Committed as passed x86 bootstrap and regression test, thanks Richard. Pan
-----Original Message----- From: Richard Biener <rguent...@suse.de> Sent: Tuesday, September 26, 2023 7:35 PM To: Juzhe-Zhong <juzhe.zh...@rivai.ai> Cc: gcc-patches@gcc.gnu.org; richard.sandif...@arm.com Subject: Re: [PATCH V3] MATCH: Optimize COND_ADD_LEN reduction pattern On Tue, 26 Sep 2023, Juzhe-Zhong wrote: > This patch leverage this commit: > https://gcc.gnu.org/git/gitweb.cgi?p=gcc.git;h=62b505a4d5fc89 > to optimize COND_LEN_ADD reduction pattern. > > We are doing optimization of VEC_COND_EXPR + COND_LEN_ADD -> COND_LEN_ADD. > > Consider thsi following case: > > #include <stdint.h> > > void > pr11594 (uint64_t *restrict a, uint64_t *restrict b, int loop_size) > { > uint64_t result = 0; > > for (int i = 0; i < loop_size; i++) > { > if (b[i] <= a[i]) > { > result += a[i]; > } > } > > a[0] = result; > } > > Before this patch: > vsetvli a7,zero,e64,m1,ta,ma > vmv.v.i v2,0 > vmv1r.v v3,v2 --- redundant > .L3: > vsetvli a5,a2,e64,m1,ta,ma > vle64.v v1,0(a3) > vle64.v v0,0(a1) > slli a6,a5,3 > vsetvli a7,zero,e64,m1,ta,ma > sub a2,a2,a5 > vmsleu.vv v0,v0,v1 > add a1,a1,a6 > vmerge.vvm v1,v3,v1,v0 ---- redundant. > add a3,a3,a6 > vsetvli zero,a5,e64,m1,tu,ma > vadd.vv v2,v2,v1 > bne a2,zero,.L3 > li a5,0 > vsetvli a4,zero,e64,m1,ta,ma > vmv.s.x v1,a5 > vredsum.vs v2,v2,v1 > vmv.x.s a5,v2 > sd a5,0(a0) > ret > > After this patch: > > vsetvli a6,zero,e64,m1,ta,ma > vmv.v.i v1,0 > .L3: > vsetvli a5,a2,e64,m1,ta,ma > vle64.v v2,0(a4) > vle64.v v0,0(a1) > slli a3,a5,3 > vsetvli a6,zero,e64,m1,ta,ma > sub a2,a2,a5 > vmsleu.vv v0,v0,v2 > add a1,a1,a3 > vsetvli zero,a5,e64,m1,tu,mu > add a4,a4,a3 > vadd.vv v1,v1,v2,v0.t > bne a2,zero,.L3 > li a5,0 > vsetivli zero,1,e64,m1,ta,ma > vmv.s.x v2,a5 > vsetvli a5,zero,e64,m1,ta,ma > vredsum.vs v1,v1,v2 > vmv.x.s a5,v1 > sd a5,0(a0) > ret > > Bootstrap && Regression is running. > > Ok for trunk when testing passes ? OK > PR tree-optimization/111594 > PR tree-optimization/110660 > > gcc/ChangeLog: > > * match.pd: Optimize COND_LEN_ADD reduction. > > gcc/testsuite/ChangeLog: > > * gcc.target/riscv/rvv/autovec/cond/cond_reduc-1.c: New test. > * gcc.target/riscv/rvv/autovec/cond/pr111594.c: New test. > > --- > gcc/match.pd | 15 ++++++++++ > .../riscv/rvv/autovec/cond/cond_reduc-1.c | 29 +++++++++++++++++++ > .../riscv/rvv/autovec/cond/pr111594.c | 22 ++++++++++++++ > 3 files changed, 66 insertions(+) > create mode 100644 > gcc/testsuite/gcc.target/riscv/rvv/autovec/cond/cond_reduc-1.c > create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/autovec/cond/pr111594.c > > diff --git a/gcc/match.pd b/gcc/match.pd > index a17778fbaa6..3ce90c3333b 100644 > --- a/gcc/match.pd > +++ b/gcc/match.pd > @@ -8866,6 +8866,21 @@ and, > (IFN_COND_ADD @0 @1 (vec_cond @2 @3 