Message ID | 6c315e9c-f86b-5704-feea-73e434ecffee@linux.ibm.com |
---|---|
State | New |
Headers | show |
Series | [v3,rs6000] Implemented f[min/max]_optab by xs[min/max]dp [PR103605] | expand |
Hi, Gentle ping this: https://gcc.gnu.org/pipermail/gcc-patches/2022-May/595164.html Thanks. On 18/5/2022 下午 4:52, HAO CHEN GUI wrote: > Hi, > This patch implements optab f[min/max]_optab by xs[min/max]dp on rs6000. > Tests show that outputs of xs[min/max]dp are consistent with the standard > of C99 fmin/max. > > This patch also binds __builtin_vsx_xs[min/max]dp to fmin/max instead > of smin/max. So the builtins always generate xs[min/max]dp on all > platforms. > > Bootstrapped and tested on ppc64 Linux BE and LE with no regressions. > Is this okay for trunk? Any recommendations? Thanks a lot. > > ChangeLog > 2022-05-18 Haochen Gui <guihaoc@linux.ibm.com> > > gcc/ > PR target/103605 > * rs6000.md (FMINMAX): New. > (minmax_op): New. > (f<minmax_op><mode>3): New pattern by UNSPEC_FMAX and UNSPEC_FMIN. > * rs6000-builtins.def (__builtin_vsx_xsmaxdp): Set pattern to fmaxdf3. > (__builtin_vsx_xsmindp): Set pattern to fmindf3. > > gcc/testsuite/ > PR target/103605 > * gcc.dg/pr103605.c: New. > > patch.diff > diff --git a/gcc/config/rs6000/rs6000-builtins.def b/gcc/config/rs6000/rs6000-builtins.def > index f4a9f24bcc5..8b735493b40 100644 > --- a/gcc/config/rs6000/rs6000-builtins.def > +++ b/gcc/config/rs6000/rs6000-builtins.def > @@ -1613,10 +1613,10 @@ > XSCVSPDP vsx_xscvspdp {} > > const double __builtin_vsx_xsmaxdp (double, double); > - XSMAXDP smaxdf3 {} > + XSMAXDP fmaxdf3 {} > > const double __builtin_vsx_xsmindp (double, double); > - XSMINDP smindf3 {} > + XSMINDP fmindf3 {} > > const double __builtin_vsx_xsrdpi (double); > XSRDPI vsx_xsrdpi {} > diff --git a/gcc/config/rs6000/rs6000.md b/gcc/config/rs6000/rs6000.md > index bf85baa5370..197de0838ee 100644 > --- a/gcc/config/rs6000/rs6000.md > +++ b/gcc/config/rs6000/rs6000.md > @@ -158,6 +158,8 @@ (define_c_enum "unspec" > UNSPEC_HASHCHK > UNSPEC_XXSPLTIDP_CONST > UNSPEC_XXSPLTIW_CONST > + UNSPEC_FMAX > + UNSPEC_FMIN > ]) > > ;; > @@ -5341,6 +5343,22 @@ (define_insn_and_split "*s<minmax><mode>3_fpr" > DONE; > }) > > + > +(define_int_iterator FMINMAX [UNSPEC_FMAX UNSPEC_FMIN]) > + > +(define_int_attr minmax_op [(UNSPEC_FMAX "max") > + (UNSPEC_FMIN "min")]) > + > +(define_insn "f<minmax_op><mode>3" > + [(set (match_operand:SFDF 0 "vsx_register_operand" "=wa") > + (unspec:SFDF [(match_operand:SFDF 1 "vsx_register_operand" "wa") > + (match_operand:SFDF 2 "vsx_register_operand" "wa")] > + FMINMAX))] > +"TARGET_VSX" > +"xs<minmax_op>dp %x0,%x1,%x2" > +[(set_attr "type" "fp")] > +) > + > (define_expand "mov<mode>cc" > [(set (match_operand:GPR 0 "gpc_reg_operand") > (if_then_else:GPR (match_operand 1 "comparison_operator") > diff --git