Cranelift: provide Gpr and Xmm newtype wrappers around Reg and use them extensively in ISLE lowering
#3685
Labels
cranelift
Issues related to the Cranelift code generator
Comments
fitzgen
added a commit
to fitzgen/wasmtime
that referenced
this issue
Feb 1, 2022
This primary motivation of this large commit (apologies for its size!) is to
introduce `Gpr` and `Xmm` newtypes over `Reg`. This should help catch
difficult-to-diagnose register class mixup bugs in x64 lowerings.
But having a newtype for `Gpr` and `Xmm` themselves isn't enough to catch all of
our operand-with-wrong-register-class bugs, because about 50% of operands on x64
aren't just a register, but a register or memory address or even an
immediate! So we have `{Gpr,Xmm}Mem[Imm]` newtypes as well.
Unfortunately, `GprMem` et al can't be `enum`s and are therefore a little bit
noisier to work with from ISLE. They need to maintain the invariant that their
registers really are of the claimed register class, so they need to encapsulate
the inner data. If they exposed the underlying `enum` variants, then anyone
could just change register classes or construct a `GprMem` that holds an XMM
register, defeating the whole point of these newtypes. So when working with
these newtypes from ISLE, we rely on external constructors like `(gpr_to_gpr_mem
my_gpr)` instead of `(GprMem.Gpr my_gpr)`.
A bit of extra lines of code are included to add support for register mapping
for all of these newtypes as well. Ultimately this is all a bit wordier than I'd
hoped it would be when I first started authoring this commit, but I think it is
all worth it nonetheless!
In the process of adding these newtypes, I didn't want to have to update both
the ISLE `extern` type definition of `MInst` and the Rust definition, so I move
the definition fully into ISLE, similar as aarch64.
Finally, this process isn't complete. I've introduced the newtypes here, and
I've made most XMM-using instructions switch from `Reg` to `Xmm`, as well as
register class-converting instructions, but I haven't moved all of the GPR-using
instructions over to the newtypes yet. I figured this commit was big enough as
it was, and I can continue the adoption of these newtypes in follow up commits.
Part of bytecodealliance#3685.
fitzgen
added a commit
to fitzgen/wasmtime
that referenced
this issue
Feb 3, 2022
This primary motivation of this large commit (apologies for its size!) is to
introduce `Gpr` and `Xmm` newtypes over `Reg`. This should help catch
difficult-to-diagnose register class mixup bugs in x64 lowerings.
But having a newtype for `Gpr` and `Xmm` themselves isn't enough to catch all of
our operand-with-wrong-register-class bugs, because about 50% of operands on x64
aren't just a register, but a register or memory address or even an
immediate! So we have `{Gpr,Xmm}Mem[Imm]` newtypes as well.
Unfortunately, `GprMem` et al can't be `enum`s and are therefore a little bit
noisier to work with from ISLE. They need to maintain the invariant that their
registers really are of the claimed register class, so they need to encapsulate
the inner data. If they exposed the underlying `enum` variants, then anyone
could just change register classes or construct a `GprMem` that holds an XMM
register, defeating the whole point of these newtypes. So when working with
these newtypes from ISLE, we rely on external constructors like `(gpr_to_gpr_mem
my_gpr)` instead of `(GprMem.Gpr my_gpr)`.
A bit of extra lines of code are included to add support for register mapping
for all of these newtypes as well. Ultimately this is all a bit wordier than I'd
hoped it would be when I first started authoring this commit, but I think it is
all worth it nonetheless!
In the process of adding these newtypes, I didn't want to have to update both
the ISLE `extern` type definition of `MInst` and the Rust definition, so I move
the definition fully into ISLE, similar as aarch64.
Finally, this process isn't complete. I've introduced the newtypes here, and
I've made most XMM-using instructions switch from `Reg` to `Xmm`, as well as
register class-converting instructions, but I haven't moved all of the GPR-using
instructions over to the newtypes yet. I figured this commit was big enough as
it was, and I can continue the adoption of these newtypes in follow up commits.
Part of bytecodealliance#3685.
fitzgen
added a commit
to fitzgen/wasmtime
that referenced
this issue
Feb 3, 2022
This primary motivation of this large commit (apologies for its size!) is to
introduce `Gpr` and `Xmm` newtypes over `Reg`. This should help catch
difficult-to-diagnose register class mixup bugs in x64 lowerings.
But having a newtype for `Gpr` and `Xmm` themselves isn't enough to catch all of
our operand-with-wrong-register-class bugs, because about 50% of operands on x64
aren't just a register, but a register or memory address or even an
immediate! So we have `{Gpr,Xmm}Mem[Imm]` newtypes as well.
Unfortunately, `GprMem` et al can't be `enum`s and are therefore a little bit
noisier to work with from ISLE. They need to maintain the invariant that their
registers really are of the claimed register class, so they need to encapsulate
the inner data. If they exposed the underlying `enum` variants, then anyone
could just change register classes or construct a `GprMem` that holds an XMM
register, defeating the whole point of these newtypes. So when working with
these newtypes from ISLE, we rely on external constructors like `(gpr_to_gpr_mem
my_gpr)` instead of `(GprMem.Gpr my_gpr)`.
A bit of extra lines of code are included to add support for register mapping
for all of these newtypes as well. Ultimately this is all a bit wordier than I'd
hoped it would be when I first started authoring this commit, but I think it is
all worth it nonetheless!
In the process of adding these newtypes, I didn't want to have to update both
the ISLE `extern` type definition of `MInst` and the Rust definition, so I move
the definition fully into ISLE, similar as aarch64.
