Patchwork [v2] bitops: unify bitops_ffsl with the one in host-utils.h, using __builtin_ffsl

login
register
mail settings
Submitter Paolo Bonzini
Date Jan. 30, 2013, 5:57 p.m.
Message ID <1359568655-23982-1-git-send-email-pbonzini@redhat.com>
Download mbox | patch
Permalink /patch/216946/
State New
Headers show

Comments

Paolo Bonzini - Jan. 30, 2013, 5:57 p.m.
Fixes the build on Mac OS X, which has ffsl.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 include/qemu/bitops.h     | 40 ++++++++++++++--------------------------
 include/qemu/hbitmap.h    |  2 +-
 include/qemu/host-utils.h | 25 -------------------------
 util/hbitmap.c            |  2 +-
 4 files changed, 16 insertions(+), 53 deletions(-)
Paolo Bonzini - Jan. 30, 2013, 6:01 p.m.
Il 30/01/2013 18:57, Paolo Bonzini ha scritto:
> @@ -30,34 +31,21 @@
>   */
>  static unsigned long bitops_ffsl(unsigned long word)
>  {
> -	int num = 0;
> +#if QEMU_GNUC_PREREQ(3, 4)
> +    return __builtin_ffsl(word) + 1;

Obviously wrong, sent before "git commit --amend".

But it also reminds me that I noticed most uses of ffsl tend to subtract
one, and GCC optimizes it a little better if it is written as ctz + 1
and inlined.

Nacked-by: Paolo Bonzini <pbonzini@redhat.com>

but v1 stands.

Paolo

> +#else
> +    if (!word) {
> +        return 0;
> +    }
>  
> -#if LONG_MAX > 0x7FFFFFFF
> -	if ((word & 0xffffffff) == 0) {
> -		num += 32;
> -		word >>= 32;
> -	}
> +    if (sizeof(long) == 4) {
> +        return ctz32(word) + 1;
> +    } else if (sizeof(long) == 8) {
> +        return ctz64(word) + 1;
> +    } else {
> +        abort();
> +    }
>  #endif
> -	if ((word & 0xffff) == 0) {
> -		num += 16;
> -		word >>= 16;
> -	}
> -	if ((word & 0xff) == 0) {
> -		num += 8;
> -		word >>= 8;
> -	}
> -	if ((word & 0xf) == 0) {
> -		num += 4;
> -		word >>= 4;
> -	}
> -	if ((word & 0x3) == 0) {
> -		num += 2;
> -		word >>= 2;
> -	}
> -	if ((word & 0x1) == 0) {
> -		num += 1;
> -        }
> -	return num;
>  }
>  
>  /**
> diff --git a/include/qemu/hbitmap.h b/include/qemu/hbitmap.h
> index 73f5d1d..abad209 100644
> --- a/include/qemu/hbitmap.h
> +++ b/include/qemu/hbitmap.h
> @@ -170,7 +170,7 @@ static inline int64_t hbitmap_iter_next(HBitmapIter *hbi)
>  
>      /* The next call will resume work from the next bit.  */
>      hbi->cur[HBITMAP_LEVELS - 1] = cur & (cur - 1);
> -    item = ((uint64_t)hbi->pos << BITS_PER_LEVEL) + ffsl(cur) - 1;
> +    item = ((uint64_t)hbi->pos << BITS_PER_LEVEL) + bitops_ffsl(cur) - 1;
>  
>      return item << hbi->granularity;
>  }
> diff --git a/include/qemu/host-utils.h b/include/qemu/host-utils.h
> index 2a32be4..40457bd 100644
> --- a/include/qemu/host-utils.h
> +++ b/include/qemu/host-utils.h
> @@ -238,29 +238,4 @@ static inline int ctpop64(uint64_t val)
>  #endif
>  }
>  
> -/* glibc does not provide an inline version of ffsl, so always define
> - * ours.  We need to give it a different name, however.
> - */
> -#ifdef __GLIBC__
> -#define ffsl qemu_ffsl
> -#endif
> -static inline int ffsl(long val)
> -{
> -    if (!val) {
> -        return 0;
> -    }
> -
> -#if QEMU_GNUC_PREREQ(3, 4)
> -    return __builtin_ctzl(val) + 1;
> -#else
> -    if (sizeof(long) == 4) {
> -        return ctz32(val) + 1;
> -    } else if (sizeof(long) == 8) {
> -        return ctz64(val) + 1;
> -    } else {
> -        abort();
> -    }
> -#endif
> -}
> -
>  #endif
> diff --git a/util/hbitmap.c b/util/hbitmap.c
> index 2aa487d..32c3d59 100644
> --- a/util/hbitmap.c
> +++ b/util/hbitmap.c
> @@ -126,7 +126,7 @@ unsigned long hbitmap_iter_skip_words(HBitmapIter *hbi)
>           * The index of this word's least significant set bit provides
>           * the low-order bits.
>           */
> -        pos = (pos << BITS_PER_LEVEL) + ffsl(cur) - 1;
> +        pos = (pos << BITS_PER_LEVEL) + bitops_ffsl(cur) - 1;
>          hbi->cur[i] = cur & (cur - 1);
>  
>          /* Set up next level for iteration.  */
>
Eric Blake - Jan. 30, 2013, 6:03 p.m.
On 01/30/2013 10:57 AM, Paolo Bonzini wrote:
> Fixes the build on Mac OS X, which has ffsl.
> 
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> ---
>  include/qemu/bitops.h     | 40 ++++++++++++++--------------------------
>  include/qemu/hbitmap.h    |  2 +-
>  include/qemu/host-utils.h | 25 -------------------------
>  util/hbitmap.c            |  2 +-
>  4 files changed, 16 insertions(+), 53 deletions(-)
> 
> diff --git a/include/qemu/bitops.h b/include/qemu/bitops.h
> index 74e14e5..27b2d42 100644
> --- a/include/qemu/bitops.h
> +++ b/include/qemu/bitops.h
> @@ -13,6 +13,7 @@
>  #define BITOPS_H
>  
>  #include "qemu-common.h"
> +#include "host-utils.h"
>  
>  #define BITS_PER_BYTE           CHAR_BIT
>  #define BITS_PER_LONG           (sizeof (unsigned long) * BITS_PER_BYTE)
> @@ -30,34 +31,21 @@
>   */
>  static unsigned long bitops_ffsl(unsigned long word)
>  {
> -	int num = 0;
> +#if QEMU_GNUC_PREREQ(3, 4)
> +    return __builtin_ffsl(word) + 1;