a/gcc/testsuite/gcc.target/powerpc/pr103605.c b/gcc/testsuite/gcc.target/powerpc/pr103605.c > new file mode 100644 > index 00000000000..e43ac40c2d1 > --- /dev/null > +++ b/gcc/testsuite/gcc.target/powerpc/pr103605.c > @@ -0,0 +1,37 @@ > +/* { dg-do compile } */ > +/* { dg-require-effective-target powerpc_vsx_ok } */ > +/* { dg-options "-O1 -mvsx" } */ > +/* { dg-final { scan-assembler-times {\mxsmaxdp\M} 3 } } */ > +/* { dg-final { scan-assembler-times {\mxsmindp\M} 3 } } */ > + > +#include <math.h> > + > +double test1 (double d0, double d1) > +{ > + return fmin (d0, d1); > +} > + > +float test2 (float d0, float d1) > +{ > + return fmin (d0, d1); > +} > + > +double test3 (double d0, double d1) > +{ > + return fmax (d0, d1); > +} > + > +float test4 (float d0, float d1) > +{ > + return fmax (d0, d1); > +} > + > +double test5 (double d0, double d1) > +{ > + return __builtin_vsx_xsmindp (d0, d1); > +} > + > +double test6 (double d0, double d1) > +{ > + return __builtin_vsx_xsmaxdp (d0, d1); > +}
Hi Haochen, on 2022/5/18 16:52, HAO CHEN GUI wrote: > Hi, > This patch implements optab f[min/max]_optab by xs[min/max]dp on rs6000. > Tests show that outputs of xs[min/max]dp are consistent with the standard > of C99 fmin/max. > > This patch also binds __builtin_vsx_xs[min/max]dp to fmin/max instead > of smin/max. So the builtins always generate xs[min/max]dp on all > platforms. > > Bootstrapped and tested on ppc64 Linux BE and LE with no regressions. > Is this okay for trunk? Any recommendations? Thanks a lot. > > ChangeLog > 2022-05-18 Haochen Gui <guihaoc@linux.ibm.com> > > gcc/ > PR target/103605 > * rs6000.md (FMINMAX): New. > (minmax_op): New. > (f<minmax_op><mode>3): New pattern by UNSPEC_FMAX and UNSPEC_FMIN. > * rs6000-builtins.def (__builtin_vsx_xsmaxdp): Set pattern to fmaxdf3. > (__builtin_vsx_xsmindp): Set pattern to fmindf3. > These changelog entries look wrong to me, they miss the relative path names. * config/rs6000/rs6000-builtins.def ... * config/rs6000/rs6000.md ... > gcc/testsuite/ > PR target/103605 > * gcc.dg/pr103605.c: New. ... and wrong path here. * gcc.target/powerpc/pr103605.c: New test. OK with the changelog above fixed. Thanks! BR, Kewen > > patch.diff > diff --git a/gcc/config/rs6000/rs6000-builtins.def b/gcc/config/rs6000/rs6000-builtins.def > index f4a9f24bcc5..8b735493b40 100644 > --- a/gcc/config/rs6000/rs6000-builtins.def > +++ b/gcc/config/rs6000/rs6000-builtins.def > @@ -1613,10 +1613,10 @@ > XSCVSPDP vsx_xscvspdp {} > > const double __builtin_vsx_xsmaxdp (double, double); > - XSMAXDP smaxdf3 {} > + XSMAXDP fmaxdf3 {} > > const double __builtin_vsx_xsmindp (double, double); > - XSMINDP smindf3 {} > + XSMINDP fmindf3 {} > > const double __builtin_vsx_xsrdpi (double); > XSRDPI vsx_xsrdpi {} > diff --git a/gcc/config/rs6000/rs6000.md b/gcc/config/rs6000/rs6000.md > index