Skip to content

Aarch64 be fix #1066

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.

Sign up for GitHub

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

Open
zeldin wants to merge 2 commits into
base: main
Choose a base branch
from
Open

Aarch64 be fix #1066

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
c43d217
Bump sljit submodule
zeldin Nov 8, 2026
eff9ba4
ARM64: Fix JIT on big endian
zeldin Oct 7, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion deps/sljit
Open in desktop
Submodule sljit updated 2 files
+0 −8 sljit_src/sljitConfigCPU.h
+101 −48 sljit_src/sljitNativeARM_64.c
84 changes: 82 additions & 2 deletions src/pcre2_jit_simd_inc.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1291,16 +1291,26 @@ fast_forward_char_simd(compiler_common *common, PCRE2_UCHAR char1, PCRE2_UCHAR c
fast_forward_char_pair_sse2_compare(compiler, compare_type, i, data_ind, cmp1_ind, cmp2_ind,
tmp_ind);

#if (defined SLJIT_LITTLE_ENDIAN && SLJIT_LITTLE_ENDIAN)
/* SHRN */
instruction = 0x0f0c8400 | (data_ind << 5) | data_ind;
#else
/* SHRN2 */
instruction = 0x4f0c8400 | (data_ind << 5) | data_ind;

@zherczeg zherczeg Nov 9, 2026

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This change is repeated many times. We should introduce a const for SHRN2.

@zeldin zeldin Nov 9, 2026

Copy link
Copy Markdown
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Good point. And of course also for SHRN, CMEQ, ORR, INS, SHL and USHR which were already repeated before.

#endif
sljit_emit_op_custom(compiler, &instruction, sizeof(sljit_u32));

sljit_emit_simd_lane_mov(compiler, SLJIT_SIMD_STORE | SLJIT_SIMD_REG_128 | SLJIT_SIMD_ELEM_64,
SLJIT_VR0, 0, TMP1, 0);

OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 2);
#if (defined SLJIT_LITTLE_ENDIAN && SLJIT_LITTLE_ENDIAN)
OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, TMP2, 0);
OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, TMP2, 0);
#else
OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, TMP2, 0);
OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, TMP2, 0);
#endif

quit = CMP(SLJIT_NOT_ZERO, TMP1, 0, SLJIT_IMM, 0);

Expand All @@ -1319,8 +1329,13 @@ fast_forward_char_simd(compiler_common *common, PCRE2_UCHAR char1, PCRE2_UCHAR c
fast_forward_char_pair_sse2_compare(compiler, compare_type, i, data_ind, cmp1_ind, cmp2_ind,
tmp_ind);

#if (defined SLJIT_LITTLE_ENDIAN && SLJIT_LITTLE_ENDIAN)
/* SHRN */
instruction = 0x0f0c8400 | (data_ind << 5) | data_ind;
#else
/* SHRN2 */
instruction = 0x4f0c8400 | (data_ind << 5) | data_ind;
#endif
sljit_emit_op_custom(compiler, &instruction, sizeof(sljit_u32));

sljit_emit_simd_lane_mov(compiler, SLJIT_SIMD_STORE | SLJIT_SIMD_REG_128 | SLJIT_SIMD_ELEM_64,
Expand All @@ -1329,7 +1344,11 @@ fast_forward_char_simd(compiler_common *common, PCRE2_UCHAR char1, PCRE2_UCHAR c

JUMPHERE(quit);

#if (defined SLJIT_LITTLE_ENDIAN && SLJIT_LITTLE_ENDIAN)
sljit_emit_op1(compiler, SLJIT_CTZ, TMP1, 0, TMP1, 0);
#else
sljit_emit_op1(compiler, SLJIT_CLZ, TMP1, 0, TMP1, 0);
#endif
OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 2);
OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);

Expand Down Expand Up @@ -1416,16 +1435,26 @@ fast_requested_char_simd(compiler_common *common, PCRE2_UCHAR char1, PCRE2_UCHAR
fast_forward_char_pair_sse2_compare(compiler, compare_type, i, data_ind, cmp1_ind, cmp2_ind,
tmp_ind);