Nope.  The +1 is wrong (it was only needed when using __builtin_ctzl()).
Stefan Weil - Jan. 30, 2013, 6:08 p.m.
Am 30.01.2013 18:57, schrieb Paolo Bonzini:
> Fixes the build on Mac OS X, which has ffsl.
>
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> ---
>  include/qemu/bitops.h     | 40 ++++++++++++++--------------------------
>  include/qemu/hbitmap.h    |  2 +-
>  include/qemu/host-utils.h | 25 -------------------------
>  util/hbitmap.c            |  2 +-
>  4 files changed, 16 insertions(+), 53 deletions(-)


For MinGW / MinGW-w64 builds:

Tested-by: Stefan Weil <sw@weilnetz.de>

Patch

diff --git a/include/qemu/bitops.h b/include/qemu/bitops.h
index 74e14e5..27b2d42 100644
--- a/include/qemu/bitops.h
+++ b/include/qemu/bitops.h
@@ -13,6 +13,7 @@ 
 #define BITOPS_H
 
 #include "qemu-common.h"
+#include "host-utils.h"
 
 #define BITS_PER_BYTE           CHAR_BIT
 #define BITS_PER_LONG           (sizeof (unsigned long) * BITS_PER_BYTE)
@@ -30,34 +31,21 @@ 
  */
 static unsigned long bitops_ffsl(unsigned long word)
 {
-	int num = 0;
+#if QEMU_GNUC_PREREQ(3, 4)
+    return __builtin_ffsl(word) + 1;
+#else
+    if (!word) {
+        return 0;
+    }
 
-#if LONG_MAX > 0x7FFFFFFF
-	if ((word & 0xffffffff) == 0) {
-		num += 32;
-		word >>= 32;
-	}
+    if (sizeof(long) == 4) {
+        return ctz32(word) + 1;
+    } else if (sizeof(long) == 8) {
+        return ctz64(word) + 1;
+    } else {
+        abort();
+    }
 #endif
-	if ((word & 0xffff) == 0) {
-		num += 16;
-		word >>= 16;
-	}
-	if ((word & 0xff) == 0) {
-		num += 8;
-		word >>= 8;
-	}
-	if ((word & 0xf) == 0) {
-		num += 4;
-		word >>= 4;
-	}
-	if ((word & 0x3) == 0) {
-		num += 2;
-		word >>= 2;
-	}
-	if ((word & 0x1) == 0) {
-		num += 1;
-        }
-	return num;
 }
 
 /**
diff --git a/include/qemu/hbitmap.h b/include/qemu/hbitmap.h
index 73f5d1d..abad209 100644
--- a/include/qemu/hbitmap.h
+++ b/include/qemu/hbitmap.h
@@ -170,7 +170,7 @@  static inline int64_t hbitmap_iter_next(HBitmapIter *hbi)
 
     /* The next call will resume work from the next bit.  */
     hbi->cur[HBITMAP_LEVELS - 1] = cur & (cur - 1);
-    item = ((uint64_t)hbi->pos << BITS_PER_LEVEL) + ffsl(cur) - 1;
+    item = ((uint64_t)hbi->pos << BITS_PER_LEVEL) + bitops_ffsl(cur) - 1;
 
     return item << hbi->granularity;
 }
diff --git a/include/qemu/host-utils.h b/include/qemu/host-utils.h
index 2a32be4..40457bd 100644
--- a/include/qemu/host-utils.h
+++ b/include/qemu/host-utils.h
@@ -238,29 +238,4 @@  static inline int ctpop64(uint64_t val)
 #endif
 }
 
-/* glibc does not provide an inline version of ffsl, so always define
- * ours.  We need to give it a different name, however.
- */
-#ifdef __GLIBC__
-#define ffsl qemu_ffsl
-#endif
-static inline int ffsl(long val)
-{
-    if (!val) {
-        return 0;
-    }
-
-#if QEMU_GNUC_PREREQ(3, 4)
-    return __builtin_ctzl(val) + 1;
-#else
-    if (sizeof(long) == 4) {
-        return ctz32(val) + 1;
-    } else if (sizeof(long) == 8) {
-        return ctz64(val) + 1;
-    } else {
-        abort();
-    }
-#endif
-}
-
 #endif
diff --git a/util/hbitmap.c b/util/hbitmap.c
index 2aa487d..32c3d59 100644
--- a/util/hbitmap.c
+++ b/util/hbitmap.c
@@ -126,7 +126,7 @@  unsigned long hbitmap_iter_skip_words(HBitmapIter *hbi)
          * The index of this word's least significant set bit provides
          * the low-order bits.
          */
-        pos = (pos << BITS_PER_LEVEL) + ffsl(cur) - 1;
+        pos = (pos << BITS_PER_LEVEL) + bitops_ffsl(cur) - 1;
         hbi->cur[i] = cur & (cur - 1);
 
         /* Set up next level for iteration.  */