maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
100,111 Commits 23 Branches 0 Tags 2.2 GiB
Commit Graph

249 Commits

Author SHA1 Message Date
Chris Lattner 865fe3b283 add a note, progress unblocked by PR8575 being fixed. 
llvm-svn: 124599
 2011-01-31 20:23:28 +00:00
Benjamin Kramer 946e1522b6 Teach DAGCombine to fold fold (sra (trunc (sr x, c1)), c2) -> (trunc (sra x, c1+c2) when c1 equals the amount of bits that are truncated off. 
This happens all the time when a smul is promoted to a larger type.

On x86-64 we now compile "int test(int x) { return x/10; }" into
  movslq  %edi, %rax
  imulq $1717986919, %rax, %rax
  movq  %rax, %rcx
  shrq  $63, %rcx
  sarq  $34, %rax <- used to be "shrq $32, %rax; sarl $2, %eax"
  addl  %ecx, %eax

This fires 96 times in gcc.c on x86-64.

llvm-svn: 124559
 2011-01-30 16:38:43 +00:00
Chris Lattner 9685603260 this isn't a memset, we do convert dest[i] to one though :) 
llvm-svn: 124097
 2011-01-24 02:32:00 +00:00
Chris Lattner b830ee5250 with recent work, we now optimize this into: 
define i32 @foo(i32 %x) nounwind readnone ssp {
entry:
  %tobool = icmp eq i32 %x, 0
  %tmp5 = select i1 %tobool, i32 2, i32 1
  ret i32 %tmp5
}

llvm-svn: 124091
 2011-01-24 01:12:18 +00:00
Anders Carlsson 773bc67eff Add a memset loop that LoopIdiomRecognize doesn't recognize. 
llvm-svn: 124082
 2011-01-23 20:31:00 +00:00
Chris Lattner a56c8279e8 add a note 
llvm-svn: 123752
 2011-01-18 07:47:48 +00:00
Anders Carlsson 6a5171ba68 Update README.txt to remove the DAE enhancement. 
llvm-svn: 123597
 2011-01-16 21:26:15 +00:00
Chris Lattner c326ebd118 add some commentary 
llvm-svn: 123572
 2011-01-16 06:39:44 +00:00
Chandler Carruth ef28abefd0 Simplify a README.txt entry significantly to expose the core issue. 
llvm-svn: 123556
 2011-01-16 01:40:23 +00:00
Chris Lattner b6c3aff1cb typo 
llvm-svn: 123406
 2011-01-13 22:11:56 +00:00
Chris Lattner b9cdf393a4 memcpy + metadata = bliss :) 
llvm-svn: 123405
 2011-01-13 22:08:15 +00:00
Chandler Carruth b1e7f557b7 Teach constant folding to perform conversions from constant floating 
point values to their integer representation through the SSE intrinsic
calls. This is the last part of a README.txt entry for which I have real
world examples.

llvm-svn: 123206
 2011-01-11 01:07:24 +00:00
Owen Anderson d490c2d2ae Fix a random missed optimization by making InstCombine more aggressive when determining which bits are demanded by 
a comparison against a constant.

llvm-svn: 123203
 2011-01-11 00:36:45 +00:00
Chris Lattner 78cdd2a6c6 +0.0 vs -0.0 differences can be handled by looking at the user of the 
operation in some cases.

llvm-svn: 123190
 2011-01-10 21:01:17 +00:00
Chris Lattner eef1455020 expand on a note 
llvm-svn: 123145
 2011-01-10 00:33:01 +00:00
Chris Lattner 5b358c6825 typo 
llvm-svn: 123142
 2011-01-09 23:48:41 +00:00
Chris Lattner 320370e3ca xref a PR # 
llvm-svn: 123141
 2011-01-09 23:42:22 +00:00
Chandler Carruth d011d5317c Add a note about the inability to model FP -> int conversions which 
perform rounding other than truncation in the IR. Common C code for this
turns into really an LLVM intrinsic call that blocks a lot of further
optimizations.

