Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

aeni error #14

Open
koolhazz opened this issue Apr 2, 2024 · 1 comment
Open

aeni error #14

koolhazz opened this issue Apr 2, 2024 · 1 comment

Comments

@koolhazz
Copy link

koolhazz commented Apr 2, 2024

image

#pragma once

#include <stdint.h>
#include <stdbool.h>
#include <immintrin.h>

#define MIX(x) do { \
		m = m; \
		a = _mm_aesenc_si128(x, a); \
		a = _mm_aesenc_si128(a, m); \
		b = _mm_shuffle_epi8(_mm_xor_si128(x, b), s); \
		b = _mm_shuffle_epi8(_mm_aesdec_si128(b, m), s); \
} while (0)


__m128i AESNI_Hash128(const uint8_t* msg, unsigned len, uint32_t seed) {
	__m128i a = _mm_set1_epi32(seed);
	__m128i b = _mm_set1_epi32(len);
	__m128i m = _mm_set_epi32(0xdeadbeef, 0xffff0000, 0x01234567, 0x89abcdef);
	__m128i s = _mm_set_epi8(3, 7, 11, 15, 2, 6, 10, 14, 1, 5, 9, 13, 0, 4, 8, 12);

	bool greed = (((uintptr_t)msg + (len-1)) & 0xfffUL) >= 15; // do not cross page
	if (len > 80) {
		__m128i c = _mm_aesenc_si128(b, m);
		__m128i d = _mm_aesdec_si128(a, m);
		a = _mm_aesenc_si128(a, m);
		b = _mm_aesdec_si128(b, m);
		do {
			a = _mm_xor_si128(a, _mm_lddqu_si128((const __m128i*)msg));
			b = _mm_xor_si128(b, _mm_lddqu_si128((const __m128i*)(msg + 16)));
			c = _mm_xor_si128(c, _mm_lddqu_si128((const __m128i*)(msg + 32)));
			d = _mm_xor_si128(d, _mm_lddqu_si128((const __m128i*)(msg + 48)));
			a = _mm_shuffle_epi8(_mm_aesenc_si128(a, m), s);
			b = _mm_shuffle_epi8(_mm_aesdec_si128(b, m), s);
			c = _mm_shuffle_epi8(_mm_aesenc_si128(c, m), s);
			d = _mm_shuffle_epi8(_mm_aesdec_si128(d, m), s);
			msg += 64;
			len -= 64;
		} while (len > 80);
		c = _mm_aesenc_si128(a, c);
		d = _mm_aesdec_si128(b, d);
		a = _mm_aesenc_si128(c, d);
		b = _mm_aesdec_si128(d, c);
	}

	while (len >= 16) {
		MIX(_mm_lddqu_si128((const __m128i*)msg));
		msg += 16;
		len -= 16;
	}

	if (greed) {
#define GREEDILY_READ(n, addr) \
        _mm_bsrli_si128(_mm_bslli_si128(_mm_lddqu_si128((const __m128i*)addr), (16-(n))), (16-(n)))

		switch (len) {
			case 15: MIX(GREEDILY_READ(15,msg)); break;
			case 14: MIX(GREEDILY_READ(14,msg)); break;
			case 13: MIX(GREEDILY_READ(13,msg)); break;
			case 12: MIX(GREEDILY_READ(12,msg)); break;
			case 11: MIX(GREEDILY_READ(11,msg)); break;
			case 10: MIX(GREEDILY_READ(10,msg)); break;
			case 9: MIX(GREEDILY_READ(9,msg)); break;
			case 8: MIX((__m128i)_mm_load_sd((const double*)msg)); break;
			case 7: MIX(GREEDILY_READ(7,msg)); break;
			case 6: MIX(GREEDILY_READ(6,msg)); break;
			case 5: MIX(GREEDILY_READ(5,msg)); break;
			case 4: MIX((__m128i)_mm_load_ss((const float*)msg)); break;
			case 3: MIX(GREEDILY_READ(3,msg)); break;
			case 2: MIX(GREEDILY_READ(2,msg)); break;
			case 1: MIX(GREEDILY_READ(1,msg)); break;
			case 0:
			default: // try to keep m & s from register spilling
				a = _mm_add_epi8(a, s);
				b = _mm_add_epi8(b, m);
		}
#undef GREEDILY_READ
		return _mm_aesenc_si128(a, b);
	}

	uint64_t x = 0;
	switch (len) {
		case 15:
			x |= ((uint64_t)msg[14]) << 48U;
		case 14:
			x |= ((uint64_t)msg[13]) << 40U;
		case 13:
			x |= ((uint64_t)msg[12]) << 32U;
		case 12:
			x |= *(const uint32_t*)(msg + 8);
			MIX(_mm_set_epi64x(x, *(const uint64_t*)msg));
			break;
		case 11:
			x |= ((uint32_t)msg[10]) << 16U;
		case 10:
			x |= ((uint32_t)msg[9]) << 8U;
		case 9:
			x |= msg[8];
		case 8:
			MIX(_mm_set_epi64x(x, *(const uint64_t*)msg));
			break;
		case 7:
			x |= ((uint64_t)msg[6]) << 48U;
		case 6:
			x |= ((uint64_t)msg[5]) << 40U;
		case 5:
			x |= ((uint64_t)msg[4]) << 32U;
		case 4:
			x |= *(const uint32_t*)msg;
			MIX(_mm_set_epi64x(0, x));
			break;
		case 3:
			x |= ((uint32_t)msg[2]) << 16U;
		case 2:
			x |= ((uint32_t)msg[1]) << 8U;
		case 1:
			x |= msg[0];
			MIX(_mm_set_epi64x(0, x));
			break;
		case 0:
		default: // try to keep m & s from register spilling
			a = _mm_add_epi8(a, s);
			b = _mm_add_epi8(b, m);

	}
	return _mm_aesenc_si128(a, b);
}

uint64_t AESNI_Hash64(const uint8_t* msg, unsigned len, uint32_t seed) {
	union {
		uint64_t x[2];
		__m128i v;
	} t;
	t.v = AESNI_Hash128(msg, len, seed);
	return t.x[0];
}

@koolhazz
Copy link
Author

koolhazz commented Apr 2, 2024

执行 goat 时,报 segv 错误

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant