Review: further formatting fixes, add assert and use CPU version of fp32->fp16

Author: am17an

ggml/src/ggml-cpu/ops.cpp

@@ -6546,11 +6546,11 @@ void ggml_compute_forward_im2col_back_f32(
     }
 }
 
-static void ggml_call_mul_mat(ggml_type T, const ggml_compute_params * params, int64_t m, int64_t n, int64_t k,
-                              void * a, void * b, void * c) {
-    const ggml_type_traits * traits = ggml_get_type_traits(T);
+static void ggml_call_mul_mat(ggml_type type, const ggml_compute_params * params, int64_t m, int64_t n, int64_t k,
+                              const void * a, const void * b, float * c) {
+    const ggml_type_traits * traits = ggml_get_type_traits(type);
     struct ggml_tensor src1 = {};
-    src1.type  = T;
+    src1.type  = type;
     src1.ne[0] = k;
     src1.ne[1] = m;
     src1.ne[2] = 1;
@@ -6559,10 +6559,10 @@ static void ggml_call_mul_mat(ggml_type T, const ggml_compute_params * params, i
     src1.nb[1] = k * traits->type_size;
     src1.nb[2] = src1.nb[1];
     src1.nb[3] = src1.nb[2];
-    src1.data  = a;
+    src1.data  = (void *)a;
 
     struct ggml_tensor src0 = {};
-    src0.type  = T;
+    src0.type  = type;
     src0.ne[0] = k;
     src0.ne[1] = n;
     src0.ne[2] = 1;
@@ -6571,7 +6571,7 @@ static void ggml_call_mul_mat(ggml_type T, const ggml_compute_params * params, i
     src0.nb[1] = k * traits->type_size;
     src0.nb[2] = src0.nb[1];
     src0.nb[3] = src0.nb[2];
-    src0.data  = b;
+    src0.data  = (void *)b;
 
     struct ggml_tensor dst = {};
     dst.ne[0] = n;
@@ -6598,6 +6598,7 @@ static void ggml_compute_forward_conv_2d_impl(const ggml_compute_params * params
                                               ggml_type                   kernel_type) {
 
     GGML_ASSERT(ggml_is_contiguous(kernel));
+    GGML_ASSERT(kernel_type == GGML_TYPE_F16 || kernel_type == GGML_TYPE_F32);
     GGML_ASSERT(kernel->type == kernel_type);
 
     const ggml_type_traits * traits = ggml_get_type_traits(kernel_type);
@@ -6620,9 +6621,9 @@ static void ggml_compute_forward_conv_2d_impl(const ggml_compute_params * params
     const int64_t dst_w = dst->ne[0];
     const int64_t dst_h = dst->ne[1];
 
-    float * src_data = (float*) src->data;
-    void  * knl_data = kernel->data;
-    float * dst_data = (float*) dst->data;
+    const float * src_data = (float *) src->data;
+    void  * knl_data       = kernel->data;
+    float * dst_data       = (float *) dst->data;
 
     const int64_t knl_n           = knl_w * knl_h * c_in;
     const int64_t patch_total     = dst->ne[3] * dst_w * dst_h;
@@ -6676,7 +6677,7 @@ static void ggml_compute_forward_conv_2d_impl(const ggml_compute_params * params
                         if (kernel_type == GGML_TYPE_F32) {
                             *(float *) element_ptr = src_val;
                         } else if (kernel_type == GGML_TYPE_F16) {
-                            *(ggml_fp16_t *) element_ptr = GGML_FP32_TO_FP16(src_val);
+                            *(ggml_fp16_t *) element_ptr = GGML_CPU_FP32_TO_FP16(src_val);
                         }
                     }
                 }

ggml/src/ggml.c

@@ -987,7 +987,7 @@ static const char * GGML_OP_NAME[GGML_OP_COUNT] = {
     "GLU",
 };
 
-static_assert(GGML_OP_COUNT == 85, "GGML_OP_COUNT != 85");
+static_assert(GGML_OP_COUNT == 86, "GGML_OP_COUNT != 86");
 
 static const char * GGML_OP_SYMBOL[GGML_OP_COUNT] = {
     "none",
@@ -1087,7 +1087,7 @@ static const char * GGML_OP_SYMBOL[GGML_OP_COUNT] = {
     "glu(x)",
 };
 
-static_assert(GGML_OP_COUNT == 85, "GGML_OP_COUNT != 85");
+static_assert(GGML_OP_COUNT == 86, "GGML_OP_COUNT != 86");
 
 static_assert(GGML_OP_POOL_COUNT == 2, "GGML_OP_POOL_COUNT != 2");