8 Commits

Author SHA1 Message Date
Justin Bogner
aa07f92210
[DirectX][SPIRV] Consistent names for HLSL resource intrinsics (#120466)
Rename HLSL resource-related intrinsics to be consistent with the naming
conventions discussed in [wg-hlsl:0014].

This is an entirely mechanical change, consisting of the following
commands and automated formatting.

```sh
git grep -l handle.fromBinding | xargs perl -pi -e \
  's/(dx|spv)(.)handle.fromBinding/$1$2resource$2handlefrombinding/g'
git grep -l typedBufferLoad_checkbit | xargs perl -pi -e \
  's/(dx|spv)(.)typedBufferLoad_checkbit/$1$2resource$2loadchecked$2typedbuffer/g'
git grep -l typedBufferLoad | xargs perl -pi -e \
  's/(dx|spv)(.)typedBufferLoad/$1$2resource$2load$2typedbuffer/g'
git grep -l typedBufferStore | xargs perl -pi -e \
  's/(dx|spv)(.)typedBufferStore/$1$2resource$2store$2typedbuffer/g'
git grep -l bufferUpdateCounter | xargs perl -pi -e \
  's/(dx|spv)(.)bufferUpdateCounter/$1$2resource$2updatecounter/g'
git grep -l cast_handle | xargs perl -pi -e \
  's/(dx|spv)(.)cast.handle/$1$2resource$2casthandle/g'
```

[wg-hlsl:0014]: https://github.com/llvm/wg-hlsl/blob/main/proposals/0014-consistent-naming-for-dx-intrinsics.md
2024-12-19 12:17:21 -07:00
Justin Bogner
3eca15cbb9
[DirectX] Split resource info into type and binding info. NFC (#119773)
This splits the DXILResourceAnalysis pass into TypeAnalysis and
BindingAnalysis passes. The type analysis pass is made immutable and
populated lazily so that it can be used earlier in the pipeline without
needing to carefully maintain the invariants of the binding analysis.

Fixes #118400
2024-12-18 09:02:28 -07:00
Justin Bogner
482237e884
[DirectX] Get resource information via TargetExtType (#119772)
Instead of storing an auxilliary structure with the information from the
DXIL resource target extension types duplicated, access the information
that we can via the type itself.

This also means we need to handle some of the target extension types we
haven't fully defined yet, like Texture and CBuffer. For now we make an
educated guess to what those should look like based on llvm/wg-hlsl#76,
and we can update them fairly easily when we've defined them more
thoroughly.

First part of #118400
2024-12-16 16:04:25 -07:00
Justin Bogner
782bc4f669
[DXIL][Analysis] Uniquify duplicate resources in DXILResourceAnalysis
If a resources is used multiple times, we should only have one resource record
for it. This comes up most prominantly with arrays of resources like so:

```hlsl
RWBuffer<float4> BufferArray[10] : register(u0, space4);
RWBuffer<float4> B1 = BufferArray[0];
RWBuffer<float4> B2 = BufferArray[SomeIndex];
RWBuffer<float4> B3 = BufferArray[3];
```

In this case, there's only one resource, but we'll generate 3 different
`dx.handle.fromBinding` calls to access different slices.

Note that this adds some API that won't be used until #104447 later in the
stack. Trying to avoid that results in unnecessary churn.

Fixes #105143

Pull Request: https://github.com/llvm/llvm-project/pull/105602
2024-08-23 12:06:53 -07:00
Justin Bogner
1a2a18fdc0
[DXIL][Analysis] Update test to match comment. NFC (#105409)
The mismatch between the comment on this test and the test itself was
pointed out in
https://github.com/llvm/llvm-project/pull/100699#discussion_r1715835841,
but apparently I failed to actually commit the fix.
2024-08-20 15:09:38 -07:00
Justin Bogner
51ede55ee2
Re-Apply "[DXIL][Analysis] Implement enough of DXILResourceAnalysis for buffers" (#104517)
Some build configs allow `llvm_unreachable` in a constexpr context, but
not all, so these functions that map a fully covered enum to a string
can't be constexpr. This version fixes that by dropping constexpr from
those functions.

This reverts commit fcc318ff7960d7de8cbac56eb4f32b44b5261677, reapplying
28d577ecefa1557f5dea5566bf33b885c563d14b.

Original message follows:

This implements the DXILResourceAnalysis pass for `dx.TypedBuffer` and
`dx.RawBuffer` types. This should be sufficient to lower
`dx.handle.fromBinding` for this set of types, but it leaves a number of
TODOs around for other resource types.

This also includes a straightforward `print` method in `ResourceInfo` to
make the analysis testable. This is deliberately different than the
printer in `lib/Target/DirectX/DXILResource.cpp`, which attempts to
print bindings in a format compatible with the comments `dxc` prints. We
will eventually want to make that functionality driven by this analysis
pass, but it isn't sufficient for testing so we need both.
2024-08-16 11:36:41 -07:00
Mehdi Amini
fcc318ff79
Revert "[DXIL][Analysis] Implement enough of DXILResourceAnalysis for buffers" (#104504)
Reverts llvm/llvm-project#100699

This broke a few bots unfortunately.
2024-08-15 22:48:43 +02:00
Justin Bogner
28d577ecef
[DXIL][Analysis] Implement enough of DXILResourceAnalysis for buffers
This implements the DXILResourceAnalysis pass for `dx.TypedBuffer` and
`dx.RawBuffer` types. This should be sufficient to lower
`dx.handle.fromBinding` for this set of types, but it leaves a number
of TODOs around for other resource types.

This also includes a straightforward `print` method in `ResourceInfo`
to make the analysis testable. This is deliberately different than the
printer in `lib/Target/DirectX/DXILResource.cpp`, which attempts to
print bindings in a format compatible with the comments `dxc` prints.
We will eventually want to make that functionality driven by this
analysis pass, but it isn't sufficient for testing so we need both.

Pull Request: https://github.com/llvm/llvm-project/pull/100699
2024-08-15 00:24:55 +03:00