integer_zerop) @1) > (IFN_COND_ADD (bit_and @0 @2) @1 @3 @1)) > > +/* Detect simplication for a conditional length reduction where > + > + a = mask ? b : 0 > + c = i < len + bias ? d + a : d > + > + is turned into > + > + c = mask && i < len + bias ? d + b : d. */ > +(simplify > + (IFN_COND_LEN_ADD integer_truep @0 (vec_cond @1 @2 zerop@5) @0 @3 @4) > + (if (ANY_INTEGRAL_TYPE_P (type) > + || (FLOAT_TYPE_P (type) > + && fold_real_zero_addition_p (type, NULL_TREE, @5, 0))) > + (IFN_COND_LEN_ADD @1 @0 @2 @0 @3 @4))) > + > /* For pointers @0 and @2 and nonnegative constant offset @1, look for > expressions like: > > diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/cond/cond_reduc-1.c > b/gcc/testsuite/gcc.target/riscv/rvv/autovec/cond/cond_reduc-1.c > new file mode 100644 > index 00000000000..db6f9d1ec6c > --- /dev/null > +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/cond/cond_reduc-1.c > @@ -0,0 +1,29 @@ > +/* { dg-do compile } */ > +/* { dg-additional-options "-march=rv64gcv_zvfh -mabi=lp64d > -fno-vect-cost-model -ffast-math -fdump-tree-optimized" } */ > + > +#include <stdint-gcc.h> > + > +#define COND_REDUCTION(TYPE) > \ > + TYPE foo##TYPE (TYPE *restrict a, TYPE *restrict b, int loop_size) > \ > + { > \ > + TYPE result = 0; > \ > + for (int i = 0; i < loop_size; i++) > \ > + if (b[i] <= a[i]) > \ > + result += a[i]; \ > + return result; > \ > + } > + > +COND_REDUCTION (int8_t) > +COND_REDUCTION (int16_t) > +COND_REDUCTION (int32_t) > +COND_REDUCTION (int64_t) > +COND_REDUCTION (uint8_t) > +COND_REDUCTION (uint16_t) > +COND_REDUCTION (uint32_t) > +COND_REDUCTION (uint64_t) > +COND_REDUCTION (_Float16) > +COND_REDUCTION (float) > +COND_REDUCTION (double) > + > +/* { dg-final { scan-tree-dump-not "VCOND_MASK" "optimized" } } */ > +/* { dg-final { scan-tree-dump-times "COND_LEN_ADD" 11 "optimized" } } */ > diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/cond/pr111594.c > b/gcc/testsuite/gcc.target/riscv/rvv/autovec/cond/pr111594.c > new file mode 100644 > index 00000000000..6d81b26fbd0 > --- /dev/null > +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/cond/pr111594.c > @@ -0,0 +1,22 @@ > +/* { dg-do compile } */ > +/* { dg-additional-options "-march=rv64gcv -mabi=lp64d -fno-vect-cost-model > -ffast-math" } */ > + > +#include <stdint-gcc.h> > + > +void > +pr11594 (uint64_t *restrict a, uint64_t *restrict b, int loop_size) > +{ > + uint64_t result = 0; > + > + for (int i = 0; i < loop_size; i++) > + { > + if (b[i] <= a[i]) > + { > + result += a[i]; > + } > + } > + > + a[0] = result; > +} > + > +/* { dg-final { scan-assembler-not {vmerge} } } */ > -- Richard Biener <rguent...@suse.de> SUSE Software Solutions Germany GmbH, Frankenstrasse 146, 90461 Nuernberg, Germany; GF: Ivo Totev, Andrew McDonald, Werner Knoblich; (HRB 36809, AG Nuernberg)
