zend_types: Convert HT_*() macros to inline functions

[TimWolla]

No description provided.

[TimWolla]

For this one the benchmarking job shows a meaningful increase in instructions. It's not clear to me why 🤔

[github-actions]

AWS x86_64 (c6id.metal)

Attribute	Value
Environment	aws
Instance type	c6id.metal
Architecture	x86_64
CPU	Intel(R) Xeon(R) Platinum 8375C CPU @ 2.90GHz, 64 cores @ 2900 MHz
CPU settings	disabled deeper C-states, disabled turbo boost, disabled hyper-threading
RAM	251 GB
Kernel	6.18.20-20.229.amzn2023.x86_64
OS	Amazon Linux 2023.11.20260427
GCC	14.2.1
Time	2026-06-14 17:01:29 UTC
Job details	https://github.com/php/php-src/actions/runs/27505874389 (Artifacts)
Changeset	https://github.com/php/php-src/compare/951680712b..e41b1af644

Laravel 12.11.0 demo app - 50 iterations, 50 warmups, 100 requests (sec)

PHP	Min	Max	Std dev	Rel std dev %	Mean	Mean diff %	Median	Median diff %	Skewness	Z-stat	P-value	Memory
PHP - baseline@9516807	0.38628	0.38721	0.00020	0.05%	0.38673	0.00%	0.38671	0.00%	0.321	0.000	1.000	25.82 MB
PHP - zend-types-ht-demacro	0.38783	0.38910	0.00028	0.07%	0.38818	0.38%	0.38813	0.37%	1.582	-8.614	0.000	25.82 MB

Symfony 2.8.0 demo app - 50 iterations, 50 warmups, 100 requests (sec)

PHP	Min	Max	Std dev	Rel std dev %	Mean	Mean diff %	Median	Median diff %	Skewness	Z-stat	P-value	Memory
PHP - baseline@9516807	0.66771	0.66945	0.00034	0.05%	0.66816	0.00%	0.66810	0.00%	1.530	0.000	1.000	25.92 MB
PHP - zend-types-ht-demacro	0.67150	0.67614	0.00071	0.11%	0.67230	0.62%	0.67215	0.60%	3.838	-8.614	0.000	25.88 MB

Wordpress 6.9 main page - 50 iterations, 20 warmups, 20 requests (sec)

PHP	Min	Max	Std dev	Rel std dev %	Mean	Mean diff %	Median	Median diff %	Skewness	Z-stat	P-value	Memory
PHP - baseline@9516807	0.58724	0.59001	0.00068	0.12%	0.58789	0.00%	0.58769	0.00%	1.813	0.000	1.000	26.32 MB
PHP - zend-types-ht-demacro	0.59001	0.59452	0.00080	0.14%	0.59073	0.48%	0.59054	0.49%	3.487	-8.607	0.000	25.90 MB

bench.php - 50 iterations, 20 warmups, 2 requests (sec)

PHP	Min	Max	Std dev	Rel std dev %	Mean	Mean diff %	Median	Median diff %	Skewness	Z-stat	P-value	Memory
PHP - baseline@9516807	0.44958	0.45289	0.00061	0.13%	0.45083	0.00%	0.45080	0.00%	0.851	0.000	1.000	26.32 MB
PHP - zend-types-ht-demacro	0.44957	0.45320	0.00071	0.16%	0.45085	0.00%	0.45080	0.00%	0.810	0.000	1.000	25.90 MB

micro_bench.php - 50 iterations, 5 warmups, 1 request (sec)

PHP	Min	Max	Std dev	Rel std dev %	Mean	Mean diff %	Median	Median diff %	Skewness	Z-stat	P-value	Memory
PHP - baseline@9516807	1.27251	1.31261	0.00688	0.54%	1.28142	0.00%	1.27996	0.00%	2.909	0.000	1.000	26.32 MB
PHP - zend-types-ht-demacro	1.27062	1.28890	0.00370	0.29%	1.28029	-0.09%	1.28015	0.01%	-0.336	-0.052	0.959	25.90 MB

[ndossche]

With macros, functions like zend_hash_real_init (& co) can use the specialised emalloc_ functions.
These functions can only be used if the size expression is a constant; otherwise it degrades to a generic emalloc call.
__builtin_constant_p fails for the function call because... well it isn't a constant. And the optimizer eventually makes it a constant but that's too late for __builtin_constant_p to kick in.
You can validate that by looking at the assembly for those functions.
Similar effects can happen with efree.
EDIT: pure or const attribute would work: https://godbolt.org/z/T4n8jjc9Y