986ee2466f14e23e0331f1a3b742b3eb07e783fe - sw/vec_iree - Git at Google

commit: 986ee2466f14e23e0331f1a3b742b3eb07e783fe
[log]
author: Cindy Liu <hcindyl@google.com>
Thu Jul 29 23:23:17 2021 -0700
committer: Cindy Liu <hcindyl@google.com>
Wed Aug 11 15:54:45 2021 -0700
tree: 0d8e8e685f52008284e56fb0d4a3e63a06976a6a
parent: 48f6baf58cab4921e340cf2e1970340375a58348 [diff]

Add samples to build/run vector multiplication with RVV instructions

Add both int32 and float examples. mhlo.dot MLIR (i8,i8)->i32
(int8 dot product to int32 out) doesn't support RVV codegen
at clang trunk.

The binary can run successfully on qemu-system. renode will hit illegal
instuction for the float example.

Change-Id: I6c87211542fea485eda4d3de7da8250ec60e2a01

CMakeLists.txt[diff]
samples/CMakeLists.txt[Added - diff]
samples/simple_vec_mul/CMakeLists.txt[Added - diff]
samples/simple_vec_mul/device_embedded_sync.c[Added - diff]
samples/simple_vec_mul/float_vec.c[Added - diff]
samples/simple_vec_mul/int_vec.c[Added - diff]
samples/simple_vec_mul/simple_float_mul.mlir[Added - diff]
samples/simple_vec_mul/simple_int_mul.mlir[Added - diff]
samples/simple_vec_mul/simple_vec_mul.c[Added - diff]

9 files changed

tree: 0d8e8e685f52008284e56fb0d4a3e63a06976a6a