Re: [PATCH 1/2] x86/bitops: implement __test_bit

[Ingo Molnar]

* Ingo Molnar <mingo@xxxxxxxxxx> wrote:

> > Disassembly of section .text:
> > 
> > 00000000 <__variable_test_bit>:
> > __variable_test_bit():
> >    0:   8b 54 24 08             mov    0x8(%esp),%edx
> >    4:   8b 44 24 04             mov    0x4(%esp),%eax
> >    8:   0f a3 02                bt     %eax,(%edx)
> >    b:   19 c0                   sbb    %eax,%eax
> >    d:   c3                      ret    
> >    e:   66 90                   xchg   %ax,%ax
> > 
> > 00000010 <__constant_test_bit>:
> > __constant_test_bit():
> >   10:   8b 4c 24 04             mov    0x4(%esp),%ecx
> >   14:   8b 44 24 08             mov    0x8(%esp),%eax
> >   18:   89 ca                   mov    %ecx,%edx
> >   1a:   c1 fa 04                sar    $0x4,%edx
> >   1d:   8b 04 90                mov    (%eax,%edx,4),%eax
> >   20:   d3 e8                   shr    %cl,%eax
> >   22:   83 e0 01                and    $0x1,%eax
> >   25:   c3                      ret    
> 
> But that's due to the forced interface of generating a return code. Please 
> compare it at an inlined usage site, where GCC is free to do the comparison 
> directly and use the result in flags.

So I was thinking about the patch below on top of yours.

But turns out GCC indeed generates worse code even under the best of 
circumstances. For example the nested_vmx_disable_intercept_for_msr() change:

@@ -4275,24 +4275,24 @@ static void nested_vmx_disable_intercept
         */
        if (msr <= 0x1fff) {
                if (type & MSR_TYPE_R &&
-                  !test_bit(msr, msr_bitmap_l1 + 0x000 / f))
+                  !__test_bit(msr, msr_bitmap_l1 + 0x000 / f))
                        /* read-low */
                        __clear_bit(msr, msr_bitmap_nested + 0x000 / f);

before (i.e. your series):

ffffffff818b1082:       89 d0                   mov    %edx,%eax
ffffffff818b1084:       48 0f a3 07             bt     %rax,(%rdi)
ffffffff818b1088:       45 19 c0                sbb    %r8d,%r8d
ffffffff818b108b:       45 85 c0                test   %r8d,%r8d
ffffffff818b108e:       75 04                   jne    ffffffff818b1094 <nested_vmx_disable_intercept_for_msr+0x43>

after (with my 'optimization' patch applied):

ffffffff818b1091:       89 d0                   mov    %edx,%eax
ffffffff818b1093:       49 89 c0                mov    %rax,%r8
ffffffff818b1096:       49 c1 f8 06             sar    $0x6,%r8
ffffffff818b109a:       4e 8b 04 c7             mov    (%rdi,%r8,8),%r8
ffffffff818b109e:       49 0f a3 d0             bt     %rdx,%r8
ffffffff818b10a2:       72 04                   jb     ffffffff818b10a8 <nested_vmx_disable_intercept_for_msr+0x48>

So GCC when left to its own devices, generates one more instruction and 4 more 
bytes. Why does GCC do that? Why doesn't it use BT directly and use the flag, i.e. 
something like [pseudocode]:

ffffffff818b1082:       89 d0                   mov    %edx,%eax
ffffffff818b1084:       48 0f a3 07             bt     %rax,(%rdi)
ffffffff818b108e:       75 04                   jne    ffffffff818b1094 <nested_vmx_disable_intercept_for_msr+0x43>

?

In any case I take back my objection:

  Acked-by: Ingo Molnar <mingo@xxxxxxxxxx>

Thanks,

	Ingo


---
 arch/x86/include/asm/bitops.h |   19 +------------------
 1 file changed, 1 insertion(+), 18 deletions(-)

Index: tip/arch/x86/include/asm/bitops.h
===================================================================
--- tip.orig/arch/x86/include/asm/bitops.h
+++ tip/arch/x86/include/asm/bitops.h
@@ -323,24 +323,12 @@ static inline int variable_test_bit(long
 	return oldbit;
 }
 
-static __always_inline int __constant_test_bit(long nr, const unsigned long *addr)
+static __always_inline int __test_bit(long nr, const unsigned long *addr)
 {
 	return ((1UL << (nr & (BITS_PER_LONG-1))) &
 		(addr[nr >> _BITOPS_LONG_SHIFT])) != 0;
 }
 
-static inline int __variable_test_bit(long nr, const unsigned long *addr)
-{
-	int oldbit;
-
-	asm volatile("bt %2,%1\n\t"
-		     "sbb %0,%0"
-		     : "=r" (oldbit)
-		     : "m" (*addr), "Ir" (nr));
-
-	return oldbit;
-}
-
 #if 0 /* Fool kernel-doc since it doesn't do macros yet */
 /**
  * test_bit - Determine whether a bit is set
@@ -362,11 +350,6 @@ static int __test_bit(int nr, const vola
 	 ? constant_test_bit((nr), (addr))	\
 	 : variable_test_bit((nr), (addr)))
 
-#define __test_bit(nr, addr)			\
-	(__builtin_constant_p((nr))		\
-	 ? __constant_test_bit((nr), (addr))	\
-	 : __variable_test_bit((nr), (addr)))
-
 /**
  * __ffs - find first set bit in word
  * @word: The word to search

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/