Difference between revisions of "X86 Todo List"
From gem5
(Created page with '= Highest priority = * Flesh out and debug 64-bit modern ISA (what's needed by users) * Complete x87 support * AVX support * Performance correlation ** With real hardware and/or …') |
|||
| (2 intermediate revisions by 2 users not shown) | |||
| Line 1: | Line 1: | ||
= Highest priority = | = Highest priority = | ||
* Flesh out and debug 64-bit modern ISA (what's needed by users) | * Flesh out and debug 64-bit modern ISA (what's needed by users) | ||
| − | * | + | * FS-mode core timing issues |
| − | * | + | ** Debug TimingSimpleCPU issues? |
| + | ** In-order pipeline core model support | ||
| + | ** Out-of-order core model (O3) support | ||
| + | * Multiprocessor timing support: need to enforce atomicity of locked load/op/store sequences in timing cache models | ||
| + | ** Ruby and M5 classic? | ||
* Performance correlation | * Performance correlation | ||
** With real hardware and/or existing correlated simulator | ** With real hardware and/or existing correlated simulator | ||
** Micro-op counts for functional implementation | ** Micro-op counts for functional implementation | ||
** Timing for out-of-order core (requires O3 support) | ** Timing for out-of-order core (requires O3 support) | ||
| − | * | + | * Complete x87 support |
| − | * | + | * AVX support |
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
= Useful but not strictly necessary = | = Useful but not strictly necessary = | ||
| Line 26: | Line 25: | ||
= Could be done but might never happen = | = Could be done but might never happen = | ||
* other OS support (OSX, Windows, ??) | * other OS support (OSX, Windows, ??) | ||
| − | * | + | * complete real mode support |
| − | |||
| − | |||
Latest revision as of 01:49, 4 May 2010
Contents
Highest priority
- Flesh out and debug 64-bit modern ISA (what's needed by users)
- FS-mode core timing issues
- Debug TimingSimpleCPU issues?
- In-order pipeline core model support
- Out-of-order core model (O3) support
- Multiprocessor timing support: need to enforce atomicity of locked load/op/store sequences in timing cache models
- Ruby and M5 classic?
- Performance correlation
- With real hardware and/or existing correlated simulator
- Micro-op counts for functional implementation
- Timing for out-of-order core (requires O3 support)
- Complete x87 support
- AVX support
Useful but not strictly necessary
- Split up ISA output for faster compiling
- Improve ISA description language support for x86
To do eventually but not right away
- ACPI support
- KVM-based fast functional CPU model
- Virtualization extensions support (AMD SVM, etc.)
Could be done but might never happen
- other OS support (OSX, Windows, ??)
- complete real mode support
