diff options
| author | Matt Arsenault <Matthew.Arsenault@amd.com> | 2016-02-12 02:40:47 +0000 |
|---|---|---|
| committer | Matt Arsenault <Matthew.Arsenault@amd.com> | 2016-02-12 02:40:47 +0000 |
| commit | 24ee0785dd1bb1a2f5cb6cb398d697080a92803b (patch) | |
| tree | c7ec2cfa1749a3f65c017c05897d620f2ae7207b /llvm/lib/Target/AMDGPU/AMDKernelCodeT.h | |
| parent | 9d14e4bb1523f9d29ea1f26af7d0294013682cea (diff) | |
| download | bcm5719-llvm-24ee0785dd1bb1a2f5cb6cb398d697080a92803b.tar.gz bcm5719-llvm-24ee0785dd1bb1a2f5cb6cb398d697080a92803b.zip | |
AMDGPU: Set element_size in private resource descriptor
Introduce a subtarget feature for this, and leave the default with
the current behavior which assumes up to 16-byte loads/stores can
be used. The field also seems to have the ability to be set to 2 bytes,
but I'm not sure what that would be used for.
llvm-svn: 260651
Diffstat (limited to 'llvm/lib/Target/AMDGPU/AMDKernelCodeT.h')
| -rw-r--r-- | llvm/lib/Target/AMDGPU/AMDKernelCodeT.h | 9 |
1 files changed, 9 insertions, 0 deletions
diff --git a/llvm/lib/Target/AMDGPU/AMDKernelCodeT.h b/llvm/lib/Target/AMDGPU/AMDKernelCodeT.h index a9ba60c8cba..425261c15f9 100644 --- a/llvm/lib/Target/AMDGPU/AMDKernelCodeT.h +++ b/llvm/lib/Target/AMDGPU/AMDKernelCodeT.h @@ -44,6 +44,15 @@ enum amd_code_version_t { AMD_CODE_VERSION_MINOR = 1 }; +// Sets val bits for specified mask in specified dst packed instance. +#define AMD_HSA_BITS_SET(dst, mask, val) \ + dst &= (~(1 << mask ## _SHIFT) & ~mask); \ + dst |= (((val) << mask ## _SHIFT) & mask) + +// Gets bits for specified mask from specified src packed instance. +#define AMD_HSA_BITS_GET(src, mask) \ + ((src & mask) >> mask ## _SHIFT) \ + /// The values used to define the number of bytes to use for the /// swizzle element size. enum amd_element_byte_size_t { |
