The vast majority of the following (very common) opcodes were always
called with identical arguments:
- `GIM_CheckType` for the root
- `GIM_CheckRegBankForClass` for the root
- `GIR_Copy` between the old and new root
- `GIR_ConstrainSelectedInstOperands` on the new root
- `GIR_BuildMI` to create the new root
I added overloaded version of each opcode specialized for the root
instructions. It always saves between 1 and 2 bytes per instance
depending on the number of arguments specialized into the opcode. Some
of these opcodes had between 5 and 15k occurences in the AArch64
GlobalISel Match Table.
Additionally, the following opcodes are almost always used in the same
sequence:
- `GIR_EraseFromParent 0` + `GIR_Done`
- `GIR_EraseRootFromParent_Done` has been created to do both. Saves 2
bytes per occurence.
- `GIR_IsSafeToFold` was *always* called for each InsnID except 0.
- Changed the opcode to take the number of instructions to check after
`MI[0]`
The savings from these are pretty neat. For `AArch64GenGlobalISel.inc`:
- `AArch64InstructionSelector.cpp.o` goes down from 772kb to 704kb (-10%
code size)
- Self-reported MatchTable size goes from 420380 bytes to 352426 bytes
(~ -17%)
A smaller match table means a faster match table because we spend less
time iterating and decoding.
I don't have a solid measurement methodology for GlobalISel performance
so I don't have precise numbers but I saw a few % of improvements in a
simple testcase.
This tries to fix a bug by resolving a few FIXMEs. The bug is that
`EraseInstAction` is emitted after emitting the _first_ `BuildMIAction`,
which is too early because the erased instruction may still be used by
subsequent `BuildMIAction`s (in particular, by `CopyRenderer`).
An example of the bug (from `match-table-operand-types.td`):
```
def InstTest0 : GICombineRule<
(defs root:$a),
(match (G_MUL i32:$x, i32:$b, i32:$c),
(G_MUL $a, i32:$b, i32:$x)),
(apply (G_ADD i64:$tmp, $b, i32:$c),
(G_ADD i8:$a, $b, i64:$tmp))>;
GIR_EraseFromParent, /*InsnID*/0,
GIR_BuildMI, /*InsnID*/1, /*Opcode*/GIMT_Encode2(TargetOpcode::G_ADD),
GIR_Copy, /*NewInsnID*/1, /*OldInsnID*/0, /*OpIdx*/0, // a
GIR_Copy, /*NewInsnID*/1, /*OldInsnID*/0, /*OpIdx*/1, // b
GIR_AddSimpleTempRegister, /*InsnID*/1, /*TempRegID*/0,
```
Here, the root instruction is destroyed before copying its operands ('a'
and 'b') to the new instruction.
The solution is to emit `EraseInstAction` for the root instruction as
the last action in the emission pipeline.
When importing instruction selection patterns into GlobalISel, the
operands matched in the "source" DAG are copied into corresponding
operands of the "destination" DAG according to their names (such as Rd).
If multiple operands in the source DAG share the same name, a
GIM_CheckIsSameOperand predicate makes instruction selector check the
corresponding operands for equality (at compiler run-time) as part of
matching the source pattern.
The Def operands of the root node of the destination DAG are handled
specially. The operands of the instruction corresponding to the root
node are taken and GIM_CheckRegBankForClass predicates are
tablegen-erated accordingly. If by coincidence the Def operand in
question has the same name as one of the named operands in the pattern,
a GIM_CheckIsSameOperand predicate is automatically added that is likely
to prevent matching the source of otherwise applicable selection pattern
at compiler run-time.
This patch mangles the Def operand names taken from the instruction
corresponding to the root of the destination DAG (for example, "Rd"
becomes "DstI[Rd]") preventing unexpected name clashes with pattern's
named operands.
The patch consists of three sets of changes:
* changes to the GlobalISelEmitter.cpp file are the actual fix
* a test case is added to GlobalISelEmitter.td file as a regression test
* everything else is the biggest and least interesting part - updates to
the existing test cases: renames of the form Rd -> DstI[Rd] inside the
inline comments in tablegen-erated code
