mirror/glibc
2
0
Fork 0
mirror of git://sourceware.org/git/glibc.git synced 2025-02-23 13:09:58 +08:00
Code Issues Packages Projects Releases Wiki Activity

Fix comment formatting in new optimized sparc math files.

Browse Source
This commit is contained in:
David S. Miller 2012-02-27 15:24:09 -08:00
parent 0e7727f71d
commit d674b76d0e
9 changed files with 123 additions and 120 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
ChangeLog
View File

@ -1,3 +1,14 @@
2012-02-27 David S. Miller <davem@davemloft.net>
* sysdeps/sparc/sparc32/sparcv9/fpu/s_ceil.S: Fix comment formatting.
* sysdeps/sparc/sparc32/sparcv9/fpu/s_ceilf.S: Likewise.
* sysdeps/sparc/sparc32/sparcv9/fpu/s_rint.S: Likewise.
* sysdeps/sparc/sparc32/sparcv9/fpu/s_rintf.S: Likewise.
* sysdeps/sparc/sparc64/fpu/s_ceil.S: Likewise.
* sysdeps/sparc/sparc64/fpu/s_ceilf.S: Likewise.
* sysdeps/sparc/sparc64/fpu/s_rint.S: Likewise.
* sysdeps/sparc/sparc64/fpu/s_rintf.S: Likewise.
2012-02-27 Joseph Myers <joseph@codesourcery.com>
* configure.in (CC): Restrict allowed GCC versions to 4.3 and

45
sysdeps/sparc/sparc32/sparcv9/fpu/s_ceil.S
View File

@ -20,29 +20,28 @@
#include <sysdep.h>
/* Since changing the rounding mode is extremely expensive, we
* try to round up using a method that is rounding mode
* agnostic.
*
* We add then subtract (or subtract than add if the initial
* value was negative) 2**23 to the value, then subtract it
* back out.
*
* This will clear out the fractional portion of the value.
* One of two things will happen for non-whole initial values.
* Either the rounding mode will round it up, or it will be
* rounded down. If the value started out whole, it will be
* equal after the addition and subtraction. This means we
* can accurately detect with one test whether we need to add
* another 1.0 to round it up properly.
*
* We pop constants into the FPU registers using the incoming
* argument stack slots, since this avoid having to use any PIC
* references. We also thus avoid having to allocate a register
* window.
*
* VIS instructions are used to facilitate the formation of
* easier constants, and the propagation of the sign bit.
*/
try to round up using a method that is rounding mode
agnostic.
We add then subtract (or subtract than add if the initial
value was negative) 2**23 to the value, then subtract it
back out.
This will clear out the fractional portion of the value.
One of two things will happen for non-whole initial values.
Either the rounding mode will round it up, or it will be
rounded down. If the value started out whole, it will be
equal after the addition and subtraction. This means we
can accurately detect with one test whether we need to add
another 1.0 to round it up properly.
We pop constants into the FPU registers using the incoming
argument stack slots, since this avoid having to use any
PIC references. We also thus avoid having to allocate a
register window.
VIS instructions are used to facilitate the formation of
easier constants, and the propagation of the sign bit. */
#define TWO_FIFTYTWO 0x43300000 /* 2**52 */
#define ONE_DOT_ZERO 0x3ff00000 /* 1.0 */

45
sysdeps/sparc/sparc32/sparcv9/fpu/s_ceilf.S
View File

@ -20,29 +20,28 @@
#include <sysdep.h>
/* Since changing the rounding mode is extremely expensive, we
* try to round up using a method that is rounding mode
* agnostic.
*
* We add then subtract (or subtract than add if the initial
* value was negative) 2**23 to the value, then subtract it
* back out.
*
* This will clear out the fractional portion of the value.
* One of two things will happen for non-whole initial values.
* Either the rounding mode will round it up, or it will be
* rounded down. If the value started out whole, it will be
* equal after the addition and subtraction. This means we
* can accurately detect with one test whether we need to add
* another 1.0 to round it up properly.
*
* We pop constants into the FPU registers using the incoming
* argument stack slots, since this avoid having to use any PIC
* references. We also thus avoid having to allocate a register
* window.
*
* VIS instructions are used to facilitate the formation of
* easier constants, and the propagation of the sign bit.
*/
try to round up using a method that is rounding mode
agnostic.
We add then subtract (or subtract than add if the initial
value was negative) 2**23 to the value, then subtract it
back out.
This will clear out the fractional portion of the value.
One of two things will happen for non-whole initial values.
Either the rounding mode will round it up, or it will be
rounded down. If the value started out whole, it will be
equal after the addition and subtraction. This means we
can accurately detect with one test whether we need to add
another 1.0 to round it up properly.
We pop constants into the FPU registers using the incoming
argument stack slots, since this avoid having to use any
PIC references. We also thus avoid having to allocate a
register window.
VIS instructions are used to facilitate the formation of
easier constants, and the propagation of the sign bit. */
#define TWO_TWENTYTHREE 0x4b000000 /* 2**23 */
#define ONE_DOT_ZERO 0x3f800000 /* 1.0 */

13
sysdeps/sparc/sparc32/sparcv9/fpu/s_rint.S
View File

@ -20,13 +20,12 @@
#include <sysdep.h>
/* We pop constants into the FPU registers using the incoming
* argument stack slots, since this avoid having to use any PIC
* references. We also thus avoid having to allocate a register
* window.
*
* VIS instructions are used to facilitate the formation of
* easier constants, and the propagation of the sign bit.
*/
argument stack slots, since this avoid having to use any PIC
references. We also thus avoid having to allocate a register
window.
VIS instructions are used to facilitate the formation of
easier constants, and the propagation of the sign bit. */
#define TWO_FIFTYTWO 0x43300000 /* 2**52 */

