Commit Graph

17 Commits

Author SHA1 Message Date
Elena Demikhovsky 52e4a0e109 AVX-512: Added more intrinsics for convert and min/max.
Removed vzeroupper from AVX-512 mode - our optimization gude does not recommend to insert vzeroupper at all.

llvm-svn: 198557
2014-01-05 10:46:09 +00:00
Michael Liao 9a0e3f4823 [X86] Check YMM31/ZMM31 as well
- No test case as there's no calling convention preserve YMM31/ZMM31 only

llvm-svn: 196391
2013-12-04 17:44:22 +00:00
Michael Liao 14b02848a3 Enhance the fix of PR17631
- The fix to PR17631 fixes part of the cases where 'vzeroupper' should
  not be issued before 'call' insn. There're other cases where helper
  calls will be inserted not limited to epilog. These helper calls do
  not follow the standard calling convention and won't clobber any YMM
  registers. (So far, all call conventions will clobber any or part of
  YMM registers.)
  This patch enhances the previous fix to cover more cases 'vzerosupper' should
  not be inserted by checking if that function call won't clobber any YMM
  registers and skipping it if so.

llvm-svn: 196261
2013-12-03 09:17:32 +00:00
Michael Liao c7bf44d7bb Fix PR17631
- Skip instructions added in prolog. For specific targets, prolog may
  insert helper function calls (e.g. _chkstk will be called when
  there're more than 4K bytes allocated on stack). However, these
  helpers don't use/def YMM/XMM registers.

llvm-svn: 193261
2013-10-23 18:32:43 +00:00
Elena Demikhovsky 8cfb43f73b I'm starting to commit KNL backend. I'll push patches one-by-one. This patch includes support for the extended register set XMM16-31, YMM16-31, ZMM0-31.
The full ISA you can see here: http://software.intel.com/en-us/intel-isa-extensions

llvm-svn: 187030
2013-07-24 11:02:47 +00:00
Elena Demikhovsky 9e0df7cb01 Prevent insertion of "vzeroupper" before call that preserves YMM registers, since a caller uses preserved registers across the call.
llvm-svn: 175043
2013-02-13 08:02:04 +00:00
Jakob Stoklund Olesen a10c09804d Check for empty YMM use-def lists in X86VZeroUpper.
The previous MRI.isPhysRegUsed(YMM0) would also return true when the
function contains a call to a function that may clobber YMM0. That's
most of them.

Checking the use-def chains allows us to skip functions that don't
explicitly mention YMM registers.

llvm-svn: 166110
2012-10-17 17:52:35 +00:00
Craig Topper 056dfcccb7 Don't cache the MBB in the class. Its only used by one function. Change a for loop over operands to use unsigned instead of int.
llvm-svn: 162344
2012-08-22 05:59:59 +00:00
Craig Topper 455bcafa3b Mark a function as static since it doesn't use anything in the class.
llvm-svn: 162342
2012-08-22 05:36:44 +00:00
Chad Rosier 24c19d20c0 Whitespace.
llvm-svn: 161122
2012-08-01 18:39:17 +00:00
Benjamin Kramer bde9176663 Fix typos found by http://github.com/lyda/misspell-check
llvm-svn: 157885
2012-06-02 10:20:22 +00:00
Craig Topper abadc660e0 Convert some uses of XXXRegisterClass to &XXXRegClass. No functional change since they are equivalent.
llvm-svn: 155186
2012-04-20 06:31:50 +00:00
Craig Topper 760b134ffa Make all pointers to TargetRegisterClass const since they are all pointers to static data that should not be modified.
llvm-svn: 151134
2012-02-22 05:59:10 +00:00
Evan Cheng 7f8e563a69 Add bundle aware API for querying instruction properties and switch the code
generator to it. For non-bundle instructions, these behave exactly the same
as the MC layer API.

For properties like mayLoad / mayStore, look into the bundle and if any of the
bundled instructions has the property it would return true.
For properties like isPredicable, only return true if *all* of the bundled
instructions have the property.
For properties like canFoldAsLoad, isCompare, conservatively return false for
bundles.

llvm-svn: 146026
2011-12-07 07:15:52 +00:00
Eli Friedman 8f249600e7 Enhanced vzeroupper insertion pass that avoids inserting vzeroupper where it is unnecessary through local analysis. Patch from Bruno Cardoso Lopes, with some additional changes.
I'm going to wait for any review comments and perform some additional testing before turning this on by default.

llvm-svn: 143750
2011-11-04 23:46:11 +00:00
Benjamin Kramer 7859d2e148 Use internal storage for command line option.
llvm-svn: 139079
2011-09-03 03:45:06 +00:00
Bruno Cardoso Lopes 2a3ffb5d97 Introduce a pass to insert vzeroupper instructions to avoid AVX to
SSE transition penalty. The pass is enabled through the "x86-use-vzeroupper"
llc command line option. This is only the first step (very naive and
conservative one) to sketch out the idea, but proper DFA is coming next
to allow smarter decisions. Comments and ideas now and in further commits
will be very appreciated.

llvm-svn: 138317
2011-08-23 01:14:17 +00:00