| Misha Brukman | 1be339a | 2004-07-23 22:37:22 +0000 | [diff] [blame] | 1 | Currently unimplemented: |
| 2 | * cast fp to bool |
| Misha Brukman | bb966a4 | 2004-07-26 16:23:55 +0000 | [diff] [blame] | 3 | * signed right shift of long by reg |
| Misha Brukman | 1be339a | 2004-07-23 22:37:22 +0000 | [diff] [blame] | 4 | |
| 5 | Current bugs: |
| Misha Brukman | 8b88d90 | 2004-07-28 19:16:10 +0000 | [diff] [blame] | 6 | * ulong to double. ahhh, here's the problem: |
| 7 | floatdidf assumes signed longs. so if the high but of a ulong |
| 8 | just happens to be set, you get the wrong sign. The fix for this |
| 9 | is to call cmpdi2 to compare against zero, if so shift right by one, |
| 10 | convert to fp, and multiply by (add to itself). the sequence would |
| 11 | look like: |
| 12 | {r3:r4} holds ulong a; |
| 13 | li r5, 0 |
| 14 | li r6, 0 (set r5:r6 to ulong 0) |
| 15 | call cmpdi2 ==> sets r3 <, =, > 0 |
| 16 | if r3 > 0 |
| 17 | call floatdidf as usual |
| 18 | else |
| 19 | shift right ulong a, 1 (we could use emitShift) |
| 20 | call floatdidf |
| 21 | fadd f1, f1, f1 (fp left shift by 1) |
| 22 | * linking llvmg++ .s files with gcc instead of g++ |
| Misha Brukman | bb966a4 | 2004-07-26 16:23:55 +0000 | [diff] [blame] | 23 | |
| 24 | Codegen improvements needed: |
| Chris Lattner | 3ea9346 | 2004-08-06 06:58:50 +0000 | [diff] [blame] | 25 | * PowerPCPEI.cpp needs to save/restore regs in the opposite order |
| Misha Brukman | bb966a4 | 2004-07-26 16:23:55 +0000 | [diff] [blame] | 26 | * setCondInst needs to know branchless versions of seteq/setne/etc |
| 27 | * cast elimination pass (uint -> sbyte -> short, kill the byte -> short) |
| Misha Brukman | 4ce5ce2 | 2004-07-27 18:43:04 +0000 | [diff] [blame] | 28 | * should hint to the branch select pass that it doesn't need to print the |
| 29 | second unconditional branch, so we don't end up with things like: |
| Misha Brukman | 4ce5ce2 | 2004-07-27 18:43:04 +0000 | [diff] [blame] | 30 | b .LBBl42__2E_expand_function_8_674 ; loopentry.24 |
| 31 | b .LBBl42__2E_expand_function_8_42 ; NewDefault |
| 32 | b .LBBl42__2E_expand_function_8_42 ; NewDefault |
| Misha Brukman | bb966a4 | 2004-07-26 16:23:55 +0000 | [diff] [blame] | 33 | |
| 34 | Current hacks: |
| 35 | * lazy insert of GlobalBaseReg definition at front of first MBB |
| 36 | A prime candidate for sabre's "slightly above ISel" passes. |
| 37 | * cast code is huge, unwieldy. Should probably be broken up into |
| 38 | smaller pieces. |
| 39 | * visitLoadInst is getting awfully cluttered as well. |
| Misha Brukman | 1be339a | 2004-07-23 22:37:22 +0000 | [diff] [blame] | 40 | |
| 41 | Currently failing tests: |
| Misha Brukman | 1be339a | 2004-07-23 22:37:22 +0000 | [diff] [blame] | 42 | * SingleSource |
| Misha Brukman | b9b8ba5 | 2004-07-28 00:55:12 +0000 | [diff] [blame] | 43 | `- Regression |
| Misha Brukman | 88f8f92 | 2004-07-30 15:53:09 +0000 | [diff] [blame] | 44 | | `- casts (ulong to fp failure) |
| Misha Brukman | 1be339a | 2004-07-23 22:37:22 +0000 | [diff] [blame] | 45 | `- Benchmarks |
| 46 | | `- Shootout-C++ : most programs fail, miscompilations |
| 47 | `- UnitTests |
| Misha Brukman | 88f8f92 | 2004-07-30 15:53:09 +0000 | [diff] [blame] | 48 | | `- C++Catch |
| 49 | | `- SimpleC++Test |
| 50 | | `- ConditionalExpr (also C++) |
| Misha Brukman | 1be339a | 2004-07-23 22:37:22 +0000 | [diff] [blame] | 51 | * MultiSource |
| 52 | |- Applications |
| 53 | | `- burg: miscompilation |
| 54 | | `- siod: llc bus error |
| 55 | | `- hbd: miscompilation |
| 56 | | `- d (make_dparser): miscompilation |
| 57 | `- Benchmarks |
| Misha Brukman | 88f8f92 | 2004-07-30 15:53:09 +0000 | [diff] [blame] | 58 | `- MallocBench/make: miscompilation |