You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
with spv1.5 (in glslc or glslvalidator) results in OpExecutionMode LocalSize 128 1 1 being emitted, which seems fine.
However compiling it with spv1.6, it results in OpExecutionModeId LocalSizeId 128 1 1.
If I understand correctly, LocalSizeId is used, when you want to use specialization constants to specify the Workgroup Size. The numbers behind LocalSizeId are the Ids used to identify the specialization constants. This doesn't make any sense, because I explicitely specified local_size_x with a constant size (not an Id) instead of local_size_x_id.
If I understood the comments, LocalSize with a constant was to be deprecated in 1.6 (#2934 (comment)), but later was not deprecated (#3351 (comment)). This pullrequest then fixed LocalSizeId to be only created, when needed, in compute shaders. Other shaders were not touched.
This issue (google/shaderc#1414) might also be related to this (and I also commented there).
The text was updated successfully, but these errors were encountered:
Compiling this shader
with spv1.5 (in glslc or glslvalidator) results in
OpExecutionMode LocalSize 128 1 1
being emitted, which seems fine.However compiling it with spv1.6, it results in
OpExecutionModeId LocalSizeId 128 1 1
.If I understand correctly, LocalSizeId is used, when you want to use specialization constants to specify the Workgroup Size. The numbers behind LocalSizeId are the Ids used to identify the specialization constants. This doesn't make any sense, because I explicitely specified
local_size_x
with a constant size (not an Id) instead oflocal_size_x_id
.If I understood the comments, LocalSize with a constant was to be deprecated in 1.6 (#2934 (comment)), but later was not deprecated (#3351 (comment)). This pullrequest then fixed LocalSizeId to be only created, when needed, in compute shaders. Other shaders were not touched.
This issue (google/shaderc#1414) might also be related to this (and I also commented there).
The text was updated successfully, but these errors were encountered: