doc: doesn't -> does not

doc/advanced/design/understanding_memory_formats.md

@@ -312,7 +312,7 @@ So we came up with another solution using zero-padding. The idea is to round the
 channels up to make them multiples of the block size and pad the resulting tail
 with zeros (in the example above, `24 = div_up(17, 8) * 8`). Then primitives
 like convolutions might work with a rounded-up number of channels instead of
-the original ones and compute the correct result (adding zeros doesn't change
+the original ones and compute the correct result (adding zeros does not change
 the result).
 
 That enables supporting an arbitrary number of channels with almost no changes

doc/advanced/int8_computations.md

@@ -143,7 +143,7 @@ represented by the u8 data type, the non-symmetric quantization is assumed.
 Namely, the formula is:
 - \f$data_{f32}[:] = \frac{1}{scale}(data_{u8}[:] - shift)\f$.
 
-But similarly to the other primitives, the LSTM primitive doesn't handle
+But similarly to the other primitives, the LSTM primitive does not handle
 potential overflows automatically. It is up to the user to specify the
 appropriate quantization parameters (see
 dnnl::primitive_attr::set_rnn_data_qparams() and
@@ -264,7 +264,7 @@ the implementations are given below:
 
 2. **s8/s8 GEMM** (dnnl_gemm_s8s8s32()) does nothing to handle the overflow
    issue. It is up to the user to prepare the data so that the
-   overflow/saturation doesn't occur. For instance, the user can specify
+   overflow/saturation does not occur. For instance, the user can specify
    s7 `[-64, 63]` instead of s8 for the second input.
 
    @warning

doc/advanced/primitive_cache.md

@@ -12,7 +12,7 @@ primitive implementations, etc. It can significantly reduce primitive creation
 overhead, especially when an application or a framework creates primitives
 for every instance of inference or iteration of training process.
 
-The primitive cache is global hence a user doesn't have to maintain any
+The primitive cache is global hence a user does not have to maintain any
 persistent oneDNN resources to benefit from the primitive cache.
 
 ## Managing Memory Consumption

doc/advanced/transition-to-v1.md

@@ -283,7 +283,7 @@ In Intel MKL-DNN v1.0, constructing a memory object using special value
 `MKLDNN_MEMORY_ALLOCATE` for a handle results in the buffer being allocated by
 the library. This makes the behavior of the C API memory object constructor
 aligned with its C++ API `mkldnn::memory` counterpart. Note that the C++ API
-memory object class still has an extra constructor that doesn't take a handle
+memory object class still has an extra constructor that does not take a handle
 at all, and asks the library to allocate the buffer (that is, the same behavior
 as calling with the handle equal to `MKLDNN_MEMORY_ALLOCATE`).
 
@@ -346,7 +346,7 @@ These changes bring new data types and functions, such as:
     // create a memory
     // native_handle can:
     //  - point to the user allocated memory, i.e. valid handle. In this case the
-    //    library doesn't own allocated memory.
+    //    library does not own allocated memory.
     //  - be MKLDNN_NATIVE_HANDLE_ALLOCATE to ask the library to allocate and
     //    attach memory. In this case the library owns allocated memory.
     //  - be MKLDNN_NATIVE_HANDLE_NONE to create mkldnn_memory w/o attached memory.

doc/primitives/concat.md

@@ -16,9 +16,9 @@ standard @ref dev_guide_conventions):
 
 where \f$c = C_1 + .. + C_{i-1} {}_{} + c'\f$.
 
-The concat primitive doesn't have a notion of forward or backward propagations.
-The backward propagation for the concatenation operation is simply an identity
-operation.
+The concat primitive does not have a notion of forward or backward
+propagations.  The backward propagation for the concatenation operation is
+simply an identity operation.
 
 ## Execution Arguments
 When executed, the inputs and outputs should be mapped to an execution argument index as specified by the following table.
@@ -55,7 +55,7 @@ meaning associated with any logical dimensions.
 
 ### Post-ops and Attributes
 
-The concat primitive doesn't support any post-ops or attributes.
+The concat primitive does not support any post-ops or attributes.
 
