1//===- README.txt - Notes for improving CellSPU-specific code gen ---------===//
2
3This code was contributed by a team from the Computer Systems Research
4Department in The Aerospace Corporation:
5
6- Scott Michel (head bottle washer and much of the non-floating point
7  instructions)
8- Mark Thomas (floating point instructions)
9- Michael AuYeung (intrinsics)
10- Chandler Carruth (LLVM expertise)
11- Nehal Desai (debugging, i32 operations, RoadRunner SPU expertise)
12
13Some minor fixes added by Kalle Raiskila.
14
15THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
17MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, NONINFRINGEMENT, OR
18OTHERWISE.  IN NO EVENT SHALL THE AEROSPACE CORPORATION BE LIABLE FOR DAMAGES
19OF ANY KIND OR NATURE WHETHER BASED IN CONTRACT, TORT, OR OTHERWISE ARISING
20OUT OF OR IN CONNECTION WITH THE USE OF THE SOFTWARE INCLUDING, WITHOUT
21LIMITATION, DAMAGES RESULTING FROM LOST OR CONTAMINATED DATA, LOST PROFITS OR
22REVENUE, COMPUTER MALFUNCTION, OR FOR ANY SPECIAL, INCIDENTAL, CONSEQUENTIAL,
23OR PUNITIVE  DAMAGES, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGES OR
24SUCH DAMAGES ARE FORESEEABLE.
25
26---------------------------------------------------------------------------
27--WARNING--:
28--WARNING--: The CellSPU work is work-in-progress and "alpha" quality code.
29--WARNING--:
30
31If you are brave enough to try this code or help to hack on it, be sure
32to add 'spu' to configure's --enable-targets option, e.g.:
33
34        ./configure <your_configure_flags_here> \
35           --enable-targets=x86,x86_64,powerpc,spu
36
37---------------------------------------------------------------------------
38
39TODO:
40* In commit r142152 vector legalization was set to element promotion per
41  default. This breaks half vectors (e.g. v2i32) badly as they get element
42  promoted to much slower types (v2i64).
43
44* Many CellSPU specific codegen tests only grep & count the number of 
45  instructions, not checking their place with FileCheck. There have also
46  been some commits that change the CellSPU checks, some of which might
47  have not been thoroughly scrutinized w.r.t. to the changes they cause in SPU
48  assembly. (especially since about the time of r142152)  
49
50* Some of the i64 math have huge tablegen rules, which sometime cause
51  tablegen to run out of memory. See e.g. bug 8850. i64 arithmetics 
52  should probably be done with libraries.
53
54* Create a machine pass for performing dual-pipeline scheduling specifically
55  for CellSPU, and insert branch prediction instructions as needed.
56
57* i32 instructions:
58
59  * i32 division (work-in-progress)
60
61* i64 support (see i64operations.c test harness):
62
63  * shifts and comparison operators: done
64  * sign and zero extension: done
65  * addition: done
66  * subtraction: needed
67  * multiplication: done
68
69* i128 support:
70
71  * zero extension, any extension: done
72  * sign extension: done
73  * arithmetic operators (add, sub, mul, div): needed
74  * logical operations (and, or, shl, srl, sra, xor, nor, nand): needed
75
76    * or: done
77
78* f64 support
79
80  * Comparison operators:
81    SETOEQ              unimplemented
82    SETOGT              unimplemented
83    SETOGE              unimplemented
84    SETOLT              unimplemented
85    SETOLE              unimplemented
86    SETONE              unimplemented
87    SETO                done (lowered)
88    SETUO               done (lowered)
89    SETUEQ              unimplemented
90    SETUGT              unimplemented
91    SETUGE              unimplemented
92    SETULT              unimplemented
93    SETULE              unimplemented
94    SETUNE              unimplemented
95
96* LLVM vector suport
97
98  * VSETCC needs to be implemented. It's pretty straightforward to code, but
99    needs implementation.
100
101* Intrinsics
102
103  * spu.h instrinsics added but not tested. Need to have an operational
104    llvm-spu-gcc in order to write a unit test harness.
105
106===-------------------------------------------------------------------------===
107