Tue, 15 Jan 2013 18:27:01 +0100 |
Thomas Wuerthinger |
Added DivRemOp that produces both the division result and the remainder.
|
Tue, 15 Jan 2013 17:53:31 +0100 |
Thomas Wuerthinger |
Canonicalize ((a - a % b) / b) into (a / b).
|
Tue, 15 Jan 2013 15:44:04 +0100 |
Doug Simon |
an intrinsic graph must be cloned before being compiled
|
Tue, 15 Jan 2013 11:10:59 +0100 |
Gilles Duboscq |
Long.bitCount should also not be intrinsified if popcnt is missing
|
Tue, 15 Jan 2013 00:51:29 +0100 |
Thomas Wuerthinger |
Merge.
|
Tue, 15 Jan 2013 00:51:12 +0100 |
Thomas Wuerthinger |
New experiment with LSRA order. Remove old block order calculation.
|
Mon, 14 Jan 2013 18:41:26 +0100 |
Doug Simon |
reworked mechanism for omitting intrinsics if the underlying hardware support is not available and used it to make intrinsification of Integer.bitCount() dependent on whether the POPCNT instruction is available
|
Mon, 14 Jan 2013 16:56:54 +0100 |
Thomas Wuerthinger |
Merge.
|
Mon, 14 Jan 2013 16:52:59 +0100 |
Thomas Wuerthinger |
Merge.
|
Mon, 14 Jan 2013 16:52:44 +0100 |
Thomas Wuerthinger |
Experiment with new block order for LSRA.
|
Mon, 14 Jan 2013 14:19:49 +0100 |
Thomas Wuerthinger |
Backed out changeset b5280041f59e. Alignment for branch targets has no positive effect. Remove to keep code cache smaller.
|
Mon, 14 Jan 2013 16:29:08 +0100 |
Lukas Stadler |
new facility to add branch probabilities for snippets, probabilities for frequently-used snippets
|
Mon, 14 Jan 2013 15:23:13 +0100 |
Gilles Duboscq |
Make sure that compressed oops are not used in benchmarks
|
Mon, 14 Jan 2013 14:26:00 +0100 |
Doug Simon |
fixed bugs in AESCrypt intrinsifications and enabled them by default
|