vulkan: use prealloc intermediate reuse for mmvq path

Author: 0cc4m

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

@@ -4638,7 +4638,7 @@ static vk_pipeline ggml_vk_get_dequantize_mul_mat_vec(ggml_backend_vk_context *
 
     // heuristic to choose workgroup size
     uint32_t dmmv_wg = DMMV_WG_SIZE_SUBGROUP;
-    if (ctx->device->vendor_id == VK_VENDOR_ID_NVIDIA || ctx->device->vendor_id == VK_VENDOR_ID_INTEL) {
+    if (ctx->device->vendor_id == VK_VENDOR_ID_NVIDIA) {
         // Prefer larger workgroups when M is small, to spread the work out more
         // and keep more SMs busy.
         // q6_k seems to prefer small workgroup size even for "medium" values of M.
@@ -4654,7 +4654,7 @@ static vk_pipeline ggml_vk_get_dequantize_mul_mat_vec(ggml_backend_vk_context *
     }
 
     if (b_type == GGML_TYPE_Q8_1) {
-        return ctx->device->pipeline_dequant_mul_mat_vec_q8_1_f32[DMMV_WG_SIZE_SUBGROUP][a_type][num_cols-1];
+        return ctx->device->pipeline_dequant_mul_mat_vec_q8_1_f32[dmmv_wg][a_type][num_cols-1];
     }
 
     return b_type == GGML_TYPE_F32 ? ctx->device->pipeline_dequant_mul_mat_vec_f32_f32[dmmv_wg][a_type][num_cols-1] : ctx->device->pipeline_dequant_mul_mat_vec_f16_f32[dmmv_wg][a_type][num_cols-1];
@@ -6114,7 +6114,12 @@ static void ggml_vk_mul_mat_vec_q_f16(ggml_backend_vk_context * ctx, vk_context&
         }
     }
     if (quantize_y) {
-        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, true);
+        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, true);
+            ctx->prealloc_y_last_pipeline_used = to_q8_1.get();
+            ctx->prealloc_y_last_tensor_used = src1;
+        }
     }
 
     // For batch_n, the A matrix is the same for each batch, and B/D use the row stride as the batch stride