Stream: git-wasmtime

Topic: wasmtime / PR #5904 Add filetest for unexpected `imm12_fr...

view this post on Zulip Wasmtime GitHub notifications bot (Mar 01 2023 at 20:12):

avanhatt opened PR #5904 from imm12neg to main:

Add a precise-output filetest for the unexpected behavior described in https://github.com/bytecodealliance/wasmtime/issues/5903 (immediates not being used for narrow widths).

view this post on Zulip Wasmtime GitHub notifications bot (Mar 01 2023 at 20:23):

jameysharp submitted PR review.

view this post on Zulip Wasmtime GitHub notifications bot (Mar 01 2023 at 20:23):

jameysharp has enabled auto merge for PR #5904.

view this post on Zulip Wasmtime GitHub notifications bot (Mar 01 2023 at 21:08):

jameysharp merged PR #5904.

view this post on Zulip Wasmtime GitHub notifications bot (Mar 01 2023 at 23:11):

bjorn3 submitted PR review.

view this post on Zulip Wasmtime GitHub notifications bot (Mar 01 2023 at 23:11):

bjorn3 created PR review comment:

Why doesn't the immediate here match the vcode?

view this post on Zulip Wasmtime GitHub notifications bot (Mar 02 2023 at 00:00):

jameysharp submitted PR review.

view this post on Zulip Wasmtime GitHub notifications bot (Mar 02 2023 at 00:00):

jameysharp created PR review comment:

Good question! I had to stare at this a minute after seeing your question. The vcode printed the instruction as movn, so storing bitwise-not of 0 into w2. I guess the disassembler special-cased that to print a more "friendly" mnemonic for the instruction, where the bitwise-not is folded into the immediate value. I believe both are correct though, at least given a sufficiently clever assembler.

Last updated: Dec 23 2024 at 13:07 UTC