[Bug target/94135] New: PPC: subfic instead of neg used for rotate right

[Bug target/94135] New: PPC: subfic instead of neg used for rotate right

[jens.seifert at de dot ibm.com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94135

            Bug ID: 94135
           Summary: PPC: subfic instead of neg used for rotate right
           Product: gcc
           Version: 8.3.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: jens.seifert at de dot ibm.com
  Target Milestone: ---

Input:

unsigned int rotr32(unsigned int v, unsigned int r)
{
   return (v>>r)|(v<<(32-r));
}

unsigned long long rotr64(unsigned long long v, unsigned long long r)
{
   return (v>>r)|(v<<(64-r));
}

Command line:
gcc -O2 -save-temps rotr.C

Output:
_Z6rotr32jj:
.LFB0:
        .cfi_startproc
        subfic 4,4,32
        rotlw 3,3,4
        blr
        .long 0
        .byte 0,9,0,0,0,0,0,0
        .cfi_endproc

_Z6rotr64yy:
.LFB1:
        .cfi_startproc
        subfic 4,4,64
        rotld 3,3,4
        blr
        .long 0
        .byte 0,9,0,0,0,0,0,0
        .cfi_endproc

subfic is a 2 cycle instruction, but can be replaced by 1 cycle instruction
neg.
rotr32(v,r) = rotl32(v,32-r) = rotl32(v,(32-r)%32) = rotl32(v,(-r)%32))=
rotl32(v,-r) as long as you have a modulo rotate like rotlw/rlwnm.

Same for 64-bit.

[Bug target/94135] PPC: subfic instead of neg used for rotate right

[segher at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94135

Segher Boessenkool <segher at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |segher at gcc dot gnu.org

--- Comment #1 from Segher Boessenkool <segher at gcc dot gnu.org> ---
On what CPU do subfic and neg execute at different speed?

(neg is better, of course, it doesn't write CA).

GCC does not know rotates work for any masking of the amount (with 1's in
the low 5 (resp. 6) bits); the rs6000 target code does not know about any
masking (the SHIFT_COUNT_TRUNCATED macro cannot be used, but we could have
more patterns, and then combine can do this in many cases).

[Bug target/94135] PPC: subfic instead of neg used for rotate right

[segher at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94135

Segher Boessenkool <segher at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2020-03-11
             Status|UNCONFIRMED                 |NEW

[Bug target/94135] PPC: subfic instead of neg used for rotate right

[jens.seifert at de dot ibm.com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94135

--- Comment #2 from Jens Seifert <jens.seifert at de dot ibm.com> ---
POWER8 Processor User’s Manual for the Single-Chip Module:

addi addis add add. subf subf. addic subfic adde addme subfme addze. subfze neg
neg. nego

1 - 2 cycles (GPR)
2 cycles (XER)
5 cycles (CR)

6/cycle, 2/cycle (with XER or CR updates)

CA is part of XER.

1-2 cycles versus 2 cycles.

[Bug target/94135] PPC: subfic instead of neg used for rotate right

[segher at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94135

--- Comment #3 from Segher Boessenkool <segher at gcc dot gnu.org> ---
Both subfic and neg are 1-2 if run on the integer units.  neg can run on
more units, but it is always 2 cycles then!  (And the conditions where you
*can* have 1 cycle are not very often satisfied, anyway).

[Bug target/94135] PPC: subfic instead of neg used for rotate right

[jens.seifert at de dot ibm.com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94135

--- Comment #4 from Jens Seifert <jens.seifert at de dot ibm.com> ---
Setting CA in XER increases issue to issue latency by 1 on Power8.
See:
Table 10-14. Issue-to-Issue Latencies

In addition, setting the CA restricts instruction reordering.

[Bug target/94135] PPC: subfic instead of neg used for rotate right

[segher at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94135

--- Comment #5 from Segher Boessenkool <segher at gcc dot gnu.org> ---
Please try it out on hardware (or on a cycle-accurate simulator) if you don't
believe me ;-)