Re: [PATCH 4/5] shrink-wrap: Shrink-wrapping for separate components

[Segher Boessenkool <segher@kernel.crashing.org>]

[-- Attachment #1: Type: text/plain, Size: 4205 bytes --]

[ whoops, message too big, resending with the attachment compressed ]

On Tue, Sep 27, 2016 at 03:14:51PM -0600, Jeff Law wrote:
> With transposition issue addressed, the only blocker I see are some 
> simple testcases we can add to the suite.  They don't have to be real 
> extensive.  And one motivating example for the list archives, ideally 
> the glibc malloc case.

And here is the malloc testcase.

A very important (for performance) function is _int_malloc, which starts
with


static void *
_int_malloc (mstate av, size_t bytes)
{
// [ variable declarations culled ]

  if (((unsigned long) (bytes) >= (unsigned long) (size_t) (-2 * (unsigned long)((((
 __builtin_offsetof (
 struct malloc_chunk
 ,-
 fd_nextsize
 )
 )+((2 *(sizeof(size_t)) < __alignof__ (long double) ? __alignof__ (long double) : 2 *(sizeof(size_t))) - 1)) & ~((2 *(sizeof(size_t)) < __alignof__ (long double) ? __alignof__ (long double) : 2 *(sizeof(size_t))) - 1)))))) { (__libc_errno = (
 12
 )); return 0; }

  if (__builtin_expect ((av ==-
     ((
     void-
     *)0)
     ), 0))
    {
      void *p = sysmalloc (nb, av);
      if (p !=-
              ((
              void-
              *)0)
                  )
 alloc_perturb (p, bytes);
      return p;
    }


which without separate shrink-wrapping ends up as (reordered the blocks):


.L._int_malloc:
        mflr 0
        li 9,-65
        std 14,-144(1)
        std 15,-136(1)
        cmpld 7,4,9
        std 16,-128(1)
        std 17,-120(1)
        std 18,-112(1)
        std 19,-104(1)
        std 20,-96(1)
        std 21,-88(1)
        std 22,-80(1)
        std 23,-72(1)
        std 0,16(1)
        std 24,-64(1)
        std 25,-56(1)
        std 26,-48(1)
        std 27,-40(1)
        std 28,-32(1)
        std 29,-24(1)
        std 30,-16(1)
        std 31,-8(1)
        stdu 1,-288(1)
        bgt 7,.L768
        addi 14,4,23
        mr 15,3
        cmpldi 0,14,31
        ble 0,.L769

# ...

.L768:
        addis 27,2,__libc_errno@got@tprel@ha
        li 19,12
        ld 28,__libc_errno@got@tprel@l(27)
        li 3,0
        add 17,28,__libc_errno@tls
        stw 19,0(17)
        b .L631

# ...

.L631:
        addi 1,1,288
        ld 29,16(1)
        ld 14,-144(1)
        ld 15,-136(1)
        ld 16,-128(1)
        ld 17,-120(1)
        ld 18,-112(1)
        ld 19,-104(1)
        ld 20,-96(1)
        ld 21,-88(1)
        ld 22,-80(1)
        ld 23,-72(1)
        ld 24,-64(1)
        mtlr 29
        ld 25,-56(1)
        ld 26,-48(1)
        ld 27,-40(1)
        ld 28,-32(1)
        ld 29,-24(1)
        ld 30,-16(1)
        ld 31,-8(1)
        blr
# ...

.L769:
        cmpdi 1,3,0
        beq 1,.L715

# ...

.L715:
        li 14,32
.L635:
        li 4,0
.L762:
        addi 1,1,288
        mr 3,14
        ld 14,16(1)
        ld 15,-136(1)
        ld 16,-128(1)
        ld 17,-120(1)
        ld 18,-112(1)
        ld 19,-104(1)
        ld 20,-96(1)
        ld 21,-88(1)
        ld 22,-80(1)
        ld 23,-72(1)
        ld 24,-64(1)
        ld 25,-56(1)
        mtlr 14
        ld 26,-48(1)
        ld 14,-144(1)
        ld 27,-40(1)
        ld 28,-32(1)
        ld 29,-24(1)
        ld 30,-16(1)
        ld 31,-8(1)
        b sysmalloc


[ I see have regrename on by default still; doesn't matter much for
this test, it's just less readable ]


With separate shrink-wrapping we get instead


.L._int_malloc:
        li 9,-65
        stdu 1,-288(1)
        cmpld 7,4,9
        bgt 7,.L811
        std 14,144(1)
        addi 14,4,23
        std 15,152(1)
        mr 15,3
        cmpldi 0,14,31
        std 25,232(1)
        std 30,272(1)
        ble 0,.L812

# ...

.L811:
        addis 6,2,__libc_errno@got@tprel@ha
        li 11,12
        ld 10,__libc_errno@got@tprel@l(6)
        li 3,0
        add 12,10,__libc_errno@tls
        stw 11,0(12)
        b .L673

# ...

.L673:
        addi 1,1,288
        blr

# ...

.L812:
        cmpdi 1,3,0
        beq 1,.L757

# ...

.L757:
        li 14,32
.L677:
        mr 3,14
        ld 15,152(1)
        ld 14,144(1)
        ld 25,232(1)
        ld 30,272(1)
        li 4,0
        addi 1,1,288
        b sysmalloc


I'm attaching the full testcase (pre-processed for powerpc64-linux).


Segher

[-- Attachment #2: malloc.i.gz --]
[-- Type: application/x-gzip, Size: 74379 bytes --]