Closed mtsokol closed 3 months ago
This PR adds support for multiple architectures, namely:
x64_64
haswell
sandybridge
skylakex
zen
arm64
armv8
xGEMM_UNROLL_x
CI runs three jobs: ubuntu-latest (which is haswell or zen), macos13 (looks like it's using sandybridge), and macos-latest (an M1 armv8).
ubuntu-latest
macos13
macos-latest
This is fantastic, thanks for seeing this through.
This PR adds support for multiple architectures, namely:
x64_64
arch withhaswell
,sandybridge
,skylakex
, andzen
subarchs.arm64
arch witharmv8
subarch.xGEMM_UNROLL_x
configs to determine loop unroll size instead of hardcoding them.CI runs three jobs:
ubuntu-latest
(which ishaswell
orzen
),macos13
(looks like it's usingsandybridge
), andmacos-latest
(an M1 armv8).