Before PCREL in POWER10, we were not allowed to perform sibcalls to longcall functions since callee's return would skip the TOC restore in the caller. However, with PCREL we can now safely perform a sibling call to longcall functions. The problem with the current code in rs6000_sibcall_aix is that it asserts we do not have a longcall and fixing that, it generates a direct call to a PLT stub, when it should generate an indirect sibcall due to -fno-plt. The solution here is to check for a pcrel longcall and emit an indirect sibcall using an inline plt stub in that case.
This passed bootstrap and regtesting on powerpc64le-linux with no regressions. Ok for trunk? This is marked as a GCC 11/12 regression. Ok for a GCC 11 backport after some burn-in on trunk? Peter gcc/ PR target/104894 * config/rs6000/rs6000.cc (rs6000_sibcall_aix): Handle pcrel sibcalls to longcall functions. gcc/testsuite/ PR target/104894 * gcc.target/powerpc/pr104894.c: New test. * gcc.target/powerpc/pr104894-2.c: New test. diff --git a/gcc/config/rs6000/rs6000.cc b/gcc/config/rs6000/rs6000.cc index cb18db06a2d..d38a1d61cfe 100644 --- a/gcc/config/rs6000/rs6000.cc +++ b/gcc/config/rs6000/rs6000.cc @@ -25659,11 +25659,21 @@ rs6000_sibcall_aix (rtx value, rtx func_desc, rtx tlsarg, rtx cookie) rtx r12 = NULL_RTX; rtx func_addr = func_desc; - gcc_assert (INTVAL (cookie) == 0); - if (global_tlsarg) tlsarg = global_tlsarg; + /* Handle longcall attributes. */ + if ((INTVAL (cookie) & CALL_LONG) != 0 + && GET_CODE (func_desc) == SYMBOL_REF) + { + /* Only PCREL can do a sibling call to a longcall function + because we don't need to restore the TOC register. */ + gcc_assert (rs6000_pcrel_p ()); + func_desc = rs6000_longcall_ref (func_desc, tlsarg); + } + else + gcc_assert (INTVAL (cookie) == 0); + /* For ELFv2, r12 and CTR need to hold the function address for an indirect call. */ if (GET_CODE (func_desc) != SYMBOL_REF && DEFAULT_ABI == ABI_ELFv2) diff --git a/gcc/testsuite/gcc.target/powerpc/pr104894.c b/gcc/testsuite/gcc.target/powerpc/pr104894.c new file mode 100644 index 00000000000..f46fe88168f --- /dev/null +++ b/gcc/testsuite/gcc.target/powerpc/pr104894.c @@ -0,0 +1,20 @@ +/* PR target/104894 */ +/* { dg-require-effective-target powerpc_elfv2 } */ +/* { dg-require-effective-target power10_ok } */ +/* { dg-options "-O2 -mdejagnu-cpu=power10 -fno-plt" } */ + +/* Verify we do not ICE on the following test case and that we emit an + indirect sibcall, with r12 and CTR containing the function address. */ + +void foo (void); + +void +bar (void) +{ + foo (); +} + +/* { dg-final { scan-assembler-times {\mmtctr 12\M} 1 } } */ +/* { dg-final { scan-assembler-times {\mbctr\M} 1 } } */ +/* { dg-final { scan-assembler-not {\mbl\M} } } */ +/* { dg-final { scan-assembler-not {\mbctrl\M} } } */ diff --git a/gcc/testsuite/gcc.target/powerpc/pr104894-2.c b/gcc/testsuite/gcc.target/powerpc/pr104894-2.c new file mode 100644 index 00000000000..d1a011ef4d9 --- /dev/null +++ b/gcc/testsuite/gcc.target/powerpc/pr104894-2.c @@ -0,0 +1,22 @@ +/* PR target/104894 */ +/* { dg-require-effective-target powerpc_elfv2 } */ +/* { dg-require-effective-target power10_ok } */ +/* { dg-options "-O2 -mdejagnu-cpu=power10 -fno-plt" } */ + +/* Verify we do not ICE on the following test case and that we emit one + indirect call and one indirect sibcall, with r12 and CTR containing + the function addresses. */ + +void foo (void); + +void +bar (void) +{ + foo (); + foo (); +} + +/* { dg-final { scan-assembler-times {\mmtctr 12\M} 2 } } */ +/* { dg-final { scan-assembler-times {\mbctrl\M} 1 } } */ +/* { dg-final { scan-assembler-times {\mbctr\M} 1 } } */ +/* { dg-final { scan-assembler-not {\mbl\M} } } */