wasmtime / issue #5919 Codegen fix atomic_cas with samll... · git-wasmtime

Stream: git-wasmtime

Topic: wasmtime / issue #5919 Codegen fix atomic_cas with samll...

Wasmtime GitHub notifications bot (Mar 03 2023 at 03:17):

yuyang-ok commented on issue #5919:

The test failure indicate a s930x has a wrong resule too.
I think target interpreter and x86_64 can pass.

Wasmtime GitHub notifications bot (Mar 03 2023 at 03:18):

yuyang-ok edited a comment on issue #5919:

The test failure indicates a s930x has a wrong resule too ???
I think target interpreter and x86_64 can pass.

Wasmtime GitHub notifications bot (Mar 03 2023 at 04:02):

yuyang-ok edited a comment on issue #5919:

The test failure indicates a s930x has a wrong result too ???
I think target interpreter and x86_64 can pass.

Wasmtime GitHub notifications bot (Mar 03 2023 at 04:45):

yuyang-ok edited a comment on issue #5919:

The test failure indicates s930x has a wrong result too ???
I think target interpreter and x86_64 can pass.

Wasmtime GitHub notifications bot (Mar 03 2023 at 05:30):

yuyang-ok edited a comment on issue #5919:

The test failure indicates s390x has a wrong result too ???
I think target interpreter and x86_64 can pass.

Wasmtime GitHub notifications bot (Mar 03 2023 at 12:32):

afonso360 commented on issue #5919:

Oh that's right, this test is not endianness independent! If you change the store to store little and the atomic_cas to atomic_cas little it should start passing on all arches!

<details>
<summary>New Testcase</summary>

test interpret
test run
target aarch64
target s390x
target x86_64
target riscv64

function %a(i8) -> i8 sext system_v {
    ss2 = explicit_slot 2

block0(v0: i8):
    v2 = iconst.i16 0x00ff
    v3 = stack_addr.i64 ss2
    store little v2, v3

    v4 = stack_addr.i64 ss2+1
    v5 = atomic_cas little v4, v0, v0
    return v5
}

; run: %a(0) == 0