test. >
On 30/11/16 16:47, Kyrill Tkachov wrote: > Hi all, > > In this awkward ICE we have a *load_multiple pattern that is being > transformed in reload from: > (insn 55 67 151 3 (parallel [ > (set (reg:SI 0 r0) > (mem/u/c:SI (reg/f:SI 147) [2 c+0 S4 A32])) > (set (reg:SI 158 [ c+4 ]) > (mem/u/c:SI (plus:SI (reg/f:SI 147) > (const_int 4 [0x4])) [2 c+4 S4 A32])) > ]) arm-crash.c:25 393 {*load_multiple} > (expr_list:REG_UNUSED (reg:SI 0 r0) > (nil))) > > > into the invalid: > (insn 55 67 70 3 (parallel [ > (set (reg:SI 0 r0) > (mem/u/c:SI (reg/f:SI 5 r5 [147]) [2 c+0 S4 A32])) > (set (mem/c:SI (plus:SI (reg/f:SI 102 sfp) > (const_int -4 [0xfffffffffffffffc])) [4 %sfp+-12 > S4 A32]) > (mem/u/c:SI (plus:SI (reg/f:SI 5 r5 [147]) > (const_int 4 [0x4])) [2 c+4 S4 A32])) > ]) arm-crash.c:25 393 {*load_multiple} > (nil)) > > The operands of *load_multiple are not validated through constraints > like LRA is used to, but rather through > a match_parallel predicate which ends up calling ldm_stm_operation_p to > validate the multiple sets. > But this means that LRA cannot reason about the constraints properly. > This two-regiseter load should not have used *load_multiple anyway, it > should have used *ldm2_ from ldmstm.md > and indeed it did until the loop2_invariant pass which copied the ldm2_ > pattern: > (insn 27 23 28 4 (parallel [ > (set (reg:SI 0 r0) > (mem/u/c:SI (reg/f:SI 147) [2 c+0 S4 A32])) > (set (reg:SI 1 r1) > (mem/u/c:SI (plus:SI (reg/f:SI 147) > (const_int 4 [0x4])) [2 c+4 S4 A32])) > ]) "ldm.c":25 385 {*ldm2_} > (nil)) > > into: > (insn 55 19 67 3 (parallel [ > (set (reg:SI 0 r0) > (mem/u/c:SI (reg/f:SI 147) [2 c+0 S4 A32])) > (set (reg:SI 158) > (mem/u/c:SI (plus:SI (reg/f:SI 147) > (const_int 4 [0x4])) [2 c+4 S4 A32])) > ]) "ldm.c":25 404 {*load_multiple} > (expr_list:REG_UNUSED (reg:SI 0 r0) > (nil))) > > Note that it now got recognised as load_multiple because the second > register is not a hard register but the pseudo 158. > In any case, the solution suggested in the PR (and I agree with it) is > to restrict *load_multiple to after reload. > The similar pattern *load_multiple_with_writeback also has a similar > condition and the comment above *load_multiple says that > it's used to generate epilogues, which is done after reload anyway. For > pre-reload load-multiples the patterns in ldmstm.md > should do just fine. > > Bootstrapped and tested on arm-none-linux-gnueabihf. > > Ok for trunk? > I don't think this is right. Firstly, these patterns look to me like the ones used for memcpy expansion, so not recognizing them could lead to compiler aborts. Secondly, the bug is when we generate (insn 55 67 70 3 (parallel [ (set (reg:SI 0 r0) (mem/u/c:SI (reg/f:SI 5 r5 [147]) [2 c+0 S4 A32])) (set (mem/c:SI (plus:SI (reg/f:SI 102 sfp) (const_int -4 [0xfffffffffffffffc])) [4 %sfp+-12 S4 A32]) (mem/u/c:SI (plus:SI (reg/f:SI 5 r5 [147]) (const_int 4 [0x4])) [2 c+4 S4 A32])) ]) arm-crash.c:25 393 {*load_multiple} (nil)) These patterns are supposed to enforce that the load (store) target register is a hard register that is higher numbered than the hard register in the memory slot that precedes it (thus satisfying the constraints for a normal ldm/stm instruction. The real question is why did ldm_stm_operation_p permit this modified pattern through, or was the pattern validation bypassed incorrectly by the loop2 invariant code when it copied the insn and made