[all-commits] [llvm/llvm-project] e408c7: [MLIR][XeGPU] Extend 8-bit load_nd support in XeVM...

[Sang Ik Lee via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: e408c75f8617b314a9cc30a19f61988004eecb85
      https://github.com/llvm/llvm-project/commit/e408c75f8617b314a9cc30a19f61988004eecb85
  Author: Sang Ik Lee <sang.ik.lee at intel.com>
  Date:   2026-06-10 (Wed, 10 Jun 2026)

  Changed paths:
    M mlir/lib/Conversion/XeGPUToXeVM/XeGPUToXeVM.cpp
    M mlir/test/Conversion/XeGPUToXeVM/loadstore_nd.mlir

  Log Message:
  -----------
  [MLIR][XeGPU] Extend 8-bit load_nd support in XeVM lowering (#201645)

2D block load on 8bit element type has a shape 32x16 supported by OpenCL
API
```
 void intel_sub_group_2d_block_read_transform_8b_32r16x1c(   // reads eight uints
    global void* base_address,
    int width, int height, int pitch, int2 coord, private uint* destination);
```
The API is for load with transform/VNNI request.
OpenCL does not provide a load API for the same vector type and no
transform request. But value returned is identical for this special
vector type. <32x16x"8b">
The PR adds support for this vector type with no transform request.



To unsubscribe from these emails, change your notification settings at https://github.com/llvm/llvm-project/settings/notifications