#if (defined SLJIT_LITTLE_ENDIAN && SLJIT_LITTLE_ENDIAN)
/* SHRN */
instruction = 0x0f0c8400 | (data_ind << 5) | data_ind;
#else
/* SHRN2 */
instruction = 0x4f0c8400 | (data_ind << 5) | data_ind;
#endif
sljit_emit_op_custom(compiler, &instruction, sizeof(sljit_u32));

sljit_emit_simd_lane_mov(compiler, SLJIT_SIMD_STORE | SLJIT_SIMD_REG_128 | SLJIT_SIMD_ELEM_64,
SLJIT_VR0, 0, TMP1, 0);

OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 2);
#if (defined SLJIT_LITTLE_ENDIAN && SLJIT_LITTLE_ENDIAN)
OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, TMP2, 0);
OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, TMP2, 0);
#else
OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, TMP2, 0);
OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, TMP2, 0);
#endif

quit = CMP(SLJIT_NOT_ZERO, TMP1, 0, SLJIT_IMM, 0);

Expand All @@ -1443,8 +1472,13 @@ fast_requested_char_simd(compiler_common *common, PCRE2_UCHAR char1, PCRE2_UCHAR
fast_forward_char_pair_sse2_compare(compiler, compare_type, i, data_ind, cmp1_ind, cmp2_ind,
tmp_ind);

#if (defined SLJIT_LITTLE_ENDIAN && SLJIT_LITTLE_ENDIAN)
/* SHRN */
instruction = 0x0f0c8400 | (data_ind << 5) | data_ind;
#else
/* SHRN2 */
instruction = 0x4f0c8400 | (data_ind << 5) | data_ind;
#endif
sljit_emit_op_custom(compiler, &instruction, sizeof(sljit_u32));

sljit_emit_simd_lane_mov(compiler, SLJIT_SIMD_STORE | SLJIT_SIMD_REG_128 | SLJIT_SIMD_ELEM_64,
Expand All @@ -1453,7 +1487,11 @@ fast_requested_char_simd(compiler_common *common, PCRE2_UCHAR char1, PCRE2_UCHAR

JUMPHERE(quit);

#if (defined SLJIT_LITTLE_ENDIAN && SLJIT_LITTLE_ENDIAN)
sljit_emit_op1(compiler, SLJIT_CTZ, TMP1, 0, TMP1, 0);
#else
sljit_emit_op1(compiler, SLJIT_CLZ, TMP1, 0, TMP1, 0);
#endif
OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 2);
OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, TMP3, 0);
add_jump(compiler, &not_found, CMP(SLJIT_GREATER_EQUAL, TMP1, 0, STR_END, 0));
Expand Down Expand Up @@ -1570,28 +1608,51 @@ fast_forward_char_pair_simd(compiler_common *common, sljit_s32 offs1, PCRE2_UCHA
if (diff >= 8)
{
/* MOV (element) */
instruction = 0x6e180400 | (data1_ind << 5) | data2_ind;
sljit_emit_op_custom(compiler, &instruction, sizeof(sljit_u32));
#if (defined SLJIT_LITTLE_ENDIAN && SLJIT_LITTLE_ENDIAN)
instruction = 0x6e180400 | (data1_ind << 5) | data2_ind;
#else
instruction = 0x6e084400 | (data1_ind << 5) | data2_ind;
#endif
sljit_emit_op_custom(compiler, &instruction, sizeof(sljit_u32));

if (diff > 8)
{
#if (defined SLJIT_LITTLE_ENDIAN && SLJIT_LITTLE_ENDIAN)
/* SHL */
instruction = 0x4f405400 | ((diff - 8) << 19) | (data2_ind << 5) | data2_ind;
#else
/* USHR */
instruction = 0x6f400400 | ((16 - diff) << 19) | (data2_ind << 5) | data2_ind;
#endif
sljit_emit_op_custom(compiler, &instruction, sizeof(sljit_u32));
}
}
else
{
/* MOV (element) */
#if (defined SLJIT_LITTLE_ENDIAN && SLJIT_LITTLE_ENDIAN)
instruction = 0x6e180400 | (data1_ind << 5) | tmp1_ind;
#else
instruction = 0x6e084400 | (data1_ind << 5) | tmp1_ind;
#endif
sljit_emit_op_custom(compiler, &instruction, sizeof(sljit_u32));

