[Kernel] Classification cleanup: unify _jit_ naming, drop empty/model groups, add elementwise (RFC #29630) (#32148)

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>

sgl-kernel/benchmark/bench_fp8_gemm.py

@@ -8,7 +8,7 @@ import torch
 import triton
 from sgl_kernel import fp8_scaled_mm as sgl_scaled_mm
 
-from sglang.kernels.ops.quantization._jit_per_tensor_quant_fp8 import (
+from sglang.kernels.ops.quantization.per_tensor_quant_fp8 import (
     per_tensor_quant_fp8,
 )
 from sglang.utils import is_in_ci

sgl-kernel/benchmark/bench_fp8_gemm_swap_ab.py

@@ -18,7 +18,7 @@ import torch
 import triton
 from sgl_kernel import fp8_scaled_mm as sgl_scaled_mm
 
-from sglang.kernels.ops.quantization._jit_per_tensor_quant_fp8 import (
+from sglang.kernels.ops.quantization.per_tensor_quant_fp8 import (
     per_tensor_quant_fp8,
 )
 from sglang.utils import is_in_ci

sgl-kernel/benchmark/bench_per_tensor_quant_fp8.py

@@ -8,7 +8,7 @@ import torch
 import triton
 import triton.testing
 
-from sglang.kernels.ops.quantization._jit_per_tensor_quant_fp8 import (
+from sglang.kernels.ops.quantization.per_tensor_quant_fp8 import (
     per_tensor_quant_fp8,
 )
 from sglang.utils import is_in_ci

sgl-kernel/tests/test_cutlass_w4a8_moe_mm.py

@@ -5,7 +5,7 @@ import torch
 from sgl_kernel import cutlass_w4a8_moe_mm
 from utils import is_hopper
 
-from sglang.kernels.ops.quantization._jit_per_tensor_quant_fp8 import (
+from sglang.kernels.ops.quantization.per_tensor_quant_fp8 import (
     per_tensor_quant_fp8,
 )