[fio.git] / crc / crc32c-intel.c

#include <inttypes.h>
#include <string.h>
#include <unistd.h>
#include <stdlib.h>
#include <signal.h>
#include <sys/types.h>
#include <sys/wait.h>
#include "crc32c.h"

/*
 * Based on a posting to lkml by Austin Zhang <austin.zhang@intel.com>
 *
 * Using hardware provided CRC32 instruction to accelerate the CRC32 disposal.
 * CRC32C polynomial:0x1EDC6F41(BE)/0x82F63B78(LE)
 * CRC32 is a new instruction in Intel SSE4.2, the reference can be found at:
 * http://www.intel.com/products/processor/manuals/
 * Intel(R) 64 and IA-32 Architectures Software Developer's Manual
 * Volume 2A: Instruction Set Reference, A-M
 */

#ifdef ARCH_HAVE_SSE4_2

#if BITS_PER_LONG == 64
#define REX_PRE "0x48, "
#define SCALE_F 8
#else
#define REX_PRE
#define SCALE_F 4
#endif

uint32_t crc32c_intel_le_hw_byte(uint32_t crc, unsigned char const *data,
				 unsigned long length)
{
	while (length--) {
		__asm__ __volatile__(
			".byte 0xf2, 0xf, 0x38, 0xf0, 0xf1"
			:"=S"(crc)
			:"0"(crc), "c"(*data)
		);
		data++;
	}

	return crc;
}

/*
 * Steps through buffer one byte at at time, calculates reflected 
 * crc using table.
 */
uint32_t crc32c_intel(unsigned char const *data, unsigned long length)
{
	unsigned int iquotient = length / SCALE_F;
	unsigned int iremainder = length % SCALE_F;
#if BITS_PER_LONG == 64
	uint64_t *ptmp = (uint64_t *) data;
#else
	uint32_t *ptmp = (uint32_t *) data;
#endif
	uint32_t crc = ~0;

	while (iquotient--) {
		__asm__ __volatile__(
			".byte 0xf2, " REX_PRE "0xf, 0x38, 0xf1, 0xf1;"
			:"=S"(crc)
			:"0"(crc), "c"(*ptmp)
		);
		ptmp++;
	}

	if (iremainder)
		crc = crc32c_intel_le_hw_byte(crc, (unsigned char *)ptmp,
				 iremainder);

	return crc;
}

static void do_cpuid(unsigned int *eax, unsigned int *ebx, unsigned int *ecx,
		     unsigned int *edx)
{
	int id = *eax;

	asm("movl %4, %%eax;"
	    "cpuid;"
	    "movl %%eax, %0;"
	    "movl %%ebx, %1;"
	    "movl %%ecx, %2;"
	    "movl %%edx, %3;"
		: "=r" (*eax), "=r" (*ebx), "=r" (*ecx), "=r" (*edx)
		: "r" (id)
		: "eax", "ebx", "ecx", "edx");
}

int crc32c_intel_works(void)
{
	unsigned int eax, ebx, ecx, edx;

	eax = 1;

	do_cpuid(&eax, &ebx, &ecx, &edx);
	return (ecx & (1 << 20)) != 0;
}

#endif /* ARCH_HAVE_SSE */
Commit	Line	Data
3845591f	1	#include <inttypes.h>
5d7c5d34 JA	2	#include <string.h>
	3	#include <unistd.h>
	4	#include <stdlib.h>
	5	#include <signal.h>
	6	#include <sys/types.h>
	7	#include <sys/wait.h>
419484b9	8	#include "crc32c.h"
3845591f JA	9
	10	/*
	11	* Based on a posting to lkml by Austin Zhang <austin.zhang@intel.com>
	12	*
	13	* Using hardware provided CRC32 instruction to accelerate the CRC32 disposal.
	14	* CRC32C polynomial:0x1EDC6F41(BE)/0x82F63B78(LE)
	15	* CRC32 is a new instruction in Intel SSE4.2, the reference can be found at:
	16	* http://www.intel.com/products/processor/manuals/
	17	* Intel(R) 64 and IA-32 Architectures Software Developer's Manual
	18	* Volume 2A: Instruction Set Reference, A-M
	19	*/
	20
2f68124f	21	#ifdef ARCH_HAVE_SSE4_2
419484b9	22
3845591f JA	23	#if BITS_PER_LONG == 64
	24	#define REX_PRE "0x48, "
	25	#define SCALE_F 8
	26	#else
	27	#define REX_PRE
	28	#define SCALE_F 4
	29	#endif
	30
	31	uint32_t crc32c_intel_le_hw_byte(uint32_t crc, unsigned char const *data,
	32	unsigned long length)
	33	{
	34	while (length--) {
	35	__asm__ __volatile__(
	36	".byte 0xf2, 0xf, 0x38, 0xf0, 0xf1"
	37	:"=S"(crc)
	38	:"0"(crc), "c"(*data)
	39	);
	40	data++;
	41	}
	42
	43	return crc;
	44	}
	45
	46	/*
	47	* Steps through buffer one byte at at time, calculates reflected
	48	* crc using table.
	49	*/
	50	uint32_t crc32c_intel(unsigned char const *data, unsigned long length)
	51	{
	52	unsigned int iquotient = length / SCALE_F;
	53	unsigned int iremainder = length % SCALE_F;
	54	#if BITS_PER_LONG == 64
	55	uint64_t ptmp = (uint64_t ) data;
	56	#else
	57	uint32_t ptmp = (uint32_t ) data;
	58	#endif
	59	uint32_t crc = ~0;
	60
	61	while (iquotient--) {
	62	__asm__ __volatile__(
	63	".byte 0xf2, " REX_PRE "0xf, 0x38, 0xf1, 0xf1;"
	64	:"=S"(crc)
	65	:"0"(crc), "c"(*ptmp)
	66	);
	67	ptmp++;
	68	}
	69
	70	if (iremainder)
	71	crc = crc32c_intel_le_hw_byte(crc, (unsigned char *)ptmp,
	72	iremainder);
	73
	74	return crc;
	75	}
419484b9	76
e0ab5f97 JA	77	static void do_cpuid(unsigned int eax, unsigned int ebx, unsigned int *ecx,
e0ab5f97 JA	78	unsigned int *edx)
5d7c5d34	79	{
e0ab5f97 JA	80	int id = *eax;
	81
	82	asm("movl %4, %%eax;"
	83	"cpuid;"
	84	"movl %%eax, %0;"
	85	"movl %%ebx, %1;"
	86	"movl %%ecx, %2;"
	87	"movl %%edx, %3;"
	88	: "=r" (eax), "=r" (ebx), "=r" (ecx), "=r" (edx)
	89	: "r" (id)
	90	: "eax", "ebx", "ecx", "edx");
5d7c5d34 JA	91	}
	92
	93	int crc32c_intel_works(void)
	94	{
e0ab5f97	95	unsigned int eax, ebx, ecx, edx;
419484b9	96
e0ab5f97 JA	97	eax = 1;
	98
	99	do_cpuid(&eax, &ebx, &ecx, &edx);
	100	return (ecx & (1 << 20)) != 0;
5d7c5d34 JA	101	}
	102
	103	#endif /* ARCH_HAVE_SSE */