 ## Implementation Limitations
 

doc/primitives/convolution.md

@@ -220,7 +220,7 @@ if the scales are not known at the primitive descriptor creation stage.
 In this case, the user must provide the scales as an additional input memory
 object with argument `DNNL_ARG_ATTR_OUTPUT_SCALES` during the execution stage.
 
-@note The library doesn't prevent using post-ops in training, but note that
+@note The library does not prevent using post-ops in training, but note that
 not all post-ops are feasible for training usage. For instance, using ReLU
 with non-zero negative slope parameter as a post-op would not produce an
 additional output `workspace` that is required to compute backward propagation

doc/primitives/eltwise.md

@@ -126,7 +126,7 @@ meaning associated with any logical dimensions.
 
 ### Post-ops and Attributes
 
-The eltwise primitive doesn't support any post-ops or attributes.
+The eltwise primitive does not support any post-ops or attributes.
 
 @anchor dg_eltwise_impl_limits
 ## Implementation Limitations
@@ -151,7 +151,7 @@ The eltwise primitive doesn't support any post-ops or attributes.
 
     - Moreover, such operations can be fused as a
       [post-op](@ref dev_guide_attributes) with the previous operation if that
-      operation doesn't require its \dst to compute the backward
+      operation does not require its \dst to compute the backward
       propagation (e.g., if the convolution operation satisfies these
       conditions).
 

doc/primitives/logsoftmax.md

@@ -76,7 +76,7 @@ argument index as specified by the following table.
 
 ### Post-ops and Attributes
 
-The logsoftmax primitive doesn't support any post-ops or attributes.
+The logsoftmax primitive does not support any post-ops or attributes.
 
 ### Data Type Support
 

doc/primitives/lrn.md

@@ -113,7 +113,7 @@ of any preceding compute-intensive primitive.
 
 ### Post-ops and Attributes
 
-The LRN primitive doesn't support any post-ops or attributes.
+The LRN primitive does not support any post-ops or attributes.
 
 
 @anchor dg_lrn_impl_limits

doc/primitives/matmul.md

@@ -134,7 +134,7 @@ argument with index set to
 
 1. Check @ref dev_guide_data_types.
 
-2. The CPU engine doesn't support `u8` data type for weights.
+2. The CPU engine does not support `u8` data type for weights.
 
 ## Performance Tips
 

doc/primitives/pooling.md

@@ -47,7 +47,7 @@ Average pooling supports two algorithms:
 #### Difference Between Forward Training and Forward Inference
 
 - Max pooling requires a `workspace` for the #dnnl_forward_training propagation
-  kind, and doesn't require it for #dnnl_forward_inference (see details below).
+  kind, and does not require it for #dnnl_forward_inference (see details below).
 
 ### Backward
 

doc/primitives/resampling.md

@@ -120,7 +120,7 @@ source and destination memory objects:
 
 ### Post-ops and Attributes
 
-The resampling primitive doesn't support any post-ops or attributes.
+The resampling primitive does not support any post-ops or attributes.
 
 ## Implementation Limitations
 

doc/primitives/shuffle.md

@@ -106,7 +106,7 @@ of any preceding compute-intensive primitive.
 
 ### Post-ops and Attributes
 
-The shuffle primitive doesn't support any post-ops or attributes.
+The shuffle primitive does not support any post-ops or attributes.
 
 
 @anchor dg_shuffle_impl_limits

doc/primitives/softmax.md

@@ -67,7 +67,7 @@ argument index as specified by the following table.
 
 ### Post-ops and Attributes
 
-The softmax primitive doesn't support any post-ops or attributes.
+The softmax primitive does not support any post-ops or attributes.
 
 ### Data Type Support
 

doc/primitives/sum.md

@@ -15,7 +15,7 @@ The sum primitive sums \f$N\f$ tensors (the variable names follow the standard
         \src_i(\overline{x})
 \f]
 
-The sum primitive doesn't have a notion of forward or backward propagations.
+The sum primitive does not have a notion of forward or backward propagations.
 The backward propagation for the sum operation is simply an identity operation.
 
