| sewardj | 034dc33 | 2004-11-15 13:30:26 +0000 | [diff] [blame] | 1 | |
| 2 | Last updated 15 Nov 04 |
| 3 | ~~~~~~~~~~~~~~~~~~~~~~ |
| 4 | |
| 5 | Critical (correctness) |
| 6 | ~~~~~~~~~~~~~~~~~~~~~~ |
| sewardj | 034dc33 | 2004-11-15 13:30:26 +0000 | [diff] [blame] | 7 | x86 isel: should free up all fp reg tags when calling a helper. |
| sewardj | 791446e | 2005-03-10 15:03:52 +0000 | [diff] [blame] | 8 | And save FP and SSE insns across the helper. |
| sewardj | 034dc33 | 2004-11-15 13:30:26 +0000 | [diff] [blame] | 9 | |
| 10 | iropt: reconsider precise exceptions |
| 11 | |
| sewardj | 14d3ddf | 2004-11-16 02:07:18 +0000 | [diff] [blame] | 12 | x86 guest: look at FP accuracy |
| 13 | |
| sewardj | 034dc33 | 2004-11-15 13:30:26 +0000 | [diff] [blame] | 14 | |
| 15 | Test |
| 16 | ~~~~ |
| 17 | Test adc/sbb carry dependency in memcheck |
| 18 | |
| 19 | |
| 20 | Optimisation opportunities |
| 21 | ~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 22 | Improved isel for memcheck artefacts on x86 (generate neg ; sbbl) |
| 23 | |
| 24 | Assess tt_fast miss rates |
| 25 | |
| 26 | Better register allocation - spilling reduction |
| 27 | |
| 28 | iropt less conservative re precise exns |
| 29 | |
| 30 | improve stack-update pass |
| 31 | |
| 32 | proper profiling machinery |
| 33 | |
| 34 | do not CSE exprs :: Ity_Bit |
| 35 | |
| sewardj | 791446e | 2005-03-10 15:03:52 +0000 | [diff] [blame] | 36 | x86 iselIntExpr_RMI: actually generate the M case if possible |
| sewardj | 034dc33 | 2004-11-15 13:30:26 +0000 | [diff] [blame] | 37 | |
| 38 | JIT speedups |
| 39 | ~~~~~~~~~~~~ |
| 40 | Ensure incremental flatness throughout |
| 41 | |
| sewardj | bc4853b | 2005-03-10 13:54:15 +0000 | [diff] [blame] | 42 | Profile again with cachegrind/calltree |
| sewardj | 034dc33 | 2004-11-15 13:30:26 +0000 | [diff] [blame] | 43 | |
| 44 | change IRTemp to 16 bits? |
| 45 | |
| 46 | |
| 47 | Integration |
| 48 | ~~~~~~~~~~~ |
| 49 | Get rid of sloppy-malloc |
| 50 | |
| 51 | Get rid of partial-loads-ok |
| 52 | |
| 53 | Optimisation after first instrumentation rather than 2nd ? |
| 54 | |
| 55 | disallow dirty helpers from writing SP/IP |
| 56 | |
| 57 | make instrumentation work at no optimisation |
| 58 | |
| 59 | write API doc, clarify IR semantics |
| 60 | |
| 61 | make IR utils module |
| 62 | |
| 63 | tt/tc simplification, also profiling |
| 64 | |
| 65 | generic stack pointer identification at startup? |
| 66 | |
| 67 | New memstack_k: old or new sp? |
| 68 | |
| 69 | Set host FPU mode before starting. Also set Dflag = 0 |
| 70 | |
| 71 | skin_panic -> tool_panic |
| 72 | |