#if (defined SLJIT_LITTLE_ENDIAN && SLJIT_LITTLE_ENDIAN)
/* SHL */
instruction = 0x4f405400 | (diff << 19) | (data1_ind << 5) | data2_ind;
#else
/* USHR */
instruction = 0x6f400400 | ((8 - diff) << 19) | (data1_ind << 5) | data2_ind;
#endif
sljit_emit_op_custom(compiler, &instruction, sizeof(sljit_u32));

#if (defined SLJIT_LITTLE_ENDIAN && SLJIT_LITTLE_ENDIAN)
/* USHR */
instruction = 0x6f400400 | (diff << 19) | (tmp1_ind << 5) | tmp1_ind;
#else
/* SHL */
instruction = 0x4f405400 | ((8 - diff) << 19) | (tmp1_ind << 5) | tmp1_ind;
#endif
sljit_emit_op_custom(compiler, &instruction, sizeof(sljit_u32));

sljit_emit_simd_op2(compiler, SLJIT_SIMD_OP2_OR | SLJIT_SIMD_REG_128, SLJIT_VR1, SLJIT_VR1,
Expand All @@ -1613,17 +1674,27 @@ fast_forward_char_pair_simd(compiler_common *common, sljit_s32 offs1, PCRE2_UCHA
sljit_emit_simd_op2(compiler, SLJIT_SIMD_OP2_AND | SLJIT_SIMD_REG_128, SLJIT_VR0, SLJIT_VR0,
SLJIT_VR1, 0);

#if (defined SLJIT_LITTLE_ENDIAN && SLJIT_LITTLE_ENDIAN)
/* SHRN */
instruction = 0x0f0c8400 | (data1_ind << 5) | data1_ind;
#else
/* SHRN2 */
instruction = 0x4f0c8400 | (data1_ind << 5) | data1_ind;
#endif
sljit_emit_op_custom(compiler, &instruction, sizeof(sljit_u32));

sljit_emit_simd_lane_mov(compiler, SLJIT_SIMD_STORE | SLJIT_SIMD_REG_128 | SLJIT_SIMD_ELEM_64,
SLJIT_VR0, 0, TMP1, 0);

/* Ignore matches before the first STR_PTR. */
OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 2);
#if (defined SLJIT_LITTLE_ENDIAN && SLJIT_LITTLE_ENDIAN)
OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, TMP2, 0);
OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, TMP2, 0);
#else
OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, TMP2, 0);
OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, TMP2, 0);
#endif

jump[0] = CMP(SLJIT_NOT_ZERO, TMP1, 0, SLJIT_IMM, 0);

Expand All @@ -1649,8 +1720,13 @@ fast_forward_char_pair_simd(compiler_common *common, sljit_s32 offs1, PCRE2_UCHA
sljit_emit_simd_op2(compiler, SLJIT_SIMD_OP2_AND | SLJIT_SIMD_REG_128, SLJIT_VR0, SLJIT_VR0,
SLJIT_VR1, 0);

#if (defined SLJIT_LITTLE_ENDIAN && SLJIT_LITTLE_ENDIAN)
/* SHRN */
instruction = 0x0f0c8400 | (data1_ind << 5) | data1_ind;
#else
/* SHRN2 */
instruction = 0x4f0c8400 | (data1_ind << 5) | data1_ind;
#endif
sljit_emit_op_custom(compiler, &instruction, sizeof(sljit_u32));

sljit_emit_simd_lane_mov(compiler, SLJIT_SIMD_STORE | SLJIT_SIMD_REG_128 | SLJIT_SIMD_ELEM_64,
Expand All @@ -1660,7 +1736,11 @@ fast_forward_char_pair_simd(compiler_common *common, sljit_s32 offs1, PCRE2_UCHA

JUMPHERE(jump[0]);

#if (defined SLJIT_LITTLE_ENDIAN && SLJIT_LITTLE_ENDIAN)
sljit_emit_op1(compiler, SLJIT_CTZ, TMP1, 0, TMP1, 0);
#else
sljit_emit_op1(compiler, SLJIT_CLZ, TMP1, 0, TMP1, 0);
#endif
OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 2);
OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);

Expand Down
Loading