The long-promised update to the ARM code-generator is now in Git master. This builds on the old version by adding the register allocation strategy borrowed from the X86 version as well as low-level peephole optimisation. On my Apple M1 processor it now seems to be faster than X86 code with Rosetta.
There are still some improvements to be made, particularly with floating point, but no major changes are anticipated. It includes one optimisation that isn't present on the X86 at the moment: small tuples are returned in registers rather than on the stack.
Please try it and let me know how it goes.
David
I?m getting random SIGSEGVs while building HOL with latest git on an Apple Silicon MacBook Pro. These don?t always occur, but when they do, they have (so far) always been in one of two places.
Michael
On 16 Jan 2022, at 10:51 pm, David Matthews <David.Matthews at prolingua.co.uk> wrote:
The long-promised update to the ARM code-generator is now in Git master. This builds on the old version by adding the register allocation strategy borrowed from the X86 version as well as low-level peephole optimisation. On my Apple M1 processor it now seems to be faster than X86 code with Rosetta.
There are still some improvements to be made, particularly with floating point, but no major changes are anticipated. It includes one optimisation that isn't present on the X86 at the moment: small tuples are returned in registers rather than on the stack.
Please try it and let me know how it goes.
David
polyml mailing list polyml at inf.ed.ac.uk https://aus01.safelinks.protection.outlook.com/?url=http%3A%2F%2Flists.inf.e...
I've pushed some more updates to git master. These are mostly improvements to the code, particularly floating point, but there are some fixes to bugs that could have produced bus errors or segmentation faults.
David
On 17/01/2022 23:44, Michael Norrish via polyml wrote:
I?m getting random SIGSEGVs while building HOL with latest git on an Apple Silicon MacBook Pro. These don?t always occur, but when they do, they have (so far) always been in one of two places.
Michael
On 16/01/2022 12:51, David Matthews wrote:
The long-promised update to the ARM code-generator is now in Git master. ?This builds on the old version by adding the register allocation strategy borrowed from the X86 version as well as low-level peephole optimisation.? On my Apple M1 processor it now seems to be faster than X86 code with Rosetta.
There are still some improvements to be made, particularly with floating point, but no major changes are anticipated.? It includes one optimisation that isn't present on the X86 at the moment: small tuples are returned in registers rather than on the stack.
This is really great.
I have now started to make some tests with Isabelle + AFP, and will come back eventually with reports (and potential problems).
Makarius
-
David.Matthews＠prolingua.co.uk -
makarius＠sketis.net -
michael.norrish＠anu.edu.au