[OpPerf] Fix axis_shape and function mismatch for LTS (#17894)

benchmark/opperf/README.md

@@ -76,6 +76,8 @@ python incubator-mxnet/benchmark/opperf/opperf.py --output-format json --output-
 4. **profiler** : `native` or `python`. By default, 'native'. You can override and set the global profiler for all operator benchmarks. Example: --profiler 'python'.
 Native profiler uses MXNet C++ based built-in profiler. Python profiler uses Python package time. Generally, native profiler is used by developers and python profiler is used by users.
 
+5. **int64-tensor** : `on` or `off`. By default, 'off'. You can override and set the large tensor flag to ON. Example: --int64-tensor ON
+
 ## Usecase 2 - Run benchmarks for all the operators in a specific category
 
 For example, you want to run benchmarks for all NDArray Broadcast Binary Operators, Ex: broadcast_add, broadcast_mod, broadcast_pow etc., You just run the following python script.
@@ -199,7 +201,7 @@ By default, MXNet profiler is used as the profiler engine.
 
 All contributions are welcome. Below is the list of desired features:
 
-1. Cover all MXNet operators.
+1. ~~Cover all MXNet operators~~.
 2. Enhance MXNet profiler with additional APIs to programmatically fetch and process profiler data.
 3. Integration with CI/CD system to run operator benchmarks for PR builds, nightly builds.
 4. Dashboards and other modes of presentation of results for analyzing and planning tasks such as operator performance improvements.

benchmark/opperf/nd_operations/array_manipulation_operators.py

@@ -97,7 +97,7 @@ def run_rearrange_operators_benchmarks(ctx=mx.cpu(), dtype='float32', profiler='
     mx_rearrange_ops = get_all_rearrange_operators()
 
     # Run benchmarks
-    mx_rearrange_op_results = run_op_benchmarks(mx_rearrange_ops, dtype, ctx, profiler, warmup, runs)
+    mx_rearrange_op_results = run_op_benchmarks(mx_rearrange_ops, dtype, ctx, profiler, int64_tensor, warmup, runs)
     return mx_rearrange_op_results
 
 
@@ -129,7 +129,7 @@ def run_shape_operators_benchmarks(ctx=mx.cpu(), dtype='float32', profiler='nati
     mx_shape_ops = get_all_shape_operators()
 
     # Run benchmarks
-    mx_shape_op_results = run_op_benchmarks(mx_shape_ops, dtype, ctx, profiler, warmup, runs)
+    mx_shape_op_results = run_op_benchmarks(mx_shape_ops, dtype, ctx, profiler, int64_tensor, warmup, runs)
     return mx_shape_op_results
 
 
@@ -161,7 +161,7 @@ def run_expanding_operators_benchmarks(ctx=mx.cpu(), dtype='float32', profiler='
     mx_expanding_ops = get_all_expanding_operators()
 
     # Run benchmarks
-    mx_expanding_op_results = run_op_benchmarks(mx_expanding_ops, dtype, ctx, profiler, warmup, runs)
+    mx_expanding_op_results = run_op_benchmarks(mx_expanding_ops, dtype, ctx, profiler, int64_tensor, warmup, runs)
     return mx_expanding_op_results
 
 
@@ -193,7 +193,7 @@ def run_rounding_operators_benchmarks(ctx=mx.cpu(), dtype='float32', profiler='n
     mx_rounding_ops = get_all_rounding_operators()
 
     # Run benchmarks
-    mx_rounding_op_results = run_op_benchmarks(mx_rounding_ops, dtype, ctx, profiler, warmup, runs)
+    mx_rounding_op_results = run_op_benchmarks(mx_rounding_ops, dtype, ctx, profiler, int64_tensor, warmup, runs)
     return mx_rounding_op_results
 
 

benchmark/opperf/rules/default_params.py

@@ -603,7 +603,6 @@
                                 "p": DEFAULT_P,
                                 "k_nd": DEFAULT_K_ND_LARGE_TENSOR,
                                 "p_nd": DEFAULT_P_ND_LARGE_TENSOR,
-                                "axis_shape": DEFAULT_AXIS_SHAPE,
                                 "axis": DEFAULT_AXIS,
                                 "weight" : DEFAULT_WEIGHT_LARGE_TENSOR,
                                 "weight32" : DEFAULT_WEIGHT_LARGE_TENSOR,