[QNN] Change in Pass Context for lookup table calculation #13660
Conversation
|
Thanks for contributing to TVM! Please refer to the contributing guidelines https://tvm.apache.org/docs/contribute/ for useful information and tips. Please request code reviews from Reviewers by @-ing them in a comment.
Generated by tvm-bot |
Motivation: It is possible to disable specific passes through the "disabled_pass" parameter in the Pass Context. These "disabled" passes can be optional for one target and mandatory for another one. Since lookup table for some QNN operations (tanh, round and etc.) is calculated on the host and some of disabled passes can be required for the host, no need to disable these passes. This constant calculation/ evaluation is orthogonal to the compilation process for specific target. What was changed: This commit creates its own compilation Pass Context for lookup table calculation and evaluation (for elemwise QNN ops: tanh, sqrt ...).
ibsidorenko
force-pushed
the
qnn-const-expression
branch
from
a63729a to
304a225
Compare
ibsidorenko
changed the title
|
@tvm-bot rerun |
|
|
||
| If a number of passes are disabled in the current Pass Context, then there is no need to disable | ||
| these passes for const expression evaluation as well. That's why we use empty list | ||
| "disabled_pass=[]", all other arguments are inherited from the current Pass Context. |
There was a problem hiding this comment.
Do you mean, without disabled_pass=[], the original disabled_pass list in the current context will be ignored during const folding?
There was a problem hiding this comment.
I put here disabled_pass=[] to ignore the original disabled_pass list from the current context. Otherwise VM inherits non-empty disabled_pass list and fails.
Note, we call this code only for several qnn ops: tanh, sqrt, erf, exp + some other ops. It does not affect global constant folding.
For clarification, here is my example to illustrate this issue:
I am trying to compile subgraph with qnn.tanh operation for Hexagon target without QNN canonicalization but with QNN legalization (this is acceptable for Hexagon). For qnn.tanh we compose lookup table by means of Virtual Machine and compute it on the host. For most of users/developers host is x86 cpu. But for x86 QNN canonicalization is mandatory pass, otherwise it fails.
Looks like the simplest way to fix this issue is to create new Pass Context and ignore the original disabled_pass list for lookup table calculation and constant evaluation. Other arguments (opt_level, instruments, config ...) I inherit from the current context.
P.S. Corresponding unit test will be added in the next PR.
There was a problem hiding this comment.
I see, you do want to enable QNN legalization (which is disabled in the original ctx) here.
Motivation: It is possible to disable specific passes through the "disabled_pass" parameter in the Pass Context. These "disabled" passes can be optional for one target and mandatory for another one. Since lookup table for some QNN operations (tanh, round and etc.) is calculated on the host and some of disabled passes can be required for the host, no need to disable these passes. This constant calculation/ evaluation is orthogonal to the compilation process for specific target. What was changed: This commit creates its own compilation Pass Context for lookup table calculation and evaluation (for elemwise QNN ops: tanh, sqrt ...).
Motivation: It is possible to disable specific passes through the "disabled_pass" parameter in the Pass Context. These "disabled" passes can be optional for one target and mandatory for another one. Since lookup table for some QNN operations (tanh, round and etc.) is calculated on the host and some of disabled passes can be required for the host, no need to disable these passes. This constant calculation/ evaluation is orthogonal to the compilation process for specific target. What was changed: This commit creates its own compilation Pass Context for lookup table calculation and evaluation (for elemwise QNN ops: tanh, sqrt ...).
Motivation:
It is possible to disable specific passes through the "disabled_pass" parameter in the Pass Context. These "disabled" passes can be optional for one target and mandatory for another one.
Since lookup table for some QNN operations (tanh, round and etc.) is calculated on the host and some of disabled passes can be required for the host, no need to disable these passes. This constant calculation/ evaluation is orthogonal to the compilation process for specific target.
What was changed:
This commit creates its own compilation Pass Context for lookup table
calculation and evaluation (for elemwise QNN ops: tanh, sqrt ...).