On 8/19/21 10:54 AM, Henri Verbeet wrote:
On Tue, 17 Aug 2021 at 19:39, Zebediah Figura zfigura@codeweavers.com wrote:
@@ -1272,5 +1272,5 @@ int hlsl_emit_dxbc(struct hlsl_ctx *ctx, struct hlsl_ir_function_decl *entry_fun if (ctx->profile->major_version < 4) return hlsl_sm1_write(ctx, entry_func, out); else
return VKD3D_ERROR_NOT_IMPLEMENTED;
}return hlsl_sm4_write(ctx, entry_func, out);
It's perhaps worth pointing out that there's an interaction with the target type here as well. Shader model 1-3 bytecode can either be output as-is, which is the legacy D3D bytecode format, or inside an "Aon9" section embedded inside a DXBC container. Neither of those is technically dxbc-tpf, but it's also possible to output DXBC containers that contain both TPF and "Aon9" sections.
I was assuming TPF referred to both SM1 and SM4 formats, since they are tokenized. What's the naming convention then? and what is the API convention?