| Message ID | 1391979085-8012-1-git-send-email-imirkin@alum.mit.edu (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Mon, Feb 10, 2014 at 6:51 AM, Ilia Mirkin <imirkin@alum.mit.edu> wrote: > The ffsll function is a lot slower than the __ffs64 built-in which > compiles to a single instruction on 64-bit. It's also nice to avoid > custom versions of standard functions. Note that __ffs == ffs - 1. > > Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu> Merged both patches in the series. Thanks. > --- > > I wrote a user-space program to test these out and make sure that the > functions behaved as expected. The logic in abi16 had to be flipped around a > bit since __ffs doesn't distinguish between 0 and 1. There's a minor > difference in that init->channel is going to get returned as 0 for ENOSPC vs > -1, but I can't imagine that'd matter. > > drivers/gpu/drm/nouveau/core/core/parent.c | 2 +- > drivers/gpu/drm/nouveau/core/os.h | 11 ----------- > drivers/gpu/drm/nouveau/nouveau_abi16.c | 4 ++-- > 3 files changed, 3 insertions(+), 14 deletions(-) > > diff --git a/drivers/gpu/drm/nouveau/core/core/parent.c b/drivers/gpu/drm/nouveau/core/core/parent.c > index 313380c..dee5d12 100644 > --- a/drivers/gpu/drm/nouveau/core/core/parent.c > +++ b/drivers/gpu/drm/nouveau/core/core/parent.c > @@ -49,7 +49,7 @@ nouveau_parent_sclass(struct nouveau_object *parent, u16 handle, > > mask = nv_parent(parent)->engine; > while (mask) { > - int i = ffsll(mask) - 1; > + int i = __ffs64(mask); > > if (nv_iclass(parent, NV_CLIENT_CLASS)) > engine = nv_engine(nv_client(parent)->device); > diff --git a/drivers/gpu/drm/nouveau/core/os.h b/drivers/gpu/drm/nouveau/core/os.h > index 191e739..3cd6120 100644 > --- a/drivers/gpu/drm/nouveau/core/os.h > +++ b/drivers/gpu/drm/nouveau/core/os.h > @@ -23,17 +23,6 @@ > > #include <asm/unaligned.h> > > -static inline int > -ffsll(u64 mask) > -{ > - int i; > - for (i = 0; i < 64; i++) { > - if (mask & (1ULL << i)) > - return i + 1; > - } > - return 0; > -} > - > #ifndef ioread32_native > #ifdef __BIG_ENDIAN > #define ioread16_native ioread16be > diff --git a/drivers/gpu/drm/nouveau/nouveau_abi16.c b/drivers/gpu/drm/nouveau/nouveau_abi16.c > index 900fae0..b701117 100644 > --- a/drivers/gpu/drm/nouveau/nouveau_abi16.c > +++ b/drivers/gpu/drm/nouveau/nouveau_abi16.c > @@ -270,8 +270,8 @@ nouveau_abi16_ioctl_channel_alloc(ABI16_IOCTL_ARGS) > return nouveau_abi16_put(abi16, -EINVAL); > > /* allocate "abi16 channel" data and make up a handle for it */ > - init->channel = ffsll(~abi16->handles); > - if (!init->channel--) > + init->channel = __ffs64(~abi16->handles); > + if (~abi16->handles == 0) > return nouveau_abi16_put(abi16, -ENOSPC); > > chan = kzalloc(sizeof(*chan), GFP_KERNEL); > -- > 1.8.3.2 > > _______________________________________________ > dri-devel mailing list > dri-devel@lists.freedesktop.org > http://lists.freedesktop.org/mailman/listinfo/dri-devel
diff --git a/drivers/gpu/drm/nouveau/core/core/parent.c b/drivers/gpu/drm/nouveau/core/core/parent.c index 313380c..dee5d12 100644 --- a/drivers/gpu/drm/nouveau/core/core/parent.c +++ b/drivers/gpu/drm/nouveau/core/core/parent.c @@ -49,7 +49,7 @@ nouveau_parent_sclass(struct nouveau_object *parent, u16 handle, mask = nv_parent(parent)->engine; while (mask) { - int i = ffsll(mask) - 1; + int i = __ffs64(mask); if (nv_iclass(parent, NV_CLIENT_CLASS)) engine = nv_engine(nv_client(parent)->device); diff --git a/drivers/gpu/drm/nouveau/core/os.h b/drivers/gpu/drm/nouveau/core/os.h index 191e739..3cd6120 100644 --- a/drivers/gpu/drm/nouveau/core/os.h +++ b/drivers/gpu/drm/nouveau/core/os.h @@ -23,17 +23,6 @@ #include <asm/unaligned.h> -static inline int -ffsll(u64 mask) -{ - int i; - for (i = 0; i < 64; i++) { - if (mask & (1ULL << i)) - return i + 1; - } - return 0; -} - #ifndef ioread32_native #ifdef __BIG_ENDIAN #define ioread16_native ioread16be diff --git a/drivers/gpu/drm/nouveau/nouveau_abi16.c b/drivers/gpu/drm/nouveau/nouveau_abi16.c index 900fae0..b701117 100644 --- a/drivers/gpu/drm/nouveau/nouveau_abi16.c +++ b/drivers/gpu/drm/nouveau/nouveau_abi16.c @@ -270,8 +270,8 @@ nouveau_abi16_ioctl_channel_alloc(ABI16_IOCTL_ARGS) return nouveau_abi16_put(abi16, -EINVAL); /* allocate "abi16 channel" data and make up a handle for it */ - init->channel = ffsll(~abi16->handles); - if (!init->channel--) + init->channel = __ffs64(~abi16->handles); + if (~abi16->handles == 0) return nouveau_abi16_put(abi16, -ENOSPC); chan = kzalloc(sizeof(*chan), GFP_KERNEL);
The ffsll function is a lot slower than the __ffs64 built-in which compiles to a single instruction on 64-bit. It's also nice to avoid custom versions of standard functions. Note that __ffs == ffs - 1. Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu> --- I wrote a user-space program to test these out and make sure that the functions behaved as expected. The logic in abi16 had to be flipped around a bit since __ffs doesn't distinguish between 0 and 1. There's a minor difference in that init->channel is going to get returned as 0 for ENOSPC vs -1, but I can't imagine that'd matter. drivers/gpu/drm/nouveau/core/core/parent.c | 2 +- drivers/gpu/drm/nouveau/core/os.h | 11 ----------- drivers/gpu/drm/nouveau/nouveau_abi16.c | 4 ++-- 3 files changed, 3 insertions(+), 14 deletions(-)