amd64: micro-optimize vptopte()/vtopde() further

[kib]

Summary

Eliminate shlq $3,address shift after masking of the va is done, which
is needed to convert pt_entry_t[] array index into byte offset.
Do it by preshifting the mask, and compensating the right shift of va.

[alc]

I'll take a look at this tomorrow.

As an aside, have you seen how the compiler implements the static variable pg_g? It's implementing pg_g using a single byte of storage and performing a size extending load of the byte followed by a left shift by 8 wherever the variable is used to recreate a value of the declared type, pt_entry. Removing static from the variable definition reduces the code size by 240 bytes.

[kib]

In D33786#764217, @alc wrote:

As an aside, have you seen how the compiler implements the static variable pg_g? It's implementing pg_g using a single byte of storage and performing a size extending load of the byte followed by a left shift by 8 wherever the variable is used to recreate a value of the declared type, pt_entry. Removing static from the variable definition reduces the code size by 240 bytes.

Will you commit the change? Or do you want me to add it to the series?

[kib]

In D33786#764218, @kib wrote:

In D33786#764217, @alc wrote:

As an aside, have you seen how the compiler implements the static variable pg_g? It's implementing pg_g using a single byte of storage and performing a size extending load of the byte followed by a left shift by 8 wherever the variable is used to recreate a value of the declared type, pt_entry. Removing static from the variable definition reduces the code size by 240 bytes.

Will you commit the change? Or do you want me to add it to the series?

Ping?