Building a jitless V8 for unsupported architectures
32 views
Skip to first unread message
John Paul Adrian Glaubitz
Dec 26, 2022, 7:45:06 AM12/26/22
to v8-dev
Hello!
I was wondering whether it would be possible to build a jitless V8 on unsupported architectures such as sparc64. I have seen that V8 generally supports a jitless interpretor-only mode [1] but it's not clear to me whether it would also support unsupported architectures, i.e. those for which exists no native codegen backend.
Background is that the nodejs in Debian is not available on a number of less common architectures such as alpha, hppa, powerpc32 and sparc64 etc [2] despite machines being fast enough to run an interpreter.
Does anyone know whether building a jitless V8 on an unsupported architecture would generally work or does the jitless code still require the codegen backends?
Thanks,
Adrian
Jakob Gruber
Dec 27, 2022, 2:18:32 AM12/27/22
to v8-...@googlegroups.com
On Mon, Dec 26, 2022 at 1:45 PM 'John Paul Adrian Glaubitz' via v8-dev <v8-...@googlegroups.com> wrote:
Hello!I was wondering whether it would be possible to build a jitless V8 on unsupported architectures such as sparc64. I have seen that V8 generally supports a jitless interpretor-only mode [1] but it's not clear to me whether it would also support unsupported architectures, i.e. those for which exists no native codegen backend.Background is that the nodejs in Debian is not available on a number of less common architectures such as alpha, hppa, powerpc32 and sparc64 etc [2] despite machines being fast enough to run an interpreter.Does anyone know whether building a jitless V8 on an unsupported architecture would generally work or does the jitless code still require the codegen backends?
Hi Adrian,
unfortunately yes, --jitless still requires codegen at V8-compile-time since it relies heavily on builtins and bytecode handlers, both of which are (mostly) generated by the CodeStubAssembler.
Thanks,Adrian
--
--
v8-dev mailing list
v8-...@googlegroups.com
http://groups.google.com/group/v8-dev
---
You received this message because you are subscribed to the Google Groups "v8-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to v8-dev+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/v8-dev/3b6ed0f0-0991-4395-9cf1-fa915140bd5bn%40googlegroups.com.
John Paul Adrian Glaubitz
Dec 27, 2022, 7:37:14 AM12/27/22
to v8-...@googlegroups.com
Hi Jakob!
On 12/27/22 08:18, Jakob Gruber wrote:
> unfortunately yes, --jitless still requires codegen at V8-compile-time since it relies
> heavily on builtins and bytecode handlers, both of which are (mostly) generated by the
> CodeStubAssembler.
That's a pity. Maybe that could change in the future with libgccjit which could be used
to generate bytecode for more architectures. The Rust compiler has now the capability to
use libgccjit as a codegen backend, maybe V8 can follow in the future ;-).
In the mean time, I'm trying to get V8 to build on older PowerPC targets. I have started
with my first contribution just yesterday [1] and will send more patches once I got the
hang of it ;-).
Adrian
> [1] https://chromium-review.googlesource.com/c/v8/v8/+/4124995
On 12/27/22 08:18, Jakob Gruber wrote:
> unfortunately yes, --jitless still requires codegen at V8-compile-time since it relies
> heavily on builtins and bytecode handlers, both of which are (mostly) generated by the
> CodeStubAssembler.
to generate bytecode for more architectures. The Rust compiler has now the capability to
use libgccjit as a codegen backend, maybe V8 can follow in the future ;-).
In the mean time, I'm trying to get V8 to build on older PowerPC targets. I have started
with my first contribution just yesterday [1] and will send more patches once I got the
hang of it ;-).
Adrian
> [1] https://chromium-review.googlesource.com/c/v8/v8/+/4124995
Reply all
Reply to author
Forward
0 new messages