Repository navigation
[ET-VK][binary] Select binary ops with a specialization constant - #24717
Merged
Merged
Conversation
## Ulterior Motive
Preparing ET-VK binary ops for operator fusion, such as folding DSR's
`y + x * scale` residual into the preceding convolution. A fused shader needs a
shared implementation of every binary op it can apply, instead of a separate
shader per op.
## Rationale
**What**: Every binary op of a storage type and dtype now dispatches one shader.
The `op_type` specialization constant selects the op, and `should_broadcast`
selects between broadcast indexing and direct indexing. The op implementations
move to `binary_op.glslh` as `apply_binary_op()`. `q8ta_add` uses the same
header and constant.
**Why**: Fusion shaders can include `binary_op.glslh` to apply any binary op.
Generated binary shaders drop from 80 to 14. Equal-shape inputs skip the
broadcast index math.
## Details
- Shader variants
- **Before**: one per op, storage and dtype (`binary_add_texture3d_half`,
`binary_mul_buffer_float`, ...), each with the op expression baked in by
codegen.
- **After**: `binary_op_{texture3d,buffer}_{half,float,int32,uint8}` for
arithmetic, `swiglu` and bitwise ops (bitwise for uint8 only), and
`binary_cmp_{texture3d,buffer}_{half,float,int32}` for comparisons, which
write uint8.
- Op selection
- **Before**: the kernel name (`"binary_" + op + storage + dtype`).
- **After**: `op_type` (specialization constant 8), from `BinaryOpType` in
`BinaryOp.h`; values must match `binary_op.glslh`. The driver folds the
switch at pipeline creation; every op/flag combination checked compiles to
the same Adreno ISA as the old per-op shader.
- Equal-shape inputs
- **Before**: texture shaders always converted positions through tensor
indices; buffer shaders compared input metadata at runtime.
- **After**: `should_broadcast` (specialization constant 9) is 0 when `out`,
`in` and `other` share sizes and hashed layout, so all three share texel
positions or buffer indices. It is fixed when the graph is built, like the
existing packed-dim broadcast flags.
- q8ta add
- **Before**: `q8ta_add_buffer`, with `X + Y` baked in by codegen.
- **After**: `q8ta_binary_buffer`, which calls `apply_binary_op()` with
`op_type` (specialization constant 7). `add_q8ta_binary_node()` takes a
`BinaryOpType` instead of an op name; the enum moves to `BinaryOp.h` so
both files share it.
- Kernel names in profiles: `binary_op_*` / `binary_cmp_*` for every op, and
`q8ta_binary_buffer` for q8ta add; the op is no longer part of the name.
- Benchmark references, example configs and tests use the new shader names.
A fusion shader declares `op_type`, defines the operand types, and includes the
header:
```glsl
#define IN_T VEC4_T
#define OUT_T VEC4_T
#define ALPHA_T float
#define BINARY_OP_VEC4
#define BINARY_OP_IS_FLOAT
${layout_declare_spec_const(C, "int", "op_type", "0")}
#include "binary_op.glslh"
...
out_texel = apply_binary_op(conv_texel, residual_texel, alpha);
```
Differential Revision: [D130014628](https://our.internmc.facebook.com/intern/diff/D130014628/)
ghstack-source-id: 466197009
Pull-Request: #24695
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/24717
Note: Links to docs will display an error until the docs builds have been completed. ❌ 135 New Failures, 7 Unrelated Failures, 10 Unclassified FailuresAs of commit 4b2ce2c with merge base 728e13f ( NEW FAILURES - The following jobs have failed:
UNCLASSIFIED FAILURES - DrCI could not classify the following jobs because the workflow did not run on the merge base. The failures may be pre-existing on trunk or introduced by this PR:
FLAKY - The following jobs failed but were likely due to flakiness present on trunk:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
Bot
added
the
CLA Signed
SS-JIA
approved these changes
Nov 6, 2026
This branch had an error being deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This PR was created by the merge bot to help merge the original PR into the main branch.
ghstack PR number: #24695 by @SS-JIA
^ Please use this as the source of truth for the PR details, comments, and reviews
ghstack PR base: https://github.com/pytorch/executorch/tree/gh/SS-JIA/690/base
ghstack PR head: https://github.com/pytorch/executorch/tree/gh/SS-JIA/690/head
Merge bot PR base: https://github.com/pytorch/executorch/tree/gh/SS-JIA/689/orig
Merge bot PR head: https://github.com/pytorch/executorch/tree/gh/SS-JIA/690/orig
Differential Revision: D130014628
@diff-train-skip-merge