shz_mem.h File Reference

#include "shz_cdefs.h"
#include <stdbool.h>
#include <stddef.h>
#include "inline/shz_mem.inl.h"

Include dependency graph for shz_mem.h:

This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Macros
Barriers
Macros for preventing GCC from reordering instructions.
#define	SHZ_MEMORY_BARRIER_SOFT()
#define	SHZ_MEMORY_BARRIER_HARD()

Functions
void	shz_dcache_alloc_line (void *src) SHZ_NOEXCEPT
void *	shz_memcpy (void *SHZ_RESTRICT dst, const void *SHZ_RESTRICT src, size_t bytes) SHZ_NOEXCEPT
Specializations
Specialized routines for specific sizes + alignments.
void *	shz_memcpy1 (void *SHZ_RESTRICT dst, const void *SHZ_RESTRICT src, size_t bytes) SHZ_NOEXCEPT
void *	shz_memcpy2 (void *SHZ_RESTRICT dst, const void *SHZ_RESTRICT src, size_t bytes) SHZ_NOEXCEPT
void *	shz_memcpy4 (void *SHZ_RESTRICT dst, const void *SHZ_RESTRICT src, size_t bytes) SHZ_NOEXCEPT
void *	shz_memcpy8 (void *SHZ_RESTRICT dst, const void *SHZ_RESTRICT src, size_t bytes) SHZ_NOEXCEPT
void *	shz_memset8 (void *dst, uint64_t value, size_t bytes) SHZ_NOEXCEPT
void *	shz_memcpy32 (void *SHZ_RESTRICT dst, const void *SHZ_RESTRICT src, size_t bytes) SHZ_NOEXCEPT
void *	shz_sq_memcpy32 (void *SHZ_RESTRICT dst, const void *SHZ_RESTRICT src, size_t bytes) SHZ_NOEXCEPT
void *	shz_sq_memcpy32_xmtrx (void *SHZ_RESTRICT dst, const void *SHZ_RESTRICT src, size_t bytes) SHZ_NOEXCEPT
void *	shz_memcpy64 (void *SHZ_RESTRICT dst, const void *SHZ_RESTRICT src, size_t bytes) SHZ_NOEXCEPT
void *	shz_memcpy128 (void *SHZ_RESTRICT dst, const void *SHZ_RESTRICT src, size_t bytes) SHZ_NOEXCEPT
Constant-sized Operations
Specialized routines for operating on statically sized buffers.
void	shz_memcpy2_16 (void *SHZ_RESTRICT dst, const void *SHZ_RESTRICT src) SHZ_NOEXCEPT
void	shz_memset2_16 (void *dst, uint16_t value) SHZ_NOEXCEPT
void	shz_memcpy4_16 (void *SHZ_RESTRICT dst, const void *SHZ_RESTRICT src) SHZ_NOEXCEPT
void	shz_memcpy32_1 (void *SHZ_RESTRICT dst, const void *SHZ_RESTRICT src) SHZ_NOEXCEPT
void	shz_memswap32_1 (void *SHZ_RESTRICT p1, void *SHZ_RESTRICT p2) SHZ_NOEXCEPT
void	shz_memswap32_1_xmtrx (void *SHZ_RESTRICT p1, void *SHZ_RESTRICT p2) SHZ_NOEXCEPT
void *	shz_sq_memcpy32_1 (void *SHZ_RESTRICT dst, const void *SHZ_RESTRICT src) SHZ_NOEXCEPT
void *	shz_sq_memcpy32_1_xmtrx (void *SHZ_RESTRICT dst, const void *SHZ_RESTRICT src) SHZ_NOEXCEPT

Detailed Description

Memory API.

API built around copying, assigning, and working with memory.

Todo

• shz_macw()
• shz_memset2()
• shz_memset4()
• shz_memset32()
• shz_memset()
• shz_memmove()

Author

2025, 2026 Falco Girgis

2020 MoopTheHedgehog

Copyright

MIT License

Definition in file shz_mem.h.

Macro Definition Documentation

SHZ_MEMORY_BARRIER_SOFT

#define SHZ_MEMORY_BARRIER_SOFT

(

)

Creates a software memory barrier beyond which any loads or stores may not be reordered.

Definition at line 51 of file shz_mem.h.

SHZ_MEMORY_BARRIER_HARD

#define SHZ_MEMORY_BARRIER_HARD

(

)

Creates a hardware memory barrier beyond which any loads or stores may not be reordered.

Definition at line 53 of file shz_mem.h.

Function Documentation

shz_dcache_alloc_line()

void shz_dcache_alloc_line

(

void *

src

)

Intrinsic around the SH4 MOVCA.L instruction.

Preallocates the cache-line containing src.

Zero-initializes all 32-bytes within the src cache-line, setting the valid bit to 1.

shz_memcpy()

void * shz_memcpy	(	void *SHZ_RESTRICT	dst,
		const void *SHZ_RESTRICT	src,
		size_t	bytes )

Generic drop-in fast memcpy() replacement.

Copies bytes from src to dst, determining the most efficient specialization to call into at run-time, returning dst.

There are no alignment or size requirements for this routine.

Note

When you know of and can control the src and dst alignments and batch sizes, you can micro-optimize by calling into the most specific memcpy() specialization for your given scenario, over just using this generic implementation, which must choose which one to use at run-time.

Warning

dst and src buffers should not be overlapping.

See also

shz_memcpy1(), shz_memcpy2(), shz_memcpy4(), shz_memcpy8(), shz_memcpy32(), shz_memcpy64(), shz_memcpy128()

shz_memcpy1()

void * shz_memcpy1	(	void *SHZ_RESTRICT	dst,
		const void *SHZ_RESTRICT	src,
		size_t	bytes )

Copies an unaligned buffer to another one byte at a time.

The dst pointer is returned.

Note

Typically, unless you know you are copying a tiny number of definitely unaligned bytes, you want to use shz_memcpy(), which automatically handles arbitrary alignment for you, potentially more efficiently than copying byte-by-byte.

Warning

dst and src buffers should not be overlapping.

See also

shz_memcpy()

shz_memcpy2()

void * shz_memcpy2	(	void *SHZ_RESTRICT	dst,
		const void *SHZ_RESTRICT	src,
		size_t	bytes )

Copies from one 2-byte aligned buffer to another two bytes at a time.

The dst pointer is returned.

Warning

dst and src must both be aligned by at least 2 bytes, and bytes must be a multiple of 2.

dst and src buffers should not be overlapping.

shz_memcpy4()

void * shz_memcpy4	(	void *SHZ_RESTRICT	dst,
		const void *SHZ_RESTRICT	src,
		size_t	bytes )

Copies a from one 4-byte aligned buffer to another 4 bytes at a time.

The dst buffer is returned.

Warning

dst and src must both be aligned by at least 4 bytes, and bytes must be a multiple of 4.

dst and src buffers should not be overlapping.

shz_memcpy8()

void * shz_memcpy8	(	void *SHZ_RESTRICT	dst,
		const void *SHZ_RESTRICT	src,
		size_t	bytes )

Copies a from one 8-byte aligned buffer to another 8 bytes at a time.

The dst buffer is returned.

Warning

dst and src must both be aligned by at least 8 bytes, and bytes must be a multiple of 8.

src and dst should not overlap.

shz_memset8()

void * shz_memset8	(	void *	dst,
		uint64_t	value,
		size_t	bytes )

Assigns the given 8-byte value to the bytes in dst.

Warning

dst should be at least 8-byte aligned, and bytes should be a multiple of 8!

shz_memcpy32()

void * shz_memcpy32	(	void *SHZ_RESTRICT	dst,
		const void *SHZ_RESTRICT	src,
		size_t	bytes )

Copies bytes from the src to the dst buffer in 32-byte chunks.

Transfers from 8-byte aligned buffer, src to 32-byte aligned buffer, dst, 32 bytes at a time. Returns the dst address.

Warning

dst must be 32-byte aligned, while src can be only 8-byte aligned. bytes must be a multiple of 32.

src and dst buffers must not overlap.

Note

This is the quickest way to move 32-byte chunks of data around within memory, but the shz_sq_memcpy32() will be faster when writing through the cache to external memory.

See also

shz_sq_memcpy32()

shz_sq_memcpy32()

void * shz_sq_memcpy32	(	void *SHZ_RESTRICT	dst,
		const void *SHZ_RESTRICT	src,
		size_t	bytes )

Copies bytes from src to dst in 32-byte chunks, using the Store Queues.

Transfers from 8-byte aligned buffer, src to 4-byte aligned address, dst, 32 bytes at a time, writing through the cache, using the SH4's Store Queues. Returns the dst address.

Warning

src must be at least 8-byte aligned, while dst can be only 4-byte aligned. bytes must be a multiple of 32.

