Bug 1825285 - Do not use implicit arch parameter when loading r=RyanVM

Otherwise we end up with potentially invalid configuration when
generating code for specific architectures.

Differential Revision: https://phabricator.services.mozilla.com/D173965
This commit is contained in:
serge-sans-paille 2023-03-29 15:02:20 +00:00
parent 50df903910
commit 5fd3f3868e
2 changed files with 4 additions and 4 deletions

View File

@ -968,9 +968,9 @@ AddBias::operator()(
std::tuple<xsimd::batch<float, Arch>, xsimd::batch<float, Arch>> total,
size_t, size_t col_idx, size_t) {
return std::make_tuple(
std::get<0>(total) + xsimd::load_aligned(bias_addr + col_idx + 0),
std::get<0>(total) + xsimd::batch<float, Arch>::load_aligned(bias_addr + col_idx + 0),
std::get<1>(total) +
xsimd::load_aligned(bias_addr + col_idx +
xsimd::batch<float, Arch>::load_aligned(bias_addr + col_idx +
xsimd::batch<float, Arch>::size));
}

View File

@ -10,8 +10,8 @@ origin:
url: https://github.com/serge-sans-paille/gemmology
release: e1167c52cbbfd989390e4d9515c84c88878bfe80 (2023-03-28T11:32:43Z).
revision: e1167c52cbbfd989390e4d9515c84c88878bfe80
release: 3c20b313a6c49ca674a830630c0ef5ea5663b91e (2023-03-29T14:50:11Z).
revision: 3c20b313a6c49ca674a830630c0ef5ea5663b91e
license: MIT