Message ID | 547CA43A.2080005@arm.com |
---|---|
State | New |
Headers | show |
CCing release maintainers as well as they need to approve this backport if Vlad is happy with it. Vlad - is this ok to go back as it fixes a bug for ARM in the 4.9 tree that came up in building bits of debian. Ramana On Mon, Dec 1, 2014 at 5:24 PM, Renlin Li <renlin.li@arm.com> wrote: > On 01/12/14 15:58, H.J. Lu wrote: >> >> On Thu, Nov 27, 2014 at 8:38 AM, Renlin Li <renlin.li@arm.com> wrote: >>> >>> On 27/11/14 15:37, H.J. Lu wrote: >>>> >>>> On Thu, Nov 27, 2014 at 7:32 AM, Renlin Li <renlin.li@arm.com> wrote: >>>>> >>>>> On 26/11/14 18:12, H.J. Lu wrote: >>>>>> >>>>>> On Wed, Nov 26, 2014 at 10:09 AM, Renlin Li <renlin.li@arm.com> wrote: >>>>>>> >>>>>>> On 26/11/14 12:16, H.J. Lu wrote: >>>>>>>> >>>>>>>> On Wed, Nov 26, 2014 at 4:07 AM, Renlin Li <renlin.li@arm.com> >>>>>>>> wrote: >>>>>>>>> >>>>>>>>> On 20/11/14 16:17, Renlin Li wrote: >>>>>>>>>> >>>>>>>>>> Hi all, >>>>>>>>>> >>>>>>>>>> This is a backport for gcc-4_9-branch of the patch "[PR63762]GCC >>>>>>>>>> generates >>>>>>>>>> UNPREDICTABLE STR with Rn = Rt for arm" posted in: >>>>>>>>>> https://gcc.gnu.org/ml/gcc-patches/2014-11/msg02253.html >>>>>>>>>> >>>>>>>>>> arm-none-eabi has been test on the model, no new issues. >>>>>>>>>> bootstrapping >>>>>>>>>> and >>>>>>>>>> regression tested on x86, no new issues. >>>>>>>>>> >>>>>>>>>> Is it Okay for gcc-4_9-branch? >>>>>>>>>> >>>>>>>>>> gcc/ChangeLog: >>>>>>>>>> >>>>>>>>>> 2014-11-20 Renlin Li <Renlin.Li@arm.com> >>>>>>>>>> >>>>>>>>>> PR middle-end/63762 >>>>>>>>>> * ira.c (ira): Update preferred class. >>>>>>>>>> >>>>>>>>>> gcc/testsuite/ChangeLog: >>>>>>>>>> >>>>>>>>>> 2014-11-20 Renlin Li <Renlin.Li@arm.com> >>>>>>>>>> >>>>>>>>>> PR middle-end/63762 >>>>>>>>>> * gcc.dg/pr63762.c: New. >>>>>>>>> >>>>>>>>> Ping for it. >>>>>>>>> >>>>>>>> Please verify if it is the real fix for >>>>>>>> >>>>>>>> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63661 >>>>>>>> >>>>>>>> If yes, please add a testcase for PR 63661 and mention it in >>>>>>>> your ChangeLog entry. >>>>>>>> >>>>>>>> Thanks. >>>>>>>> >>>>>>>> >>>>>>> Hi H.J. >>>>>>> >>>>>>> Yes, I have verified that, this patch additionally fixes PR 63661. >>>>>>> >>>>>>> I observed the same behaviour as I saw on arm backend. It will be >>>>>>> great >>>>>>> if >>>>>>> you can double check they are caused by exactly the same reason. >>>>>> >>>>>> I will ask our people to take a look. >>>>>> >>>>>>> A new testcase has been added, ChangeLog has been updated to reflect >>>>>>> the >>>>>>> change. Updated patch has bee attached. >>>>>>> Okay for gcc-4_9-branch? >>>>>>> >>>>>>> Regards, >>>>>>> Renlin Li >>>>>>> >>>>>>> >>>>>>> gcc/ChangeLog: >>>>>>> >>>>>>> 2014-11-26 Renlin Li <Renlin.Li@arm.com> >>>>>>> >>>>>>> PR middle-end/63762 >>>>>>> PR middle-end/63661 >>>>>>> * ira.c (ira): Update preferred class. >>>>>>> >>>>>>> gcc/testsuite/ChangeLog: >>>>>>> >>>>>>> 2014-11-26 Renlin Li <Renlin.Li@arm.com> >>>>>>> >>>>>>> PR middle-end/63661 >>>>>>> PR middle-end/63762 >>>>>>> * testsuite/gcc.dg/pr63661.c: New. >>>>>>> * testsuite/gcc.dg/pr63762.c: New. >>>>>>> >>>>>>> >>>>>> pr63661.c should be moved to gcc.target/i386 and run it >>>>>> on PIC target. >>>>>> >>>>>> Thanks. >>>>>> >>>>> Hi H.J. >>>>> >>>>> The patch has been adjusted according to your suggestion. >>>>> >>>>> >>>>> gcc/ChangeLog: >>>>> >>>>> 2014-11-27 Renlin Li <Renlin.Li@arm.com> >>>>> >>>>> PR middle-end/63762 >>>>> PR target/63661 >>>>> * ira.c (ira): Update preferred class. >>>>> >>>>> gcc/testsuite/ChangeLog: >>>>> >>>>> 2014-11-27 Renlin Li <Renlin.Li@arm.com> >>>>> >>>>> PR middle-end/63762 >>>>> PR target/63661 >>>>> * testsuite/gcc.dg/pr63762.c: New. >>>>> * testsuite/gcc.target/i386/pr63661.c: New. >>>> >>>> ^^^^^^^^^^^ No testsuite/ >>> >>> >>> Hi H.J. >>>> >>>> gcc.target/i386/pr63661.c should be checked into trunk first. >>>> >>> A separate patch is sent to mailing list for this. >>> >>> >>> ChangeLog has been corrected. >>> >>> Regards, >>> Renlin Li >>> >>> gcc/ChangeLog: >>> >>> 2014-11-27 Renlin Li <Renlin.Li@arm.com> >>> >>> PR middle-end/63762 >>> PR target/63661 >>> * ira.c (ira): Update preferred class. >>> >>> gcc/testsuite/ChangeLog: >>> >>> 2014-11-27 Renlin Li <Renlin.Li@arm.com> >>> >>> PR middle-end/63762 >>> PR target/63661 >>> * gcc.dg/pr63762.c: New. >>> * gcc.target/i386/pr63661.c: New. >>> >>> >> You need to update gcc.target/i386/pr63661.c from trunk. >> >> > > Update the test case for pr63661 from trunk. > > Okay for gcc-4_9-branch? > > gcc/ChangeLog: > > 2014-12-01 Renlin Li <Renlin.Li@arm.com> > > PR middle-end/63762 > PR target/63661 > * ira.c (ira): Update preferred class. > > gcc/testsuite/ChangeLog: > > 2014-12-01 Renlin Li <Renlin.Li@arm.com> > H.J Lu <hongjiu.lu@intel.com> > > > PR middle-end/63762 > PR target/63661 > * gcc.dg/pr63762.c: New. > * gcc.target/i386/pr63661.c: New.
On 2014-12-01 12:24 PM, Renlin Li wrote: > On 01/12/14 15:58, H.J. Lu wrote: >> On Thu, Nov 27, 2014 at 8:38 AM, Renlin Li <renlin.li@arm.com> wrote: >>> On 27/11/14 15:37, H.J. Lu wrote: >>>> On Thu, Nov 27, 2014 at 7:32 AM, Renlin Li <renlin.li@arm.com> wrote: >>>>> On 26/11/14 18:12, H.J. Lu wrote: >>>>>> On Wed, Nov 26, 2014 at 10:09 AM, Renlin Li <renlin.li@arm.com> >>>>>> wrote: >>>>>>> On 26/11/14 12:16, H.J. Lu wrote: >>>>>>>> On Wed, Nov 26, 2014 at 4:07 AM, Renlin Li <renlin.li@arm.com> >>>>>>>> wrote: >>>>>>>>> On 20/11/14 16:17, Renlin Li wrote: >>>>>>>>>> Hi all, >>>>>>>>>> >>>>>>>>>> This is a backport for gcc-4_9-branch of the patch "[PR63762]GCC >>>>>>>>>> generates >>>>>>>>>> UNPREDICTABLE STR with Rn = Rt for arm" posted in: >>>>>>>>>> https://gcc.gnu.org/ml/gcc-patches/2014-11/msg02253.html >>>>>>>>>> >>>>>>>>>> arm-none-eabi has been test on the model, no new issues. >>>>>>>>>> bootstrapping >>>>>>>>>> and >>>>>>>>>> regression tested on x86, no new issues. >>>>>>>>>> >>>>>>>>>> Is it Okay for gcc-4_9-branch? >>>>>>>>>> >>>>>>>>>> gcc/ChangeLog: >>>>>>>>>> >>>>>>>>>> 2014-11-20 Renlin Li <Renlin.Li@arm.com> >>>>>>>>>> >>>>>>>>>> PR