llvm-svn: 123135
 2011-01-09 22:36:18 +00:00
Chandler Carruth 0c68a668fa Add a note about a missed FP optimization. 
llvm-svn: 123126
 2011-01-09 21:00:19 +00:00
Chandler Carruth 82e6f6a325 Another missed memset in std::vector initialization. 
llvm-svn: 123116
 2011-01-09 11:29:57 +00:00
Chandler Carruth 43f6d1b67e Fix a cut-paste-o so that the sample code is correct for my last note. 
Also, switch to a more clear 'sink' function with its declaration to
avoid any confusion about 'g'. Thanks for the suggestion Frits.

llvm-svn: 123113
 2011-01-09 10:10:59 +00:00
Chandler Carruth ad6e1f0501 Another missed optimization of trivial vector code. 
llvm-svn: 123112
 2011-01-09 09:58:36 +00:00
Chandler Carruth f32619300a Add a note about vector's size-constructor producing dead stores. 
llvm-svn: 123111
 2011-01-09 09:58:33 +00:00
Chandler Carruth 5d684c17a7 Add a note about a missed memset optimization from std::fill. 
llvm-svn: 123103
 2011-01-09 01:32:55 +00:00
Benjamin Kramer 134cde912a Revert 122959, it needs more thought. Add it back to README.txt with additional notes. 
llvm-svn: 123030
 2011-01-07 20:42:20 +00:00
Chris Lattner 84184b7207 With Benjamin's recent amazing patches, we should be able to do even better things :) 
llvm-svn: 122978
 2011-01-06 22:25:00 +00:00
Benjamin Kramer 1e01ade2e8 Add a note from llvmdev, this time with more info. 
llvm-svn: 122966
 2011-01-06 17:35:50 +00:00
Benjamin Kramer 605f21a6c8 EarlyCSE does this now (and GVN always did it). 
llvm-svn: 122960
 2011-01-06 13:19:46 +00:00
Benjamin Kramer 799b011276 InstCombine: If we call llvm.objectsize on a malloc call we can replace it with the size passed to malloc. 
llvm-svn: 122959
 2011-01-06 13:11:05 +00:00
Chris Lattner 245de78e06 add a note about object size from drystone, add a poorly optimized loop from 179.art. 
llvm-svn: 122954
 2011-01-06 07:41:22 +00:00
Chris Lattner 73552c2cce add a trivial instcombine missed in Dhrystone 
llvm-svn: 122953
 2011-01-06 07:09:23 +00:00
Chris Lattner 51415d26f1 update a bunch of entries. 
llvm-svn: 122700
 2011-01-02 18:31:38 +00:00
Chris Lattner ddf58010bd Allow loop-idiom to run on multiple BB loops, but still only scan the loop 
header for now for memset/memcpy opportunities.  It turns out that loop-rotate
is successfully rotating loops, but *DOESN'T MERGE THE BLOCKS*, turning "for 
loops" into 2 basic block loops that loop-idiom was ignoring.

With this fix, we form many *many* more memcpy and memsets than before, including
on the "history" loops in the viterbi benchmark, which look like this:

        for (j=0; j<MAX_history; ++j) {
          history_new[i][j+1] = history[2*i][j];
        }

Transforming these loops into memcpy's speeds up the viterbi benchmark from
11.98s to 3.55s on my machine.  Woo.

llvm-svn: 122685
 2011-01-02 07:58:36 +00:00
Chris Lattner 6c3fc0a52d a missed __builtin_object_size case. 
llvm-svn: 122676
 2011-01-01 22:57:31 +00:00
Chris Lattner e5d5a41a58 various updates. 
llvm-svn: 122675
 2011-01-01 22:52:11 +00:00
Duncan Sands 772749aea1 Revert commit 122654 at the request of Chris, who reckons that instsimplify 
is the wrong hammer for this nail, and is probably right.

