public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
* [Bug target/106167] New: Missed optimization (memory vs register read)
@ 2022-07-02 10:50 vluchits at gmail dot com
  0 siblings, 0 replies; only message in thread
From: vluchits at gmail dot com @ 2022-07-02 10:50 UTC (permalink / raw)
  To: gcc-bugs

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106167

            Bug ID: 106167
           Summary: Missed optimization (memory vs register read)
           Product: gcc
           Version: 9.3.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: vluchits at gmail dot com
  Target Milestone: ---
            Target: sh

Created attachment 53238
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=53238&action=edit
Test case

Hello,

in the provided example, this piece of C code:

        if (newclipbounds)
        {
            int newfloorclipx = floorclipx;
            int newceilingclipx = ceilingclipx;
            uint16_t newclip;

            // rewrite clipbounds
            if (actionbits & AC_NEWFLOOR)
                newfloorclipx = low;
            if (actionbits & AC_NEWCEILING)
                newceilingclipx = high;

            newclip = (newceilingclipx << 8) + newfloorclipx;
            clipbounds[x] = newclip;
            newclipbounds[x] = newclip;
        }

is compiled to the following assembler code:

        if (newclipbounds)
 16c:   54 f7           mov.l   @(28,r15),r4
 16e:   24 48           tst     r4,r4
 170:   8d 10           bt.s    194 <_R_SegLoop+0x194>
 172:   e0 54           mov     #84,r0
            if (actionbits & AC_NEWFLOOR)
 174:   05 fe           mov.l   @(r0,r15),r5
 176:   25 58           tst     r5,r5
 178:   8f 01           bf.s    17e <_R_SegLoop+0x17e>
 17a:   e0 58           mov     #88,r0
        floorclipx = ceilingclipx & 0x00ff;
 17c:   67 a3           mov     r10,r7
            if (actionbits & AC_NEWCEILING)
 17e:   00 fe           mov.l   @(r0,r15),r0
 180:   20 08           tst     r0,r0
 182:   8f 01           bf.s    188 <_R_SegLoop+0x188>
 184:   50 fb           mov.l   @(44,r15),r0
            int newceilingclipx = ceilingclipx;
 186:   66 93           mov     r9,r6
            newclip = (newceilingclipx << 8) + newfloorclipx;
 188:   46 18           shll8   r6
 18a:   37 6c           add     r6,r7
 18c:   67 7d           extu.w  r7,r7
            clipbounds[x] = newclip;
 18e:   08 75           mov.w   r7,@(r0,r8)
            newclipbounds[x] = newclip;


The following lines are of particular interest:
        if (newclipbounds)
 16c:   54 f7           mov.l   @(28,r15),r4
 16e:   24 48           tst     r4,r4
...
            newclipbounds[x] = newclip;
 190:   50 f7           mov.l   @(28,r15),r0
 192:   08 75           mov.w   r7,@(r0,r8)

The compiler fails to notice that it's already holding the value of @(28,r15)
in r4 and re-reads the value from stack instead of doing the mov r4,r0.

CFLAGS: -c -std=c11 -g -m2 -mb -Os -fomit-frame-pointer -Wall -Wextra -pedantic
-Wno-unused-parameter -Wimplicit-fallthrough=0 -Wno-missing-field-initializers
-Wnonnull

This may or may not be relevant to bug #106161

^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2022-07-02 10:50 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-07-02 10:50 [Bug target/106167] New: Missed optimization (memory vs register read) vluchits at gmail dot com

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).