On Wed, Feb 06, 2019 at 05:45:17AM -0600, Segher Boessenkool wrote: > On Wed, Feb 06, 2019 at 04:43:18PM +1030, Alan Modra wrote: > > Segher, you'll recognize these as your patches from pr88343. All I've > > done here is give you a testcase for the legitimize_tls_address change > > (which you said you had no idea what the patch was for!) > > I don't see a testcase, and the rs6000_legitimize_tls_address part is new > to me.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88343#c21 > Oh you mean the stuff in the commit message below here :-) Yes. > > I'll make real testcases from that, let's see if things are fixed. Do you > know if it helps the glibs libm stuff? The patch fixes the libm testcase that Joseph attached to the bug. I believe we might have more places where we need to set uses_pic_offset_table for Linux ABIs, builtin_setjmp_receiver and the use of RS6000_PIC_OFFSET_TABLE_REGNUM in rs6000_longcall_ref. (I'd looked at the latter case and decided it wasn't strictly necessary because we'd get use_pic_offset_table set when emitting the call, but that isn't true for old -mbss-plt code.) > > Another question below: > > > ---- > > Fixes lack of r30 save/restore on powerpc-linux. > > > > // -m32 -fpic -ftls-model=initial-exec > > __thread char* p; > > char** f1 (void) { return &p; } > > > > and > > > > // -m32 -fpic -msecure-plt > > extern int foo (int); > > int f1 (int x) { return foo (x); } > > > > PR target/88343 > > * config/rs6000/rs6000.c (rs6000_legitimize_tls_address), > > (rs6000_call_sysv): Set uses_pic_offset_table. > > > > diff --git a/gcc/config/rs6000/rs6000.c b/gcc/config/rs6000/rs6000.c > > index 883361cabbe..ab01dee9b68 100644 > > --- a/gcc/config/rs6000/rs6000.c > > +++ b/gcc/config/rs6000/rs6000.c > > @@ -8705,7 +8705,10 @@ rs6000_legitimize_tls_address (rtx addr, enum > > tls_model model) > > else > > { > > if (flag_pic == 1) > > - got = gen_rtx_REG (Pmode, RS6000_PIC_OFFSET_TABLE_REGNUM); > > + { > > + crtl->uses_pic_offset_table = 1; > > + got = gen_rtx_REG (Pmode, RS6000_PIC_OFFSET_TABLE_REGNUM); > > This is the same as just > > got = pic_offset_table_rtx; Yes, I wondered about that too, then decided that it wasn't a stage 4 fix. It's true if rs6000_legitimize_tls_address isn't used by powerpc darwin. I believe you could use pic_offset_table_rtx for the TARGET_64BIT "got = gen_rtx_REG (Pmode, 2);" too. And if we're tidying the function it would be nice to use DEFAULT_ABI tests rather than some of the TARGET_64BIT tests. Register selection for local-exec model is an ABI thing, as is the GOT register choice. -- Alan Modra Australia Development Lab, IBM