bf85baa5370..197de0838ee 100644 > --- a/gcc/config/rs6000/rs6000.md > +++ b/gcc/config/rs6000/rs6000.md > @@ -158,6 +158,8 @@ (define_c_enum "unspec" > UNSPEC_HASHCHK > UNSPEC_XXSPLTIDP_CONST > UNSPEC_XXSPLTIW_CONST > + UNSPEC_FMAX > + UNSPEC_FMIN > ]) > > ;; > @@ -5341,6 +5343,22 @@ (define_insn_and_split "*s<minmax><mode>3_fpr" > DONE; > }) > > + > +(define_int_iterator FMINMAX [UNSPEC_FMAX UNSPEC_FMIN]) > + > +(define_int_attr minmax_op [(UNSPEC_FMAX "max") > + (UNSPEC_FMIN "min")]) > + > +(define_insn "f<minmax_op><mode>3" > + [(set (match_operand:SFDF 0 "vsx_register_operand" "=wa") > + (unspec:SFDF [(match_operand:SFDF 1 "vsx_register_operand" "wa") > + (match_operand:SFDF 2 "vsx_register_operand" "wa")] > + FMINMAX))] > +"TARGET_VSX" > +"xs<minmax_op>dp %x0,%x1,%x2" > +[(set_attr "type" "fp")] > +) > + > (define_expand "mov<mode>cc" > [(set (match_operand:GPR 0 "gpc_reg_operand") > (if_then_else:GPR (match_operand 1 "comparison_operator") > diff --git a/gcc/testsuite/gcc.target/powerpc/pr103605.c b/gcc/testsuite/gcc.target/powerpc/pr103605.c > new file mode 100644 > index 00000000000..e43ac40c2d1 > --- /dev/null > +++ b/gcc/testsuite/gcc.target/powerpc/pr103605.c > @@ -0,0 +1,37 @@ > +/* { dg-do compile } */ > +/* { dg-require-effective-target powerpc_vsx_ok } */ > +/* { dg-options "-O1 -mvsx" } */ > +/* { dg-final { scan-assembler-times {\mxsmaxdp\M} 3 } } */ > +/* { dg-final { scan-assembler-times {\mxsmindp\M} 3 } } */ > + > +#include <math.h> > + > +double test1 (double d0, double d1) > +{ > + return fmin (d0, d1); > +} > + > +float test2 (float d0, float d1) > +{ > + return fmin (d0, d1); > +} > + > +double test3 (double d0, double d1) > +{ > + return fmax (d0, d1); > +} > + > +float test4 (float d0, float d1) > +{ > + return fmax (d0, d1); > +} > + > +double test5 (double d0, double d1) > +{ > + return __builtin_vsx_xsmindp (d0, d1); > +} > + > +double test6 (double d0, double d1) > +{ > + return __builtin_vsx_xsmaxdp (d0, d1); > +}
Hi! Some more nitpicking... On Wed, May 18, 2022 at 04:52:26PM +0800, HAO CHEN GUI wrote: > const double __builtin_vsx_xsmaxdp (double, double); > - XSMAXDP smaxdf3 {} > + XSMAXDP fmaxdf3 {} > > const double __builtin_vsx_xsmindp (double, double); > - XSMINDP smindf3 {} > + XSMINDP fmindf3 {} Are s{min,max}df3 still used after this? > + UNSPEC_FMAX > + UNSPEC_FMIN Pity we have to do this as an unspec still, this should be handled by some generic code, with some new operator (fmin/fmax would be obvious names :-) ) > +(define_insn "f<minmax_op><mode>3" > + [(set (match_operand:SFDF 0 "vsx_register_operand" "=wa") > + (unspec:SFDF [(match_operand:SFDF 1 "vsx_register_operand" "wa") > + (match_operand:SFDF 2 "vsx_register_operand" "wa")] > + FMINMAX))] > +"TARGET_VSX" > +"xs<minmax_op>dp %x0,%x1,%x2" > +[(set_attr "type" "fp")] > +) Indentation is