llvm-svn: 122661
 2011-01-01 20:08:02 +00:00
Duncan Sands e3c539581c Fix a README item by having InstructionSimplify do a mild form of value 
numbering, in which it considers (for example) "%a = add i32 %x, %y" and
"%b = add i32 %x, %y" to be equal because the operands are equal and the
result of the instructions only depends on the values of the operands.
This has almost no effect (it removes 4 instructions from gcc-as-one-file),
and perhaps slows down compilation: I measured a 0.4% slowdown on the large
gcc-as-one-file testcase, but it wasn't statistically significant.

llvm-svn: 122654
 2011-01-01 16:12:09 +00:00
Chris Lattner 102bc01900 add a note from llvmdev 
llvm-svn: 122603
 2010-12-28 18:45:02 +00:00
Benjamin Kramer dfa40f8f19 Remove/fix invalid README entries. The well thought out strcpy function doesn't return a pointer to the end of the string. 
llvm-svn: 122496
 2010-12-23 15:32:07 +00:00
Chris Lattner 5e0c0c72e9 recognize an unsigned add with overflow idiom into uadd. 
This resolves a README entry and technically resolves PR4916,
but we still get poor code for the testcase in that PR because
GVN isn't CSE'ing uadd with add, filed as PR8817.

Previously we got:

_test7:                                 ## @test7
	addq	%rsi, %rdi
	cmpq	%rdi, %rsi
	movl	$42, %eax
	cmovaq	%rsi, %rax
	ret

Now we get:

_test7:                                 ## @test7
	addq	%rsi, %rdi
	movl	$42, %eax
	cmovbq	%rsi, %rax
	ret

llvm-svn: 122182
 2010-12-19 19:37:52 +00:00
Chris Lattner 5174921b5b add another overflow idiom 
llvm-svn: 121854
 2010-12-15 07:28:58 +00:00
Chris Lattner 2e33985300 add a note about overflow idiom recognition. 
llvm-svn: 121853
 2010-12-15 07:25:55 +00:00
Chris Lattner 27ecda1efd add a shift/imul missed optimization 
llvm-svn: 121850
 2010-12-15 07:10:43 +00:00
Chris Lattner aded09f27f add a note about a SPEC hack that gcc mainline does. 
llvm-svn: 121849
 2010-12-15 06:38:24 +00:00
Chris Lattner 14cb11ddb2 add a note 
llvm-svn: 121656
 2010-12-13 00:15:25 +00:00
Benjamin Kramer c4169cebe3 Generalize the and-icmp-select instcombine further by allowing selects of the form 
(x & 2^n) ? 2^m+C : C

we can offset both arms by C to get the "(x & 2^n) ? 2^m : 0" form, optimize the
select to a shift and apply the offset afterwards.

llvm-svn: 121609
 2010-12-11 10:49:22 +00:00
Benjamin Kramer 94a622af4c The srem -> urem transform is not safe for any divisor that's not a power of two. 
E.g. -5 % 5 is 0 with srem and 1 with urem.

Also addresses Frits van Bommel's comments.

llvm-svn: 120049
 2010-11-23 20:33:57 +00:00
Benjamin Kramer b5afa65b0a InstCombine: Reduce "X shift (A srem B)" to "X shift (A urem B)" iff B is positive. 
This allows to transform the rem in "1 << ((int)x % 8);" to an and.

llvm-svn: 120028
 2010-11-23 18:52:42 +00:00
Benjamin Kramer f1ebb63161 InstCombine: Implement X - A*-B -> X + A*B. 
llvm-svn: 119984
 2010-11-22 20:31:27 +00:00
Benjamin Kramer 24656c9583 Implement the "if (X == 6 || X == 4)" -> "if ((X|2) == 6)" optimization. 
This currently only catches the most basic case, a two-case switch, but can be
extended later.

llvm-svn: 119964
 2010-11-22 09:45:38 +00:00