Finally, this process isn't complete. I've introduced the newtypes here, and
I've made most XMM-using instructions switch from `Reg` to `Xmm`, as well as
register class-converting instructions, but I haven't moved all of the GPR-using
instructions over to the newtypes yet. I figured this commit was big enough as
it was, and I can continue the adoption of these newtypes in follow up commits.
Part of bytecodealliance#3685.
fitzgen
added a commit
to fitzgen/wasmtime
that referenced
this issue
Feb 3, 2022
This primary motivation of this large commit (apologies for its size!) is to
introduce `Gpr` and `Xmm` newtypes over `Reg`. This should help catch
difficult-to-diagnose register class mixup bugs in x64 lowerings.
But having a newtype for `Gpr` and `Xmm` themselves isn't enough to catch all of
our operand-with-wrong-register-class bugs, because about 50% of operands on x64
aren't just a register, but a register or memory address or even an
immediate! So we have `{Gpr,Xmm}Mem[Imm]` newtypes as well.
Unfortunately, `GprMem` et al can't be `enum`s and are therefore a little bit
noisier to work with from ISLE. They need to maintain the invariant that their
registers really are of the claimed register class, so they need to encapsulate
the inner data. If they exposed the underlying `enum` variants, then anyone
could just change register classes or construct a `GprMem` that holds an XMM
register, defeating the whole point of these newtypes. So when working with
these newtypes from ISLE, we rely on external constructors like `(gpr_to_gpr_mem
my_gpr)` instead of `(GprMem.Gpr my_gpr)`.
A bit of extra lines of code are included to add support for register mapping
for all of these newtypes as well. Ultimately this is all a bit wordier than I'd
hoped it would be when I first started authoring this commit, but I think it is
all worth it nonetheless!
In the process of adding these newtypes, I didn't want to have to update both
the ISLE `extern` type definition of `MInst` and the Rust definition, so I move
the definition fully into ISLE, similar as aarch64.
Finally, this process isn't complete. I've introduced the newtypes here, and
I've made most XMM-using instructions switch from `Reg` to `Xmm`, as well as
register class-converting instructions, but I haven't moved all of the GPR-using
instructions over to the newtypes yet. I figured this commit was big enough as
it was, and I can continue the adoption of these newtypes in follow up commits.
Part of bytecodealliance#3685.
|
I believe this was implemented in #3752, so closing. |
|
This isn't complete yet, #3752 was just the first half. I'll have a PR up for the other half sometime soon ish. |
fitzgen
added a commit
to fitzgen/wasmtime
that referenced
this issue
Feb 14, 2022
We already defined the `Gpr` newtype and used it in a few places, and we already defined the `Xmm` newtype and used it extensively. This finishes the transition to using the newtypes extensively in lowering by making use of `Gpr` in more places. Fixes bytecodealliance#3685
fitzgen
added a commit
that referenced
this issue
Feb 14, 2022
We already defined the `Gpr` newtype and used it in a few places, and we already defined the `Xmm` newtype and used it extensively. This finishes the transition to using the newtypes extensively in lowering by making use of `Gpr` in more places. Fixes #3685
mpardesh
pushed a commit
to avanhatt/wasmtime
that referenced
this issue
Mar 17, 2022
This primary motivation of this large commit (apologies for its size!) is to
introduce `Gpr` and `Xmm` newtypes over `Reg`. This should help catch
difficult-to-diagnose register class mixup bugs in x64 lowerings.
But having a newtype for `Gpr` and `Xmm` themselves isn't enough to catch all of
our operand-with-wrong-register-class bugs, because about 50% of operands on x64
aren't just a register, but a register or memory address or even an
immediate! So we have `{Gpr,Xmm}Mem[Imm]` newtypes as well.
Unfortunately, `GprMem` et al can't be `enum`s and are therefore a little bit
noisier to work with from ISLE. They need to maintain the invariant that their
registers really are of the claimed register class, so they need to encapsulate
the inner data. If they exposed the underlying `enum` variants, then anyone
could just change register classes or construct a `GprMem` that holds an XMM
register, defeating the whole point of these newtypes. So when working with
these newtypes from ISLE, we rely on external constructors like `(gpr_to_gpr_mem
my_gpr)` instead of `(GprMem.Gpr my_gpr)`.
A bit of extra lines of code are included to add support for register mapping
for all of these newtypes as well. Ultimately this is all a bit wordier than I'd
hoped it would be when I first started authoring this commit, but I think it is
all worth it nonetheless!
In the process of adding these newtypes, I didn't want to have to update both
the ISLE `extern` type definition of `MInst` and the Rust definition, so I move
the definition fully into ISLE, similar as aarch64.
Finally, this process isn't complete. I've introduced the newtypes here, and
I've made most XMM-using instructions switch from `Reg` to `Xmm`, as well as
register class-converting instructions, but I haven't moved all of the GPR-using
instructions over to the newtypes yet. I figured this commit was big enough as
it was, and I can continue the adoption of these newtypes in follow up commits.
Part of bytecodealliance#3685.
mpardesh
pushed a commit
to avanhatt/wasmtime
that referenced
this issue
Mar 17, 2022
…ance#3798) We already defined the `Gpr` newtype and used it in a few places, and we already defined the `Xmm` newtype and used it extensively. This finishes the transition to using the newtypes extensively in lowering by making use of `Gpr` in more places. Fixes bytecodealliance#3685
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
We've talked a bit
I think adding newtype wrappers around
Regfor general purpose registers vs XMM registers, etc... would help both these things a lot. Basically every register class should have its own newtype wrapper.I'm spending too much time debugging dynamic errors where I'm accidentally implicitly moving between register classes (either via mov mitosis or otherwise) and it would be way easier to fix this kind of thing if it was a compile time type error that specified exactly where in the sources I'm doing the wrong thing.
The text was updated successfully, but these errors were encountered: