Modified: tomcat/jk/trunk/native/iis/pcre/sljit/sljitNativeX86_32.c URL: http://svn.apache.org/viewvc/tomcat/jk/trunk/native/iis/pcre/sljit/sljitNativeX86_32.c?rev=1838867&r1=1838866&r2=1838867&view=diff ============================================================================== --- tomcat/jk/trunk/native/iis/pcre/sljit/sljitNativeX86_32.c (original) +++ tomcat/jk/trunk/native/iis/pcre/sljit/sljitNativeX86_32.c Fri Aug 24 14:36:12 2018 @@ -64,29 +64,28 @@ static sljit_u8* generate_far_jump_code( } SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit_emit_enter(struct sljit_compiler *compiler, - sljit_s32 options, sljit_s32 args, sljit_s32 scratches, sljit_s32 saveds, + sljit_s32 options, sljit_s32 arg_types, sljit_s32 scratches, sljit_s32 saveds, sljit_s32 fscratches, sljit_s32 fsaveds, sljit_s32 local_size) { - sljit_s32 size; + sljit_s32 args, size; sljit_u8 *inst; CHECK_ERROR(); - CHECK(check_sljit_emit_enter(compiler, options, args, scratches, saveds, fscratches, fsaveds, local_size)); - set_emit_enter(compiler, options, args, scratches, saveds, fscratches, fsaveds, local_size); + CHECK(check_sljit_emit_enter(compiler, options, arg_types, scratches, saveds, fscratches, fsaveds, local_size)); + set_emit_enter(compiler, options, arg_types, scratches, saveds, fscratches, fsaveds, local_size); + args = get_arg_count(arg_types); compiler->args = args; -#if (defined SLJIT_X86_32_FASTCALL && SLJIT_X86_32_FASTCALL) - /* [esp+0] for saving temporaries and third argument for calls. */ - compiler->saveds_offset = 1 * sizeof(sljit_sw); -#else - /* [esp+0] for saving temporaries and space for maximum three arguments. */ - if (scratches <= 1) - compiler->saveds_offset = 1 * sizeof(sljit_sw); - else - compiler->saveds_offset = ((scratches == 2) ? 2 : 3) * sizeof(sljit_sw); + /* [esp+0] for saving temporaries and function calls. */ + compiler->stack_tmp_size = 2 * sizeof(sljit_sw); + +#if !(defined SLJIT_X86_32_FASTCALL && SLJIT_X86_32_FASTCALL) + if (scratches > 3) + compiler->stack_tmp_size = 3 * sizeof(sljit_sw); #endif + compiler->saveds_offset = compiler->stack_tmp_size; if (scratches > 3) compiler->saveds_offset += ((scratches > (3 + 6)) ? 6 : (scratches - 3)) * sizeof(sljit_sw); @@ -124,34 +123,38 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit #if (defined SLJIT_X86_32_FASTCALL && SLJIT_X86_32_FASTCALL) if (args > 0) { - *inst++ = MOV_r_rm; - *inst++ = MOD_REG | (reg_map[SLJIT_S0] << 3) | reg_map[SLJIT_R2]; + inst[0] = MOV_r_rm; + inst[1] = MOD_REG | (reg_map[SLJIT_S0] << 3) | reg_map[SLJIT_R2]; + inst += 2; } if (args > 1) { - *inst++ = MOV_r_rm; - *inst++ = MOD_REG | (reg_map[SLJIT_S1] << 3) | reg_map[SLJIT_R1]; + inst[0] = MOV_r_rm; + inst[1] = MOD_REG | (reg_map[SLJIT_S1] << 3) | reg_map[SLJIT_R1]; + inst += 2; } if (args > 2) { - *inst++ = MOV_r_rm; - *inst++ = MOD_DISP8 | (reg_map[SLJIT_S2] << 3) | 0x4 /* esp */; - *inst++ = 0x24; - *inst++ = sizeof(sljit_sw) * (3 + 2); /* saveds >= 3 as well. */ + inst[0] = MOV_r_rm; + inst[1] = MOD_DISP8 | (reg_map[SLJIT_S2] << 3) | 0x4 /* esp */; + inst[2] = 0x24; + inst[3] = sizeof(sljit_sw) * (3 + 2); /* saveds >= 3 as well. */ } #else if (args > 0) { - *inst++ = MOV_r_rm; - *inst++ = MOD_DISP8 | (reg_map[SLJIT_S0] << 3) | reg_map[TMP_REG1]; - *inst++ = sizeof(sljit_sw) * 2; + inst[0] = MOV_r_rm; + inst[1] = MOD_DISP8 | (reg_map[SLJIT_S0] << 3) | reg_map[TMP_REG1]; + inst[2] = sizeof(sljit_sw) * 2; + inst += 3; } if (args > 1) { - *inst++ = MOV_r_rm; - *inst++ = MOD_DISP8 | (reg_map[SLJIT_S1] << 3) | reg_map[TMP_REG1]; - *inst++ = sizeof(sljit_sw) * 3; + inst[0] = MOV_r_rm; + inst[1] = MOD_DISP8 | (reg_map[SLJIT_S1] << 3) | reg_map[TMP_REG1]; + inst[2] = sizeof(sljit_sw) * 3; + inst += 3; } if (args > 2) { - *inst++ = MOV_r_rm; - *inst++ = MOD_DISP8 | (reg_map[SLJIT_S2] << 3) | reg_map[TMP_REG1]; - *inst++ = sizeof(sljit_sw) * 4; + inst[0] = MOV_r_rm; + inst[1] = MOD_DISP8 | (reg_map[SLJIT_S2] << 3) | reg_map[TMP_REG1]; + inst[2] = sizeof(sljit_sw) * 4; } #endif @@ -171,17 +174,36 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit compiler->local_size = local_size; #ifdef _WIN32 - if (local_size > 1024) { -#if (defined SLJIT_X86_32_FASTCALL && SLJIT_X86_32_FASTCALL) - FAIL_IF(emit_do_imm(compiler, MOV_r_i32 + reg_map[SLJIT_R0], local_size)); -#else - /* Space for a single argument. This amount is excluded when the stack is allocated below. */ - local_size -= sizeof(sljit_sw); - FAIL_IF(emit_do_imm(compiler, MOV_r_i32 + reg_map[SLJIT_R0], local_size)); - FAIL_IF(emit_non_cum_binary(compiler, SUB_r_rm, SUB_rm_r, SUB, SUB_EAX_i32, - SLJIT_SP, 0, SLJIT_SP, 0, SLJIT_IMM, sizeof(sljit_sw))); -#endif - FAIL_IF(sljit_emit_ijump(compiler, SLJIT_CALL1, SLJIT_IMM, SLJIT_FUNC_OFFSET(sljit_grow_stack))); + if (local_size > 0) { + if (local_size <= 4 * 4096) { + if (local_size > 4096) + EMIT_MOV(compiler, TMP_REG1, 0, SLJIT_MEM1(SLJIT_SP), -4096); + if (local_size > 2 * 4096) + EMIT_MOV(compiler, TMP_REG1, 0, SLJIT_MEM1(SLJIT_SP), -4096 * 2); + if (local_size > 3 * 4096) + EMIT_MOV(compiler, TMP_REG1, 0, SLJIT_MEM1(SLJIT_SP), -4096 * 3); + } + else { + EMIT_MOV(compiler, SLJIT_R0, 0, SLJIT_SP, 0); + EMIT_MOV(compiler, SLJIT_R1, 0, SLJIT_IMM, (local_size - 1) >> 12); + + SLJIT_ASSERT (reg_map[SLJIT_R0] == 0); + + EMIT_MOV(compiler, TMP_REG1, 0, SLJIT_MEM1(SLJIT_R0), -4096); + FAIL_IF(emit_non_cum_binary(compiler, BINARY_OPCODE(SUB), + SLJIT_R0, 0, SLJIT_R0, 0, SLJIT_IMM, 4096)); + FAIL_IF(emit_non_cum_binary(compiler, BINARY_OPCODE(SUB), + SLJIT_R1, 0, SLJIT_R1, 0, SLJIT_IMM, 1)); + + inst = (sljit_u8*)ensure_buf(compiler, 1 + 2); + FAIL_IF(!inst); + + INC_SIZE(2); + inst[0] = JNE_i8; + inst[1] = (sljit_s8) -16; + } + + EMIT_MOV(compiler, TMP_REG1, 0, SLJIT_MEM1(SLJIT_SP), -local_size); } #endif @@ -192,12 +214,12 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit EMIT_MOV(compiler, TMP_REG1, 0, SLJIT_SP, 0); /* Some space might allocated during sljit_grow_stack() above on WIN32. */ - FAIL_IF(emit_non_cum_binary(compiler, SUB_r_rm, SUB_rm_r, SUB, SUB_EAX_i32, + FAIL_IF(emit_non_cum_binary(compiler, BINARY_OPCODE(SUB), SLJIT_SP, 0, SLJIT_SP, 0, SLJIT_IMM, local_size + sizeof(sljit_sw))); #if defined _WIN32 && !(defined SLJIT_X86_32_FASTCALL && SLJIT_X86_32_FASTCALL) if (compiler->local_size > 1024) - FAIL_IF(emit_cum_binary(compiler, ADD_r_rm, ADD_rm_r, ADD, ADD_EAX_i32, + FAIL_IF(emit_cum_binary(compiler, BINARY_OPCODE(ADD), TMP_REG1, 0, TMP_REG1, 0, SLJIT_IMM, sizeof(sljit_sw))); #endif @@ -213,31 +235,29 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit return emit_mov(compiler, SLJIT_MEM1(SLJIT_SP), compiler->local_size, TMP_REG1, 0); } #endif - return emit_non_cum_binary(compiler, SUB_r_rm, SUB_rm_r, SUB, SUB_EAX_i32, + return emit_non_cum_binary(compiler, BINARY_OPCODE(SUB), SLJIT_SP, 0, SLJIT_SP, 0, SLJIT_IMM, local_size); } SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit_set_context(struct sljit_compiler *compiler, - sljit_s32 options, sljit_s32 args, sljit_s32 scratches, sljit_s32 saveds, + sljit_s32 options, sljit_s32 arg_types, sljit_s32 scratches, sljit_s32 saveds, sljit_s32 fscratches, sljit_s32 fsaveds, sljit_s32 local_size) { CHECK_ERROR(); - CHECK(check_sljit_set_context(compiler, options, args, scratches, saveds, fscratches, fsaveds, local_size)); - set_set_context(compiler, options, args, scratches, saveds, fscratches, fsaveds, local_size); + CHECK(check_sljit_set_context(compiler, options, arg_types, scratches, saveds, fscratches, fsaveds, local_size)); + set_set_context(compiler, options, arg_types, scratches, saveds, fscratches, fsaveds, local_size); - compiler->args = args; + compiler->args = get_arg_count(arg_types); -#if (defined SLJIT_X86_32_FASTCALL && SLJIT_X86_32_FASTCALL) - /* [esp+0] for saving temporaries and third argument for calls. */ - compiler->saveds_offset = 1 * sizeof(sljit_sw); -#else - /* [esp+0] for saving temporaries and space for maximum three arguments. */ - if (scratches <= 1) - compiler->saveds_offset = 1 * sizeof(sljit_sw); - else - compiler->saveds_offset = ((scratches == 2) ? 2 : 3) * sizeof(sljit_sw); + /* [esp+0] for saving temporaries and function calls. */ + compiler->stack_tmp_size = 2 * sizeof(sljit_sw); + +#if !(defined SLJIT_X86_32_FASTCALL && SLJIT_X86_32_FASTCALL) + if (scratches > 3) + compiler->stack_tmp_size = 3 * sizeof(sljit_sw); #endif + compiler->saveds_offset = compiler->stack_tmp_size; if (scratches > 3) compiler->saveds_offset += ((scratches > (3 + 6)) ? 6 : (scratches - 3)) * sizeof(sljit_sw); @@ -278,10 +298,10 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit if (compiler->options & SLJIT_F64_ALIGNMENT) EMIT_MOV(compiler, SLJIT_SP, 0, SLJIT_MEM1(SLJIT_SP), compiler->local_size) else - FAIL_IF(emit_cum_binary(compiler, ADD_r_rm, ADD_rm_r, ADD, ADD_EAX_i32, + FAIL_IF(emit_cum_binary(compiler, BINARY_OPCODE(ADD), SLJIT_SP, 0, SLJIT_SP, 0, SLJIT_IMM, compiler->local_size)); #else - FAIL_IF(emit_cum_binary(compiler, ADD_r_rm, ADD_rm_r, ADD, ADD_EAX_i32, + FAIL_IF(emit_cum_binary(compiler, BINARY_OPCODE(ADD), SLJIT_SP, 0, SLJIT_SP, 0, SLJIT_IMM, compiler->local_size)); #endif @@ -418,7 +438,7 @@ static sljit_u8* emit_x86_instruction(st if ((flags & EX86_BIN_INS) && (a & SLJIT_IMM)) *inst = (flags & EX86_BYTE_ARG) ? GROUP_BINARY_83 : GROUP_BINARY_81; - if ((a & SLJIT_IMM) || (a == 0)) + if (a & SLJIT_IMM) *buf_ptr = 0; else if (!(flags & EX86_SSE2_OP1)) *buf_ptr = reg_map[a] << 3; @@ -490,42 +510,324 @@ static sljit_u8* emit_x86_instruction(st /* Call / return instructions */ /* --------------------------------------------------------------------- */ -static SLJIT_INLINE sljit_s32 call_with_args(struct sljit_compiler *compiler, sljit_s32 type) +#if (defined SLJIT_X86_32_FASTCALL && SLJIT_X86_32_FASTCALL) + +static sljit_s32 c_fast_call_get_stack_size(sljit_s32 arg_types, sljit_s32 *word_arg_count_ptr) { - sljit_u8 *inst; + sljit_s32 stack_size = 0; + sljit_s32 word_arg_count = 0; -#if (defined SLJIT_X86_32_FASTCALL && SLJIT_X86_32_FASTCALL) - inst = (sljit_u8*)ensure_buf(compiler, type >= SLJIT_CALL3 ? 1 + 2 + 1 : 1 + 2); - FAIL_IF(!inst); - INC_SIZE(type >= SLJIT_CALL3 ? 2 + 1 : 2); + arg_types >>= SLJIT_DEF_SHIFT; + + while (arg_types) { + switch (arg_types & SLJIT_DEF_MASK) { + case SLJIT_ARG_TYPE_F32: + stack_size += sizeof(sljit_f32); + break; + case SLJIT_ARG_TYPE_F64: + stack_size += sizeof(sljit_f64); + break; + default: + word_arg_count++; + if (word_arg_count > 2) + stack_size += sizeof(sljit_sw); + break; + } - if (type >= SLJIT_CALL3) + arg_types >>= SLJIT_DEF_SHIFT; + } + + if (word_arg_count_ptr) + *word_arg_count_ptr = word_arg_count; + + return stack_size; +} + +static sljit_s32 c_fast_call_with_args(struct sljit_compiler *compiler, + sljit_s32 arg_types, sljit_s32 stack_size, sljit_s32 word_arg_count, sljit_s32 swap_args) +{ + sljit_u8 *inst; + sljit_s32 float_arg_count; + + if (stack_size == sizeof(sljit_sw) && word_arg_count == 3) { + inst = (sljit_u8*)ensure_buf(compiler, 1 + 1); + FAIL_IF(!inst); + INC_SIZE(1); PUSH_REG(reg_map[SLJIT_R2]); - *inst++ = MOV_r_rm; - *inst++ = MOD_REG | (reg_map[SLJIT_R2] << 3) | reg_map[SLJIT_R0]; + } + else if (stack_size > 0) { + if (word_arg_count >= 4) + EMIT_MOV(compiler, TMP_REG1, 0, SLJIT_MEM1(SLJIT_SP), compiler->saveds_offset - sizeof(sljit_sw)); + + FAIL_IF(emit_non_cum_binary(compiler, BINARY_OPCODE(SUB), + SLJIT_SP, 0, SLJIT_SP, 0, SLJIT_IMM, stack_size)); + + stack_size = 0; + arg_types >>= SLJIT_DEF_SHIFT; + word_arg_count = 0; + float_arg_count = 0; + while (arg_types) { + switch (arg_types & SLJIT_DEF_MASK) { + case SLJIT_ARG_TYPE_F32: + float_arg_count++; + FAIL_IF(emit_sse2_store(compiler, 1, SLJIT_MEM1(SLJIT_SP), stack_size, float_arg_count)); + stack_size += sizeof(sljit_f32); + break; + case SLJIT_ARG_TYPE_F64: + float_arg_count++; + FAIL_IF(emit_sse2_store(compiler, 0, SLJIT_MEM1(SLJIT_SP), stack_size, float_arg_count)); + stack_size += sizeof(sljit_f64); + break; + default: + word_arg_count++; + if (word_arg_count == 3) { + EMIT_MOV(compiler, SLJIT_MEM1(SLJIT_SP), stack_size, SLJIT_R2, 0); + stack_size += sizeof(sljit_sw); + } + else if (word_arg_count == 4) { + EMIT_MOV(compiler, SLJIT_MEM1(SLJIT_SP), stack_size, TMP_REG1, 0); + stack_size += sizeof(sljit_sw); + } + break; + } + + arg_types >>= SLJIT_DEF_SHIFT; + } + } + + if (word_arg_count > 0) { + if (swap_args) { + inst = (sljit_u8*)ensure_buf(compiler, 1 + 1); + FAIL_IF(!inst); + INC_SIZE(1); + + *inst++ = XCHG_EAX_r | reg_map[SLJIT_R2]; + } + else { + inst = (sljit_u8*)ensure_buf(compiler, 1 + 2); + FAIL_IF(!inst); + INC_SIZE(2); + + *inst++ = MOV_r_rm; + *inst++ = MOD_REG | (reg_map[SLJIT_R2] << 3) | reg_map[SLJIT_R0]; + } + } + + return SLJIT_SUCCESS; +} + +#endif + +static sljit_s32 cdecl_call_get_stack_size(struct sljit_compiler *compiler, sljit_s32 arg_types, sljit_s32 *word_arg_count_ptr) +{ + sljit_s32 stack_size = 0; + sljit_s32 word_arg_count = 0; + + arg_types >>= SLJIT_DEF_SHIFT; + + while (arg_types) { + switch (arg_types & SLJIT_DEF_MASK) { + case SLJIT_ARG_TYPE_F32: + stack_size += sizeof(sljit_f32); + break; + case SLJIT_ARG_TYPE_F64: + stack_size += sizeof(sljit_f64); + break; + default: + word_arg_count++; + stack_size += sizeof(sljit_sw); + break; + } + + arg_types >>= SLJIT_DEF_SHIFT; + } + + if (word_arg_count_ptr) + *word_arg_count_ptr = word_arg_count; + + if (stack_size <= compiler->stack_tmp_size) + return 0; + +#if defined(__APPLE__) + return ((stack_size - compiler->stack_tmp_size + 15) & ~15); #else - inst = (sljit_u8*)ensure_buf(compiler, 1 + 4 * (type - SLJIT_CALL0)); + return stack_size - compiler->stack_tmp_size; +#endif +} + +static sljit_s32 cdecl_call_with_args(struct sljit_compiler *compiler, + sljit_s32 arg_types, sljit_s32 stack_size, sljit_s32 word_arg_count) +{ + sljit_s32 float_arg_count = 0; + + if (word_arg_count >= 4) + EMIT_MOV(compiler, TMP_REG1, 0, SLJIT_MEM1(SLJIT_SP), compiler->saveds_offset - sizeof(sljit_sw)); + + if (stack_size > 0) + FAIL_IF(emit_non_cum_binary(compiler, BINARY_OPCODE(SUB), + SLJIT_SP, 0, SLJIT_SP, 0, SLJIT_IMM, stack_size)); + + stack_size = 0; + word_arg_count = 0; + arg_types >>= SLJIT_DEF_SHIFT; + + while (arg_types) { + switch (arg_types & SLJIT_DEF_MASK) { + case SLJIT_ARG_TYPE_F32: + float_arg_count++; + FAIL_IF(emit_sse2_store(compiler, 1, SLJIT_MEM1(SLJIT_SP), stack_size, float_arg_count)); + stack_size += sizeof(sljit_f32); + break; + case SLJIT_ARG_TYPE_F64: + float_arg_count++; + FAIL_IF(emit_sse2_store(compiler, 0, SLJIT_MEM1(SLJIT_SP), stack_size, float_arg_count)); + stack_size += sizeof(sljit_f64); + break; + default: + word_arg_count++; + EMIT_MOV(compiler, SLJIT_MEM1(SLJIT_SP), stack_size, (word_arg_count >= 4) ? TMP_REG1 : word_arg_count, 0); + stack_size += sizeof(sljit_sw); + break; + } + + arg_types >>= SLJIT_DEF_SHIFT; + } + + return SLJIT_SUCCESS; +} + +static sljit_s32 post_call_with_args(struct sljit_compiler *compiler, + sljit_s32 arg_types, sljit_s32 stack_size) +{ + sljit_u8 *inst; + sljit_s32 single; + + if (stack_size > 0) + FAIL_IF(emit_cum_binary(compiler, BINARY_OPCODE(ADD), + SLJIT_SP, 0, SLJIT_SP, 0, SLJIT_IMM, stack_size)); + + if ((arg_types & SLJIT_DEF_MASK) < SLJIT_ARG_TYPE_F32) + return SLJIT_SUCCESS; + + single = ((arg_types & SLJIT_DEF_MASK) == SLJIT_ARG_TYPE_F32); + + inst = (sljit_u8*)ensure_buf(compiler, 1 + 3); FAIL_IF(!inst); - INC_SIZE(4 * (type - SLJIT_CALL0)); + INC_SIZE(3); + inst[0] = single ? FSTPS : FSTPD; + inst[1] = (0x03 << 3) | 0x04; + inst[2] = (0x04 << 3) | reg_map[SLJIT_SP]; + + return emit_sse2_load(compiler, single, SLJIT_FR0, SLJIT_MEM1(SLJIT_SP), 0); +} - *inst++ = MOV_rm_r; - *inst++ = MOD_DISP8 | (reg_map[SLJIT_R0] << 3) | 0x4 /* SIB */; - *inst++ = (0x4 /* none*/ << 3) | reg_map[SLJIT_SP]; - *inst++ = 0; - if (type >= SLJIT_CALL2) { - *inst++ = MOV_rm_r; - *inst++ = MOD_DISP8 | (reg_map[SLJIT_R1] << 3) | 0x4 /* SIB */; - *inst++ = (0x4 /* none*/ << 3) | reg_map[SLJIT_SP]; - *inst++ = sizeof(sljit_sw); - } - if (type >= SLJIT_CALL3) { - *inst++ = MOV_rm_r; - *inst++ = MOD_DISP8 | (reg_map[SLJIT_R2] << 3) | 0x4 /* SIB */; - *inst++ = (0x4 /* none*/ << 3) | reg_map[SLJIT_SP]; - *inst++ = 2 * sizeof(sljit_sw); +SLJIT_API_FUNC_ATTRIBUTE struct sljit_jump* sljit_emit_call(struct sljit_compiler *compiler, sljit_s32 type, + sljit_s32 arg_types) +{ + struct sljit_jump *jump; + sljit_s32 stack_size = 0; + sljit_s32 word_arg_count; + + CHECK_ERROR_PTR(); + CHECK_PTR(check_sljit_emit_call(compiler, type, arg_types)); + +#if (defined SLJIT_X86_32_FASTCALL && SLJIT_X86_32_FASTCALL) + if ((type & 0xff) == SLJIT_CALL) { + stack_size = c_fast_call_get_stack_size(arg_types, &word_arg_count); + PTR_FAIL_IF(c_fast_call_with_args(compiler, arg_types, stack_size, word_arg_count, 0)); + +#if (defined SLJIT_VERBOSE && SLJIT_VERBOSE) \ + || (defined SLJIT_ARGUMENT_CHECKS && SLJIT_ARGUMENT_CHECKS) + compiler->skip_checks = 1; +#endif + + jump = sljit_emit_jump(compiler, type); + PTR_FAIL_IF(jump == NULL); + + PTR_FAIL_IF(post_call_with_args(compiler, arg_types, 0)); + return jump; } #endif - return SLJIT_SUCCESS; + + stack_size = cdecl_call_get_stack_size(compiler, arg_types, &word_arg_count); + PTR_FAIL_IF(cdecl_call_with_args(compiler, arg_types, stack_size, word_arg_count)); + +#if (defined SLJIT_VERBOSE && SLJIT_VERBOSE) \ + || (defined SLJIT_ARGUMENT_CHECKS && SLJIT_ARGUMENT_CHECKS) + compiler->skip_checks = 1; +#endif + + jump = sljit_emit_jump(compiler, type); + PTR_FAIL_IF(jump == NULL); + + PTR_FAIL_IF(post_call_with_args(compiler, arg_types, stack_size)); + return jump; +} + +SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit_emit_icall(struct sljit_compiler *compiler, sljit_s32 type, + sljit_s32 