Remove table64 lowering #22452
Remove table64 lowering #22452
Conversation
sbc100
force-pushed
the
use_table64
branch
6 times, most recently
from
1400cf6 to
7375406
Compare
| @@ -772,6 +784,7 @@ jobs: | |||
| executor: bionic | |||
| environment: | |||
| EMTEST_SKIP_NODE_CANARY: "1" | |||
| EMTEST_SKIP_WASM64: "1" | |||
There was a problem hiding this comment.
is it necessary to have both this and the node canary update?
There was a problem hiding this comment.
This builder does not have node canary installed. So it can no longer run wasm64 tests after this change. We could install node canary, but general we only configure one version of node per bot.
| @@ -82,6 +83,12 @@ class Feature(IntEnum): | |||
| 'safari': 140000, | |||
| 'node': 150000, | |||
| }, | |||
| Feature.MEMORY64: { | |||
| 'chrome': 128, | |||
There was a problem hiding this comment.
Is it premature to add this? we don't know which versions will ship this unflagged.
There was a problem hiding this comment.
Anyone targeting memory64 is already seeing an experimental warning, so that we are really targetting here is some arbitrary interim version of memory64. Once we remove the experimental warning we will pin the final values here I imagine?
|
|
||
| if settings.MEMORY64: | ||
| passes += ['--table64-lowering'] | ||
| passes += ['--memory64-lowering', '--table64-lowering'] |
There was a problem hiding this comment.
should --table64-lowering go away? I guess we can just roll the passes together into memory64-lowering now?
There was a problem hiding this comment.
Yes, eventually, but I guess that can/should be a followup.
sbc100
force-pushed
the
use_table64
branch
3 times, most recently
from
a2a73ea to
7ac7fc9
Compare
|
Do we only see table64 lowering as useful for testing? Not for users that want to only target wasm64 but are waiting for full browser adoption? |
Are there any such targets? Now that v8 and spidermonkey both adopted table64 I don't see it as being very useful. Am I missing something? I guess if you were not able to update to the latest browser you might be effected.. but in that case you can't really expect to target experimental features can you? |
WebKit/Safari still lacks support, I think? So for people that wanted to use this in production it could be a problem. But I honestly don't know if there are such people (that are ok with a 32-bit lowering of 64-bit code). |
|
Don't confuse table64 lowering with full memory64 lowering (as I did when I discussed this with Sam earlier 😅) IMO keeping full memory64 lowering around for a while longer is useful and I think that is what you're saying. But table64 lowering is just for the interim stage when Chrome and Firefox supported the rest of memory64 but not table64. Presumably Safari and others won't have such an interim state, they will just implement the whole proposal at once. |
|
Thanks, I was confused about exactly that. lgtm |
sbc100
force-pushed
the
use_table64
branch
2 times, most recently
from
84d9a0c to
ee8e9f0
Compare
This requires an update the version of node canary that use for testing.
This requires an update the version of node canary that use for testing.