Skip to content

lib/string.c cleanup

Jesper Juhl lkml 1 version In review

Get series with b4
b4 am [email protected]

Versions newest first

  1. v1 Latest3 patches 2 repliesIn review

Patches in v1 No applies recorded yet

#Subject and filesSizeRepliesApplied
1/3lib/string.c cleanup : whitespace and CodingStyle cleanupslib/string.c+53 -602 replies
2/3lib/string.c cleanup : remove pointless register keywordlib/string.c+2 -20 replies
3/3lib/string.c cleanup : remove pointless explicit castslib/string.c+11 -100 replies

Reviews trailers per patch

No Reviewed-by, Acked-by or Tested-by trailers for v1 yet.

Read the v1 thread

Cover letter [PATCH 0/3] lib/string.c cleanup

Here are 3 small cleanup patches for lib/string.c against 2.6.14-rc2-git3

The first patch does whitespace and CodingStyle cleanups. It removes some blank lines, removes some trailing whitespace, adds spaces after commas and a few similar changes.

The second patch removes a few pointless register keywords. register is merely a compiler hint that access to the variable should be optimized, but gcc (3.3.6 in my case) generates the exact same code with and without the keyword, and even if gcc did something different with register present I think it is doubtful we would want to optimize access to these variables - especially since this is generic library code and there are supposed to be optimized versions in asm/ for anything that really matters speed wise.

The third patch removes a few pointless explicit casts. The first two hunks of the patch really belongs in patch 1, but I missed them on the first pass and instead of redoing all 3 patches I stuck them in this one.

One odd sideeffect of patch 1 is that gcc generates an object file that is 4 bytes smaller with the patch applied, but it's not obvious to me how it manages this saving since the patch only makes whitespace changes. I tried doing an objdump -D of string.o before and after the patch and compared the output with diff, but that didn't reveal the cause of the 4 saved bytes - odd. I've included the objdump diff below - as you can see it really shouldn't make any difference, but the fact remains that the sizes differ :

  juhl@dragon:~/download/kernel/linux-2.6.14-rc2-git3$ ls -l lib/string.o.orig
  -rw-r--r--  1 juhl users 60912 2005-09-23 22:25 lib/string.o.orig
  juhl@dragon:~/download/kernel/linux-2.6.14-rc2-git3$ ls -l lib/string.o
  -rw-r--r--  1 juhl users 60908 2005-09-23 23:05 lib/string.o

Disassembly of section .text:

@@ -9,46 +9,46 @@

    3:  57                      push   %edi
    4:  89 d7                   mov    %edx,%edi
    6:  56                      push   %esi
-   7:  89 ce                   mov    %ecx,%esi
-   9:  31 c9                   xor    %ecx,%ecx
+   7:  31 d2                   xor    %edx,%edx
+   9:  89 ce                   mov    %ecx,%esi
    b:  53                      push   %ebx
    c:  31 db                   xor    %ebx,%ebx
    e:  83 ec 04                sub    $0x4,%esp
-  11:  85 f6                   test   %esi,%esi
+  11:  85 c9                   test   %ecx,%ecx
   13:  89 45 f0                mov    %eax,0xfffffff0(%ebp)
   16:  74 48                   je     60 <strnicmp+0x60>
   18:  8b 45 f0                mov    0xfffffff0(%ebp),%eax
   1b:  0f b6 1f                movzbl (%edi),%ebx
   1e:  47                      inc    %edi
-  1f:  0f b6 08                movzbl (%eax),%ecx
+  1f:  0f b6 10                movzbl (%eax),%edx
   22:  40                      inc    %eax
   23:  89 45 f0                mov    %eax,0xfffffff0(%ebp)
-  26:  84 c9                   test   %cl,%cl
+  26:  84 d2                   test   %dl,%dl
   28:  74 36                   je     60 <strnicmp+0x60>
   2a:  84 db                   test   %bl,%bl
   2c:  74 32                   je     60 <strnicmp+0x60>
-  2e:  38 d9                   cmp    %bl,%cl
+  2e:  38 da                   cmp    %bl,%dl
   30:  74 2a                   je     5c <strnicmp+0x5c>
-  32:  0f b6 c1                movzbl %cl,%eax
-  35:  89 ca                   mov    %ecx,%edx
+  32:  0f b6 c2                movzbl %dl,%eax
+  35:  89 d1                   mov    %edx,%ecx
   37:  f6 80 00 00 00 00 01    testb  $0x1,0x0(%eax)
   3e:  74 03                   je     43 <strnicmp+0x43>
-  40:  8d 51 20                lea    0x20(%ecx),%edx
+  40:  8d 4a 20                lea    0x20(%edx),%ecx
   43:  0f b6 c3                movzbl %bl,%eax
-  46:  89 d1                   mov    %edx,%ecx
-  48:  89 da                   mov    %ebx,%edx
+  46:  89 ca                   mov    %ecx,%edx
+  48:  89 d9                   mov    %ebx,%ecx
   4a:  f6 80 00 00 00 00 01    testb  $0x1,0x0(%eax)
   51:  74 03                   je     56 <strnicmp+0x56>
-  53:  8d 53 20                lea    0x20(%ebx),%edx
-  56:  38 d1                   cmp    %dl,%cl
-  58:  89 d3                   mov    %edx,%ebx
+  53:  8d 4b 20                lea    0x20(%ebx),%ecx
+  56:  38 ca                   cmp    %cl,%dl
+  58:  89 cb                   mov    %ecx,%ebx
   5a:  75 04                   jne    60 <strnicmp+0x60>
   5c:  4e                      dec    %esi
   5d:  75 b9                   jne    18 <strnicmp+0x18>
   5f:  90                      nop
   60:  83 c4 04                add    $0x4,%esp
-  63:  0f b6 d3                movzbl %bl,%edx
-  66:  0f b6 c1                movzbl %cl,%eax
+  63:  0f b6 c2                movzbl %dl,%eax
+  66:  0f b6 d3                movzbl %bl,%edx
   69:  5b                      pop    %ebx
   6a:  5e                      pop    %esi
   6b:  5f                      pop    %edi

Anyway, the patches have been compile tested, and a kernel with all 3 applied boots just fine with no observable defects.

Signature · 1 line
Jesper Juhl <[email protected]>