libvecpf - The Vector Printf Library
#include <stdio.h>
int printf(const char *format, ...);
int fprintf(FILE *stream, const char *format,
...);
int sprintf(char *str, const char *format,
...);
int snprintf(char *str, size_t size, const char
*format, ...);
This library extends the printf() function, via a
printf-hooks callback mechanism in GLIBC, so that it may format Vector data
types for display.
The following additional length modifiers are supported on
platforms that have the Altivec/VMX (Power ISA 2.01 with Vector extensions).
Since the VSX facility is predicated on VMX support these additional length
modifiers are also supported on platforms that have the VSX facility. In
this section the term, `integer conversion' stands for the d,
i, o, u, x, X, and c conversions.
The term, `floating point conversion' stands for the f, F,
e, E, g, G, a, and A
conversions.
- vl, lv
- (vee-ell,ell-vee) A following integer conversion corresponds to a 16 byte
vector composed of four vector unsigned ints or vector
signed ints.
- vh, hv
- A following integer conversions corresponds to a 16 byte vector composed
of eight vector unsigned shorts or vector signed
shorts.
- v
- A following integer conversion corresponds to a 16 byte vector composed of
sixteen vector unsigned chars or vector signed
chars. A following floating point conversion corresponds to a 16
byte vector composed of four vector float single-precision floating
point values.
The following additional length modifiers are ONLY
supported on platforms that have the VSX facility. In this section the term,
`floating point conversion' stands for the f, F, e,
E, g, G, a, and A conversions.
- vv
- (vee-vee) A following floating point conversion corresponds to a 16 byte
vector composed of two vector double double-precision floating
point values.
The following integer printf conversion specifiers are supported
by libvecpf:
d, i, o, u, x, X, c
The following floating point printf conversion specifiers are
supported by libvecpf:
f, F, e, E, g, G, a, A
- %vld, %lvd
- Output a vector of four signed ints in signed decimal
notation.
- %vli, %lvi
- Same as %vld and %lvd.
- %vlu, %lvu
- Output a vector of four unsigned ints in unsigned decimal
notation.
- %vlo, %lvo
- Output a vector of four unsigned ints in unsigned octal
notation.
- %vlx, %lvx
- Output a vector of four unsigned ints in unsigned
hexadecimal (x) notation where the letters abcdef are
used.
- %vlX, %lvX
- Output a vector of four unsigned ints in unsigned
hexadecimal (X) notation where the letters ABCDEF are
used.
- %vhd, %hvd
- Output a vector of eight signed shorts in signed decimal
notation.
- %vhi, %hvi
- Same as %vhd and %hvd.
- %vhu,%hvu
- Output a vector of eight unsigned shorts in unsigned decimal
notation.
- %vho, %hvo
- Output a vector of eight unsigned shorts in unsigned octal
notation.
- %vhx, %hvx
- Output a vector of eight unsigned shorts in unsigned
hexadecimal 'x' notation where the letters abcdef are
used.
- %vhX, %hvX
- Output a vector of eight unsigned shorts in unsigned
hexadecimal (X) notation where the letters ABCDEF are
used.
- %vd
- Output a vector of sixteen signed chars in signed decimal
notation.
- %vi
- Same as %vd.
- %vu
- Output a vector of sixteen unsigned chars in unsigned
decimal notation.
- %vo
- Output a vector of sixteen unsigned chars in unsigned octal
notation.
- %vx
- Output a vector of sixteen unsigned chars in unsigned
hexadecimal 'x' notation where the letters abcdef are
used.
- %vX
- Output a vector of sixteen unsigned chars in unsigned
hexadecimal (X) notation where the letters ABCDEF are
used.
- %vc
- Output a vector of sixteen unsigned chars as
characters.
- %ve, %vE
- Output a vector of four single-precision floats in the manner
indicated in man printf(3) in The conversion specifier
section, under the e and E specifiers, respectively.
- %vf, %vF
- Output a vector of four single-precision floats in the manner
indicated in man printf(3) in The conversion specifier
section, under the f and F specifiers, respectively.
- %vg, %vG
- Output a vector of four single-precision floats in the manner
indicated in man printf(3) in The conversion specifier
section, under the g and G specifiers, respectively.
- %va, %vA
- Output a vector of four single-precision floats in the manner
indicated in man printf(3) in The conversion specifier
section, under the a and A specifiers, respectively.
- %vve, %vvE
- Output a vector of two double-precision doubles in the manner
indicated in man printf(3) in The conversion specifier
section, under the e and E specifiers, respectively.
- %vvf, %vvF
- Output a vector of two double-precision doubles in the manner
indicated in man printf(3) in The conversion specifier
section, under the f and F specifiers, respectively.
- %vvg, %vvG
- Output a vector of two double-precision doubles in the manner
indicated in man printf(3) in The conversion specifier
section, under the g and G specifiers, respectively.
- %vva, %vvA
- Output a vector of two double-precision doubles in the manner
indicated in man printf(3) in The conversion specifier
section, under the a and A specifiers, respectively.
#include <stdio.h>
#include <altivec.h>
int main() {
vector double d = { -1111.12304912348f, 4567.987654f };
printf("%16.16vvf0, d);
return 0;
}
Using the additional printf length modifiers defined by this
library in a program will cause the GCC compiler to complain in the
following manner when compiled with -Wall:
warning: unknown conversion type character ‘v’ in format
This warning can be suppressed with the -Wno-format
compiler switch. Only suppress warnings with care.
"gcc -maltivec -O3 m32 -g -Wall -o foo foo.c -mcpu=power7 -lvecpf -Wno-format"
Compiling an application and linking it against the static library
libvecpf.a will expose a default feature of the link editor (ld) which
purges unused symbols. The link editor sees constructors as unused in this
context and purges them from the final archive. This will prevent the
registration of the printf-hook callback mechanism and printf will
not support Vector data types.
In order to prevent the link editor from purging the libvecpf
constructor on the final link of a static linking operation explicitly tell
it to not purge the constructor by passing
-Wl,-u,__libvecpf_init
during the link stage as in the following example:
gcc -static -maltivec -O3 -m32 -g -Wall -o foo foo.c -mcpu=power7
-Wl,-u,__libvecpf_init -lvecpf -Wno-format
GLIBC printf-hooks support didn't go into GLIBC until version
2.10.
Libvecpf only supports the default separator; as things are today
the GLIBC printf-hooks mechanism doesn't allow separator flags to be
changed.
Email bug reports to Ryan S. Arnold
<rsa@linux.vnet.ibm.com>.
This manual page was written by Ryan S. Arnold
<rsa@linux.vnet.ibm.com>.