broken here, correct is (define_insn "f<minmax_op><mode>3" [(set (match_operand:SFDF 0 "vsx_register_operand" "=wa") (unspec:SFDF [(match_operand:SFDF 1 "vsx_register_operand" "wa") (match_operand:SFDF 2 "vsx_register_operand" "wa")] FMINMAX))] "TARGET_VSX" "xs<minmax_op>dp %x0,%x1,%x2" [(set_attr "type" "fp")]) (FMINMAX has the same indent as the preceding [, its sibling; "TARGET_VSX" and the next two lines are indented like the same thing before it at the same level (the "[(set"); the finishing ) does never start a new line). Segher
Hi, On 2/6/2022 上午 5:01, Segher Boessenkool wrote: > Hi! > > Some more nitpicking... > > On Wed, May 18, 2022 at 04:52:26PM +0800, HAO CHEN GUI wrote: >> const double __builtin_vsx_xsmaxdp (double, double); >> - XSMAXDP smaxdf3 {} >> + XSMAXDP fmaxdf3 {} >> >> const double __builtin_vsx_xsmindp (double, double); >> - XSMINDP smindf3 {} >> + XSMINDP fmindf3 {} > > Are s{min,max}df3 still used after this? Expands reduc_s[min|max]_scal are still using s[min|max]df3. OPTAB_D (reduc_smax_scal_optab, "reduc_smax_scal_$a") OPTAB_D (reduc_smin_scal_optab, "reduc_smin_scal_$a") Also we could implement reduc_f[min|max]_scal after committing this patch. Thanks. Gui Haochen > >> + UNSPEC_FMAX >> + UNSPEC_FMIN > > Pity we have to do this as an unspec still, this should be handled by > some generic code, with some new operator (fmin/fmax would be obvious > names :-) ) > >> +(define_insn "f<minmax_op><mode>3" >> + [(set (match_operand:SFDF 0 "vsx_register_operand" "=wa") >> + (unspec:SFDF [(match_operand:SFDF 1 "vsx_register_operand" "wa") >> + (match_operand:SFDF 2 "vsx_register_operand" "wa")] >> + FMINMAX))] >> +"TARGET_VSX" >> +"xs<minmax_op>dp %x0,%x1,%x2" >> +[(set_attr "type" "fp")] >> +) > > Indentation is broken here, correct is > > (define_insn "f<minmax_op><mode>3" > [(set (match_operand:SFDF 0 "vsx_register_operand" "=wa") > (unspec:SFDF [(match_operand:SFDF 1 "vsx_register_operand" "wa") > (match_operand:SFDF 2 "vsx_register_operand" "wa")] > FMINMAX))] > "TARGET_VSX" > "xs<minmax_op>dp %x0,%x1,%x2" > [(set_attr "type" "fp")]) > > (FMINMAX has the same indent as the preceding [, its sibling; > "TARGET_VSX" and the next two lines are indented like the same thing > before it at the same level (the "[(set"); the finishing ) does never > start a new line). > > > Segher
Hi! On Tue, Jun 07, 2022 at 09:51:43AM +0800, HAO CHEN GUI wrote: > On 2/6/2022 上午 5:01, Segher Boessenkool wrote: > > On Wed, May 18, 2022 at 04:52:26PM +0800, HAO CHEN GUI wrote: > >> const double __builtin_vsx_xsmaxdp (double, double); > >> - XSMAXDP smaxdf3 {} > >> + XSMAXDP fmaxdf3 {} > >> > >> const double __builtin_vsx_xsmindp (double, double); > >> - XSMINDP smindf3 {} > >> + XSMINDP fmindf3 {} > > > > Are s{min,max}df3 still used after this? > > Expands reduc_s[min|max]_scal are still using s[min|max]df3. Ah, and reduction is undefined for float unless fast-math anyway. Okay. > Also we could implement reduc_f[min|max]_scal after committing > this patch. Because fmin and fmax treat NaNs as missing data. But you have to be careful with SNaNs here, probably just skip most of this when -fsignaling-nans is used. It also still needs -fno-signed-zeros btw, but not full -ffast-math indeed. Segher
