don't use shared pointer for prealloc_y_last_pipeline_used

Author: jeffbolznv

ggml/src/ggml-vulkan/ggml-vulkan.cpp

@@ -1194,8 +1194,8 @@ struct ggml_backend_vk_context {
     bool almost_ready_fence_pending {};
 
     // Cache most recent tensor that was converted into prealloc_y, and what pipeline it used to convert.
-    vk_pipeline prealloc_y_last_pipeline_used;
-    const ggml_tensor *prealloc_y_last_tensor_used;
+    vk_pipeline_struct * prealloc_y_last_pipeline_used {};
+    const ggml_tensor * prealloc_y_last_tensor_used {};
 
     vk_buffer buffer_pool[MAX_VK_BUFFERS];
 
@@ -5655,18 +5655,18 @@ static void ggml_vk_mul_mat_q_f16(ggml_backend_vk_context * ctx, vk_context& sub
         ggml_vk_dispatch_pipeline(ctx, subctx, to_fp16_vk_0, { vk_subbuffer{ d_Qx, qx_buf_offset, qx_sz * ne02 * ne03 }, vk_subbuffer{ d_X, 0, x_sz * ne02 * ne03 } }, pc, { (uint32_t)(x_ne * ne02 * ne03), 1, 1});
     }
     if (y_non_contig) {
-        if (ctx->prealloc_y_last_pipeline_used != to_fp16_vk_1 ||
+        if (ctx->prealloc_y_last_pipeline_used != to_fp16_vk_1.get() ||
             ctx->prealloc_y_last_tensor_used != src1) {
             ggml_vk_cpy_to_contiguous(ctx, subctx, to_fp16_vk_1, src1, { d_Qy, qy_buf_offset, VK_WHOLE_SIZE }, { d_Y, 0, VK_WHOLE_SIZE });
-            ctx->prealloc_y_last_pipeline_used = to_fp16_vk_1;
+            ctx->prealloc_y_last_pipeline_used = to_fp16_vk_1.get();
             ctx->prealloc_y_last_tensor_used = src1;
         }
     }
     if (quantize_y) {
-        if (ctx->prealloc_y_last_pipeline_used != to_q8_1 ||
+        if (ctx->prealloc_y_last_pipeline_used != to_q8_1.get() ||
             ctx->prealloc_y_last_tensor_used != src1) {
             ggml_vk_quantize_q8_1(ctx, subctx, { d_Qy, qy_buf_offset, VK_WHOLE_SIZE }, { d_Y, 0, VK_WHOLE_SIZE }, y_ne * ne12 * ne13);
-            ctx->prealloc_y_last_pipeline_used = to_q8_1;
+            ctx->prealloc_y_last_pipeline_used = to_q8_1.get();
             ctx->prealloc_y_last_tensor_used = src1;
         }
     }
@@ -5843,10 +5843,10 @@ static void ggml_vk_mul_mat_vec_q_f16(ggml_backend_vk_context * ctx, vk_context&
     }
     if (y_non_contig) {
         GGML_ASSERT(y_sz == ggml_type_size(src1->type) * y_ne);
-        if (ctx->prealloc_y_last_pipeline_used != to_fp16_vk_1 ||
+        if (ctx->prealloc_y_last_pipeline_used != to_fp16_vk_1.get() ||
             ctx->prealloc_y_last_tensor_used != src1) {
             ggml_vk_cpy_to_contiguous(ctx, subctx, to_fp16_vk_1, src1, { d_Qy, qy_buf_offset, VK_WHOLE_SIZE }, { d_Y, 0, VK_WHOLE_SIZE });
-            ctx->prealloc_y_last_pipeline_used = to_fp16_vk_1;
+            ctx->prealloc_y_last_pipeline_used = to_fp16_vk_1.get();
             ctx->prealloc_y_last_tensor_used = src1;
         }
     }
@@ -6278,10 +6278,10 @@ static void ggml_vk_mul_mat_id_q_f16(ggml_backend_vk_context * ctx, vk_context&
             { vk_subbuffer{ d_Qx, qx_buf_offset, qx_sz * ne02 * ne03 }, vk_subbuffer{ d_X, 0, x_sz * ne02 * ne03 } }, pc, { (uint32_t)(x_ne * ne02 * ne03), 1, 1});
     }
     if (y_non_contig) {
-        if (ctx->prealloc_y_last_pipeline_used != to_fp16_vk_1 ||
+        if (ctx->prealloc_y_last_pipeline_used != to_fp16_vk_1.get() ||
             ctx->prealloc_y_last_tensor_used != src1) {
             ggml_vk_cpy_to_contiguous(ctx, subctx, to_fp16_vk_1, src1, { d_Qy, qy_buf_offset, VK_WHOLE_SIZE }, { d_Y, 0, VK_WHOLE_SIZE });
-            ctx->prealloc_y_last_pipeline_used = to_fp16_vk_1;
+            ctx->prealloc_y_last_pipeline_used = to_fp16_vk_1.get();
             ctx->prealloc_y_last_tensor_used = src1;
         }
     }
@@ -6471,10 +6471,10 @@ static void ggml_vk_mul_mat_vec_id_q_f16(ggml_backend_vk_context * ctx, vk_conte
     }
     if (y_non_contig) {
         GGML_ASSERT(y_sz == ggml_type_size(src1->type) * y_ne);
-        if (ctx->prealloc_y_last_pipeline_used != to_fp16_vk_1 ||
+        if (ctx->prealloc_y_last_pipeline_used != to_fp16_vk_1.get() ||
             ctx->prealloc_y_last_tensor_used != src1) {
             ggml_vk_cpy_to_contiguous(ctx, subctx, to_fp16_vk_1, src1, { d_Qy, qy_buf_offset, VK_WHOLE_SIZE }, { d_Y, 0, VK_WHOLE_SIZE });
-            ctx->prealloc_y_last_pipeline_used = to_fp16_vk_1;
+            ctx->prealloc_y_last_pipeline_used = to_fp16_vk_1.get();
             ctx->prealloc_y_last_tensor_used = src1;
         }
     }
@@ -10383,7 +10383,7 @@ static void ggml_vk_cleanup(ggml_backend_vk_context * ctx) {
     ggml_vk_destroy_buffer(ctx->prealloc_x);
     ggml_vk_destroy_buffer(ctx->prealloc_y);
     ggml_vk_destroy_buffer(ctx->prealloc_split_k);
-    ctx->prealloc_y_last_pipeline_used = {};
+    ctx->prealloc_y_last_pipeline_used = nullptr;
 
     for (auto& buffer : ctx->buffer_pool) {
         ggml_vk_destroy_buffer(buffer);
@@ -10932,7 +10932,7 @@ static ggml_status ggml_backend_vk_graph_compute(ggml_backend_t backend, ggml_cg
         compute_ctx->s->buffer.writeTimestamp(vk::PipelineStageFlagBits::eAllCommands, ctx->device->query_pool, 0);
     }
 
-    ctx->prealloc_y_last_pipeline_used = {};
+    ctx->prealloc_y_last_pipeline_used = nullptr;
     ctx->prealloc_y_last_tensor_used = nullptr;
 
     // Submit after enough work has accumulated, to overlap CPU cmdbuffer generation with GPU execution.