 ## Execution Arguments
@@ -40,7 +40,7 @@ argument index as specified by the following table.
 
 ### Post-ops and Attributes
 
-The sum primitive doesn't support any post-ops or attributes.
+The sum primitive does not support any post-ops or attributes.
 
 ### Data Types Support
 

doc/programming_model/attributes.md

@@ -53,7 +53,7 @@ dnnl_primitive_attr_set_SOMETHING_ELSE(attr, other_params);
 dnnl_primitive_desc_t op_pd; // operation primitive descriptor
 dnnl_primitive_desc_create(&op_pd, &op_d, attr, engine, hint_fwd_pd);
 
-// changing attr object here doesn't have any effect on op_pd
+// changing attr object here does not have any effect on op_pd
 
 // once attr is no more used we can immediately destroy it
 dnnl_primitive_attr_destroy(attr);
@@ -96,7 +96,7 @@ create a primitive descriptor with the attributes they set, it might happen
 that there is no primitive implementation that supports such a configuration.
 In this case, the library will return #dnnl_unimplemented in the case of the C
 API or throw a corresponding @ref dnnl::error exception in the case of the C++
-API. Unfortunately, the library doesn't currently provide any hints about what
+API. Unfortunately, the library does not currently provide any hints about what
 exactly is going wrong in this case. The corresponding section of the
 documentation simply documents the primitives' capabilities.
 

doc/programming_model/attributes_post_ops.md

@@ -50,7 +50,7 @@ primitive::primitive_desc op_pd(params, attr); // create a pd with the attr
 @warning
     Different primitives may have different post-ops support.  Moreover, the
     support might also depend on the actual implementation of a primitive. For
-    instance, the library generally doesn't support post-ops for reference
+    instance, the library generally does not support post-ops for reference
     primitives (which are typically very slow, so there is no point in doing
     the actual fusion). Robust code should handle errors accordingly. See the
     [section on attributes error handling](@ref dev_guide_attributes_error_handling).

doc/programming_model/attributes_quantization.md

@@ -45,7 +45,7 @@ For the rest of this guide, we will assume that \f$0_{x\_int8} = 0\f$.
 Depending on the architecture, the behavior of int8 computations might slightly
 vary. For more details, refer to @ref dev_guide_int8_computations.
 
-This guide doesn't cover how the appropriate scaling factor can be found.
+This guide does not cover how the appropriate scaling factor can be found.
 Refer to the materials in the [Introduction](@ref dgaq_intro).
 
 ### Example: Convolution Quantization Workflow
@@ -80,7 +80,7 @@ where
 - \f$downconvert\_f32\_to\_s8()\f$ converts an `f32` value to `s8` with
   potential saturation if the values are out of the range of the INT8 data type.
 
-Note that in order to perform the operation, one doesn't need to know the
+Note that in order to perform the operation, one does not need to know the
 exact scaling factors for all the tensors; it is enough to know only the
 \f$output\_scale\f$. The library utilizes this fact: a user needs to provide
 only this one extra parameter to the convolution primitive (see the [Output
@@ -140,7 +140,7 @@ where
 ## API
 
 The library API to support for INT8 was designed for the model described above.
-However, it doesn't require users to follow exactly this model. As long as
+However, it does not require users to follow exactly this model. As long as
 users can fit their model into the given functionality everything should work
 fine. Having this in mind we tried to design a minimal and simple yet powerful
 enough quantization API.

doc/programming_model/attributes_scratchpad.md

@@ -136,7 +136,7 @@ dnnl::primitive::primitive_desc op_pd(op_d, attr, engine);
 // Query the scratchpad memory descriptor
 dnnl::memory::desc scratchpad_md = op_pd.scratchpad_desc();
 
-// Note, that a primitive doesn't consume memory in this configuration:
+// Note, that a primitive does not consume memory in this configuration:
 assert(op_pd.query_s64(dnnl::query::memory_consumption_s64) == 0);
 
 // Create a primitive