Hi, Rainer, thanks a lot to report the issues with mips and sparc platform.
Yes, looks like even on the assembly level, the string scanning still not reliable on different platforms. I agree with Jeff’s suggestion to apply different search result for different platforms. I will update the testcase with this approach soon. Qing > On Aug 20, 2018, at 3:57 AM, Rainer Orth <r...@cebitec.uni-bielefeld.de> > wrote: > > Hi Jeff, > >> On 08/17/2018 09:43 PM, Paul Hua wrote: >>> Hi Qing: >>> >>>> >>>> the change has been committed as: >>>> https://gcc.gnu.org/viewcvs/gcc?view=revision&revision=263563 >>>> <https://gcc.gnu.org/viewcvs/gcc?view=revision&revision=263563> >>>> >>>> Qing >>>> >>> >>> The strcmpopt_6.c test still fails on mips64el target. >>> >>> gcc.dg/strcmpopt_6.c: memcmp found 4 times >>> FAIL: gcc.dg/strcmpopt_6.c scan-assembler-times memcmp 2 >>> >>> >>> The mips asm output have ".reloc" info. >>> >>> ------------------------------------------------------------------------- >>> ld $5,%got_page(.LC0)($28) >>> ld $25,%call16(memcmp)($28) >>> li $6,3 # 0x3 >>> sd $31,8($sp) >>> .reloc 1f,R_MIPS_JALR,memcmp >>> 1: jalr $25 >>> daddiu $5,$5,%got_ofst(.LC0) >>> ---------------------------------------------------------------------------- >>> >>> scan-assembler find "4" times. >> Ugh. :( There's probably other targets that are going to have similar >> properties. I'm not really sure how to write a suitable assembler test >> for this. >> >> Maybe we just need a different search result for MIPS (and potentially >> other targets). > > we certainly do: I've reported a similar issue on sparc in PR > testsuite/86519, but powerpc and s390x are equally affected. > > Rainer > > -- > ----------------------------------------------------------------------------- > Rainer Orth, Center for Biotechnology, Bielefeld University