middle-end/63762 >>>>>>>>>> * ira.c (ira): Update preferred class. >>>>>>>>>> >>>>>>>>>> gcc/testsuite/ChangeLog: >>>>>>>>>> >>>>>>>>>> 2014-11-20 Renlin Li <Renlin.Li@arm.com> >>>>>>>>>> >>>>>>>>>> PR middle-end/63762 >>>>>>>>>> * gcc.dg/pr63762.c: New. >>>>>>>>> Ping for it. >>>>>>>>> >>>>>>>> Please verify if it is the real fix for >>>>>>>> >>>>>>>> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63661 >>>>>>>> >>>>>>>> If yes, please add a testcase for PR 63661 and mention it in >>>>>>>> your ChangeLog entry. >>>>>>>> >>>>>>>> Thanks. >>>>>>>> >>>>>>>> >>>>>>> Hi H.J. >>>>>>> >>>>>>> Yes, I have verified that, this patch additionally fixes PR 63661. >>>>>>> >>>>>>> I observed the same behaviour as I saw on arm backend. It will be >>>>>>> great >>>>>>> if >>>>>>> you can double check they are caused by exactly the same reason. >>>>>> I will ask our people to take a look. >>>>>> >>>>>>> A new testcase has been added, ChangeLog has been updated to reflect >>>>>>> the >>>>>>> change. Updated patch has bee attached. >>>>>>> Okay for gcc-4_9-branch? >>>>>>> >>>>>>> Regards, >>>>>>> Renlin Li >>>>>>> >>>>>>> >>>>>>> gcc/ChangeLog: >>>>>>> >>>>>>> 2014-11-26 Renlin Li <Renlin.Li@arm.com> >>>>>>> >>>>>>> PR middle-end/63762 >>>>>>> PR middle-end/63661 >>>>>>> * ira.c (ira): Update preferred class. >>>>>>> >>>>>>> gcc/testsuite/ChangeLog: >>>>>>> >>>>>>> 2014-11-26 Renlin Li <Renlin.Li@arm.com> >>>>>>> >>>>>>> PR middle-end/63661 >>>>>>> PR middle-end/63762 >>>>>>> * testsuite/gcc.dg/pr63661.c: New. >>>>>>> * testsuite/gcc.dg/pr63762.c: New. >>>>>>> >>>>>>> >>>>>> pr63661.c should be moved to gcc.target/i386 and run it >>>>>> on PIC target. >>>>>> >>>>>> Thanks. >>>>>> >>>>> Hi H.J. >>>>> >>>>> The patch has been adjusted according to your suggestion. >>>>> >>>>> >>>>> gcc/ChangeLog: >>>>> >>>>> 2014-11-27 Renlin Li <Renlin.Li@arm.com> >>>>> >>>>> PR middle-end/63762 >>>>> PR target/63661 >>>>> * ira.c (ira): Update preferred class. >>>>> >>>>> gcc/testsuite/ChangeLog: >>>>> >>>>> 2014-11-27 Renlin Li <Renlin.Li@arm.com> >>>>> >>>>> PR middle-end/63762 >>>>> PR target/63661 >>>>> * testsuite/gcc.dg/pr63762.c: New. >>>>> * testsuite/gcc.target/i386/pr63661.c: New. >>>> ^^^^^^^^^^^ No testsuite/ >>> >>> Hi H.J. >>>> gcc.target/i386/pr63661.c should be checked into trunk first. >>>> >>> A separate patch is sent to mailing list for this. >>> >>> >>> ChangeLog has been corrected. >>> >>> Regards, >>> Renlin Li >>> >>> gcc/ChangeLog: >>> >>> 2014-11-27 Renlin Li <Renlin.Li@arm.com> >>> >>> PR middle-end/63762 >>> PR target/63661 >>> * ira.c (ira): Update preferred class. >>> >>> gcc/testsuite/ChangeLog: >>> >>> 2014-11-27 Renlin Li <Renlin.Li@arm.com> >>> >>> PR middle-end/63762 >>> PR target/63661 >>> * gcc.dg/pr63762.c: New. >>> * gcc.target/i386/pr63661.c: New. >>> >>> >> You need to update gcc.target/i386/pr63661.c from trunk. >> >> > > Update the test case for pr63661 from trunk. > > Okay for gcc-4_9-branch? > Yes. Thanks for working on this, Renlin.
