| Message ID | b620e634-3adf-dba7-51a1-25220a150d6c@suse.com |
|---|---|
| State | New |
| Headers | show |
| Series | [v3] x86: make better use of VBROADCASTSS / VPBROADCASTD | expand |
> -----Original Message----- > From: Jan Beulich <jbeulich@suse.com> > Sent: Tuesday, July 11, 2023 2:04 PM > To: gcc-patches@gcc.gnu.org > Cc: Kirill Yukhin <kirill.yukhin@gmail.com>; Liu, Hongtao > <hongtao.liu@intel.com> > Subject: [PATCH v3] x86: make better use of VBROADCASTSS / > VPBROADCASTD > > ... in vec_dupv4sf / *vec_dupv4si. The respective broadcast insns are never > longer (yet sometimes shorter) than the corresponding VSHUFPS / VPSHUFD, > due to the immediate operand of the shuffle insns balancing the > (uniform) need for VEX3 in the broadcast ones. When EVEX encoding is > respective the broadcast insns are always shorter. > > Add new alternatives to cover the AVX2 and AVX512 cases as appropriate. > > While touching this anyway, switch to consistently using "sseshuf1" in the > "type" attributes for all shuffle forms. > > gcc/ > > * config/i386/sse.md (vec_dupv4sf): Make first alternative use > vbroadcastss for AVX2. New AVX512F alternative. > (*vec_dupv4si): New AVX2 and AVX512F alternatives using > vpbroadcastd. Replace sselog1 by sseshuf1 in "type" attribute. > > gcc/testsuite/ > > * gcc.target/i386/avx2-dupv4sf.c: New test. > * gcc.target/i386/avx2-dupv4si.c: Likewise. > * gcc.target/i386/avx512f-dupv4sf.c: Likewise. > * gcc.target/i386/avx512f-dupv4si.c: Likewise. > --- > Note that unlike originally intended, "prefix_extra" isn't dropped: > "length_vex" uses it to determine whether 2-byte VEX encoding is possible > (which it isn't for VBROADCASTSS / VPBROADCASTD). "length" > itself specifically does not use it for VEX/EVEX encoded insns. > > Especially with the added "enabled" attribute I didn't really see how to > (further) fold alternatives 0 and 1. Instead *vec_dupv4si might benefit from > using sse2_noavx2 instead of sse2 for alternative 2, except that there is no > sse2_noavx2, only sse2_noavx. > > I'm working from the assumption that the isa attributes to the original 1st and > 2nd alternatives don't need further restricting (to sse2_noavx2 or > avx_noavx2 as applicable), as the new earlier alternatives cover all operand > forms already when at least AVX2 is enabled. Yes, the patch LGTM. > --- > v3: Testcases for new alternatives. "type" and "prefix_extra" > adjustments. > v2: Correct operand constraints. Respect -mprefer-vector-width=. Fold > two alternatives of vec_dupv4sf. > > --- a/gcc/config/i386/sse.md > +++ b/gcc/config/i386/sse.md > @@ -25969,41 +25969,64 @@ > (const_int 1)))]) > > (define_insn "vec_dupv4sf" > - [(set (match_operand:V4SF 0 "register_operand" "=v,v,x") > + [(set (match_operand:V4SF 0 "register_operand" "=v,v,v,x") > (vec_duplicate:V4SF > - (match_operand:SF 1 "nonimmediate_operand" "Yv,m,0")))] > + (match_operand:SF 1 "nonimmediate_operand" "Yv,v,m,0")))] > "TARGET_SSE" > "@ > - vshufps\t{$0, %1, %1, %0|%0, %1, %1, 0} > + * return TARGET_AVX2 ? \"vbroadcastss\t{%1, %0|%0, %1}\" : > \"vshufps\t{$0, %d1, %0|%0, %d1, 0}\"; > + vbroadcastss\t{%1, %g0|%g0, %1} > vbroadcastss\t{%1, %0|%0, %1} > shufps\t{$0, %0, %0|%0, %0, 0}" > - [(set_attr "isa" "avx,avx,noavx") > - (set_attr "type" "sseshuf1,ssemov,sseshuf1") > - (set_attr "length_immediate" "1,0,1") > - (set_attr "prefix_extra" "0,1,*") > - (set_attr "prefix" "maybe_evex,maybe_evex,orig") > - (set_attr "mode" "V4SF")]) > + [(set_attr "isa" "avx,*,avx,noavx") > + (set (attr "type") > + (cond [(and (eq_attr "alternative" "0") > + (match_test "!TARGET_AVX2")) > + (const_string "sseshuf1") > + (eq_attr "alternative" "3") > + (const_string "sseshuf1") > + ] > + (const_string "ssemov"))) > + (set (attr "length_immediate") > + (if_then_else (eq_attr "type" "sseshuf1") > + (const_string "1") > + (const_string "0"))) > + (set_attr "prefix_extra" "0,1,1,*") > + (set_attr "prefix" "maybe_evex,evex,maybe_evex,orig") > + (set_attr "mode" "V4SF,V16SF,V4SF,V4SF") > + (set (attr "enabled") > + (if_then_else (eq_attr "alternative" "1") > + (symbol_ref "TARGET_AVX512F && !TARGET_AVX512VL > + && !TARGET_PREFER_AVX256") > + (const_string "*")))]) > > (define_insn "*vec_dupv4si" > - [(set (match_operand:V4SI 0 "register_operand" "=v,v,x") > + [(set (match_operand:V4SI 0 "register_operand" "=v,v,v,v,x") > (vec_duplicate:V4SI > - (match_operand:SI 1 "nonimmediate_operand" "Yv,m,0")))] > + (match_operand:SI 1 "nonimmediate_operand" "Yvm,v,Yv,m,0")))] > "TARGET_SSE" > "@ > + vpbroadcastd\t{%1, %0|%0, %1} > + vpbroadcastd\t{%1, %g0|%g0, %1} > %vpshufd\t{$0, %1, %0|%0, %1, 0} > vbroadcastss\t{%1, %0|%0, %1} > shufps\t{$0, %0, %0|%0, %0, 0}" > - [(set_attr "isa" "sse2,avx,noavx") > - (set_attr "type" "sselog1,ssemov,sselog1") > - (set_attr "length_immediate" "1,0,1") > - (set_attr "prefix_extra" "0,1,*") > - (set_attr "prefix" "maybe_vex,maybe_evex,orig") > - (set_attr "mode" "TI,V4SF,V4SF") > + [(set_attr "isa" "avx2,*,sse2,avx,noavx") > + (set_attr "type" "ssemov,ssemov,sseshuf1,ssemov,sseshuf1") > + (set_attr "length_immediate" "0,0,1,0,1") > + (set_attr "prefix_extra" "1,1,0,1,*") > + (set_attr "prefix" "maybe_evex,evex,maybe_vex,maybe_evex,orig") > + (set_attr "mode" "TI,XI,TI,V4SF,V4SF") > (set (attr "preferred_for_speed") > - (cond [(eq_attr "alternative" "1") > + (cond [(eq_attr "alternative" "3") > (symbol_ref "!TARGET_INTER_UNIT_MOVES_TO_VEC") > ] > - (symbol_ref "true")))]) > + (symbol_ref "true"))) > + (set (attr "enabled") > + (if_then_else (eq_attr "alternative" "1") > + (symbol_ref "TARGET_AVX512F && !TARGET_AVX512VL > + && !TARGET_PREFER_AVX256") > + (const_string "*")))]) > > (define_insn "*vec_dupv2di" > [(set (match_operand:V2DI 0 "register_operand" "=x,v,v,v,x") > --- /dev/null > +++ b/gcc/testsuite/gcc.target/i386/avx2-dupv4sf.c > @@ -0,0 +1,18 @@ > +/* { dg-do compile } */ > +/* { dg-options "-mavx2 -O2" } */ > +/* { dg-final { scan-assembler-times "vbroadcastss" 2 } } */ > + > +typedef float __attribute__ ((vector_size (16))) v4sf; > + > +v4sf bcst_reg (float f) > +{ > + register float x asm ("xmm7") = f; > + > + asm ("" : "+v" (x)); > + return (v4sf) {x, x, x, x}; > +} > + > +v4sf bcst_mem (const float *f) > +{ > + return (v4sf) {*f, *f, *f, *f}; > +} > --- /dev/null > +++ b/gcc/testsuite/gcc.target/i386/avx2-dupv4si.c > @@ -0,0 +1,18 @@ > +/* { dg-do compile } */ > +/* { dg-options "-mavx2 -O2" } */ > +/* { dg-final { scan-assembler-times "vpbroadcastd" 2 } } */ > + > +typedef int __attribute__ ((vector_size (16))) v4si; > + > +v4si bcst_reg (int i) > +{ > + register int x asm ("xmm7") = i; > + > + asm ("" : "+v" (x)); > + return (v4si) {x, x, x, x}; > +} > + > +v4si bcst_mem (const int *i) > +{ > + return (v4si) {*i, *i, *i, *i}; > +} > --- /dev/null > +++ b/gcc/testsuite/gcc.target/i386/avx512f-dupv4sf.c > @@ -0,0 +1,13 @@ > +/* { dg-do compile { target { ! ia32 } } } */ > +/* { dg-options "-mavx512f -mno-avx512vl -mprefer-vector-width=512 -O2" > +} */ > +/* { dg-final { scan-assembler "vbroadcastss\[^\n\]*%xmm17, *%zmm" } } > +*/ > + > +typedef float __attribute__ ((vector_size (16))) v4sf; > + > +v4sf bcst (float f) > +{ > + register float x asm ("xmm17") = f; > + > + asm ("" : "+v" (x)); > + return (v4sf) {x, x, x, x}; > +} > --- /dev/null > +++ b/gcc/testsuite/gcc.target/i386/avx512f-dupv4si.c > @@ -0,0 +1,13 @@ > +/* { dg-do compile { target { ! ia32 } } } */ > +/* { dg-options "-mavx512f -mno-avx512vl -mprefer-vector-width=512 -O2" > +} */ > +/* { dg-final { scan-assembler "vpbroadcastd\[^\n\]*%xmm17, *%zmm" } } > +*/ > + > +typedef int __attribute__ ((vector_size (16))) v4si; > + > +v4si bcst (int i) > +{ > + register int x asm ("xmm17") = i; > + > + asm ("" : "+v" (x)); > + return (v4si) {x, x, x, x}; > +}
--- a/gcc/config/i386/sse.md +++ b/gcc/config/i386/sse.md @@ -25969,41 +25969,64 @@ (const_int 1)))]) (define_insn "vec_dupv4sf" - [(set (match_operand:V4SF 0 "register_operand" "=v,v,x") + [(set (match_operand:V4SF 0 "register_operand" "=v,v,v,x") (vec_duplicate:V4SF - (match_operand:SF 1 "nonimmediate_operand" "Yv,m,0")))] + (match_operand:SF 1 "nonimmediate_operand" "Yv,v,m,0")))] "TARGET_SSE" "@ - vshufps\t{$0, %1, %1, %0|%0, %1, %1, 0} + * return TARGET_AVX2 ? \"vbroadcastss\t{%1, %0|%0, %1}\" : \"vshufps\t{$0, %d1, %0|%0, %d1, 0}\"; + vbroadcastss\t{%1, %g0|%g0, %1} vbroadcastss\t{%1, %0|%0, %1} shufps\t{$0, %0, %0|%0, %0, 0}" - [(set_attr "isa" "avx,avx,noavx") - (set_attr "type" "sseshuf1,ssemov,sseshuf1") - (set_attr "length_immediate" "1,0,1") - (set_attr "prefix_extra" "0,1,*") - (set_attr "prefix" "maybe_evex,maybe_evex,orig") - (set_attr "mode" "V4SF")]) + [(set_attr "isa" "avx,*,avx,noavx") + (set (attr "type") + (cond [(and (eq_attr "alternative" "0") + (match_test "!TARGET_AVX2")) + (const_string "sseshuf1") + (eq_attr "alternative" "3") + (const_string "sseshuf1") + ] + (const_string "ssemov"))) + (set (attr "length_immediate") + (if_then_else (eq_attr "type" "sseshuf1") + (const_string "1") + (const_string "0"))) + (set_attr "prefix_extra" "0,1,1,*") + (set_attr "prefix" "maybe_evex,evex,maybe_evex,orig") + (set_attr "mode" "V4SF,V16SF,V4SF,V4SF") + (set (attr "enabled") + (if_then_else (eq_attr "alternative" "1") + (symbol_ref "TARGET_AVX512F && !TARGET_AVX512VL + && !TARGET_PREFER_AVX256") + (const_string "*")))]) (define_insn "*vec_dupv4si" - [(set (match_operand:V4SI 0 "register_operand" "=v,v,x") + [(set (match_operand:V4SI 0 "register_operand" "=v,v,v,v,x") (vec_duplicate:V4SI - (match_operand:SI 1 "nonimmediate_operand" "Yv,m,0")))] + (match_operand:SI 1 "nonimmediate_operand" "Yvm,v,Yv,m,0")))] "TARGET_SSE" "@ + vpbroadcastd\t{%1, %0|%0, %1} + vpbroadcastd\t{%1, %g0|%g0, %1} %vpshufd\t{$0, %1, %0|%0, %1, 0} vbroadcastss\t{%1, %0|%0, %1} shufps\t{$0, %0, %0|%0, %0, 0}" - [(set_attr "isa" "sse2,avx,noavx") - (set_attr "type" "sselog1,ssemov,sselog1") - (set_attr "length_immediate" "1,0,1") - (set_attr "prefix_extra" "0,1,*") - (set_attr "prefix" "maybe_vex,maybe_evex,orig") - (set_attr "mode" "TI,V4SF,V4SF") + [(set_attr "isa" "avx2,*,sse2,avx,noavx") + (set_attr "type" "ssemov,ssemov,sseshuf1,ssemov,sseshuf1") + (set_attr "length_immediate" "0,0,1,0,1") + (set_attr "prefix_extra" "1,1,0,1,*") + (set_attr "prefix" "maybe_evex,evex,maybe_vex,maybe_evex,orig") + (set_attr "mode" "TI,XI,TI,V4SF,V4SF") (set (attr "preferred_for_speed") - (cond [(eq_attr "alternative" "1") + (cond [(eq_attr "alternative" "3") (symbol_ref "!TARGET_INTER_UNIT_MOVES_TO_VEC") ] - (symbol_ref "true")))]) + (symbol_ref "true"))) + (set (attr "enabled") + (if_then_else (eq_attr "alternative" "1") + (symbol_ref "TARGET_AVX512F && !TARGET_AVX512VL + && !TARGET_PREFER_AVX256") + (const_string "*")))]) (define_insn "*vec_dupv2di" [(set (match_operand:V2DI 0 "register_operand" "=x,v,v,v,x") --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx2-dupv4sf.c @@ -0,0 +1,18 @@ +/* { dg-do compile } */ +/* { dg-options "-mavx2 -O2" } */ +/* { dg-final { scan-assembler-times "vbroadcastss" 2 } } */ + +typedef float __attribute__ ((vector_size (16))) v4sf; + +v4sf bcst_reg (float f) +{ + register float x asm ("xmm7") = f; + + asm ("" : "+v" (x)); + return (v4sf) {x, x, x, x}; +} + +v4sf bcst_mem (const float *f) +{ + return (v4sf) {*f, *f, *f, *f}; +} --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx2-dupv4si.c @@ -0,0 +1,18 @@ +/* { dg-do compile } */ +/* { dg-options "-mavx2 -O2" } */ +/* { dg-final { scan-assembler-times "vpbroadcastd" 2 } } */ + +typedef int __attribute__ ((vector_size (16))) v4si; + +v4si bcst_reg (int i) +{ + register int x asm ("xmm7") = i; + + asm ("" : "+v" (x)); + return (v4si) {x, x, x, x}; +} + +v4si bcst_mem (const int *i) +{ + return (v4si) {*i, *i, *i, *i}; +} --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512f-dupv4sf.c @@ -0,0 +1,13 @@ +/* { dg-do compile { target { ! ia32 } } } */ +/* { dg-options "-mavx512f -mno-avx512vl -mprefer-vector-width=512 -O2" } */ +/* { dg-final { scan-assembler "vbroadcastss\[^\n\]*%xmm17, *%zmm" } } */ + +typedef float __attribute__ ((vector_size (16))) v4sf; + +v4sf bcst (float f) +{ + register float x asm ("xmm17") = f; + + asm ("" : "+v" (x)); + return (v4sf) {x, x, x, x}; +} --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512f-dupv4si.c @@ -0,0 +1,13 @@ +/* { dg-do compile { target { ! ia32 } } } */ +/* { dg-options "-mavx512f -mno-avx512vl -mprefer-vector-width=512 -O2" } */ +/* { dg-final { scan-assembler "vpbroadcastd\[^\n\]*%xmm17, *%zmm" } } */ + +typedef int __attribute__ ((vector_size (16))) v4si; + +v4si bcst (int i) +{ + register int x asm ("xmm17") = i; + + asm ("" : "+v" (x)); + return (v4si) {x, x, x, x}; +}
... in vec_dupv4sf / *vec_dupv4si. The respective broadcast insns are never longer (yet sometimes shorter) than the corresponding VSHUFPS / VPSHUFD, due to the immediate operand of the shuffle insns balancing the (uniform) need for VEX3 in the broadcast ones. When EVEX encoding is respective the broadcast insns are always shorter. Add new alternatives to cover the AVX2 and AVX512 cases as appropriate. While touching this anyway, switch to consistently using "sseshuf1" in the "type" attributes for all shuffle forms. gcc/ * config/i386/sse.md (vec_dupv4sf): Make first alternative use vbroadcastss for AVX2. New AVX512F alternative. (*vec_dupv4si): New AVX2 and AVX512F alternatives using vpbroadcastd. Replace sselog1 by sseshuf1 in "type" attribute. gcc/testsuite/ * gcc.target/i386/avx2-dupv4sf.c: New test. * gcc.target/i386/avx2-dupv4si.c: Likewise. * gcc.target/i386/avx512f-dupv4sf.c: Likewise. * gcc.target/i386/avx512f-dupv4si.c: Likewise. --- Note that unlike originally intended, "prefix_extra" isn't dropped: "length_vex" uses it to determine whether 2-byte VEX encoding is possible (which it isn't for VBROADCASTSS / VPBROADCASTD). "length" itself specifically does not use it for VEX/EVEX encoded insns. Especially with the added "enabled" attribute I didn't really see how to (further) fold alternatives 0 and 1. Instead *vec_dupv4si might benefit from using sse2_noavx2 instead of sse2 for alternative 2, except that there is no sse2_noavx2, only sse2_noavx. I'm working from the assumption that the isa attributes to the original 1st and 2nd alternatives don't need further restricting (to sse2_noavx2 or avx_noavx2 as applicable), as the new earlier alternatives cover all operand forms already when at least AVX2 is enabled. --- v3: Testcases for new alternatives. "type" and "prefix_extra" adjustments. v2: Correct operand constraints. Respect -mprefer-vector-width=. Fold two alternatives of vec_dupv4sf.