arg_types, + sljit_s32 src, sljit_sw srcw) +{ + sljit_s32 stack_size = 0; + sljit_s32 word_arg_count; +#if (defined SLJIT_X86_32_FASTCALL && SLJIT_X86_32_FASTCALL) + sljit_s32 swap_args; +#endif + + CHECK_ERROR(); + CHECK(check_sljit_emit_icall(compiler, type, arg_types, src, srcw)); + +#if (defined SLJIT_X86_32_FASTCALL && SLJIT_X86_32_FASTCALL) + SLJIT_ASSERT(reg_map[SLJIT_R0] == 0 && reg_map[SLJIT_R2] == 1 && SLJIT_R0 == 1 && SLJIT_R2 == 3); + + if ((type & 0xff) == SLJIT_CALL) { + stack_size = c_fast_call_get_stack_size(arg_types, &word_arg_count); + swap_args = 0; + + if (word_arg_count > 0) { + if ((src & REG_MASK) == SLJIT_R2 || OFFS_REG(src) == SLJIT_R2) { + swap_args = 1; + if (((src & REG_MASK) | 0x2) == SLJIT_R2) + src ^= 0x2; + if ((OFFS_REG(src) | 0x2) == SLJIT_R2) + src ^= TO_OFFS_REG(0x2); + } + } + + FAIL_IF(c_fast_call_with_args(compiler, arg_types, stack_size, word_arg_count, swap_args)); + + compiler->saveds_offset += stack_size; + compiler->locals_offset += stack_size; + +#if (defined SLJIT_VERBOSE && SLJIT_VERBOSE) \ + || (defined SLJIT_ARGUMENT_CHECKS && SLJIT_ARGUMENT_CHECKS) + compiler->skip_checks = 1; +#endif + FAIL_IF(sljit_emit_ijump(compiler, type, src, srcw)); + + compiler->saveds_offset -= stack_size; + compiler->locals_offset -= stack_size; + + return post_call_with_args(compiler, arg_types, 0); + } +#endif + + stack_size = cdecl_call_get_stack_size(compiler, arg_types, &word_arg_count); + FAIL_IF(cdecl_call_with_args(compiler, arg_types, stack_size, word_arg_count)); + + compiler->saveds_offset += stack_size; + compiler->locals_offset += stack_size; + +#if (defined SLJIT_VERBOSE && SLJIT_VERBOSE) \ + || (defined SLJIT_ARGUMENT_CHECKS && SLJIT_ARGUMENT_CHECKS) + compiler->skip_checks = 1; +#endif + FAIL_IF(sljit_emit_ijump(compiler, type, src, srcw)); + + compiler->saveds_offset -= stack_size; + compiler->locals_offset -= stack_size; + + return post_call_with_args(compiler, arg_types, stack_size); } SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit_emit_fast_enter(struct sljit_compiler *compiler, sljit_s32 dst, sljit_sw dstw) @@ -576,7 +878,7 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit INC_SIZE(1 + 1); PUSH_REG(reg_map[src]); } - else if (src & SLJIT_MEM) { + else { inst = emit_x86_instruction(compiler, 1, 0, 0, src, srcw); FAIL_IF(!inst); *inst++ = GROUP_FF; @@ -586,16 +888,6 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit FAIL_IF(!inst); INC_SIZE(1); } - else { - /* SLJIT_IMM. */ - inst = (sljit_u8*)ensure_buf(compiler, 1 + 5 + 1); - FAIL_IF(!inst); - - INC_SIZE(5 + 1); - *inst++ = PUSH_i32; - sljit_unaligned_store_sw(inst, srcw); - inst += sizeof(sljit_sw); - } RET(); return SLJIT_SUCCESS;
Modified: tomcat/jk/trunk/native/iis/pcre/sljit/sljitNativeX86_64.c URL: http://svn.apache.org/viewvc/tomcat/jk/trunk/native/iis/pcre/sljit/sljitNativeX86_64.c?rev=1838867&r1=1838866&r2=1838867&view=diff ============================================================================== --- tomcat/jk/trunk/native/iis/pcre/sljit/sljitNativeX86_64.c (original) +++ tomcat/jk/trunk/native/iis/pcre/sljit/sljitNativeX86_64.c Fri Aug 24 14:36:12 2018 @@ -41,24 +41,31 @@ static sljit_s32 emit_load_imm64(struct static sljit_u8* generate_far_jump_code(struct sljit_jump *jump, sljit_u8 *code_ptr, sljit_s32 type) { + int short_addr = !(jump->flags & SLJIT_REWRITABLE_JUMP) && !(jump->flags & JUMP_LABEL) && (jump->u.target <= 0xffffffff); + + /* The relative jump below specialized for this case. */ + SLJIT_ASSERT(reg_map[TMP_REG2] >= 8); + if (type < SLJIT_JUMP) { /* Invert type. */ *code_ptr++ = get_jump_code(type ^ 0x1) - 0x10; - *code_ptr++ = 10 + 3; + *code_ptr++ = short_addr ? (6 + 3) : (10 + 3); } - *code_ptr++ = REX_W | ((reg_map[TMP_REG2] <= 7) ? 0 : REX_B); + *code_ptr++ = short_addr ? REX_B : (REX_W | REX_B); *code_ptr++ = MOV_r_i32 | reg_lmap[TMP_REG2]; jump->addr = (sljit_uw)code_ptr; if (jump->flags & JUMP_LABEL) jump->flags |= PATCH_MD; + else if (short_addr) + sljit_unaligned_store_s32(code_ptr, (sljit_s32)jump->u.target); else sljit_unaligned_store_sw(code_ptr, jump->u.target); - code_ptr += sizeof(sljit_sw); - if (reg_map[TMP_REG2] >= 8) - *code_ptr++ = REX_B; + code_ptr += short_addr ? sizeof(sljit_s32) : sizeof(sljit_sw); + + *code_ptr++ = REX_B; *code_ptr++ = GROUP_FF; *code_ptr++ = MOD_REG | (type >= SLJIT_FAST_CALL ? CALL_rm : JMP_rm) | reg_lmap[TMP_REG2]; @@ -66,15 +73,17 @@ static sljit_u8* generate_far_jump_code( } SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit_emit_enter(struct sljit_compiler *compiler, - sljit_s32 options, sljit_s32 args, sljit_s32 scratches, sljit_s32 saveds, + sljit_s32 options, sljit_s32 arg_types, sljit_s32 scratches, sljit_s32 saveds, sljit_s32 fscratches, sljit_s32 fsaveds, sljit_s32 local_size) { - sljit_s32 i, tmp, size, saved_register_size; + sljit_s32 args, i, tmp, size, saved_register_size; sljit_u8 *inst; CHECK_ERROR(); - CHECK(check_sljit_emit_enter(compiler, options, args, scratches, saveds, fscratches, fsaveds, local_size)); - set_emit_enter(compiler, options, args, scratches, saveds, fscratches, fsaveds, local_size); + CHECK(check_sljit_emit_enter(compiler, options, arg_types, scratches, saveds, fscratches, fsaveds, local_size)); + set_emit_enter(compiler, options, arg_types, scratches, saveds, fscratches, fsaveds, local_size); + + compiler->mode32 = 0; #ifdef _WIN64 /* Two/four register slots for parameters plus space for xmm6 register if needed. */ @@ -108,6 +117,8 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit PUSH_REG(reg_lmap[i]); } + args = get_arg_count(arg_types); + if (args > 0) { size = args * 3; inst = (sljit_u8*)ensure_buf(compiler, 1 + size); @@ -117,35 +128,39 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit #ifndef _WIN64 if (args > 0) { - *inst++ = REX_W; - *inst++ = MOV_r_rm; - *inst++ = MOD_REG | (reg_map[SLJIT_S0] << 3) | 0x7 /* rdi */; + inst[0] = REX_W; + inst[1] = MOV_r_rm; + inst[2] = MOD_REG | (reg_map[SLJIT_S0] << 3) | 0x7 /* rdi */; + inst += 3; } if (args > 1) { - *inst++ = REX_W | REX_R; - *inst++ = MOV_r_rm; - *inst++ = MOD_REG | (reg_lmap[SLJIT_S1] << 3) | 0x6 /* rsi */; + inst[0] = REX_W | REX_R; + inst[1] = MOV_r_rm; + inst[2] = MOD_REG | (reg_lmap[SLJIT_S1] << 3) | 0x6 /* rsi */; + inst += 3; } if (args > 2) { - *inst++ = REX_W | REX_R; - *inst++ = MOV_r_rm; - *inst++ = MOD_REG | (reg_lmap[SLJIT_S2] << 3) | 0x2 /* rdx */; + inst[0] = REX_W | REX_R; + inst[1] = MOV_r_rm; + inst[2] = MOD_REG | (reg_lmap[SLJIT_S2] << 3) | 0x2 /* rdx */; } #else if (args > 0) { - *inst++ = REX_W; - *inst++ = MOV_r_rm; - *inst++ = MOD_REG | (reg_map[SLJIT_S0] << 3) | 0x1 /* rcx */; + inst[0] = REX_W; + inst[1] = MOV_r_rm; + inst[2] = MOD_REG | (reg_map[SLJIT_S0] << 3) | 0x1 /* rcx */; + inst += 3; } if (args > 1) { - *inst++ = REX_W; - *inst++ = MOV_r_rm; - *inst++ = MOD_REG | (reg_map[SLJIT_S1] << 3) | 0x2 /* rdx */; + inst[0] = REX_W; + inst[1] = MOV_r_rm; + inst[2] = MOD_REG | (reg_map[SLJIT_S1] << 3) | 0x2 /* rdx */; + inst += 3; } if (args > 2) { - *inst++ = REX_W | REX_B; - *inst++ = MOV_r_rm; - *inst++ = MOD_REG | (reg_map[SLJIT_S2] << 3) | 0x0 /* r8 */; + inst[0] = REX_W | REX_B; + inst[1] = MOV_r_rm; + inst[2] = MOD_REG | (reg_map[SLJIT_S2] << 3) | 0x0 /* r8 */; } #endif } @@ -154,58 +169,42 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit compiler->local_size = local_size; #ifdef _WIN64 - if (local_size > 1024) { - /* Allocate stack for the callback, which grows the stack. */ - inst = (sljit_u8*)ensure_buf(compiler, 1 + 4 + (3 + sizeof(sljit_s32))); - FAIL_IF(!inst); - INC_SIZE(4 + (3 + sizeof(sljit_s32))); - *inst++ = REX_W; - *inst++ = GROUP_BINARY_83; - *inst++ = MOD_REG | SUB | reg_map[SLJIT_SP]; - /* Allocated size for registers must be divisible by 8. */ - SLJIT_ASSERT(!(saved_register_size & 0x7)); - /* Aligned to 16 byte. */ - if (saved_register_size & 0x8) { - *inst++ = 5 * sizeof(sljit_sw); - local_size -= 5 * sizeof(sljit_sw); - } else { - *inst++ = 4 * sizeof(sljit_sw); - local_size -= 4 * sizeof(sljit_sw); - } - /* Second instruction */ - SLJIT_ASSERT(reg_map[SLJIT_R0] < 8); - *inst++ = REX_W; - *inst++ = MOV_rm_i32; - *inst++ = MOD_REG | reg_lmap[SLJIT_R0]; - sljit_unaligned_store_s32(inst, local_size); -#if (defined SLJIT_VERBOSE && SLJIT_VERBOSE) \ - || (defined SLJIT_ARGUMENT_CHECKS && SLJIT_ARGUMENT_CHECKS) - compiler->skip_checks = 1; -#endif - FAIL_IF(sljit_emit_ijump(compiler, SLJIT_CALL1, SLJIT_IMM, SLJIT_FUNC_OFFSET(sljit_grow_stack))); - } -#endif - if (local_size > 0) { - if (local_size <= 127) { - inst = (sljit_u8*)ensure_buf(compiler, 1 + 4); - FAIL_IF(!inst); - INC_SIZE(4); - *inst++ = REX_W; - *inst++ = GROUP_BINARY_83; - *inst++ = MOD_REG | SUB | reg_map[SLJIT_SP]; - *inst++ = local_size; + if (local_size <= 4 * 4096) { + if (local_size > 4096) + EMIT_MOV(compiler, TMP_REG1, 0, SLJIT_MEM1(SLJIT_SP), -4096); + if (local_size > 2 * 4096) + EMIT_MOV(compiler, TMP_REG1, 0, SLJIT_MEM1(SLJIT_SP), -4096 * 2); + if (local_size > 3 * 4096) + EMIT_MOV(compiler, TMP_REG1, 0, SLJIT_MEM1(SLJIT_SP), -4096 * 3); } else { - inst = (sljit_u8*)ensure_buf(compiler, 1 + 7); + EMIT_MOV(compiler, SLJIT_R0, 0, SLJIT_SP, 0); + EMIT_MOV(compiler, TMP_REG1, 0, SLJIT_IMM, (local_size - 1) >> 12); + + SLJIT_ASSERT (reg_map[SLJIT_R0] == 0); + + EMIT_MOV(compiler, TMP_REG2, 0, SLJIT_MEM1(SLJIT_R0), -4096); + FAIL_IF(emit_non_cum_binary(compiler, BINARY_OPCODE(SUB), + SLJIT_R0, 0, SLJIT_R0, 0, SLJIT_IMM, 4096)); + FAIL_IF(emit_non_cum_binary(compiler, BINARY_OPCODE(SUB), + TMP_REG1, 0, TMP_REG1, 0, SLJIT_IMM, 1)); + + inst = (sljit_u8*)ensure_buf(compiler, 1 + 2); FAIL_IF(!inst); - INC_SIZE(7); - *inst++ = REX_W; - *inst++ = GROUP_BINARY_81; - *inst++ = MOD_REG | SUB | reg_map[SLJIT_SP]; - sljit_unaligned_store_s32(inst, local_size); - inst += sizeof(sljit_s32); + + INC_SIZE(2); + inst[0] = JNE_i8; + inst[1] = (sljit_s8) -19; } + + EMIT_MOV(compiler, TMP_REG1, 0, SLJIT_MEM1(SLJIT_SP), -local_size); + } +#endif + + if (local_size > 0) { + FAIL_IF(emit_non_cum_binary(compiler, BINARY_OPCODE(SUB), + SLJIT_SP, 0, SLJIT_SP, 0, SLJIT_IMM, local_size)); } #ifdef _WIN64 @@ -223,14 +222,14 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit } SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit_set_context(struct sljit_compiler *compiler, - sljit_s32 options, sljit_s32 args, sljit_s32 scratches, sljit_s32 saveds, + sljit_s32 options, sljit_s32 arg_types, sljit_s32 scratches, sljit_s32 saveds, sljit_s32 fscratches, sljit_s32 fsaveds, sljit_s32 local_size) { sljit_s32 saved_register_size; CHECK_ERROR(); - CHECK(check_sljit_set_context(compiler, options, args, scratches, saveds, fscratches, fsaveds, local_size)); - set_set_context(compiler, options, args, scratches, saveds, fscratches, fsaveds, local_size); + CHECK(check_sljit_set_context(compiler, options, arg_types, scratches, saveds, fscratches, fsaveds, local_size)); + set_set_context(compiler, options, arg_types, scratches, saveds, fscratches, fsaveds, local_size); #ifdef _WIN64 /* Two/four register slots for parameters plus space for xmm6 register if needed. */ @@ -414,7 +413,11 @@ static sljit_u8* emit_x86_instruction(st } } } - else if (!(flags & EX86_SSE2_OP2) && reg_map[b] >= 8) + else if (!(flags & EX86_SSE2_OP2)) { + if (reg_map[b] >= 8) + rex |= REX_B; + } + else if (freg_map[b] >= 8) rex |= REX_B; if (a & SLJIT_IMM) { @@ -441,7 +444,11 @@ static sljit_u8* emit_x86_instruction(st else { SLJIT_ASSERT(!(flags & EX86_SHIFT_INS) || a == SLJIT_PREF_SHIFT_REG); /* reg_map[SLJIT_PREF_SHIFT_REG] is less than 8. */ - if (!(flags & EX86_SSE2_OP1) && reg_map[a] >= 8) + if (!