Note

This is the quickest way to move 32-byte chunks of data to external memory. When copying to cached memory, you must invalidate the cache lines containing dst before initiating the copy... Which means this routine becomes slower than doing memory-to-memory copies with shz_memcpy32().

See also

shz_memcpy32(), shz_sq_memcpy32_1()

shz_sq_memcpy32_xmtrx()

void * shz_sq_memcpy32_xmtrx	(	void *SHZ_RESTRICT	dst,
		const void *SHZ_RESTRICT	src,
		size_t	bytes )

Copies bytes from src to dst in 32-byte chunks, using the Store Queues and XMTRX.

Equiavalent to shz_sq_memcpy32(), except copying is done through XMTRX.

Warning

This routine clobbers XMTRX.

shz_memcpy64()

void * shz_memcpy64	(	void *SHZ_RESTRICT	dst,
		const void *SHZ_RESTRICT	src,
		size_t	bytes )

Specialized memcpy() variant for copying multiples of 64-bytes.

Copies a from an 8-byte aligned buffer to a 32-byte aligned buffer, 64 bytes at a time. Returns the dst address.

Warning

src and dst buffers must not overlap.

dst must be 32-byte aligned, while src can be only 8-byte aligned. bytes must be a multiple of 64.

shz_memcpy128()

void * shz_memcpy128	(	void *SHZ_RESTRICT	dst,
		const void *SHZ_RESTRICT	src,
		size_t	bytes )

Specialized memcpy() variant for copying multiples of 128 bytes.

Copies a from an 8-byte aligned buffer to a 32-byte aligned buffer, 128 bytes at a time. Returns the dst address.

Warning

src and dst buffers must not overlap.

dst must be 32-byte aligned, while src can be only 8-byte aligned. bytes must be a multiple of 128.

shz_memcpy2_16()

void shz_memcpy2_16	(	void *SHZ_RESTRICT	dst,
		const void *SHZ_RESTRICT	src )

Copies 16 shorts from src to dst.

Warning

src and dst buffers must not overlap.

dst and src must both be aligned by at least two bytes.

shz_memset2_16()

void shz_memset2_16	(	void *	dst,
		uint16_t	value )

Sets the values of the 16 shorts pointed to by dst to the given value.

Warning

dst must be aligned by at least two bytes.

shz_memcpy4_16()

void shz_memcpy4_16	(	void *SHZ_RESTRICT	dst,
		const void *SHZ_RESTRICT	src )

Copies 16 4-byte, long values from src to dst.

Warning

src and dst buffers must not overlap.

The src and dst buffers must both be at least 4-byte aligned.

shz_memcpy32_1()

void shz_memcpy32_1	(	void *SHZ_RESTRICT	dst,
		const void *SHZ_RESTRICT	src )

Copies 32 bytes from p1 to p2 as a single chunk.

Warning

dst must be 32-byte aligned, while src can be only 8-byte aligned.

shz_memswap32_1()

void shz_memswap32_1	(	void *SHZ_RESTRICT	p1,
		void *SHZ_RESTRICT	p2 )

Swaps the values within the given 32-byte buffers.

Warning

p1 and p2 must be at least 8-byte aligned.

shz_memswap32_1_xmtrx()

void shz_memswap32_1_xmtrx	(	void *SHZ_RESTRICT	p1,
		void *SHZ_RESTRICT	p2 )

Swaps the values within the given 32-byte buffers, using XMTRX.

Equivalent to shz_memcpy32_1(), except copying is done through XMTRX.

Warning

This routine clobbers XMTRX!

shz_sq_memcpy32_1()

void * shz_sq_memcpy32_1	(	void *SHZ_RESTRICT	dst,
		const void *SHZ_RESTRICT	src )

Copies src to dst in a single 32-byte transaction using the Store Queues.

Note

The Store Queues bypass the SH4's data-cache! They are typically used to transfer to external memory and are slower for memory-to-memory transactions.

Warning

dst must be at least 4-byte aligned, while src must be at least 8-byte aligned.

See also

shz_memcpy32()

shz_sq_memcpy32_1_xmtrx()

void * shz_sq_memcpy32_1_xmtrx	(	void *SHZ_RESTRICT	dst,
		const void *SHZ_RESTRICT	src )

Copies src to dst in a single 32-byte transaction using the Store Queues and XMTRX.

Equivalent to shz_sq_memcpy32_1(), except copying is done through XMTRX.

Warning

This routine clobberx XMTRX.

See also

shz_memcpy32()