1// This section is included inside the Pipelines chapter (pipelines.txt) 2 3[[pipelines-shader-information]] 4== Pipeline Shader Information 5 6[open,refpage='vkGetShaderInfoAMD',desc='Get information about a shader in a pipeline',type='protos'] 7-- 8 9Information about a particular shader that has been compiled as part of a 10pipeline object can be extracted by calling: 11 12include::../api/protos/vkGetShaderInfoAMD.txt[] 13 14 * pname:device is the device that created pname:pipeline. 15 * pname:pipeline is the target of the query. 16 * pname:shaderStage identifies the particular shader within the pipeline 17 about which information is being queried. 18 * pname:infoType describes what kind of information is being queried. 19 * pname:pInfoSize is a pointer to a value related to the amount of data 20 the query returns, as described below. 21 * pname:pInfo is either NULL or a pointer to a buffer. 22 23If pname:pInfo is `NULL`, then the maximum size of the information that can: 24be retrieved about the shader, in bytes, is returned in pname:pInfoSize. 25Otherwise, pname:pInfoSize must: point to a variable set by the user to the 26size of the buffer, in bytes, pointed to by pname:pInfo, and on return the 27variable is overwritten with the amount of data actually written to 28pname:pInfo. 29 30If pname:pInfoSize is less than the maximum size that can: be retrieved by 31the pipeline cache, then at most pname:pInfoSize bytes will be written to 32pname:pInfo, and fname:vkGetShaderInfoAMD will return ename:VK_INCOMPLETE. 33 34Not all information is available for every shader and implementations may 35not support all kinds of information for any shader. 36When a certain type of information is unavailable, the function returns 37ename:VK_ERROR_FEATURE_NOT_PRESENT. 38 39If information is successfully and fully queried, the function will return 40ename:VK_SUCCESS. 41 42For ename:VK_SHADER_INFO_TYPE_STATISTICS_AMD, an instance of 43sname:VkShaderStatisticsInfoAMD will be written to the buffer pointed to by 44pname:pInfo. 45This structure will be populated with statistics regarding the physical 46device resources used by that shader along with other miscellaneous 47information and is described in further detail below. 48 49For ename:VK_SHADER_INFO_TYPE_DISASSEMBLY_AMD, pname:pInfo points to a UTF-8 50null-terminated string containing human-readable disassembly. 51The exact formatting and contents of the disassembly string are 52vendor-specific. 53 54The formatting and contents of all other types of information, including 55ename:VK_SHADER_INFO_TYPE_BINARY_AMD, are left to the vendor and are not 56further specified by this extension. 57 58include::../validity/protos/vkGetShaderInfoAMD.txt[] 59-- 60 61[open,refpage='VkShaderStatisticsInfoAMD',desc='Statistical information about a particular shader within a pipeline',type='structs'] 62-- 63 64The sname:VkShaderStatisticsInfoAMD structure is defined as: 65 66include::../api/structs/VkShaderStatisticsInfoAMD.txt[] 67 68 * pname:shaderStageMask are the combination of logical shader stages 69 contained within this shader. 70 * pname:resourceUsage is an instance of slink:VkShaderResourceUsageAMD 71 describing internal physical device resources used by this shader. 72 * pname:numPhysicalVgprs is the maximum number of vector instruction 73 general-purpose registers (VGPRs) available to the physical device. 74 * pname:numPhysicalSgprs is the maximum number of scalar instruction 75 general-purpose registers (SGPRs) available to the physical device. 76 * pname:numAvailableVgprs is the maximum limit of VGPRs made available to 77 the shader compiler. 78 * pname:numAvailableSgprs is the maximum limit of SGPRs made available to 79 the shader compiler. 80 * pname:computeWorkGroupSize is the local workgroup size of this shader in 81 { X, Y, Z } dimensions. 82 83Some implementations may merge multiple logical shader stages together in a 84single shader. 85In such cases, pname:shaderStageMask will contain a bitmask of all of the 86stages that are active within that shader. 87Consequently, if specifying those stages as input to 88flink:vkGetShaderInfoAMD, the same output information may: be returned for 89all such shader stage queries. 90 91The number of available VGPRs and SGPRs (pname:numAvailableVgprs and 92pname:numAvailableSgprs respectively) are the shader-addressable subset of 93physical registers that is given as a limit to the compiler for register 94assignment. 95These values may: further be limited by implementations due to performance 96optimizations where register pressure is a bottleneck. 97 98include::../validity/structs/VkShaderStatisticsInfoAMD.txt[] 99-- 100 101[open,refpage='VkShaderResourceUsageAMD',desc='Resource usage information about a particular shader within a pipeline',type='structs'] 102-- 103 104The sname:VkShaderResourceUsageAMD structure is defined as: 105 106include::../api/structs/VkShaderResourceUsageAMD.txt[] 107 108 * pname:numUsedVgprs is the number of vector instruction general-purpose 109 registers used by this shader. 110 * pname:numUsedSgprs is the number of scalar instruction general-purpose 111 registers used by this shader. 112 * pname:ldsSizePerLocalWorkGroup is the maximum local data store size per 113 work group in bytes. 114 * pname:ldsUsageSizeInBytes is the LDS usage size in bytes per work group 115 by this shader. 116 * pname:scratchMemUsageInBytes is the scratch memory usage in bytes by 117 this shader. 118 119include::../validity/structs/VkShaderResourceUsageAMD.txt[] 120-- 121