bjorn3 commented on issue #4299:
The
bint
instruction returns 0 or 1. Thebmask
instruction returns all zeros or all ones. #3205 is about the actual representation in registers and memory, not about the result ofbint
/bmask
, right?
afonso360 commented on issue #4299:
I didn't realize
bint
was also a vector instruction. I'm up for clarifying the docs and implementing it for vectors or just restricting vectors from this instruction.https://github.com/bytecodealliance/wasmtime/issues/3205 is about the actual representation in registers and memory, not about the result of bint/bmask, right?
That is my understanding of it as well.
Also, as a side note, could this PR close https://github.com/bytecodealliance/wasmtime/issues/2237? Or is that issue already effectively done?
No, that issue I think is mostly related to the trampoline generating the wrong code when loading / storing booleans, and its why I had to do duplicate some code in this PR in order to not have boolean arguments to the test cases.
afonso360 edited a comment on issue #4299:
I didn't realize
bint
was also a vector instruction. I'm up for clarifying the docs and implementing it for vectors or just restricting vectors from this instruction. Let me know what you would like me to do.https://github.com/bytecodealliance/wasmtime/issues/3205 is about the actual representation in registers and memory, not about the result of bint/bmask, right?
That is my understanding of it as well.
Also, as a side note, could this PR close https://github.com/bytecodealliance/wasmtime/issues/2237? Or is that issue already effectively done?
No, that issue I think is mostly related to the trampoline generating the wrong code when loading / storing booleans, and its why I had to do duplicate some code in this PR in order to not have boolean arguments to the test cases.
afonso360 edited a comment on issue #4299:
I didn't realize
bint
was also a vector instruction. I'm up for implementing it for vectors or restricting vectors from this instruction. Let me know what you would like me to do.https://github.com/bytecodealliance/wasmtime/issues/3205 is about the actual representation in registers and memory, not about the result of bint/bmask, right?
That is my understanding of it as well.
Also, as a side note, could this PR close https://github.com/bytecodealliance/wasmtime/issues/2237? Or is that issue already effectively done?
No, that issue I think is mostly related to the trampoline generating the wrong code when loading / storing booleans, and its why I had to do duplicate some code in this PR in order to not have boolean arguments to the test cases.
abrown commented on issue #4299:
I didn't realize
bint
was also a vector instruction. I'm up for implementing it for vectors or restricting vectors from this instruction. Let me know what you would like me to do.I think restricting vectors from this instruction is the easiest path for now.
#3205 is about the actual representation in registers and memory, not about the result of bint/bmask, right?
That is my understanding of it as well.
Yes, primarily it is about representations, but @cfallin mentions
bint
in the description and I bring up an all ones/all zeroes comment ater. The main reason for me mentioning it here is that I want GitHub to keep track of all the cross-links between related issues (and semi-related issues, in this case) so that it is easier to investigate later.Also, as a side note, could this PR close #2237? Or is that issue already effectively done?
No, that issue I think is mostly related to the trampoline generating the wrong code when loading / storing booleans, and its why I had to do duplicate some code in this PR in order to not have boolean arguments to the test cases.
:+1:
abrown edited a comment on issue #4299:
I didn't realize
bint
was also a vector instruction. I'm up for implementing it for vectors or restricting vectors from this instruction. Let me know what you would like me to do.I think restricting vectors from this instruction is the easiest path for now.
#3205 is about the actual representation in registers and memory, not about the result of bint/bmask, right?
That is my understanding of it as well.
Yes, primarily it is about representations, but @cfallin mentions
bint
in the description and I bring up an all ones/all zeroes comment later. The main reason for me mentioning it here is that I want GitHub to keep track of all the cross-links between related issues (and semi-related issues, in this case) so that it is easier to investigate later.Also, as a side note, could this PR close #2237? Or is that issue already effectively done?
No, that issue I think is mostly related to the trampoline generating the wrong code when loading / storing booleans, and its why I had to do duplicate some code in this PR in order to not have boolean arguments to the test cases.
:+1:
Last updated: Dec 23 2024 at 13:07 UTC