changes? R. > Thanks, > Kyrill > > 2016-11-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com> > > PR target/71436 > * config/arm/arm.md (*load_multiple): Add reload_completed to > matching condition. > > 2016-11-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com> > > PR target/71436 > * gcc.c-torture/compile/pr71436.c: New test. > > arm-ldm.patch > > > commit 996d28e2353badd1b29ef000f94d40c7dab9010f > Author: Kyrylo Tkachov <kyrylo.tkachov@arm.com> > Date: Tue Nov 29 15:07:30 2016 +0000 > > [ARM] Restrict *load_multiple pattern till after LRA > > diff --git a/gcc/config/arm/arm.md b/gcc/config/arm/arm.md > index 74c44f3..22d2a84 100644 > --- a/gcc/config/arm/arm.md > +++ b/gcc/config/arm/arm.md > @@ -11807,12 +11807,15 @@ (define_insn "<crc_variant>" > > ;; Patterns in ldmstm.md don't cover more than 4 registers. This pattern covers > ;; large lists without explicit writeback generated for APCS_FRAME epilogue. > +;; The operands are validated through the load_multiple_operation > +;; match_parallel predicate rather than through constraints so enable it only > +;; after reload. > (define_insn "*load_multiple" > [(match_parallel 0 "load_multiple_operation" > [(set (match_operand:SI 2 "s_register_operand" "=rk") > (mem:SI (match_operand:SI 1 "s_register_operand" "rk"))) > ])] > - "TARGET_32BIT" > + "TARGET_32BIT && reload_completed" > "* > { > arm_output_multireg_pop (operands, /*return_pc=*/false, > diff --git a/gcc/testsuite/gcc.c-torture/compile/pr71436.c b/gcc/testsuite/gcc.c-torture/compile/pr71436.c > new file mode 100644 > index 0000000..ab08d5d > --- /dev/null > +++ b/gcc/testsuite/gcc.c-torture/compile/pr71436.c > @@ -0,0 +1,35 @@ > +/* PR target/71436. */ > + > +#pragma pack(1) > +struct S0 > +{ > + volatile int f0; > + short f2; > +}; > + > +void foo (struct S0 *); > +int a, d; > +static struct S0 b[5]; > +static struct S0 c; > +void fn1 (); > +void > +main () > +{ > + { > + struct S0 e; > + for (; d; fn1 ()) > + { > + { > + a = 3; > + for (; a >= 0; a -= 1) > + { > + { > + e = c; > + } > + b[a] = e; > + } > + } > + } > + } > + foo (b); > +} >
On 15/12/16 09:55, Richard Earnshaw (lists) wrote: > On 30/11/16 16:47, Kyrill Tkachov wrote: >> Hi all, >> >> In this awkward ICE we have a *load_multiple pattern that is being >> transformed in reload from: >> (insn 55 67 151 3 (parallel [ >> (set (reg:SI 0 r0) >> (mem/u/c:SI (reg/f:SI 147) [2 c+0 S4 A32])) >> (set (reg:SI 158 [ c+4 ]) >> (mem/u/c:SI (plus:SI (reg/f:SI 147) >> (const_int 4 [0x4])) [2 c+4 S4 A32])) >> ]) arm-crash.c:25 393 {*load_multiple} >> (expr_list:REG_UNUSED (reg:SI 0 r0) >> (nil))) >> >> >> into the invalid: >> (insn 55 67 70 3 (parallel [ >> (set (reg:SI 0 r0) >> (mem/u/c:SI (reg/f:SI 5 r5 [147]) [2 c+0 S4 A32])) >> (set (mem/c:SI (plus:SI (reg/f:SI 102 sfp) >> (const_int -4 [0xfffffffffffffffc])) [4 %sfp+-12 >> S4 A32]) >> (mem/u/c:SI (plus:SI (reg/f:SI 5 r5 [147]) >> (const_int 4 [0x4])) [2 c+4 S4 A32])) >> ]) arm-crash.c:25 393 {*load_multiple} >> (nil)) >> >> The operands of *load_multiple are not validated through constraints >> like LRA is used to, but rather through >> a match_parallel predicate which ends up calling ldm_stm_operation_p to >> validate the multiple sets. >> But this means that LRA cannot reason about the constraints properly. >> This two-regiseter load should not have used *load_multiple anyway, it >> should have used *ldm2_ from ldmstm.md >> and indeed it did until the loop2_invariant pass which copied the ldm2_ >> pattern: >> (insn 27 23 28 4 (parallel [ >> (set (reg:SI 0 r0) >> (mem/u/c:SI (reg/f:SI 147) [2 c+0 S4 A32])) >> (set (reg:SI 1 r1) >> (mem/u/c:SI (plus:SI (reg/f:SI 147) >> (const_int 4 [0x4])) [2 c+4 S4 A32])) >> ]) "ldm.c":25 385 {*ldm2_} >> (nil)) >> >> into: >> (insn 55 19 67 3 (parallel [ >> (set (reg:SI 0 r0) >> (mem/u/c:SI (reg/f:SI 147) [2 c+0 S4 A32])) >> (set (reg:SI 158) >> (mem/u/c:SI (plus:SI (reg/f:SI 147) >> (const_int 4 [0x4])) [2 c+4 S4 A32])) >> ]) "ldm.c":25 404 {*load_multiple} >> (expr_list:REG_UNUSED (reg:SI 0 r0) >> (nil))) >> >> Note that it now got recognised as load_multiple because the second >> register is not a hard register but the pseudo 158. >> In any case, the solution suggested in the PR (and I agree with it) is >> to restrict *load_multiple to after reload. >> The similar pattern *load_multiple_with_writeback also has a similar >> condition and the comment above *load_multiple says that >> it's used to generate epilogues, which is done after reload anyway. For >> pre-reload load-multiples the patterns in ldmstm.md >> should do just fine. >> >> Bootstrapped and tested on arm-none-linux-gnueabihf. >> >> Ok for trunk? >> > > I don't think this is right. Firstly, these patterns look to me like > the ones used for memcpy expansion, so not recognizing them could lead > to compiler aborts. > > Secondly, the bug is when we generate > > (insn 55 67 70 3 (parallel [ > (set (reg:SI 0 r0) > (mem/u/c:SI (reg/f:SI 5 r5 [147]) [2 c+0 S4 A32])) > (set (mem/c:SI (plus:SI (reg/f:SI 102 sfp) > (const_int -4 [0xfffffffffffffffc])) [4 %sfp+-12 > S4 A32]) > (mem/u/c:SI (plus:SI (reg/f:SI 5 r5 [147]) > (const_int 4 [0x4])) [2 c+4 S4 A32])) > ]) arm-crash.c:25 393 {*load_multiple} > (nil)) sorry, pasted the wrong bit of code. That should read when we generate: (insn 55 19 67 3 (parallel [ (set (reg:SI 0 r0) (mem/u/c:SI (reg/f:SI 147) [2 c+0 S4 A32])) (set (reg:SI 158) (mem/u/c:SI (plus:SI (reg/f:SI 147) (const_int 4 [0x4])) [2 c+4 S4 A32])) ]) "ldm.c":25 404 {*load_multiple} (expr_list:REG_UNUSED (reg:SI 0 r0) (nil))) ie when we put a pseudo into the register load list. > > These patterns are supposed to enforce that the load (store) target > register is a hard register that is higher numbered than the hard > register in the memory slot that precedes it (thus satisfying the > constraints for a normal ldm/stm instruction. > > The real question is why did ldm_stm_operation_p permit this modified > pattern through, or was the pattern validation bypassed incorrectly by > the loop2 invariant code when it copied the insn and made changes? > > R. > >> Thanks, >> Kyrill >> >> 2016-11-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com> >> >> PR target/71436 >> * config/arm/arm.md (*load_multiple): Add reload_completed to >> matching condition. >> >> 2016-11-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com> >> >> PR target/71436 >> * gcc.c-torture/compile/pr71436.c: New test. >> >> arm-ldm.patch >> >> >> commit 996d28e2353badd1b29ef000f94d40c7dab9010f >> Author: Kyrylo Tkachov <kyrylo.tkachov@arm.com> >> Date: Tue Nov 29 15:07:30 2016 +0000 >> >> [ARM] Restrict *load_multiple pattern till after LRA >> >> diff --git a/gcc/config/arm/arm.md b/gcc/config/arm/arm.md >> index 74c44f3..22d2a84 100644 >> --- a/gcc/config/arm/arm.md >> +++ b/gcc/config/arm/arm.md >> @@ -11807,12 +11807,15 @@ (define_insn "<crc_variant>" >> >> ;; Patterns in ldmstm.md don't cover more than 4 registers. This pattern covers >> ;; large lists without explicit writeback generated for APCS_FRAME epilogue. >> +;; The operands are validated through the load_multiple_operation >> +;; match_parallel predicate rather than through constraints so enable it only >> +;; after reload. >> (define_insn "*load_multiple" >> [(match_parallel 0 "load_multiple_operation" >> [(set (match_operand:SI 2 "s_register_operand" "=rk") >> (mem:SI (match_operand:SI 1 "s_register_operand" "rk"))) >> ])] >> - "TARGET_32BIT" >> + "TARGET_32BIT && reload_completed" >> "* >> { >> arm_output_multireg_pop (operands, /*return_pc=*/false, >> diff --git a/gcc/testsuite/gcc.c-torture/compile/pr71436.c b/gcc/testsuite/gcc.c-torture/compile/pr71436.c >> new file mode 100644 >> index 0000000..ab08d5d >> --- /dev/null >> +++ b/gcc/testsuite/gcc.c-torture/compile/pr71436.c >> @@ -0,0 +1,35 @@ >> +/* PR target/71436. */ >> + >> +#pragma pack(1) >> +struct S0 >> +{ >> + volatile int f0; >> + short f2; >> +}; >> + >> +void foo (struct S0 *); >> +int a, d; >> +static struct S0 b[5]; >> +static struct S0 c; >> +void fn1 (); >> +void >> +main () >> +{ >> + { >> + struct S0 e; >> + for (; d; fn1 ()) >> + { >> + { >> + a = 3; >> + for (; a >= 0; a -= 1) >> + { >> + { >> + e = c; >> + } >> + b[a] = e; >> + } >> + } >> + } >> + } >> + foo (b); >> +} >> >
On Thu, Dec 15, 2016 at 10:00:14AM +0000, Richard Earnshaw (lists) wrote: > sorry, pasted the wrong bit of code. > > That should read when we generate: > > (insn 55 19 67 3 (parallel [ > (set (reg:SI 0 r0) > (mem/u/c:SI (reg/f:SI 147) [2 c+0 S4 A32])) > (set (reg:SI 158) > (mem/u/c:SI (plus:SI (reg/f:SI 147) > (const_int 4 [0x4])) [2 c+4 S4 A32])) > ]) "ldm.c":25 404 {*load_multiple} > (expr_list:REG_UNUSED (reg:SI 0 r0) > (nil))) > > ie when we put a pseudo into the register load list. We put a pseudo there because the predicate on the insn allows it: (define_special_predicate "load_multiple_operation" (match_code "parallel") { return ldm_stm_operation_p (op, /*load=*/true, SImode, /*consecutive=*/false, /*return_pc=*/false); }) and the consecutive = false argument says that (almost) no verification is performed on the SET_DEST, just that it is a REG and doesn't have REGNO smaller than the first reg. That said, RA is still not able to cope with such instructions, because only the first set is represented with constraints, so if such an insn needs any kind of reloading, it just will not happen. So I think the posted patch makes lots of sense, otherwise if you use such a pattern before reload, you just have to hope no reloading will be needed on it. Jakub
On 19/12/16 14:53, Jakub Jelinek wrote: > On Thu, Dec 15, 2016 at 10:00:14AM +0000, Richard Earnshaw (lists) wrote: >> sorry, pasted the wrong bit of code. >> >> That should read when we generate: >> >> (insn 55 19 67 3 (parallel [ >> (set (reg:SI 0 r0) >> (mem/u/c:SI (reg/f:SI 147) [2 c+0 S4 A32])) >> (set (reg:SI 158) >> (mem/u/c:SI (plus:SI (reg/f:SI 147) >> (const_int 4 [0x4])) [2 c+4 S4 A32])) >> ]) "ldm.c":25 404 {*load_multiple} >> (expr_list:REG_UNUSED (reg:SI 0 r0) >> (nil))) >> >> ie when we put a pseudo into the register load list. > We put a pseudo there because the predicate on the insn allows it: > (define_special_predicate "load_multiple_operation" > (match_code "parallel") > { > return ldm_stm_operation_p (op, /*load=*/true, SImode, > /*consecutive=*/false, > /*return_pc=*/false); > }) > and the consecutive = false argument says that (almost) no verification > is performed on the SET_DEST, just that it is a REG and doesn't have > REGNO smaller than the first reg. > That said, RA is still not able to cope with such instructions, because > only the first set is represented with constraints, so if such an insn > needs any kind of reloading, it just will not happen. > So I think the posted patch makes lots of sense, otherwise if you use > such a pattern before reload, you just have to hope no reloading will be > needed on it. In that case I believe restricting the pattern till after LRA is the way to go. However, we should still add the check from [1] to ldm_stm_operation_p for when 'consecutive' is true as consecutive order has no useful meaning on pseudos here. Thanks, Kyrill [1] https://gcc.gnu.org/ml/gcc-patches/2016-12/msg01400.html > > Jakub
On 18/01/17 09:49, Kyrill Tkachov wrote: > > On 19/12/16 14:53, Jakub Jelinek wrote: >> On Thu, Dec 15, 2016 at 10:00:14AM +0000, Richard Earnshaw (lists) wrote: >>> sorry, pasted the wrong bit of code. >>> >>> That should read when we generate: >>> >>> (insn 55 19 67 3 (parallel [ >>> (set (reg:SI 0 r0) >>> (mem/u/c:SI (reg/f:SI 147) [2 c+0 S4 A32])) >>> (set (reg:SI 158) >>> (mem/u/c:SI (plus:SI (reg/f:SI 147) >>> (const_int 4 [0x4])) [2 c+4 S4 A32])) >>> ]) "ldm.c":25 404 {*load_multiple} >>> (expr_list:REG_UNUSED (reg:SI 0 r0) >>> (nil))) >>> >>> ie when we put a pseudo into the register load list. >> We put a pseudo there because the predicate on the insn allows it: >> (define_special_predicate "load_multiple_operation" >> (match_code "parallel") >> { >> return ldm_stm_operation_p (op, /*load=*/true, SImode, >> /*consecutive=*/false, >> /*return_pc=*/false); >> }) >> and the consecutive = false argument says that (almost) no verification >> is performed on the SET_DEST, just that it is a REG and doesn't have >> REGNO smaller than the first reg. >> That said, RA is still not able to cope with such instructions, because >> only the first set is represented with constraints, so if such an insn >> needs any kind of reloading, it just will not happen. >> So I think the posted patch makes lots of sense, otherwise if you use >> such a pattern before reload, you just have to hope no reloading will be >> needed on it. > > In that case I believe restricting the pattern till after LRA is the way to go. > However, we should still add the check from [1] to ldm_stm_operation_p for when > 'consecutive' is true as consecutive order has no useful meaning on pseudos here. > In the interest of fixing this PR I think the patch at https://gcc.gnu.org/ml/gcc-patches/2016-11/msg03078.html is the way to go, that is disable this pattern until after reload. It was intended to handle epilogue pops that is generated after reload anyway and all the general load/store multiple patterns are handled by ldmstmd.md so not having this before reload is not risky. I'll also propose a variant of https://gcc.gnu.org/ml/gcc-patches/2016-12/msg01400.html to tighten up the consecutivity check when 'consecutive' is passed down as true, but that is indepdendent of this PR. Is the patch at https://gcc.gnu.org/ml/gcc-patches/2016-11/msg03078.html ok for trunk? It's been in my tree for a couple of months now getting testing and there's been no fallout. Thanks, Kyrill > > > [1] https://gcc.gnu.org/ml/gcc-patches/2016-12/msg01400.html > >> >> Jakub >
