diff --git a/src/amd/vulkan/meta/radv_meta_fast_clear.c b/src/amd/vulkan/meta/radv_meta_fast_clear.c
index 77b5b1f9f95..4ccafa387ea 100644
--- a/src/amd/vulkan/meta/radv_meta_fast_clear.c
+++ b/src/amd/vulkan/meta/radv_meta_fast_clear.c
@@ -96,6 +96,26 @@ cleanup:
    return result;
 }
 
+static VkResult
+get_dcc_decompress_compute_pipeline(struct radv_device *device, VkPipeline *pipeline_out)
+{
+   struct radv_meta_state *state = &device->meta_state;
+   VkResult result = VK_SUCCESS;
+
+   mtx_lock(&state->mtx);
+   if (!state->fast_clear_flush.dcc_decompress_compute_pipeline) {
+      result = create_dcc_compress_compute(device);
+      if (result != VK_SUCCESS)
+         goto fail;
+   }
+
+   *pipeline_out = state->fast_clear_flush.dcc_decompress_compute_pipeline;
+
+fail:
+   mtx_unlock(&state->mtx);
+   return result;
+}
+
 static VkResult
 create_pipeline(struct radv_device *device, VkShaderModule vs_module_h, VkPipelineLayout layout)
 {
@@ -633,22 +653,21 @@ radv_decompress_dcc_compute(struct radv_cmd_buffer *cmd_buffer, struct radv_imag
    struct radv_meta_saved_state saved_state;
    struct radv_image_view load_iview = {0};
    struct radv_image_view store_iview = {0};
+   VkPipeline pipeline;
+   VkResult result;
+
+   result = get_dcc_decompress_compute_pipeline(device, &pipeline);
+   if (result != VK_SUCCESS) {
+      vk_command_buffer_set_error(&cmd_buffer->vk, result);
+      return;
+   }
 
    cmd_buffer->state.flush_bits |=
       radv_dst_access_flush(cmd_buffer, VK_PIPELINE_STAGE_2_COMPUTE_SHADER_BIT, VK_ACCESS_2_SHADER_READ_BIT, image);
 
-   if (!device->meta_state.fast_clear_flush.cmask_eliminate_pipeline) {
-      VkResult ret = radv_device_init_meta_fast_clear_flush_state_internal(device);
-      if (ret != VK_SUCCESS) {
-         vk_command_buffer_set_error(&cmd_buffer->vk, ret);
-         return;
-      }
-   }
-
    radv_meta_save(&saved_state, cmd_buffer, RADV_META_SAVE_DESCRIPTORS | RADV_META_SAVE_COMPUTE_PIPELINE);
 
-   radv_CmdBindPipeline(radv_cmd_buffer_to_handle(cmd_buffer), VK_PIPELINE_BIND_POINT_COMPUTE,
-                        device->meta_state.fast_clear_flush.dcc_decompress_compute_pipeline);
+   radv_CmdBindPipeline(radv_cmd_buffer_to_handle(cmd_buffer), VK_PIPELINE_BIND_POINT_COMPUTE, pipeline);
 
    for (uint32_t l = 0; l < vk_image_subresource_level_count(&image->vk, subresourceRange); l++) {
       uint32_t width, height;