From: Andreas Krebbel <krebbel@linux.vnet.ibm.com>
To: gcc-patches@gcc.gnu.org
Subject: [PATCH] S/390: PR80464: Split MEM->GPR vector moves
Date: Mon, 24 Apr 2017 17:27:00 -0000 [thread overview]
Message-ID: <20170424160430.1324-2-krebbel@linux.vnet.ibm.com> (raw)
In-Reply-To: <20170424160430.1324-1-krebbel@linux.vnet.ibm.com>
We do this already for TImode values but it was missing for vector
modes.
Bootstrapped and regression tested on s390x.
Bye,
-Andreas-
gcc/ChangeLog:
2017-04-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
* config/s390/vector.md: Split MEM->GPR vector moves for
non-s_operand addresses.
gcc/testsuite/ChangeLog:
2017-04-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
* gfortran.fortran-torture/compile/pr80464.f90: New test.
---
gcc/config/s390/vector.md | 19 +++++++++++
.../gfortran.fortran-torture/compile/pr80464.f90 | 39 ++++++++++++++++++++++
2 files changed, 58 insertions(+)
create mode 100644 gcc/testsuite/gfortran.fortran-torture/compile/pr80464.f90
diff --git a/gcc/config/s390/vector.md b/gcc/config/s390/vector.md
index 7535b9d..2952893 100644
--- a/gcc/config/s390/vector.md
+++ b/gcc/config/s390/vector.md
@@ -248,6 +248,25 @@
operands[5] = operand_subword (operands[1], 0, 0, <MODE>mode);
})
+; This is the vector equivalent to the TImode splitter in s390.md. It
+; is required if both target GPRs occur in the source address operand.
+
+; For non-s_operands at least one of the target GPRs does not conflict
+; with the address operand and one of the splitters above will take
+; over.
+(define_split
+ [(set (match_operand:V_128 0 "register_operand" "")
+ (match_operand:V_128 1 "memory_operand" ""))]
+ "TARGET_ZARCH && reload_completed
+ && !VECTOR_REG_P (operands[0])
+ && !s_operand (operands[1], VOIDmode)"
+ [(set (match_dup 0) (match_dup 1))]
+{
+ rtx addr = operand_subword (operands[0], 1, 0, <MODE>mode);
+ addr = gen_lowpart (Pmode, addr);
+ s390_load_address (addr, XEXP (operands[1], 0));
+ operands[1] = replace_equiv_address (operands[1], addr);
+})
; Moves for smaller vector modes.
diff --git a/gcc/testsuite/gfortran.fortran-torture/compile/pr80464.f90 b/gcc/testsuite/gfortran.fortran-torture/compile/pr80464.f90
new file mode 100644
index 0000000..d3a3943
--- /dev/null
+++ b/gcc/testsuite/gfortran.fortran-torture/compile/pr80464.f90
@@ -0,0 +1,39 @@
+subroutine bla(a,bar,lb,ne,nt,v,b)
+ character*8 lb
+ integer bar(20),foo(8,5)
+ real*8 a(3,*),x(3,8),v(0:3,*)
+ if(lb(4:4).eq.'3') then
+ n=8
+ elseif(lb(4:5).eq.'10') then
+ n=10
+ ns=6
+ m=4
+ endif
+ call blub(id)
+ do
+ if(id.eq.0) exit
+ if(lb(4:4).eq.'6') then
+ m=1
+ endif
+ if((n.eq.20).or.(n.eq.8)) then
+ if(b.eq.0) then
+ do i=1,ns
+ do j=1,3
+ x(j,i)=a(j,bar(foo(i,ig)))
+ enddo
+ enddo
+ else
+ do i=1,ns
+ do j=1,3
+ x(j,i)=a(j,bar(foo(i,ig)))+v(j,bar(foo(i,ig)))
+ enddo
+ enddo
+ endif
+ endif
+ do i=1,m
+ if(lb(4:5).eq.'1E') then
+ call blab(x)
+ endif
+ enddo
+ enddo
+end subroutine bla
--
2.9.1
prev parent reply other threads:[~2017-04-24 16:04 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-04-24 17:23 [PATCH] S/390: PR79895: Fix TImode constant handling Andreas Krebbel
2017-04-24 17:27 ` Andreas Krebbel [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20170424160430.1324-2-krebbel@linux.vnet.ibm.com \
--to=krebbel@linux.vnet.ibm.com \
--cc=gcc-patches@gcc.gnu.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).