diff options
| author | zack <zack@138bc75d-0d04-0410-961f-82ee72b054a4> | 2003-10-29 00:55:43 +0000 |
|---|---|---|
| committer | zack <zack@138bc75d-0d04-0410-961f-82ee72b054a4> | 2003-10-29 00:55:43 +0000 |
| commit | da4ab0bd4489694c803069ceeac240ca79d7db4b (patch) | |
| tree | ef98c209a92bd82779ac5da489d6ce4b266ce19a /libjava/gnu/java | |
| parent | fc0cc5c6e2f64624c09ad34cc68f17961e2552f5 (diff) | |
| download | ppe42-gcc-da4ab0bd4489694c803069ceeac240ca79d7db4b.tar.gz ppe42-gcc-da4ab0bd4489694c803069ceeac240ca79d7db4b.zip | |
* ia64.md (UNSPEC_SETF_EXP,UNSPEC_FR_SQRT_RECIP_APPROX): New constants.
(*sqrt_approx): New instruction pattern for approximate square roots.
(*setf_exp_xf): New instruction pattern for exponentiation.
(*maddxf4_alts_truncsf): New instruction pattern for truncation.
(sqrtsf2_internal_thr): New define_and_split implementing
throughput-optimized inline calculation of SFmode square root.
(sqrtdf2_internal_thr): Likewise for DFmode.
(sqrtxf2_internal_thr): Likewise for XFmode.
(sqrtsf2, sqrtdf2, sqrtxf2): New expanders to choose between
latency- and throughput-optimized square root algorithms.
* ia64.h (MASK_INLINE_SQRT_LAT, MASK_INLINE_SQRT_THR,
TARGET_INLINE_SQRT_LAT, TARGET_INLINE_SQRT_THR, TARGET_INLINE_SQRT):
New macros.
(TARGET_SWITCHES): Add -minline-sqrt-min-latency and
-minline-sqrt-max-throughput.
* ia64.c (ia64_override_options): If both -minline-sqrt-min-latency
and -minline-sqrt-max-throughput are given, notify the user
that both options cannot be used simultaneously.
If -minline-sqrt-min-latency is given, notify the user that
this mode is not yet implemented.
(rtx_needs_barrier): Reformat initial comment to obey
72-character width limit. Support UNSPEC_SETF_EXP and
UNSPEC_FR_SQRT_RECIP_APPROX.
git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@73027 138bc75d-0d04-0410-961f-82ee72b054a4
Diffstat (limited to 'libjava/gnu/java')
0 files changed, 0 insertions, 0 deletions