(flags & EX86_SSE2_OP1)) { + if (reg_map[a] >= 8) + rex |= REX_R; + } + else if (freg_map[a] >= 8) rex |= REX_R; } @@ -468,12 +475,12 @@ static sljit_u8* emit_x86_instruction(st if ((flags & EX86_BIN_INS) && (a & SLJIT_IMM)) *inst = (flags & EX86_BYTE_ARG) ? GROUP_BINARY_83 : GROUP_BINARY_81; - if ((a & SLJIT_IMM) || (a == 0)) + if (a & SLJIT_IMM) *buf_ptr = 0; else if (!(flags & EX86_SSE2_OP1)) *buf_ptr = reg_lmap[a] << 3; else - *buf_ptr = a << 3; + *buf_ptr = freg_lmap[a] << 3; } else { if (a & SLJIT_IMM) { @@ -487,7 +494,7 @@ static sljit_u8* emit_x86_instruction(st } if (!(b & SLJIT_MEM)) - *buf_ptr++ |= MOD_REG + ((!(flags & EX86_SSE2_OP2)) ? reg_lmap[b] : b); + *buf_ptr++ |= MOD_REG + ((!(flags & EX86_SSE2_OP2)) ? reg_lmap[b] : freg_lmap[b]); else if ((b & REG_MASK) != SLJIT_UNUSED) { if ((b & OFFS_REG_MASK) == SLJIT_UNUSED || (b & OFFS_REG_MASK) == TO_OFFS_REG(SLJIT_SP)) { if (immb != 0 || reg_lmap[b & REG_MASK] == 5) { @@ -545,45 +552,161 @@ static sljit_u8* emit_x86_instruction(st /* Call / return instructions */ /* --------------------------------------------------------------------- */ -static sljit_s32 call_with_args(struct sljit_compiler *compiler, sljit_s32 type) +#ifndef _WIN64 + +static sljit_s32 call_with_args(struct sljit_compiler *compiler, sljit_s32 arg_types, sljit_s32 *src_ptr, sljit_sw srcw) { - sljit_u8 *inst; + sljit_s32 src = src_ptr ? (*src_ptr) : 0; + sljit_s32 word_arg_count = 0; - /* After any change update IS_REG_CHANGED_BY_CALL as well. */ -#ifndef _WIN64 - SLJIT_ASSERT(reg_map[SLJIT_R1] == 6 && reg_map[SLJIT_R0] < 8 && reg_map[SLJIT_R2] < 8 && reg_map[TMP_REG1] == 2); + SLJIT_ASSERT(reg_map[SLJIT_R1] == 6 && reg_map[SLJIT_R3] == 1 && reg_map[TMP_REG1] == 2); + + compiler->mode32 = 0; + + /* Remove return value. */ + arg_types >>= SLJIT_DEF_SHIFT; + + while (arg_types) { + if ((arg_types & SLJIT_DEF_MASK) < SLJIT_ARG_TYPE_F32) + word_arg_count++; + arg_types >>= SLJIT_DEF_SHIFT; + } + + if (word_arg_count == 0) + return SLJIT_SUCCESS; + + if (src & SLJIT_MEM) { + ADJUST_LOCAL_OFFSET(src, srcw); + EMIT_MOV(compiler, TMP_REG2, 0, src, srcw); + *src_ptr = TMP_REG2; + } + else if (src == SLJIT_R2 && word_arg_count >= SLJIT_R2) + *src_ptr = TMP_REG1; + + if (word_arg_count >= 3) + EMIT_MOV(compiler, TMP_REG1, 0, SLJIT_R2, 0); + return emit_mov(compiler, SLJIT_R2, 0, SLJIT_R0, 0); +} - inst = (sljit_u8*)ensure_buf(compiler, 1 + ((type < SLJIT_CALL3) ? 3 : 6)); - FAIL_IF(!inst); - INC_SIZE((type < SLJIT_CALL3) ? 3 : 6); - if (type >= SLJIT_CALL3) { - /* Move third argument to TMP_REG1. */ - *inst++ = REX_W; - *inst++ = MOV_r_rm; - *inst++ = MOD_REG | (0x2 /* rdx */ << 3) | reg_lmap[SLJIT_R2]; - } - *inst++ = REX_W; - *inst++ = MOV_r_rm; - *inst++ = MOD_REG | (0x7 /* rdi */ << 3) | reg_lmap[SLJIT_R0]; #else - SLJIT_ASSERT(reg_map[SLJIT_R1] == 2 && reg_map[SLJIT_R0] < 8 && reg_map[SLJIT_R2] < 8 && reg_map[TMP_REG1] == 8); - inst = (sljit_u8*)ensure_buf(compiler, 1 + ((type < SLJIT_CALL3) ? 3 : 6)); - FAIL_IF(!inst); - INC_SIZE((type < SLJIT_CALL3) ? 3 : 6); - if (type >= SLJIT_CALL3) { - /* Move third argument to TMP_REG1. */ - *inst++ = REX_W | REX_R; - *inst++ = MOV_r_rm; - *inst++ = MOD_REG | (0x0 /* r8 */ << 3) | reg_lmap[SLJIT_R2]; - } - *inst++ = REX_W; - *inst++ = MOV_r_rm; - *inst++ = MOD_REG | (0x1 /* rcx */ << 3) | reg_lmap[SLJIT_R0]; -#endif +static sljit_s32 call_with_args(struct sljit_compiler *compiler, sljit_s32 arg_types, sljit_s32 *src_ptr, sljit_sw srcw) +{ + sljit_s32 src = src_ptr ? (*src_ptr) : 0; + sljit_s32 arg_count = 0; + sljit_s32 word_arg_count = 0; + sljit_s32 float_arg_count = 0; + sljit_s32 types = 0; + sljit_s32 data_trandfer = 0; + static sljit_u8 word_arg_regs[5] = { 0, SLJIT_R3, SLJIT_R1, SLJIT_R2, TMP_REG1 }; + + SLJIT_ASSERT(reg_map[SLJIT_R3] == 1 && reg_map[SLJIT_R1] == 2 && reg_map[SLJIT_R2] == 8 && reg_map[TMP_REG1] == 9); + + compiler->mode32 = 0; + arg_types >>= SLJIT_DEF_SHIFT; + + while (arg_types) { + types = (types << SLJIT_DEF_SHIFT) | (arg_types & SLJIT_DEF_MASK); + + switch (arg_types & SLJIT_DEF_MASK) { + case SLJIT_ARG_TYPE_F32: + case SLJIT_ARG_TYPE_F64: + arg_count++; + float_arg_count++; + + if (arg_count != float_arg_count) + data_trandfer = 1; + break; + default: + arg_count++; + word_arg_count++; + + if (arg_count != word_arg_count || arg_count != word_arg_regs[arg_count]) { + data_trandfer = 1; + + if (src == word_arg_regs[arg_count]) { + EMIT_MOV(compiler, TMP_REG2, 0, src, 0); + *src_ptr = TMP_REG2; + } + } + break; + } + + arg_types >>= SLJIT_DEF_SHIFT; + } + + if (!data_trandfer) + return SLJIT_SUCCESS; + + if (src & SLJIT_MEM) { + ADJUST_LOCAL_OFFSET(src, srcw); + EMIT_MOV(compiler, TMP_REG2, 0, src, srcw); + *src_ptr = TMP_REG2; + } + + while (types) { + switch (types & SLJIT_DEF_MASK) { + case SLJIT_ARG_TYPE_F32: + if (arg_count != float_arg_count) + FAIL_IF(emit_sse2_load(compiler, 1, arg_count, float_arg_count, 0)); + arg_count--; + float_arg_count--; + break; + case SLJIT_ARG_TYPE_F64: + if (arg_count != float_arg_count) + FAIL_IF(emit_sse2_load(compiler, 0, arg_count, float_arg_count, 0)); + arg_count--; + float_arg_count--; + break; + default: + if (arg_count != word_arg_count || arg_count != word_arg_regs[arg_count]) + EMIT_MOV(compiler, word_arg_regs[arg_count], 0, word_arg_count, 0); + arg_count--; + word_arg_count--; + break; + } + + types >>= SLJIT_DEF_SHIFT; + } + return SLJIT_SUCCESS; } +#endif + +SLJIT_API_FUNC_ATTRIBUTE struct sljit_jump* sljit_emit_call(struct sljit_compiler *compiler, sljit_s32 type, + sljit_s32 arg_types) +{ + CHECK_ERROR_PTR(); + CHECK_PTR(check_sljit_emit_call(compiler, type, arg_types)); + + PTR_FAIL_IF(call_with_args(compiler, arg_types, NULL, 0)); + +#if (defined SLJIT_VERBOSE && SLJIT_VERBOSE) \ + || (defined SLJIT_ARGUMENT_CHECKS && SLJIT_ARGUMENT_CHECKS) + compiler->skip_checks = 1; +#endif + + return sljit_emit_jump(compiler, type); +} + +SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit_emit_icall(struct sljit_compiler *compiler, sljit_s32 type, + sljit_s32 arg_types, + sljit_s32 src, sljit_sw srcw) +{ + CHECK_ERROR(); + CHECK(check_sljit_emit_icall(compiler, type, arg_types, src, srcw)); + + FAIL_IF(call_with_args(compiler, arg_types, &src, srcw)); + +#if (defined SLJIT_VERBOSE && SLJIT_VERBOSE) \ + || (defined SLJIT_ARGUMENT_CHECKS && SLJIT_ARGUMENT_CHECKS) + compiler->skip_checks = 1; +#endif + + return sljit_emit_ijump(compiler, type, src, srcw); +} + SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit_emit_fast_enter(struct sljit_compiler *compiler, sljit_s32 dst, sljit_sw dstw) { sljit_u8 *inst; @@ -629,11 +752,6 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit CHECK(check_sljit_emit_fast_return(compiler, src, srcw)); ADJUST_LOCAL_OFFSET(src, srcw); - if ((src & SLJIT_IMM) && NOT_HALFWORD(srcw)) { - FAIL_IF(emit_load_imm64(compiler, TMP_REG1, srcw)); - src = TMP_REG1; - } - if (FAST_IS_REG(src)) { if (reg_map[src] < 8) { inst = (sljit_u8*)ensure_buf(compiler, 1 + 1 + 1); @@ -651,7 +769,7 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit PUSH_REG(reg_lmap[src]); } } - else if (src & SLJIT_MEM) { + else { /* REX_W is not necessary (src is not immediate). */ compiler->mode32 = 1; inst = emit_x86_instruction(compiler, 1, 0, 0, src, srcw); @@ -663,23 +781,11 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit FAIL_IF(!inst); INC_SIZE(1); } - else { - SLJIT_ASSERT(IS_HALFWORD(srcw)); - /* SLJIT_IMM. */ - inst = (sljit_u8*)ensure_buf(compiler, 1 + 5 + 1); - FAIL_IF(!inst); - - INC_SIZE(5 + 1); - *inst++ = PUSH_i32; - sljit_unaligned_store_s32(inst, srcw); - inst += sizeof(sljit_s32); - } RET(); return SLJIT_SUCCESS; } - /* --------------------------------------------------------------------- */ /* Extend input */ /* --------------------------------------------------------------------- */ Modified: tomcat/jk/trunk/native/iis/pcre/sljit/sljitNativeX86_common.c URL: http://svn.apache.org/viewvc/tomcat/jk/trunk/native/iis/pcre/sljit/sljitNativeX86_common.c?rev=1838867&r1=1838866&r2=1838867&view=diff ============================================================================== --- tomcat/jk/trunk/native/iis/pcre/sljit/sljitNativeX86_common.c (original) +++ tomcat/jk/trunk/native/iis/pcre/sljit/sljitNativeX86_common.c Fri Aug 24 14:36:12 2018 @@ -26,7 +26,11 @@ SLJIT_API_FUNC_ATTRIBUTE const char* sljit_get_platform_name(void) { +#if (defined SLJIT_X86_32_FASTCALL && SLJIT_X86_32_FASTCALL) + return "x86" SLJIT_CPUINFO " ABI:fastcall"; +#else return "x86" SLJIT_CPUINFO; +#endif } /* @@ -35,7 +39,7 @@ SLJIT_API_FUNC_ATTRIBUTE const char* slj 1 - ECX 2 - EDX 3 - EBX - 4 - none + 4 - ESP 5 - EBP 6 - ESI 7 - EDI @@ -47,7 +51,7 @@ SLJIT_API_FUNC_ATTRIBUTE const char* slj 1 - RCX 2 - RDX 3 - RBX - 4 - none + 4 - RSP 5 - RBP 6 - RSI 7 - RDI @@ -92,23 +96,32 @@ static const sljit_u8 reg_map[SLJIT_NUMB #ifndef _WIN64 /* Args: rdi(=7), rsi(=6), rdx(=2), rcx(=1), r8, r9. Scratches: rax(=0), r10, r11 */ static const sljit_u8 reg_map[SLJIT_NUMBER_OF_REGISTERS + 4] = { - 0, 0, 6, 1, 7, 8, 11, 10, 12, 5, 13, 14, 15, 3, 4, 2, 9 + 0, 0, 6, 7, 1, 8, 11, 10, 12, 5, 13, 14, 15, 3, 4, 2, 9 }; /* low-map. reg_map & 0x7. */ static const sljit_u8 reg_lmap[SLJIT_NUMBER_OF_REGISTERS + 4] = { - 0, 0, 6, 1, 7, 0, 3, 2, 4, 5, 5, 6, 7, 3, 4, 2, 1 + 0, 0, 6, 7, 1, 0, 3, 2, 4, 5, 5, 6, 7, 3, 4, 2, 1 }; #else /* Args: rcx(=1), rdx(=2), r8, r9. Scratches: rax(=0), r10, r11 */ static const sljit_u8 reg_map[SLJIT_NUMBER_OF_REGISTERS + 4] = { - 0, 0, 2, 1, 10, 11, 12, 5, 13, 14, 15, 7, 6, 3, 4, 8, 9 + 0, 0, 2, 8, 1, 11, 12, 5, 13, 14, 15, 7, 6, 3, 4, 9, 10 }; /* low-map. reg_map & 0x7. */ static const sljit_u8 reg_lmap[SLJIT_NUMBER_OF_REGISTERS + 4] = { - 0, 0, 2, 1, 2, 3, 4, 5, 5, 6, 7, 7, 6, 3, 4, 0, 1 + 0, 0, 2, 0, 1, 3, 4, 5, 5, 6, 7, 7, 6, 3, 4, 1, 2 }; #endif +/* Args: xmm0-xmm3 */ +static const sljit_u8 freg_map[SLJIT_NUMBER_OF_FLOAT_REGISTERS + 1] = { + 4, 0, 1, 2, 3, 5, 6 +}; +/* low-map. freg_map & 0x7. */ +static const sljit_u8 freg_lmap[SLJIT_NUMBER_OF_FLOAT_REGISTERS + 1] = { + 4, 0, 1, 2, 3, 5, 6 +}; + #define REX_W 0x48 #define REX_R 0x44 #define REX_X 0x42 @@ -178,6 +191,8 @@ static const sljit_u8 reg_lmap[SLJIT_NUM #define CVTTSD2SI_r_xm 0x2c #define DIV (/* GROUP_F7 */ 6 << 3) #define DIVSD_x_xm 0x5e +#define FSTPS 0xd9 +#define FSTPD 0xdd #define INT3 0xcc #define IDIV (/* GROUP_F7 */ 7 << 3) #define IMUL (/* GROUP_F7 */ 5 << 3) @@ -462,11 +477,7 @@ static sljit_u8* generate_near_jump_code code_ptr += sizeof(sljit_s8); } else { jump->flags |= PATCH_MW; -#if (defined SLJIT_CONFIG_X86_32 && SLJIT_CONFIG_X86_32) - code_ptr += sizeof(sljit_sw); -#else code_ptr += sizeof(sljit_s32); -#endif } return code_ptr; @@ -613,9 +624,6 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit get_cpu_features(); return cpu_has_cmov; - case SLJIT_HAS_PREF_SHIFT_REG: - return 1; - case SLJIT_HAS_SSE2: #if (defined SLJIT_DETECT_SSE2 && SLJIT_DETECT_SSE2) if (cpu_has_sse2 == -1) @@ -634,14 +642,16 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit /* Operators */ /* --------------------------------------------------------------------- */ +#define BINARY_OPCODE(opcode) (((opcode ## _EAX_i32) << 24) | ((opcode ## _r_rm) << 16) | ((opcode ## _rm_r) << 8) | (opcode)) + static sljit_s32 emit_cum_binary(struct sljit_compiler *compiler, - sljit_u8 op_rm, sljit_u8 op_mr, sljit_u8 op_imm, sljit_u8 op_eax_imm, + sljit_u32 op_types, sljit_s32 dst, sljit_sw dstw, sljit_s32 src1, sljit_sw src1w, sljit_s32 src2, sljit_sw src2w); static sljit_s32 emit_non_cum_binary(struct sljit_compiler *compiler, - sljit_u8 op_rm, sljit_u8 op_mr, sljit_u8 op_imm, sljit_u8 op_eax_imm, + sljit_u32 op_types, sljit_s32 dst, sljit_sw dstw, sljit_s32 src1, sljit_sw src1w, sljit_s32 src2, sljit_sw src2w); @@ -653,22 +663,11 @@ static sljit_s32 emit_mov(struct sljit_c #define EMIT_MOV(compiler, dst, dstw, src, srcw) \ FAIL_IF(emit_mov(compiler, dst, dstw, src, srcw)); -#ifdef _WIN32 -#include <malloc.h> - -static void SLJIT_CALL