On 07/02/17 14:49, Kyrill Tkachov wrote: > > On 18/01/17 09:49, Kyrill Tkachov wrote: >> >> On 19/12/16 14:53, Jakub Jelinek wrote: >>> On Thu, Dec 15, 2016 at 10:00:14AM +0000, Richard Earnshaw (lists) wrote: >>>> sorry, pasted the wrong bit of code. >>>> >>>> That should read when we generate: >>>> >>>> (insn 55 19 67 3 (parallel [ >>>> (set (reg:SI 0 r0) >>>> (mem/u/c:SI (reg/f:SI 147) [2 c+0 S4 A32])) >>>> (set (reg:SI 158) >>>> (mem/u/c:SI (plus:SI (reg/f:SI 147) >>>> (const_int 4 [0x4])) [2 c+4 S4 A32])) >>>> ]) "ldm.c":25 404 {*load_multiple} >>>> (expr_list:REG_UNUSED (reg:SI 0 r0) >>>> (nil))) >>>> >>>> ie when we put a pseudo into the register load list. >>> We put a pseudo there because the predicate on the insn allows it: >>> (define_special_predicate "load_multiple_operation" >>> (match_code "parallel") >>> { >>> return ldm_stm_operation_p (op, /*load=*/true, SImode, >>> /*consecutive=*/false, >>> /*return_pc=*/false); >>> }) >>> and the consecutive = false argument says that (almost) no verification >>> is performed on the SET_DEST, just that it is a REG and doesn't have >>> REGNO smaller than the first reg. >>> That said, RA is still not able to cope with such instructions, because >>> only the first set is represented with constraints, so if such an insn >>> needs any kind of reloading, it just will not happen. >>> So I think the posted patch makes lots of sense, otherwise if you use >>> such a pattern before reload, you just have to hope no reloading will be >>> needed on it. >> >> In that case I believe restricting the pattern till after LRA is the way to go. >> However, we should still add the check from [1] to ldm_stm_operation_p for when >> 'consecutive' is true as consecutive order has no useful meaning on pseudos here. >> > > In the interest of fixing this PR I think the patch at https://gcc.gnu.org/ml/gcc-patches/2016-11/msg03078.html > is the way to go, that is disable this pattern until after reload. > It was intended to handle epilogue pops that is generated after reload anyway and all the general load/store multiple > patterns are handled by ldmstmd.md so not having this before reload is not risky. > > I'll also propose a variant of https://gcc.gnu.org/ml/gcc-patches/2016-12/msg01400.html to tighten up the consecutivity > check when 'consecutive' is passed down as true, but that is indepdendent of this PR. > > Is the patch at https://gcc.gnu.org/ml/gcc-patches/2016-11/msg03078.html ok for trunk? > It's been in my tree for a couple of months now getting testing and there's been no fallout. > Ping. Thanks, Kyrill > Thanks, > Kyrill > >> >> >> [1] https://gcc.gnu.org/ml/gcc-patches/2016-12/msg01400.html >> >>> >>> Jakub >> >
On 07/02/17 14:49, Kyrill Tkachov wrote: > > On 18/01/17 09:49, Kyrill Tkachov wrote: >> >> On 19/12/16 14:53, Jakub Jelinek wrote: >>> On Thu, Dec 15, 2016 at 10:00:14AM +0000, Richard Earnshaw (lists) >>> wrote: >>>> sorry, pasted the wrong bit of code. >>>> >>>> That should read when we generate: >>>> >>>> (insn 55 19 67 3 (parallel [ >>>> (set (reg:SI 0 r0) >>>> (mem/u/c:SI (reg/f:SI 147) [2 c+0 S4 A32])) >>>> (set (reg:SI 158) >>>> (mem/u/c:SI (plus:SI (reg/f:SI 147) >>>> (const_int 4 [0x4])) [2 c+4 S4 A32])) >>>> ]) "ldm.c":25 404 {*load_multiple} >>>> (expr_list:REG_UNUSED (reg:SI 0 r0) >>>> (nil))) >>>> >>>> ie when we put a pseudo