| Message ID | 20231106223531.3271166-1-juzhe.zhong@rivai.ai |
|---|---|
| State | New |
| Headers | show |
| Series | test: Fix XPASS of bb-slp-43.c for RVV | expand |
On 11/6/23 15:35, Juzhe-Zhong wrote: > RVV is variable length vector but also has 256 bit VLS mode vector. > This test is vectorized as: > > f: > vsetivli zero,8,e32,m2,ta,ma > vle32.v v2,0(a0) > vmv.v.i v4,1 > vle16.v v1,0(a1) > vmseq.vv v0,v2,v4 > vsetvli zero,zero,e16,m1,ta,ma > vmseq.vi v1,v1,2 > vsetvli zero,zero,e32,m2,ta,ma > vmv.v.i v2,0 > vmand.mm v0,v0,v1 > vmerge.vvm v2,v2,v4,v0 > vse32.v v2,0(a0) > ret > > Use 256 bit vector, so remove XFAIL for 256 bits vector. > > gcc/testsuite/ChangeLog: > > * gcc.dg/vect/bb-slp-43.c: Fix XPASS for RVV. So will this create a FAIL if someone runs the testsuite with the autovec preference set to scalable? Or are the fallbacks to VLS still available when we prefer scalable vectors? jeff
>> So will this create a FAIL if someone runs the testsuite with the >> autovec preference set to scalable? No, it won't. Since it is always -fno-vect-cost-model. When a scalable vector doesn't have 256bit vector, it always XFAIL, for example, ARM SVE. >> Or are the fallbacks to VLS still >> available when we prefer scalable vectors? Yes. since it is -fno-vect-cost-model. juzhe.zhong@rivai.ai From: Jeff Law Date: 2023-11-07 11:23 To: Juzhe-Zhong; gcc-patches CC: rguenther Subject: Re: [PATCH] test: Fix XPASS of bb-slp-43.c for RVV On 11/6/23 15:35, Juzhe-Zhong wrote: > RVV is variable length vector but also has 256 bit VLS mode vector. > This test is vectorized as: > > f: > vsetivli zero,8,e32,m2,ta,ma > vle32.v v2,0(a0) > vmv.v.i v4,1 > vle16.v v1,0(a1) > vmseq.vv v0,v2,v4 > vsetvli zero,zero,e16,m1,ta,ma > vmseq.vi v1,v1,2 > vsetvli zero,zero,e32,m2,ta,ma > vmv.v.i v2,0 > vmand.mm v0,v0,v1 > vmerge.vvm v2,v2,v4,v0 > vse32.v v2,0(a0) > ret > > Use 256 bit vector, so remove XFAIL for 256 bits vector. > > gcc/testsuite/ChangeLog: > > * gcc.dg/vect/bb-slp-43.c: Fix XPASS for RVV. So will this create a FAIL if someone runs the testsuite with the autovec preference set to scalable? Or are the fallbacks to VLS still available when we prefer scalable vectors? jeff
On 11/6/23 20:30, juzhe.zhong@rivai.ai wrote: > >> So will this create a FAIL if someone runs the testsuite with the >>> autovec preference set to scalable? > No, it won't. Since it is always -fno-vect-cost-model. > When a scalable vector doesn't have 256bit vector, it always XFAIL, for > example, ARM SVE. > >>> Or are the fallbacks to VLS still >>> available when we prefer scalable vectors? > Yes. since it is -fno-vect-cost-model. OK. Thanks for clarifying. This is fine for the trunk. jeff
Thanks Jeff. Just finish bootstrap +regression passed. Committed. juzhe.zhong@rivai.ai From: Jeff Law Date: 2023-11-07 11:49 To: juzhe.zhong@rivai.ai; gcc-patches CC: rguenther Subject: Re: [PATCH] test: Fix XPASS of bb-slp-43.c for RVV On 11/6/23 20:30, juzhe.zhong@rivai.ai wrote: > >> So will this create a FAIL if someone runs the testsuite with the >>> autovec preference set to scalable? > No, it won't. Since it is always -fno-vect-cost-model. > When a scalable vector doesn't have 256bit vector, it always XFAIL, for > example, ARM SVE. > >>> Or are the fallbacks to VLS still >>> available when we prefer scalable vectors? > Yes. since it is -fno-vect-cost-model. OK. Thanks for clarifying. This is fine for the trunk. jeff
diff --git a/gcc/testsuite/gcc.dg/vect/bb-slp-43.c b/gcc/testsuite/gcc.dg/vect/bb-slp-43.c index a65d9513c4d..dad2d24262d 100644 --- a/gcc/testsuite/gcc.dg/vect/bb-slp-43.c +++ b/gcc/testsuite/gcc.dg/vect/bb-slp-43.c @@ -14,4 +14,4 @@ f (int *restrict x, short *restrict y) } /* { dg-final { scan-tree-dump-not "mixed mask and nonmask" "slp2" } } */ -/* { dg-final { scan-tree-dump-not "vector operands from scalars" "slp2" { target { { vect_int && vect_bool_cmp } && { vect_unpack && vect_hw_misalign } } xfail vect_variable_length } } } */ +/* { dg-final { scan-tree-dump-not "vector operands from scalars" "slp2" { target { { vect_int && vect_bool_cmp } && { vect_unpack && vect_hw_misalign } } xfail { vect_variable_length && { ! vect256 } } } } } */
RVV is variable length vector but also has 256 bit VLS mode vector. This test is vectorized as: f: vsetivli zero,8,e32,m2,ta,ma vle32.v v2,0(a0) vmv.v.i v4,1 vle16.v v1,0(a1) vmseq.vv v0,v2,v4 vsetvli zero,zero,e16,m1,ta,ma vmseq.vi v1,v1,2 vsetvli zero,zero,e32,m2,ta,ma vmv.v.i v2,0 vmand.mm v0,v0,v1 vmerge.vvm v2,v2,v4,v0 vse32.v v2,0(a0) ret Use 256 bit vector, so remove XFAIL for 256 bits vector. gcc/testsuite/ChangeLog: * gcc.dg/vect/bb-slp-43.c: Fix XPASS for RVV. --- gcc/testsuite/gcc.dg/vect/bb-slp-43.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)