gpu: ocl: remove unecessary post_op kind

The binary and eltwise enums are guaranteed to not overlap via the API
oneapi-src · May 12, 2023 · 87fd48f · 87fd48f
1 parent 9a66ac6
commit 87fd48f
Showing 1 changed file with 28 additions and 37 deletions.
diff --git a/src/gpu/ocl/ocl_post_ops.h b/src/gpu/ocl/ocl_post_ops.h
@@ -27,30 +27,23 @@
 #include "gpu/ocl/ocl_eltwise.h"
 #include "gpu/ocl/ocl_types.h"
 
-float fwd_Xnary(unsigned kind, unsigned algorithm, float x, float y,
-        float alpha, float beta, float scale) {
-    if (kind == PO_BINARY) {
-        switch (algorithm) {
-            // binary
-            case BINARY_ADD: return x + y; break;
-            case BINARY_MUL: return x * y; break;
-            case BINARY_MIN: return x < y ? x : y; break;
-            case BINARY_MAX: return x > y ? x : y; break;
-            case BINARY_DIV: return x / y; break;
-            case BINARY_SUB: return x - y; break;
-            case BINARY_GE: return x >= y; break;
-            case BINARY_GT: return x > y; break;
-            case BINARY_LE: return x <= y; break;
-            case BINARY_LT: return x < y; break;
-            case BINARY_EQ: return x == y; break;
-            case BINARY_NE: return x != y; break;
-            case RELU: // binary && relu = prelu
-                return fwd_eltwise_common(RELU, x, y, beta, scale);
-                break;
-            default: return 0.f;
-        }
-    } else { // eltwise kind
-        return fwd_eltwise_common(algorithm, x, alpha, beta, scale);
+float fwd_Xnary(unsigned algorithm, float x, float y, float alpha, float beta,
+        float scale) {
+    switch (algorithm) {
+        // binary
+        case BINARY_ADD: return x + y; break;
+        case BINARY_MUL: return x * y; break;
+        case BINARY_MIN: return x < y ? x : y; break;
+        case BINARY_MAX: return x > y ? x : y; break;
+        case BINARY_DIV: return x / y; break;
+        case BINARY_SUB: return x - y; break;
+        case BINARY_GE: return x >= y; break;
+        case BINARY_GT: return x > y; break;
+        case BINARY_LE: return x <= y; break;
+        case BINARY_LT: return x < y; break;
+        case BINARY_EQ: return x == y; break;
+        case BINARY_NE: return x != y; break;
+        default: return fwd_eltwise_common(algorithm, x, alpha, beta, scale);
     }
 }
 
@@ -65,28 +58,26 @@ float fwd_Xnary(unsigned kind, unsigned algorithm, float x, float y,
         ret_val; \
     })
 
-#define FWD_XNARY_GENERIC_DT(po_kind, algorithm, result, result_elem_dt, \
-        arg0_ptr, arg0_len, arg1_ptr, arg1_len, alpha, beta, scale) \
+#define FWD_XNARY_GENERIC_DT(algorithm, result, result_elem_dt, arg0_ptr, \
+        arg0_len, arg1_ptr, arg1_len, alpha, beta, scale) \
     { \
         auto ty = arg0_len + arg1_len; \
         const typeof(ty) out_len \
                 = max((typeof(ty))arg0_len, (typeof(ty))arg1_len); \
         result_elem_dt *res_ptr = (result_elem_dt *)(&result); \
         unroll_for(typeof(out_len + 0) idx = 0; idx < out_len; ++idx) { \
             if (arg0_len == 1 && arg1_len == 1) { \
-                *res_ptr = fwd_Xnary(po_kind, algorithm, \
-                        convert_float(*arg0_ptr), convert_float(*arg1_ptr), \
-                        alpha, beta, scale); \
+                *res_ptr = fwd_Xnary(algorithm, convert_float(*arg0_ptr), \
+                        convert_float(*arg1_ptr), alpha, beta, scale); \
             } else if (arg0_len == 1) { \
-                res_ptr[idx] = fwd_Xnary(po_kind, algorithm, \
-                        convert_float(*arg0_ptr), \
+                res_ptr[idx] = fwd_Xnary(algorithm, convert_float(*arg0_ptr), \
                         convert_float(arg1_ptr[idx]), alpha, beta, scale); \
             } else if (arg1_len == 1) { \
-                res_ptr[idx] = fwd_Xnary(po_kind, algorithm, \
-                        convert_float(arg0_ptr[idx]), \
-                        convert_float(*arg1_ptr), alpha, beta, scale); \
+                res_ptr[idx] \
+                        = fwd_Xnary(algorithm, convert_float(arg0_ptr[idx]), \
+                                convert_float(*arg1_ptr), alpha, beta, scale); \
             } else { \
-                res_ptr[idx] = fwd_Xnary(po_kind, algorithm, \
+                res_ptr[idx] = fwd_Xnary(algorithm, \
                         convert_float(arg0_ptr[idx]), \
                         convert_float(arg1_ptr[idx]), alpha, beta, scale); \
             } \
@@ -277,7 +268,7 @@ float fwd_Xnary(unsigned kind, unsigned algorithm, float x, float y,
             REPLICATE_DATA(bin_arg_ptr, bin_arg_size, x0_s, x1_size, x2_s, \
                     x3_s, x4_s, x5_s); \
         } \
-        FWD_XNARY_GENERIC_DT(PO_BINARY, CONCAT3(PO_, idx, _ALG), accumulator, \
+        FWD_XNARY_GENERIC_DT(CONCAT3(PO_, idx, _ALG), accumulator, \
                 acc_elem_dt, ((acc_elem_dt *)(&accumulator)), \
                 (sizeof(accumulator) / sizeof(acc_elem_dt)), bin_arg_ptr, \
                 bin_arg_size, 0.0f, 0.0f, 1.0f); \
@@ -292,7 +283,7 @@ float fwd_Xnary(unsigned kind, unsigned algorithm, float x, float y,
 
 #define APPLY_PO_ELTWISE(idx, accumulator, acc_elem_dt) \
     { \
-        FWD_XNARY_GENERIC_DT(PO_ELTWISE, CONCAT3(PO_, idx, _ALG), accumulator, \
+        FWD_XNARY_GENERIC_DT(CONCAT3(PO_, idx, _ALG), accumulator, \
                 acc_elem_dt, ((acc_elem_dt *)(&accumulator)), \
                 (sizeof(accumulator) / sizeof(acc_elem_dt)), \
                 ((acc_elem_dt *)(&accumulator)), \