Add more QC8 GEMM/IGEMM JIT microkernels.
PiperOrigin-RevId: 420137137
diff --git a/BUILD.bazel b/BUILD.bazel
index 99c7fae..75e907a 100644
--- a/BUILD.bazel
+++ b/BUILD.bazel
@@ -7024,14 +7024,16 @@
"src/f32-gemm/4x8-aarch32-neon-cortex-a53.cc",
"src/f32-gemm/4x8-aarch32-neon-cortex-a55.cc",
"src/f32-gemm/4x8-aarch32-neon-cortex-a7.cc",
- "src/f32-gemm/4x8-aarch32-neon-ld64.cc",
"src/f32-gemm/4x8-aarch32-neon-cortex-a75.cc",
+ "src/f32-gemm/4x8-aarch32-neon-ld64.cc",
"src/f32-igemm/4x8-aarch32-neon-cortex-a53.cc",
"src/f32-igemm/4x8-aarch32-neon-cortex-a55.cc",
"src/f32-igemm/4x8-aarch32-neon-cortex-a7.cc",
- "src/f32-igemm/4x8-aarch32-neon-ld64.cc",
"src/f32-igemm/4x8-aarch32-neon-cortex-a75.cc",
+ "src/f32-igemm/4x8-aarch32-neon-ld64.cc",
+ "src/qc8-gemm/4x8-fp32-aarch32-neonv8-mlal-lane-ld64.cc",
"src/qc8-gemm/4x8c4-fp32-aarch32-neondot-ld64.cc",
+ "src/qc8-igemm/4x8-fp32-aarch32-neonv8-mlal-lane-ld64.cc",
"src/qc8-igemm/4x8c4-fp32-aarch32-neondot-ld64.cc",
"src/qs8-gemm/4x8-rndnu-aarch32-neon-mlal-lane-ld64.cc",
"src/qs8-gemm/4x8c4-rndnu-aarch32-neondot-ld64.cc",
@@ -11313,7 +11315,10 @@
"src/xnnpack/AlignedAllocator.h",
] + WEIGHTS_PACK_HDRS + MICROKERNEL_TEST_HDRS,
shard_count = 10,
- deps = MICROKERNEL_TEST_DEPS + [":packing"],
+ deps = MICROKERNEL_TEST_DEPS + [
+ ":packing",
+ ":jit",
+ ],
)
xnnpack_unit_test(
@@ -11325,7 +11330,10 @@
"src/xnnpack/AlignedAllocator.h",
] + WEIGHTS_PACK_HDRS + MICROKERNEL_TEST_HDRS,
shard_count = 10,
- deps = MICROKERNEL_TEST_DEPS + [":packing"],
+ deps = MICROKERNEL_TEST_DEPS + [
+ ":packing",
+ ":jit",
+ ],
)
xnnpack_unit_test(