On 9/25/20 7:58 AM, Henri Verbeet wrote:
On Fri, 25 Sep 2020 at 03:00, Zebediah Figura zfigura@codeweavers.com wrote:
+int vkd3d_shader_compile(const struct vkd3d_shader_compile_info *compile_info,
struct vkd3d_shader_code *out, char **messages)+{
- struct vkd3d_shader_message_context message_context;
- int ret;
- TRACE("compile_info %p, out %p, messages %p.\n", compile_info, out, messages);
- if (messages)
*messages = NULL;- if ((ret = vkd3d_shader_validate_compile_info(compile_info)) < 0)
return ret;- if (!vkd3d_shader_message_context_init(&message_context, compile_info->log_level, compile_info->source_name))
return VKD3D_ERROR_OUT_OF_MEMORY;- if (compile_info->source_type == VKD3D_SHADER_SOURCE_DXBC_TPF
&& compile_info->target_type == VKD3D_SHADER_TARGET_SPIRV_BINARY)ret = compile_dxbc_to_spirv(compile_info, out, &message_context);- else if (compile_info->source_type == VKD3D_SHADER_SOURCE_HLSL
&& compile_info->target_type == VKD3D_SHADER_TARGET_DXBC_TPF)ret = compile_hlsl_to_dxbc(compile_info, out, &message_context);I guess this is a rebase glitch, but VKD3D_SHADER_SOURCE_DXBC_TPF also supports VKD3D_SHADER_TARGET_SPIRV_TEXT as target. (And soon VKD3D_SHADER_TARGET_D3D_ASM.) I'd suggest to simply split based on the source type here, and then split on target type if needed in a hypothetical compile_dxbc().
Yes, this commit is pretty old...
That seems like a better structure; thanks for the suggestion.
There's probably also room for splitting some of the changes here into their own commits. For example, the "message_context" change in the vkd3d_shader_scan_context structure. (Although it's not immediately clear to me why that would be required at this stage; I guess it allows creating the message context in a single place for both compile_dxbc_to_spirv() and future compile_hlsl_to_dxbc().)
I don't think it's necessary, but in any case it does seem like a nice bit of deduplication.