On Wed, Jun 5, 2013 at 10:11 PM, Roland Scheidegger <<a href="mailto:sroland@vmware.com" target="_blank">sroland@vmware.com</a>> wrote: <div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"> Am 06.06.2013 04:16, schrieb Roland Scheidegger: <div><div class="h5">> Am 06.06.2013 02:52, schrieb Brian Paul: >> On 06/05/2013 05:44 PM, <a href="mailto:sroland@vmware.com">sroland@vmware.com</a> wrote: >>> From: Roland Scheidegger <<a href="mailto:sroland@vmware.com">sroland@vmware.com</a>> >>> >>> Also report if a shader writes the layer semantic >>> --- >>> src/gallium/auxiliary/draw/draw_context.c | 2 +- >>> src/gallium/auxiliary/tgsi/tgsi_scan.c | 5 +++++ >>> src/gallium/auxiliary/tgsi/tgsi_scan.h | 1 + >>> src/gallium/auxiliary/tgsi/tgsi_strings.c | 1 + >>> 4 files changed, 8 insertions(+), 1 deletion(-) >>> >>> diff --git a/src/gallium/auxiliary/draw/draw_context.c >>> b/src/gallium/auxiliary/draw/draw_context.c >>> index 58ce270..35063b9 100644 >>> --- a/src/gallium/auxiliary/draw/draw_context.c >>> +++ b/src/gallium/auxiliary/draw/draw_context.c >>> @@ -548,7 +548,7 @@ draw_get_shader_info(const struct draw_context *draw) >>> * function to find those attributes. >>> * >>> * -1 is returned if the attribute is not found since this is >>> - * an undefined situtation. Note, that zero is valid and can >>> + * an undefined situation. Note, that zero is valid and can >>> * be used by any of the attributes, because position is not >>> * required to be attribute 0 or even at all present. >>> */ >>> diff --git a/src/gallium/auxiliary/tgsi/tgsi_scan.c >>> b/src/gallium/auxiliary/tgsi/tgsi_scan.c >>> index 0230267..d331257 100644 >>> --- a/src/gallium/auxiliary/tgsi/tgsi_scan.c >>> +++ b/src/gallium/auxiliary/tgsi/tgsi_scan.c >>> @@ -217,6 +217,11 @@ tgsi_scan_shader(const struct tgsi_token *tokens, >>> TGSI_SEMANTIC_VIEWPORT_INDEX) { >>> info->writes_viewport_index = TRUE; >>> } >>> + if (procType == TGSI_PROCESSOR_GEOMETRY && >>> + fulldecl->Semantic.Name == >>> + TGSI_SEMANTIC_LAYER) { >>> + info->writes_layer = TRUE; >>> + } >>> } >>> >>> } >>> diff --git a/src/gallium/auxiliary/tgsi/tgsi_scan.h >>> b/src/gallium/auxiliary/tgsi/tgsi_scan.h >>> index 676abf0..a5b7024 100644 >>> --- a/src/gallium/auxiliary/tgsi/tgsi_scan.h >>> +++ b/src/gallium/auxiliary/tgsi/tgsi_scan.h >>> @@ -76,6 +76,7 @@ struct tgsi_shader_info >>> boolean pixel_center_integer; >>> boolean color0_writes_all_cbufs; >>> boolean writes_viewport_index; >>> + boolean writes_layer; >>> >>> unsigned num_written_clipdistance; >>> /** >>> diff --git a/src/gallium/auxiliary/tgsi/tgsi_strings.c >>> b/src/gallium/auxiliary/tgsi/tgsi_strings.c >>> index 6abf927..625107c 100644 >>> --- a/src/gallium/auxiliary/tgsi/tgsi_strings.c >>> +++ b/src/gallium/auxiliary/tgsi/tgsi_strings.c >>> @@ -80,6 +80,7 @@ const char *tgsi_semantic_names[TGSI_SEMANTIC_COUNT] = >>> "TEXCOORD", >>> "PCOORD", >>> "VIEWPORT_INDEX" >>> + "LAYER" >>> }; >>> >>> const char *tgsi_texture_names[TGSI_TEXTURE_COUNT] = >>> >> >> Hmm, there's a STATIC_ASSERT later in the file which should have >> detected a missing string in the array. I bet we need to remove the >> TGSI_SEMANTIC_COUNT from the array size declaration to make that work. >> Can you check into that. > That is odd indeed. The patch was actually buggy (no comma). All of > these name arrays there are indeed fixed size which looks like defeating > the purpose of the STATIC_ASSERT. Even if I change this though it never > triggers no matter what's in there. Hmm. > </div></div>So this can't work. The size of these arrays is declared elsewhere, what we put in in this file here doesn't really matter for the Elements used in STATIC_ASSERT at all. And we can't just make the size dynamic in there in tgsi_string.h as sizeof() used for Elements can't work with that. So I've no idea how to make this actually work. </blockquote><div> OK, I see the problem now. I'm going to post a patch which replaces the tgsi_file_names[] array with a function. I think this is nicer and it gives us a working static assertions. If you like it, I could create a bugzilla entry/task for someone to give the same treatment to the other tgsi string arrays. -Brian </div></div>