Replace vscaleextexp yaml with header #7408
Conversation
RahulSundarMCW
force-pushed
the
vscaleextexp
branch
from
aa79733
to
60ea4c8
Compare
| #define XNN_DEFINED_UKERNEL | ||
| #endif | ||
| #if XNN_ARCH_X86 || XNN_ARCH_X86_64 | ||
| XNN_UKERNEL_WITH_PARAMS(xnn_arch_x86_avx2, xnn_f32_vscaleextexp_ukernel__avx2_p5_u8, 8, float, struct xnn_f32_default_params, ((xnn_f32_vscaleextexp_ukernel_fn) NULL)) |
There was a problem hiding this comment.
Can you remove the cast of NULL for the init_params function? I know I did this for some unary ops, but I recently removed that as it's no longer necessary in that case.
scripts/generate-tests.sh
Outdated
| @@ -273,7 +273,7 @@ tools/generate-raddstoreexpminusmax-test.py --spec test/f16-raddstoreexpminusmax | |||
| tools/generate-raddstoreexpminusmax-test.py --spec test/f32-raddstoreexpminusmax.yaml --output test/f32-raddstoreexpminusmax.cc & | |||
|
|
|||
| ### Tests for VScaleExtExp micro-kernels | |||
| tools/generate-vscaleextexp-test.py --spec test/f32-vscaleextexp.yaml --output test/f32-vscaleextexp.cc & | |||
| tools/generate-vscaleextexp-test.py --tester VScaleExtExpMicrokernelTester --ukernel f32-vscaleextexp --output test/f32-vscaleextexp.cc & | |||
There was a problem hiding this comment.
There's only one test using this, and now that it doesn't actually generate much code, we should just remove it. Let's also move the contents of vscaleextexp-microkernel-test.* to test/f32-vscaleextexp.cc as well.
test/f32-vscaleextexp.cc
Outdated
| @@ -4,7 +4,7 @@ | |||
| // LICENSE file in the root directory of this source tree. | |||
| // | |||
| // Auto-generated file. Do not edit! | |||
There was a problem hiding this comment.
When you remove the generator script, also remove this comment, because this file will not be generated code any more.
RahulSundarMCW
force-pushed
the
vscaleextexp
branch
7 times, most recently
from
7391402
to
86c1ff6
Compare
|
Address the review comments and pushed the commit. |
RahulSundarMCW
force-pushed
the
vscaleextexp
branch
from
86c1ff6
to
b90d55c
Compare
tools/generate-vscaleextexp-test.py
Outdated
RahulSundarMCW
force-pushed
the
vscaleextexp
branch
from
b90d55c
to
21b2ffc
Compare
There was a problem hiding this comment.
I just noticed something missing for all these yaml -> header PRs: there are declarations of the microkernels that should be converted to use the header tables, e.g.:
XNNPACK/src/xnnpack/vscaleexpminusmax.h
Line 17 in 060cece
This is important to help validate that the header tables are correct.
| #endif | ||
|
|
||
| #if XNN_ARCH_X86 || XNN_ARCH_X86_64 | ||
| XNN_UKERNEL_WITH_PARAMS(xnn_arch_x86_avx2, xnn_f32_vscaleextexp_ukernel__avx2_p5_u8, 8, float, struct xnn_f32_default_params, xnn_f32_vscaleextexp_ukernel_fn) |
There was a problem hiding this comment.
Why is xnn_f32_vscaleextexp_ukernel_fn here in the last parameter? I'd expect it to be NULL.
RahulSundarMCW
force-pushed
the
vscaleextexp
branch
3 times, most recently
from
ebaecb4
to
cec6447
Compare
- Include table header and remove yaml file
RahulSundarMCW
force-pushed
the
vscaleextexp
branch
from
cec6447
to
00b875c
Compare
No description provided.