Define BFloat16 intrinsics #293
Conversation
|
Shouldn't |
|
|
eopXD
force-pushed
the
eopc/amendments-and-fixes
branch
from
68f125b to
c33f99f
Compare
Extracted reference as a separate adoc file. Signed-off-by: eop Chen <eop.chen@sifive.com>
Define (non-segment/segment) load/store intrinsics for bfloat16 values and also psuedo utility functions for manipulation across types. Signed-off-by: eop Chen <eop.chen@sifive.com>
Signed-off-by: eop Chen <eop.chen@sifive.com>
…it-commit-autogen-bf16-doc)
…ommit-autogen-bf16-test)
vfncvtbf16.f.f.w vd, vs2, vm vfwcvtbf16.f.f.v vd, vs2, vm
…it-commit-autogen-bf16-doc)
…ommit-autogen-bf16-test)
vfwmaccbf16.vv vd, vs1, vs2, vm vfwmaccbf16.vf vd, rs1, vs2, vm
…it-commit-autogen-bf16-doc)
…ommit-autogen-bf16-test)
…types Signed-off-by: eop Chen <eop.chen@sifive.com>
Signed-off-by: eop Chen <eop.chen@sifive.com>
Signed-off-by: eop Chen <eop.chen@sifive.com>
…it-commit-autogen-bf16-doc)
…ommit-autogen-bf16-test)
|
Rebase and addressed comments from @dzaima, thank you for pointing this out. |
It follows the interface defined here: riscv-non-isa/rvv-intrinsic-doc#293
|
I'll take over and proceed this patch! |
It follows the interface defined here: riscv-non-isa/rvv-intrinsic-doc#293
| @@ -0,0 +1,67 @@ | |||
| == References | |||
There was a problem hiding this comment.
Most of this file does not look bf16 specific. Only the last line is bf16. How did the references work before?
There was a problem hiding this comment.
Oh this was separated out of rvv-intrinsic-spec.adoc. I just hadn't gotten there yet.
NOTE: This PR currently is based upon #292
NOTE: These new intrinsics introduced are not part of the to-be-frozen v1.0 intrinsics specification. These new intrinsics will not be cherry-picked to the v1.0.x branch.
This pull request defines intrinsics for vector bfloat16 type manipulations and vector instructions that are introduced by the RISC-V BFloat16 instruction set extensions (
zvfbfminandzvfbfwma) [0].Intrinsics introduced for vector bfloat16 type manipulation:
vcreate,vundefined, ... etc (you may find the exact list under generated prototypes and its declarations underbfloat16_inst.py)New intrinsics are added to expose new instructions introduced:
vfwmaccbf16.vv vd, vs1, vs2, vmvfwmaccbf16.vf vd, rs1, vs2, vmvfncvtbf16.f.f.w vd, vs2, vmvfwcvtbf16.f.f.v vd, vs2, vm[0] https://github.com/riscv/riscv-bfloat16