[Intel GPU] qlinear.pointwise with mixed dtype support (pytorch#136753)

# Motivation
This PR is aimed to add mixed data type(AMP) support for `qlinear_pointwise` op. With current PR, we allow `qlinear` kernels output Tensor that is BF16, rather than FP32/INT8.

# UT verification
```bash
DNNL_VERBOSE=1 python test/inductor/test_mkldnn_pattern_matcher.py -v \
    -k test_qlinear_int8_mixed_bf16_xpu \
    -k test_qlinear_relu_int8_mixed_bf16_xpu \
    -k test_qlinear_add_int8_mixed_bf16_xpu
```

# Runtime exemplification
```bash
#qlinear+bf16 output
onednn_verbose,primitive,exec,gpu:0,matmul,ocl:gemm_with_po:any,undef,src_s8::blocked:ab::f0 wei_s8::blocked:ab::f0 bia_bf16::blocked:ab::f0_mask2 dst_bf16::blocked:ab::f0,attr-scratchpad:user attr-scales:src0:0:f32+dst:0:f32+wei:2:f32 attr-zero-points:src0:0:s32,,4x4:4x4,0.0698242
# qlinear_add + bf16 output
onednn_verbose,primitive,exec,gpu:0,matmul,ocl:gemm_with_po:any,undef,src_s8::blocked:ab::f0 wei_s8::blocked:ab::f0 bia_bf16::blocked:ab::f0_mask2 dst_bf16::blocked:ab::f0,attr-scratchpad:user attr-scales:src0:0:f32+dst:0:f32+wei:2:f32 attr-zero-points:src0:0:s32 attr-post-ops:eltwise_linear:1:-0.677141+sum:0.0132773,,4x4:4x4,0.0419922
# qlinear_add_relu + bf16 output
onednn_verbose,primitive,exec,gpu:0,matmul,ocl:gemm_with_po:any,undef,src_s8::blocked:ab::f0 wei_s8::blocked:ab::f0 bia_bf16::blocked:ab::f0_mask2 dst_bf16::blocked:ab::f0,attr-scratchpad:user attr-scales:src0:0:f32+dst:0:f32+wei:2:f32 attr-zero-points:src0:0:s32 attr-post-ops:eltwise_linear:1:0.533096+sum:0.00416481+eltwise_relu,,4x4:4x4,0.0759277
```
As shown in the oneDNN verbose, the attribute `dst_bf16::blocked:ab::f0` demonstrate that we could successfully output a bf16 tensor in int8 gemm.

Pull Request resolved: pytorch#136753
Approved by: https://github.com/EikanWang, https://github.com/guangyey, https://github.com/desertfire, https://github.com/jerryzh168
ghstack dependencies: pytorch#133307, pytorch#135189, pytorch#135337, pytorch#135465

Co-authored-by: guangyey <[email protected]>

Author: 2 people

aten/src/ATen/native/mkldnn/xpu/qlinear.cpp

@@ -7,6 +7,18 @@ using namespace at::native::onednn;
 
 namespace at::native::xpu {
 
+static inline c10::ScalarType qlinear_decide_out_dtype(
+    const at::Tensor& act,
+    const std::optional<c10::ScalarType> output_dtype) {
+  bool fp32_output = output_dtype.has_value() && (output_dtype == c10::kFloat);
+  bool bfloat16_output =
+      output_dtype.has_value() && (output_dtype == c10::kBFloat16);
+  auto dst_dtype = fp32_output
+      ? c10::kFloat
+      : (bfloat16_output ? c10::kBFloat16 : act.scalar_type());
+  return dst_dtype;
+}
+
 Tensor q_linear_pointwise(
     Tensor act,
     double act_scale,
@@ -37,9 +49,9 @@ Tensor q_linear_pointwise(
 
   std::vector<int64_t> src_dims = {M, K};
   std::vector<int64_t> dst_dims = {M, N};
-  auto out_dtype =
-      output_dtype.has_value() ? output_dtype.value() : act.scalar_type();
-  Tensor qout = at::empty(dst_dims, act.options().dtype(out_dtype));
+
+  auto dst_dtype = qlinear_decide_out_dtype(act, output_dtype);
+  Tensor qout = at::empty(dst_dims, act.options().dtype(dst_dtype));
 
   quantized_matmul(
       act.contiguous(),
@@ -96,9 +108,9 @@ Tensor q_linear_pointwise_tensor(
 
   std::vector<int64_t> src_dims = {M, K};
   std::vector<int64_t> dst_dims = {M, N};
-  auto out_dtype =
-      output_dtype.has_value() ? output_dtype.value() : act.scalar_type();
-  Tensor qout = at::empty(dst_dims, act.options().dtype(out_dtype));
+
+  auto dst_dtype = qlinear_decide_out_dtype(act, output_dtype);
+  Tensor qout = at::empty(dst_dims, act.options().dtype(dst_dtype));
 
   quantized_matmul(
       act.contiguous(),
@@ -159,9 +171,8 @@ Tensor q_linear_pointwise_binary(
 
   std::vector<int64_t> src_dims = {M, K};
   std::vector<int64_t> dst_dims = {M, N};
-  auto out_dtype =
-      output_dtype.has_value() ? output_dtype.value() : act.scalar_type();
-  Tensor qout = at::empty(dst_dims, act.options().dtype(out_dtype));
+  auto dst_dtype = qlinear_decide_out_dtype(act, output_dtype);
+  Tensor qout = at::empty(dst_dims, act.options().dtype(dst_dtype));
 
   quantized_matmul(
       act.contiguous(),
@@ -222,9 +233,8 @@ Tensor q_linear_pointwise_binary_tensor(
 
   std::vector<int64_t> src_dims = {M, K};
   std::vector<int64_t> dst_dims = {M, N};
-  auto out_dtype =
-      output_dtype.has_value() ? output_dtype.value() : act.scalar_type();
-  Tensor qout = at::empty(dst_dims, act.options().dtype(out_dtype));
+  auto dst_dtype = qlinear_decide_out_dtype(act, output_dtype);
+  Tensor qout = at::empty(dst_dims, act.options().dtype(dst_dtype));
 
   quantized_matmul(
       act.contiguous(),