[SM6.10] Implement VectorAccumDescriptor Builtin#8448
Conversation
| @@ -33,7 +33,7 @@ target triple = "dxil-ms-dx" | |||
| @"\01?SharedArr@@3PAMA" = external addrspace(3) global [64 x float], align 4 | |||
|
|
|||
| define void @mainAS() { | |||
There was a problem hiding this comment.
These tests updates still need to be moved per #8315
but I'm still kicking that can down the road:)
tex3d
left a comment
tex3d
left a comment
There was a problem hiding this comment.
I think it looks good, with just a "Matrix" nit and a question about whether we should have included an alignment parameter.
| db_dxil_param(0, "v", "", ""), | ||
| db_dxil_param(2, "$o", "vector", "vector to accumulate"), | ||
| db_dxil_param(3, "res", "handle", "buffer to accumulate into"), | ||
| db_dxil_param(4, "i32", "offset", "starting offset in the buffer"), |
There was a problem hiding this comment.
I realize the spec proposal didn't include an alignment parameter, but I can't help but wonder if that was a mistake.
There was a problem hiding this comment.
I filed an issue on the hlsl-specs proposal here:
microsoft/hlsl-specs#868
Adding the alignment parameter to the DXIL op would of course imply adding the parameter to the built-in intrinsic in this implementation as well.
| 2147483678 LinAlgConvert Convert vector components from one interpretation to another | ||
| 2147483679 ReservedE0 reserved | ||
| 2147483680 ReservedE1 reserved | ||
| 2147483679 VectorAccumulateToDescriptor Accumulates given vector to the buffer at the given offset |
There was a problem hiding this comment.
For what it's worth, as you pointed out, the naming is inconsistent here. I actually had the LinAlg prefix in a commit that didn't make it into the spec PR
4 participants
Implements VectorAccumulateToDescriptor builtin and updates the header implementation.
Fixes #8416
Co-Authored-By: @pow2clk