diff --git a/gcc/config/rs6000/rs6000-builtins.def b/gcc/config/rs6000/rs6000-builtins.def index f4a9f24bcc5..8b735493b40 100644 --- a/gcc/config/rs6000/rs6000-builtins.def +++ b/gcc/config/rs6000/rs6000-builtins.def @@ -1613,10 +1613,10 @@ XSCVSPDP vsx_xscvspdp {} const double __builtin_vsx_xsmaxdp (double, double); - XSMAXDP smaxdf3 {} + XSMAXDP fmaxdf3 {} const double __builtin_vsx_xsmindp (double, double); - XSMINDP smindf3 {} + XSMINDP fmindf3 {} const double __builtin_vsx_xsrdpi (double); XSRDPI vsx_xsrdpi {} diff --git a/gcc/config/rs6000/rs6000.md b/gcc/config/rs6000/rs6000.md index bf85baa5370..197de0838ee 100644 --- a/gcc/config/rs6000/rs6000.md +++ b/gcc/config/rs6000/rs6000.md @@ -158,6 +158,8 @@ (define_c_enum "unspec" UNSPEC_HASHCHK UNSPEC_XXSPLTIDP_CONST UNSPEC_XXSPLTIW_CONST + UNSPEC_FMAX + UNSPEC_FMIN ]) ;; @@ -5341,6 +5343,22 @@ (define_insn_and_split "*s<minmax><mode>3_fpr" DONE; }) + +(define_int_iterator FMINMAX [UNSPEC_FMAX UNSPEC_FMIN]) + +(define_int_attr minmax_op [(UNSPEC_FMAX "max") + (UNSPEC_FMIN "min")]) + +(define_insn "f<minmax_op><mode>3" + [(set (match_operand:SFDF 0 "vsx_register_operand" "=wa") + (unspec:SFDF [(match_operand:SFDF 1 "vsx_register_operand" "wa") + (match_operand:SFDF 2 "vsx_register_operand" "wa")] + FMINMAX))] +"TARGET_VSX" +"xs<minmax_op>dp %x0,%x1,%x2" +[(set_attr "type" "fp")] +) + (define_expand "mov<mode>cc" [(set (match_operand:GPR 0 "gpc_reg_operand") (if_then_else:GPR (match_operand 1 "comparison_operator") diff --git a/gcc/testsuite/gcc.target/powerpc/pr103605.c b/gcc/testsuite/gcc.target/powerpc/pr103605.c new file mode 100644 index 00000000000..e43ac40c2d1 --- /dev/null +++ b/gcc/testsuite/gcc.target/powerpc/pr103605.c @@ -0,0 +1,37 @@ +/* { dg-do compile } */ +/* { dg-require-effective-target powerpc_vsx_ok } */ +/* { dg-options "-O1 -mvsx" } */ +/* { dg-final { scan-assembler-times {\mxsmaxdp\M} 3 } } */ +/* { dg-final { scan-assembler-times {\mxsmindp\M} 3 } } */ + +#include <math.h> + +double test1 (double d0, double d1) +{ + return fmin (d0, d1); +} + +float test2 (float d0, float d1) +{ + return fmin (d0, d1); +} + +double test3 (double d0, double d1) +{ + return fmax (d0, d1); +} + +float test4 (float d0, float d1) +{ + return fmax (d0, d1); +} + +double test5 (double d0, double d1) +{ + return __builtin_vsx_xsmindp (d0, d1); +} + +double test6 (double d0, double d1) +{ + return __builtin_vsx_xsmaxdp (d0, d1); +}