ice799

#if 0
In these examples, let's assume that passing more than 2 args to the program is 
*not* the case we are optimizing this for.

With that in mind, lets build a sample program with
 and without __builtin_expect and compare assembly output. 
#endif

/*
   no predictor:

ice799

;
;  branch predictor on with -O2
;


0000000000000000 <st_lookup>:
   0:   48 89 5c 24 d8          mov    %rbx,-0x28(%rsp)
   5:   48 89 6c 24 e0          mov    %rbp,-0x20(%rsp)
   a:   48 89 fd                mov    %rdi,%rbp
   d:   4c 89 64 24 e8          mov    %r12,-0x18(%rsp)

ice799

[joe@sloth:/home/joe/code/small_malloc/bsdmalloc]% LD_PRELOAD=/home/joe/code/small_malloc/bsdmalloc/libbsdmalloc.so.1.0.1 /usr/bin/time ./a.out 
22.42user 0.01system 0:22.48elapsed 99%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (0major+138minor)pagefaults 0swaps

[joe@sloth:/home/joe/code/small_malloc/bsdmalloc]% LD_PRELOAD=/custom/lib/libtcmalloc_minimal.so.0.0.0 /usr/bin/time ./a.out 
36.89user 0.02system 0:37.00elapsed 99%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (0major+1530minor)pagefaults 0swaps


[joe@sloth:/home/joe/code/small_malloc/bsdmalloc]% /usr/bin/time ./a.out 

ice799

commit 2ec0bd00031a777237a24b7cd6a06b0f7f007430
Author: Joe Damato <ice799@gmail.com>
Date:   Wed Nov 18 16:22:06 2009 -0800

    add malloc_usable_size

diff --git a/configure.ac b/configure.ac
index 9db399a..12cc585 100644
--- a/configure.ac
+++ b/configure.ac

ice799

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/mman.h>

int rb_new_obj() {
  printf("hi\n");
  exit(1);
}

ice799

#define _GNU_SOURCE
#include <err.h>
#include <fcntl.h>
#include <gelf.h>
#include <stdio.h>
#include <stdint.h>
#include <stdlib.h>
#include <unistd.h>
#include <link.h>
#include <sysexits.h>

ice799

irb(main):021:0> Time.now.zone
=> "UTC"
irb(main):022:0> Time.now.utc?
=> false


### and also:

irb(main):005:0> Time.now.iso8601
=> "2010-03-31T23:46:20+00:00"

ice799

                                James Damato • (973) 634-2031 • jdtwelv@yahoo.com

PROFESSIONAL EXPERIENCE
KPMG, Montvale, NJ	8/09 – Present 
Research Associate – Global Risk Management 
    • Download Issuer and Securities data files daily, perform quality check to confirm file validity, 
      and post to local applications for data refresh 
    • Download Cognos reports weekly to assist with HR and Compliance Report creation process
    • Perform data file retrieval tasks and execute programs to create daily identifier change reports
    • Compile data from vendor web and place data files in established location for refresh of the

ice799

2010-05-11  Joe Damato  <ice799@gmail.com>

       * sysdeps/x86_64/dl-trampoline.S (_dl_runtime_resolve): Save and
       restore SSE registers.
---
 sysdeps/x86_64/dl-trampoline.S |   34 +++++++++++++++++++++++++++-------
 1 files changed, 27 insertions(+), 7 deletions(-)

diff --git a/sysdeps/x86_64/dl-trampoline.S b/sysdeps/x86_64/dl-trampoline.S
index 5564a11..44706d9 100644

ice799

_dl_runtime_resolve:
  subq $56,%rsp 
  cfi_adjust_cfa_offset(72) # Incorporate PLT
  movq %rax,(%rsp)  # Preserve registers otherwise clobbered.
  movq %rcx, 8(%rsp)
  movq %rdx, 16(%rsp)
  movq %rsi, 24(%rsp)
  movq %rdi, 32(%rsp)
  movq %r8, 40(%rsp)
  movq %r9, 48(%rsp)