diff --git a/gcc/ira.c b/gcc/ira.c index 4d91d21..0c703c5 100644 --- a/gcc/ira.c +++ b/gcc/ira.c @@ -5347,7 +5347,18 @@ ira (FILE *f) ira_allocno_iterator ai; FOR_EACH_ALLOCNO (a, ai) - ALLOCNO_REGNO (a) = REGNO (ALLOCNO_EMIT_DATA (a)->reg); + { + int old_regno = ALLOCNO_REGNO (a); + int new_regno = REGNO (ALLOCNO_EMIT_DATA (a)->reg); + + ALLOCNO_REGNO (a) = new_regno; + + if (old_regno != new_regno) + setup_reg_classes (new_regno, reg_preferred_class (old_regno), + reg_alternate_class (old_regno), + reg_allocno_class (old_regno)); + } + } else { diff --git a/gcc/testsuite/gcc.dg/pr63762.c b/gcc/testsuite/gcc.dg/pr63762.c new file mode 100644 index 0000000..df11067 --- /dev/null +++ b/gcc/testsuite/gcc.dg/pr63762.c @@ -0,0 +1,77 @@ +/* PR middle-end/63762 */ +/* { dg-do assemble } */ +/* { dg-options "-O2" } */ + +#include <stdlib.h> + +void *astFree (); +void *astMalloc (); +void astNegate (void *); +int astGetNegated (void *); +void astGetRegionBounds (void *, double *, double *); +int astResampleF (void *, ...); + +extern int astOK; + +int +MaskF (int inside, int ndim, const int lbnd[], const int ubnd[], + float in[], float val) +{ + + void *used_region; + float *c, *d, *out, *tmp_out; + double *lbndgd, *ubndgd; + int *lbndg, *ubndg, idim, ipix, nax, nin, nout, npix, npixg, result = 0; + if (!astOK) return result; + lbndg = astMalloc (sizeof (int)*(size_t) ndim); + ubndg = astMalloc (sizeof (int)*(size_t) ndim); + lbndgd = astMalloc (sizeof (double)*(size_t) ndim); + ubndgd = astMalloc (sizeof (double)*(size_t) ndim); + if (astOK) + { + astGetRegionBounds (used_region, lbndgd, ubndgd); + npix = 1; + npixg = 1; + for (idim = 0; idim < ndim; idim++) + { + lbndg[ idim ] = lbnd[ idim ]; + ubndg[ idim ] = ubnd[ idim ]; + npix *= (ubnd[ idim ] - lbnd[ idim ] + 1); + if (npixg >= 0) npixg *= (ubndg[ idim ] - lbndg[ idim ] + 1); + } + if (npixg <= 0 && astOK) + { + if ((inside != 0) == (astGetNegated( used_region ) != 0)) + { + c = in; + for (ipix = 0; ipix < npix; ipix++) *(c++) = val; + result = npix; + } + } + else if (npixg > 0 && astOK) + { + if ((inside != 0) == (astGetNegated (used_region) != 0)) + { + tmp_out = astMalloc (sizeof (float)*(size_t) npix); + if (tmp_out) + { + c = tmp_out; + for (ipix = 0; ipix < npix; ipix++) *(c++) = val; + result = npix - npixg; + } + out = tmp_out; + } + else + { + tmp_out = NULL; + out = in; + } + if (inside) astNegate (used_region); + result += astResampleF (used_region, ndim, lbnd, ubnd, in, NULL, + NULL, NULL, 0, 0.0, 100, val, ndim, + lbnd, ubnd, lbndg, ubndg, out, NULL); + if (inside) astNegate (used_region); + } + } + return result; +} diff --git a/gcc/testsuite/gcc.target/i386/pr63661.c b/gcc/testsuite/gcc.target/i386/pr63661.c new file mode 100644 index 0000000..a5ffd2f --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/pr63661.c @@ -0,0 +1,80 @@ +/* PR target/63661 */ +/* { dg-do run } */ +/* { dg-require-effective-target fpic } */ +/* { dg-options "-mtune=nehalem -fPIC -O2" } */ + +static void __attribute__((noinline,noclone,hot)) +foo (double a, double q, double *ff, double *gx, int e, int ni) +{ + union + { + double n; + unsigned long long o; + } punner; + double d; + + punner.n = q; + __builtin_printf("B: 0x%016llx ---- %g\n", punner.o, q); + + d = q - 5; + if(d < 0) + d = -d; + if (d > 0.1) + __builtin_abort(); +} + +static int __attribute__((noinline,noclone,hot)) +bar (int order, double q, double c[]) +{ + int ni, nn, i, e; + double g2, x2, de, s, ratio, ff; + + nn = 0; + e = order & 1; + s = 0; + ratio = 0; + x2 = 0; + g2 = 0; + + if(q == 0.0) + return 0; + + if (order < 5) + { + ratio = 1.0 / q; + nn = order; + } + + ni = -nn; + + while(1) + { + de = ratio - g2 - x2; + + foo (0, q, &ff, &g2, e, ni); + + if((int)de == 0) + break; + } + + s += 2 * nn * c[nn]; + + for (i = 0; i < 1; i++) + { + c[0] = nn; + for (; i < 10; i++) + c[i] = 0.0; + c[0] /= s; + } + + return 0; +} + +int +main () +{ + double c[1000]; + + bar (1, 5.0, c); + return 0; +}