I don't really like this... Bit I believe it is correct. The whole series (modulo validation) is Reviewed-by: Jason Ekstrand <<a href="mailto:jason@jlekstrand.net">jason@jlekstrand.net</a>> Let's cc stable on the whole thing as well. <div class="gmail_extra"> <div class="gmail_quote">On Jul 27, 2016 6:03 AM, "Emil Velikov" <<a href="mailto:emil.l.velikov@gmail.com">emil.l.velikov@gmail.com</a>> wrote: <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">From: Emil Velikov <<a href="mailto:emil.velikov@collabora.com">emil.velikov@collabora.com</a>> With version 1 of the Loader interface there is an internal/private symbol (vk_icdGetInstanceProcAddr) which is used to retrieve all the API from the Vulkan entrypoints from the ICD. Implying that exposing the Vulkan API is not recommended. Version 2 goes a step further explicitly forbiding the ICD from exposing Vulkan symbols (and adding a negotiation API) As a reference: - Nvidia 367.35 Missing negotiation API - version 1. Exposes only vk_icdGetInstanceProcAddr. - AMD <a href="tel:16.30.3.306809" value="+16303306809">16.30.3.306809</a> Have negotiation API - version 2, Exposes vk_icdGetInstanceProcAddr. Exposes a couple of Vulkan entry points - seems to be in violation with the spec. Cc: "12.0" <<a href="mailto:mesa-stable@lists.freedesktop.org">mesa-stable@lists.freedesktop.org</a>> Cc: Jason Ekstrand <<a href="mailto:jason@jlekstrand.net">jason@jlekstrand.net</a>> Cc: Christian König <<a href="mailto:christian.koenig@amd.com">christian.koenig@amd.com</a>> Signed-off-by: Emil Velikov <<a href="mailto:emil.velikov@collabora.com">emil.velikov@collabora.com</a>> --- Christian, can you ping/check with the relevant team if I'm missing something or the driver indeed goes against the spec ? Apart from the small reduction in binary size, we get rid of the ifuncs. The latter was never checked at configure time, so things could have gone very badly on platforms which lack support. --- src/intel/vulkan/anv_entrypoints_gen.py | 11 ----------- 1 file changed, 11 deletions(-) diff --git a/src/intel/vulkan/anv_entrypoints_gen.py b/src/intel/vulkan/anv_entrypoints_gen.py index a4922c2..6904d37 100644 --- a/src/intel/vulkan/anv_entrypoints_gen.py +++ b/src/intel/vulkan/anv_entrypoints_gen.py @@ -291,17 +291,6 @@ anv_resolve_entrypoint(uint32_t index) } """ -# Now output ifuncs and their resolve helpers for all entry points. The -# resolve helper calls resolve_entrypoint() with the entry point index, which -# lets the resolver look it up in the table. - -for type, name, args, num, h in entrypoints: - print_guard_start(name) - print "static void *resolve_%s(void) { return anv_resolve_entrypoint(%d); }" % (name, num) - print "%s vk%s%s\n __attribute__ ((ifunc (\"resolve_%s\"), visibility (\"default\")));\n" % (type, name, args, name) - print_guard_end(name) - - # Now generate the hash table used for entry point look up. This is a # uint16_t table of entry point indices. We use 0xffff to indicate an entry # in the hash table is empty. -- 2.9.0 </blockquote></div></div>