into the register load list. >>> We put a pseudo there because the predicate on the insn allows it: >>> (define_special_predicate "load_multiple_operation" >>> (match_code "parallel") >>> { >>> return ldm_stm_operation_p (op, /*load=*/true, SImode, >>> /*consecutive=*/false, >>> /*return_pc=*/false); >>> }) >>> and the consecutive = false argument says that (almost) no verification >>> is performed on the SET_DEST, just that it is a REG and doesn't have >>> REGNO smaller than the first reg. >>> That said, RA is still not able to cope with such instructions, because >>> only the first set is represented with constraints, so if such an insn >>> needs any kind of reloading, it just will not happen. >>> So I think the posted patch makes lots of sense, otherwise if you use >>> such a pattern before reload, you just have to hope no reloading will be >>> needed on it. >> >> In that case I believe restricting the pattern till after LRA is the >> way to go. >> However, we should still add the check from [1] to ldm_stm_operation_p >> for when >> 'consecutive' is true as consecutive order has no useful meaning on >> pseudos here. >> > > In the interest of fixing this PR I think the patch at > https://gcc.gnu.org/ml/gcc-patches/2016-11/msg03078.html > is the way to go, that is disable this pattern until after reload. > It was intended to handle epilogue pops that is generated after reload > anyway and all the general load/store multiple > patterns are handled by ldmstmd.md so not having this before reload is > not risky. > > I'll also propose a variant of > https://gcc.gnu.org/ml/gcc-patches/2016-12/msg01400.html to tighten up > the consecutivity > check when 'consecutive' is passed down as true, but that is > indepdendent of this PR. > > Is the patch at https://gcc.gnu.org/ml/gcc-patches/2016-11/msg03078.html > ok for trunk? > It's been in my tree for a couple of months now getting testing and > there's been no fallout. > We shouldn't generate any ldm / stm operations for anything other than memcpy before reload and all of those are capped at 4 integer registers (MAX_LDM_STM_OPS in arm.h) at a time and that will be handled almost entirely by the patterns in ldmstm.md. The way to generate this generically this would be through movmem expanders or through the load_multiple expander in the backend. However all of this falls through the arm_gen_load_multiple interface all of which goes through arm_gen_multiple_op and we have an assert in arm_gen_multiple_op that the number of registers is <= MAX_LDM_STM_OPS. Thus OK, please apply but be aware of any fallout Thanks, Ramana > Thanks, > Kyrill > >> >> >> [1] https://gcc.gnu.org/ml/gcc-patches/2016-12/msg01400.html >> >>> >>> Jakub >> >
commit 996d28e2353badd1b29ef000f94d40c7dab9010f
Author: Kyrylo Tkachov <kyrylo.tkachov@arm.com>
Date: Tue Nov 29 15:07:30 2016 +0000
[ARM] Restrict *load_multiple pattern till after LRA
diff --git a/gcc/config/arm/arm.md b/gcc/config/arm/arm.md
index 74c44f3..22d2a84 100644
--- a/gcc/config/arm/arm.md
+++ b/gcc/config/arm/arm.md
@@ -11807,12 +11807,15 @@ (define_insn "<crc_variant>"
;; Patterns in ldmstm.md don't cover more than 4 registers. This pattern covers
;; large lists without explicit writeback generated for APCS_FRAME epilogue.
+;; The operands are validated through the load_multiple_operation
+;; match_parallel predicate rather than through constraints so enable it only
+;; after reload.
