Shows entry point inputs & outputs
Decodes "dx.op.loadInput" and "dx.op.StoreOutput" to show the input/output name
Change identifier prefix to "_" instead of "%"
Print floats in more natural format using %#g
Print type before assignments
* These will not be in the reflection data, so we can't try to process them and
fill out their variables as there won't be anything to map to.
* Signature elements which are 'unused' by not being referenced but are present
in a pre-1.4 SPIR-V module will be considered used still.
Capture the mappings into a type usage tree
Update elements of the tree as overlapping mappings are encountered, sub-elements of a previously mapped variable i.e. single component of a vector, single array element, member in a structure.
After processing all mappings convert the finalized type usage tree into source variable mappings.
Loop over the scope downwards (parent -> child) then inside that loop iterate upwards from the current scope (child -> parent) to detect mappings which might be superseded
Generated using DirectShaderCompile python file "utils/hct/hctdb.py"
for inst in db.instr:
out = inst.name
out += "("
for i in range(2,len(inst.ops)):
if i > 2:
out += ","
op = inst.ops[i]
out += op.name
out += ")"
print(out)
Disable dxcStyleFormatting when making RD disassembly
Change float printing to use "%#g" when dxcStyleFormatting is disabled
Before (DXC style)
%7 = fmul fast float %4, 2.000000e+00
%14 = fcmp fast olt float %13, 1.000000e-06
After (RD style)
%7 = fmul fast float %4, 2.00000;
%14 = fcmp fast olt float %13, 1.00000e-06;
First pass to settle the IDs common between disassembly styles
Second pass to display the disassembly
Divide existing MakeDXCDisassemblyString() method into helper methods to be shared
Currently not exposed to the UI and code uses the existing DXC compatibile DXIL disassembly
Add bool parameter to DXBC::GetDisassembly()
Add bool parameter to DXIL::GetDisassembly()
Add void DXIL::MakeDXCDisassemblyString()
Add void DXIL::MakeRDDisassemblyString()
Fixes debug errors "exceeds end of the virtual address range of Resource" i.e.
D3D12 ERROR: ID3D12CommandList::IASetVertexBuffers: 0x0000023B9A746190 + SizeInBytes - 1 (0x000000010c36bffe) exceeds end of the virtual address range of Resource (0x0000023B9A746190:'PostVS vsoutBuffer for 162', GPU VA Range: 0x000000000c36c000 - 0x000000000c36c29f). [ EXECUTION ERROR #726: SET_VERTEX_BUFFERS_INVALID]
D3D12 ERROR: ID3D12CommandList::IASetIndexBuffer: pDesc->BufferLocation + SizeInBytes - 1 (0x000000010c36affe) exceeds end of the virtual address range of Resource (0x0000023B9A7479B0:'PostVS idxBuf for 162', GPU VA Range: 0x000000000c36b000 - 0x000000000c36b00b). [ STATE_CREATION ERROR #725: SET_INDEX_BUFFER_INVALID]
Use D3D12_RESOURCE_DESC Flags to detect depth target (D3D12_RESOURCE_FLAG_ALLOW_DEPTH_STENCIL).
Use GetDepthTypedFormat() to get correct depth copy format.
Use Pixel History IsDepthFormat() internal method instead of using DXGI_FORMAT helper function "IsDepthFormat".
Fixes issue of pixel history on typeless target cast to UNORM and used as depth i.e. R16_TYPLESS, cast to UNORM used as a shadow map
Use the same logic as TextureViewer to determine if the target is a depth target
Fixes issue of pixel history on typeless target cast to UNORM and used as depth i.e. shadow map
The buffer containing the `uint32` draw count value for indirect draws with count was not marked as used for the respective events for the following API calls: glMultiDrawArraysIndirectCount, glMultiDrawElementsIndirectCount, vkCmdDrawIndirectCount, vkCmdDrawIndexedIndirectCount, vkCmdDrawMeshTasksIndirectCountEXT.
The renderdoccmd android build assumes all files in '${ANDROID_SDK_ROOT_PATH}/platforms' are valid 'android-<version>.jar' platform libraries. Unfortunately because MacOS litters its directories with metadata files called .DS_Store, that file actually wins this sorting logic and is incorrectly selected.
This fixes the lib search logic to only consider android libraries.
Jake Turner