13
sysdeps/sparc/sparc32/sparcv9/fpu/s_rintf.S
View File

@ -20,13 +20,12 @@
#include <sysdep.h>
/* We pop constants into the FPU registers using the incoming
* argument stack slots, since this avoid having to use any PIC
* references. We also thus avoid having to allocate a register
* window.
*
* VIS instructions are used to facilitate the formation of
* easier constants, and the propagation of the sign bit.
*/
argument stack slots, since this avoid having to use any PIC
references. We also thus avoid having to allocate a register
window.
VIS instructions are used to facilitate the formation of
easier constants, and the propagation of the sign bit. */
#define TWO_TWENTYTHREE 0x4b000000 /* 2**23 */

45
sysdeps/sparc/sparc64/fpu/s_ceil.S
View File

@ -20,29 +20,28 @@
#include <sysdep.h>
/* Since changing the rounding mode is extremely expensive, we
* try to round up using a method that is rounding mode
* agnostic.
*
* We add then subtract (or subtract than add if the initial
* value was negative) 2**23 to the value, then subtract it
* back out.
*
* This will clear out the fractional portion of the value.
* One of two things will happen for non-whole initial values.
* Either the rounding mode will round it up, or it will be
* rounded down. If the value started out whole, it will be
* equal after the addition and subtraction. This means we
* can accurately detect with one test whether we need to add
* another 1.0 to round it up properly.
*
* We pop constants into the FPU registers using the incoming
* argument stack slots, since this avoid having to use any PIC
* references. We also thus avoid having to allocate a register
* window.
*
* VIS instructions are used to facilitate the formation of
* easier constants, and the propagation of the sign bit.
*/
try to round up using a method that is rounding mode
agnostic.
We add then subtract (or subtract than add if the initial
value was negative) 2**23 to the value, then subtract it
back out.
This will clear out the fractional portion of the value.
One of two things will happen for non-whole initial values.
Either the rounding mode will round it up, or it will be
rounded down. If the value started out whole, it will be
equal after the addition and subtraction. This means we
can accurately detect with one test whether we need to add
another 1.0 to round it up properly.
We pop constants into the FPU registers using the incoming
argument stack slots, since this avoid having to use any
PIC references. We also thus avoid having to allocate a
register window.
VIS instructions are used to facilitate the formation of
easier constants, and the propagation of the sign bit. */
#define TWO_FIFTYTWO 0x43300000 /* 2**52 */
#define ONE_DOT_ZERO 0x3ff00000 /* 1.0 */

45
sysdeps/sparc/sparc64/fpu/s_ceilf.S
View File

@ -20,29 +20,28 @@
#include <sysdep.h>
/* Since changing the rounding mode is extremely expensive, we
* try to round up using a method that is rounding mode
* agnostic.
*
* We add then subtract (or subtract than add if the initial
* value was negative) 2**23 to the value, then subtract it
* back out.
*
* This will clear out the fractional portion of the value.
* One of two things will happen for non-whole initial values.
* Either the rounding mode will round it up, or it will be
* rounded down. If the value started out whole, it will be
* equal after the addition and subtraction. This means we
* can accurately detect with one test whether we need to add
* another 1.0 to round it up properly.
*
* We pop constants into the FPU registers using the incoming
* argument stack slots, since this avoid having to use any PIC
* references. We also thus avoid having to allocate a register
* window.
*
* VIS instructions are used to facilitate the formation of
* easier constants, and the propagation of the sign bit.
*/
try to round up using a method that is rounding mode
agnostic.
We add then subtract (or subtract than add if the initial
value was negative) 2**23 to the value, then subtract it
back out.
This will clear out the fractional portion of the value.
One of two things will happen for non-whole initial values.
Either the rounding mode will round it up, or it will be
rounded down. If the value started out whole, it will be
equal after the addition and subtraction. This means we
can accurately detect with one test whether we need to add
another 1.0 to round it up properly.
We pop constants into the FPU registers using the incoming
argument stack slots, since this avoid having to use any
PIC references. We also thus avoid having to allocate a
register window.
VIS instructions are used to facilitate the formation of
easier constants, and the propagation of the sign bit. */
#define TWO_TWENTYTHREE 0x4b000000 /* 2**23 */
#define ONE_DOT_ZERO 0x3f800000 /* 1.0 */

13
sysdeps/sparc/sparc64/fpu/s_rint.S
View File

@ -20,13 +20,12 @@
#include <sysdep.h>
/* We pop constants into the FPU registers using the incoming
* argument stack slots, since this avoid having to use any PIC
* references. We also thus avoid having to allocate a register
* window.
*
* VIS instructions are used to facilitate the formation of
* easier constants, and the propagation of the sign bit.
*/
argument stack slots, since this avoid having to use any PIC
references. We also thus avoid having to allocate a register
window.
VIS instructions are used to facilitate the formation of
easier constants, and the propagation of the sign bit. */
#define TWO_FIFTYTWO 0x43300000 /* 2**52 */

13
sysdeps/sparc/sparc64/fpu/s_rintf.S
View File

@ -20,13 +20,12 @@
#include <sysdep.h>
/* We pop constants into the FPU registers using the incoming
* argument stack slots, since this avoid having to use any PIC
* references. We also thus avoid having to allocate a register
* window.
*
* VIS instructions are used to facilitate the formation of
* easier constants, and the propagation of the sign bit.
*/
argument stack slots, since this avoid having to use any PIC
references. We also thus avoid having to allocate a register
window.
VIS instructions are used to facilitate the formation of
easier constants, and the propagation of the sign bit. */
#define TWO_TWENTYTHREE 0x4b000000 /* 2**23 */