Rushikesh code

backends/cadence/aot/functions_hifi.yaml

@@ -51,6 +51,61 @@
   kernels:
     - arg_meta: null
       kernel_name: torch::executor::div_out_mode
+
+- op: floor_divide.out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::floor_divide_out
+
+- op: remainder.Tensor_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::remainder_Tensor_out
+
+- op: remainder.Scalar_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::remainder_Scalar_out
+
+- op: fmod.Tensor_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::fmod_Tensor_out
+
+- op: fmod.Scalar_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::fmod_Scalar_out
+
+- op: bitwise_and.Scalar_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::bitwise_and_Scalar_out
+
+- op: bitwise_and.Tensor_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::bitwise_and_Tensor_out  
+
+- op: bitwise_or.Scalar_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::bitwise_or_Scalar_out
+
+- op: bitwise_or.Tensor_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::bitwise_or_Tensor_out    
+
+- op: bitwise_xor.Scalar_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::bitwise_xor_Scalar_out
+
+- op: bitwise_xor.Tensor_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::bitwise_xor_Tensor_out
 
 - op: embedding.out
   kernels:
@@ -67,6 +122,11 @@
     - arg_meta: null
       kernel_name: torch::executor::mul_out
 
+- op: mul.Scalar_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::mul_scalar_out
+
 - op: permute_copy.out
   kernels:
     - arg_meta: null
@@ -102,10 +162,105 @@
     - arg_meta: null
       kernel_name: torch::executor::where_out
 
+- op: scalar_tensor.out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::scalar_tensor_out
+
 - op: rsqrt.out
   kernels:
     - arg_meta: null
-      kernel_name: torch::executor::rsqrt_out        
+      kernel_name: torch::executor::rsqrt_out  
+
+- op: ge.Scalar_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::ge_scalar_out
+
+- op: ge.Tensor_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::ge_tensor_out
+
+- op: gt.Scalar_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::gt_scalar_out
+
+- op: gt.Tensor_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::gt_tensor_out   
+
+- op: le.Scalar_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::le_scalar_out
+
+- op: le.Tensor_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::le_tensor_out  
+
+- op: lt.Scalar_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::lt_scalar_out
+
+- op: lt.Tensor_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::lt_tensor_out      
+
+- op: eq.Scalar_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::eq_scalar_out
+
+- op: eq.Tensor_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::eq_tensor_out        
+
+- op: ne.Scalar_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::ne_scalar_out
+
+- op: ne.Tensor_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::ne_tensor_out 
+
+- op: pow.Scalar_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::pow_Scalar_out
+
+- op: pow.Tensor_Scalar_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::pow_Tensor_Scalar_out
+
+- op: pow.Tensor_Tensor_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::pow_Tensor_Tensor_out
+
+- op: atan2.out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::atan2_out    
+
+- op: empty.out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::empty_out
+
+- op: gelu.out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::gelu_out
 
 - op: empty.out
   kernels:
@@ -139,6 +294,21 @@
   kernels:
     - arg_meta: null
       kernel_name: impl::HiFi::quantized_linear_out
+
+- func: cadence::quantized_matmul.out(Tensor X, int X_zero_point, Tensor Y, int Y_zero_point, Tensor? bias, int out_multiplier, int out_shift, int out_zero_point, bool transposed, *, Tensor(a!) out) -> Tensor(a!)
+  kernels:
+    - arg_meta: null
+      kernel_name: impl::HiFi::quantized_matmul_out
+
+- func: cadence::quantized_relu.out(Tensor X, Tensor X_zero_point, *, Tensor(a!) out) -> Tensor(a!)
+  kernels:
+    - arg_meta: null
+      kernel_name: impl::HiFi::quantized_relu_out
+
+- func: cadence::quantized_conv.out(Tensor input, Tensor weight, Tensor bias, int[] stride, SymInt[] padding, int[] dilation, int groups, int input_zero_point, Tensor weight_zero_point, Tensor bias_scale, float out_scale, int out_zero_point, Tensor out_multiplier, Tensor out_shift, bool channel_last=False, *, Tensor(a!) out) -> Tensor(a!)
+  kernels:
+    - arg_meta: null
+      kernel_name: impl::HiFi::quantized_conv_out
 
 - func: cadence::quantized_relu.out(Tensor X, Tensor X_zero_point, *, Tensor(a!) out) -> Tensor(a!)
   kernels:

backends/cadence/hifi/kernels/CMakeLists.txt

@@ -15,6 +15,14 @@ add_library(
   ${EXECUTORCH_ROOT}/backends/cadence/hifi/third-party/nnlib/xa_nn_elm_where_f32xf32_f32.c
   ${EXECUTORCH_ROOT}/backends/cadence/hifi/third-party/nnlib/xa_nn_elm_div_broadcast_f32.c
   ${EXECUTORCH_ROOT}/backends/cadence/hifi/third-party/nnlib/xa_nn_elm_floor_div_broadcast_f32.c
+  ${EXECUTORCH_ROOT}/backends/cadence/hifi/third-party/nnlib/xa_nn_concat_32.c
+  ${EXECUTORCH_ROOT}/backends/cadence/hifi/third-party/nnlib/xa_nn_broadcast_32.c
+  ${EXECUTORCH_ROOT}/backends/cadence/hifi/third-party/nnlib/xa_nn_greater_lesser_equal_f32.c
+  ${EXECUTORCH_ROOT}/backends/cadence/hifi/third-party/nnlib/xa_nn_elm_remainder_broadcast_f32.c
+  ${EXECUTORCH_ROOT}/backends/cadence/hifi/third-party/nnlib/xa_nn_elm_fmod_broadcast_f32.c
+  ${EXECUTORCH_ROOT}/backends/cadence/hifi/third-party/nnlib/xa_nn_elm_floor_div_broadcast_f32.c
+  ${EXECUTORCH_ROOT}/backends/cadence/hifi/third-party/nnlib/xa_nn_elm_logicalxor_bool_bool.c
+  ${EXECUTORCH_ROOT}/backends/cadence/hifi/third-party/nnlib/xa_nn_elm_where_f32xf32_f32.c
 )
 
 target_include_directories(