(define_insn "*load_multiple"
[(match_parallel 0 "load_multiple_operation"
[(set (match_operand:SI 2 "s_register_operand" "=rk")
(mem:SI (match_operand:SI 1 "s_register_operand" "rk")))
])]
- "TARGET_32BIT"
+ "TARGET_32BIT && reload_completed"
"*
{
arm_output_multireg_pop (operands, /*return_pc=*/false,
diff --git a/gcc/testsuite/gcc.c-torture/compile/pr71436.c b/gcc/testsuite/gcc.c-torture/compile/pr71436.c
new file mode 100644
index 0000000..ab08d5d
--- /dev/null
+++ b/gcc/testsuite/gcc.c-torture/compile/pr71436.c
@@ -0,0 +1,35 @@
+/* PR target/71436. */
+
+#pragma pack(1)
+struct S0
+{
+ volatile int f0;
+ short f2;
+};
+
+void foo (struct S0 *);
+int a, d;
+static struct S0 b[5];
+static struct S0 c;
+void fn1 ();
+void
+main ()
+{
+ {
+ struct S0 e;
+ for (; d; fn1 ())
+ {
+ {
+ a = 3;
+ for (; a >= 0; a -= 1)
+ {
+ {
+ e = c;
+ }
+ b[a] = e;
+ }
+ }
+ }
+ }
+ foo (b);
+}
Hi all, In this awkward ICE we have a *load_multiple pattern that is being transformed in reload from: (insn 55 67 151 3 (parallel [ (set (reg:SI 0 r0) (mem/u/c:SI (reg/f:SI 147) [2 c+0 S4 A32])) (set (reg:SI 158 [ c+4 ]) (mem/u/c:SI (plus:SI (reg/f:SI 147) (const_int 4 [0x4])) [2 c+4 S4 A32])) ]) arm-crash.c:25 393 {*load_multiple} (expr_list:REG_UNUSED (reg:SI 0 r0) (nil))) into the invalid: (insn 55 67 70 3 (parallel [ (set (reg:SI 0 r0) (mem/u/c:SI (reg/f:SI 5 r5 [147]) [2 c+0 S4 A32])) (set (mem/c:SI (plus:SI (reg/f:SI 102 sfp) (const_int -4 [0xfffffffffffffffc])) [4 %sfp+-12 S4 A32]) (mem/u/c:SI (plus:SI (reg/f:SI 5 r5 [147]) (const_int 4 [0x4])) [2 c+4 S4 A32])) ]) arm-crash.c:25 393 {*load_multiple} (nil)) The operands of *load_multiple are not validated through constraints like LRA is used to, but rather through a match_parallel predicate which ends up calling ldm_stm_operation_p to validate the multiple sets. But this means that LRA cannot reason about the constraints properly. This two-regiseter load should not have used *load_multiple anyway, it should have used *ldm2_ from ldmstm.md and indeed it did until the loop2_invariant pass which copied the ldm2_ pattern: (insn 27 23 28 4 (parallel [ (set (reg:SI 0 r0) (mem/u/c:SI (reg/f:SI 147) [2 c+0 S4 A32])) (set (reg:SI 1 r1) (mem/u/c:SI (plus:SI (reg/f:SI 147) (const_int 4 [0x4])) [2 c+4 S4 A32])) ]) "ldm.c":25 385 {*ldm2_} (nil)) into: (insn 55 19 67 3 (parallel [ (set (reg:SI 0 r0) (mem/u/c:SI (reg/f:SI 147) [2 c+0 S4 A32])) (set (reg:SI 158) (mem/u/c:SI (plus:SI (reg/f:SI 147) (const_int 4 [0x4])) [2 c+4 S4 A32])) ]) "ldm.c":25 404 {*load_multiple} (expr_list:REG_UNUSED (reg:SI 0 r0) (nil))) Note that it now got recognised as load_multiple because the second register is not a hard register but the pseudo 158. In any case, the solution suggested in the PR (and I agree with it) is to restrict *load_multiple to after reload. The similar pattern *load_multiple_with_writeback also has a similar condition and the comment above *load_multiple says that it's used to generate epilogues, which is done after reload anyway. For pre-reload load-multiples the patterns in ldmstm.md should do just fine. Bootstrapped and tested on arm-none-linux-gnueabihf. Ok for trunk? Thanks, Kyrill 2016-11-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com> PR target/71436 * config/arm/arm.md (*load_multiple): Add reload_completed to matching condition. 2016-11-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com> PR target/71436 * gcc.c-torture/compile/pr71436.c: New test.