sljit_grow_stack(sljit_sw local_size) -{ - /* Workaround for calling the internal _chkstk() function on Windows. - This function touches all 4k pages belongs to the requested stack space, - which size is passed in local_size. This is necessary on Windows where - the stack can only grow in 4k steps. However, this function just burn - CPU cycles if the stack is large enough. However, you don't know it in - advance, so it must always be called. I think this is a bad design in - general even if it has some reasons. */ - *(volatile sljit_s32*)alloca(local_size) = 0; -} +static SLJIT_INLINE sljit_s32 emit_sse2_store(struct sljit_compiler *compiler, + sljit_s32 single, sljit_s32 dst, sljit_sw dstw, sljit_s32 src); -#endif +static SLJIT_INLINE sljit_s32 emit_sse2_load(struct sljit_compiler *compiler, + sljit_s32 single, sljit_s32 dst, sljit_s32 src, sljit_sw srcw); #if (defined SLJIT_CONFIG_X86_32 && SLJIT_CONFIG_X86_32) #include "sljitNativeX86_32.c" @@ -1115,7 +1114,7 @@ static sljit_s32 emit_unary(struct sljit return SLJIT_SUCCESS; } - if (dst == SLJIT_UNUSED) + if (SLJIT_UNLIKELY(dst == SLJIT_UNUSED)) dst = TMP_REG1; if (FAST_IS_REG(dst)) { @@ -1182,12 +1181,6 @@ static sljit_s32 emit_clz(struct sljit_c SLJIT_UNUSED_ARG(op_flags); - if (SLJIT_UNLIKELY(src & SLJIT_IMM)) { - EMIT_MOV(compiler, TMP_REG1, 0, SLJIT_IMM, srcw); - src = TMP_REG1; - srcw = 0; - } - if (cpu_has_cmov == -1) get_cpu_features(); @@ -1242,13 +1235,9 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit sljit_s32 dst, sljit_sw dstw, sljit_s32 src, sljit_sw srcw) { - sljit_s32 update = 0; sljit_s32 op_flags = GET_ALL_FLAGS(op); #if (defined SLJIT_CONFIG_X86_32 && SLJIT_CONFIG_X86_32) sljit_s32 dst_is_ereg = 0; - sljit_s32 src_is_ereg = 0; -#else -# define src_is_ereg 0 #endif CHECK_ERROR(); @@ -1257,7 +1246,7 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit ADJUST_LOCAL_OFFSET(src, srcw); CHECK_EXTRA_REGS(dst, dstw, dst_is_ereg = 1); - CHECK_EXTRA_REGS(src, srcw, src_is_ereg = 1); + CHECK_EXTRA_REGS(src, srcw, (void)0); #if (defined SLJIT_CONFIG_X86_64 && SLJIT_CONFIG_X86_64) compiler->mode32 = op_flags & SLJIT_I32_OP; #endif @@ -1270,32 +1259,27 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit op = GET_OPCODE(op); - if (op >= SLJIT_MOV && op <= SLJIT_MOVU_P) { + if (op >= SLJIT_MOV && op <= SLJIT_MOV_P) { #if (defined SLJIT_CONFIG_X86_64 && SLJIT_CONFIG_X86_64) compiler->mode32 = 0; #endif - if (op_flags & SLJIT_I32_OP) { - if (FAST_IS_REG(src) && src == dst) { - if (!TYPE_CAST_NEEDED(op)) - return SLJIT_SUCCESS; - } -#if (defined SLJIT_CONFIG_X86_64 && SLJIT_CONFIG_X86_64) - if (op == SLJIT_MOV_S32 && (src & SLJIT_MEM)) - op = SLJIT_MOV_U32; - if (op == SLJIT_MOVU_S32 && (src & SLJIT_MEM)) - op = SLJIT_MOVU_U32; - if (op == SLJIT_MOV_U32 && (src & SLJIT_IMM)) - op = SLJIT_MOV_S32; - if (op == SLJIT_MOVU_U32 && (src & SLJIT_IMM)) - op = SLJIT_MOVU_S32; -#endif + if (FAST_IS_REG(src) && src == dst) { + if (!TYPE_CAST_NEEDED(op)) + return SLJIT_SUCCESS; } - SLJIT_COMPILE_ASSERT(SLJIT_MOV + 8 == SLJIT_MOVU, movu_offset); - if (op >= SLJIT_MOVU) { - update = 1; - op -= 8; + if (op_flags & SLJIT_I32_OP) { +#if (defined SLJIT_CONFIG_X86_64 && SLJIT_CONFIG_X86_64) + if (src & SLJIT_MEM) { + if (op == SLJIT_MOV_S32) + op = SLJIT_MOV_U32; + } + else if (src & SLJIT_IMM) { + if (op == SLJIT_MOV_U32) + op = SLJIT_MOV_S32; + } +#endif } if (src & SLJIT_IMM) { @@ -1369,28 +1353,6 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit if (SLJIT_UNLIKELY(dst_is_ereg) && dst == TMP_REG1) return emit_mov(compiler, SLJIT_MEM1(SLJIT_SP), dstw, TMP_REG1, 0); #endif - - if (SLJIT_UNLIKELY(update) && (src & SLJIT_MEM) && !src_is_ereg && (src & REG_MASK)) { - if ((src & OFFS_REG_MASK) != 0) { - FAIL_IF(emit_cum_binary(compiler, ADD_r_rm, ADD_rm_r, ADD, ADD_EAX_i32, - (src & REG_MASK), 0, (src & REG_MASK), 0, OFFS_REG(dst), 0)); - } - else if (srcw != 0) { - FAIL_IF(emit_cum_binary(compiler, ADD_r_rm, ADD_rm_r, ADD, ADD_EAX_i32, - (src & REG_MASK), 0, (src & REG_MASK), 0, SLJIT_IMM, srcw)); - } - } - - if (SLJIT_UNLIKELY(update) && (dst & SLJIT_MEM) && (dst & REG_MASK)) { - if ((dst & OFFS_REG_MASK) != 0) { - FAIL_IF(emit_cum_binary(compiler, ADD_r_rm, ADD_rm_r, ADD, ADD_EAX_i32, - (dst & REG_MASK), 0, (dst & REG_MASK), 0, OFFS_REG(dst), 0)); - } - else if (dstw != 0) { - FAIL_IF(emit_cum_binary(compiler, ADD_r_rm, ADD_rm_r, ADD, ADD_EAX_i32, - (dst & REG_MASK), 0, (dst & REG_MASK), 0, SLJIT_IMM, dstw)); - } - } return SLJIT_SUCCESS; } @@ -1408,10 +1370,6 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit } return SLJIT_SUCCESS; - -#if (defined SLJIT_CONFIG_X86_64 && SLJIT_CONFIG_X86_64) -# undef src_is_ereg -#endif } #if (defined SLJIT_CONFIG_X86_64 && SLJIT_CONFIG_X86_64) @@ -1445,12 +1403,16 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit #endif static sljit_s32 emit_cum_binary(struct sljit_compiler *compiler, - sljit_u8 op_rm, sljit_u8 op_mr, sljit_u8 op_imm, sljit_u8 op_eax_imm, + sljit_u32 op_types, sljit_s32 dst, sljit_sw dstw, sljit_s32 src1, sljit_sw src1w, sljit_s32 src2, sljit_sw src2w) { sljit_u8* inst; + sljit_u8 op_eax_imm = (op_types >> 24); + sljit_u8 op_rm = (op_types >> 16) & 0xff; + sljit_u8 op_mr = (op_types >> 8) & 0xff; + sljit_u8 op_imm = op_types & 0xff; if (dst == SLJIT_UNUSED) { EMIT_MOV(compiler, TMP_REG1, 0, src1, src1w); @@ -1561,12 +1523,16 @@ static sljit_s32 emit_cum_binary(struct } static sljit_s32 emit_non_cum_binary(struct sljit_compiler *compiler, - sljit_u8 op_rm, sljit_u8 op_mr, sljit_u8 op_imm, sljit_u8 op_eax_imm, + sljit_u32 op_types, sljit_s32 dst, sljit_sw dstw, sljit_s32 src1, sljit_sw src1w, sljit_s32 src2, sljit_sw src2w) { sljit_u8* inst; + sljit_u8 op_eax_imm = (op_types >> 24); + sljit_u8 op_rm = (op_types >> 16) & 0xff; + sljit_u8 op_mr = (op_types >> 8) & 0xff; + sljit_u8 op_imm = op_types & 0xff; if (dst == SLJIT_UNUSED) { EMIT_MOV(compiler, TMP_REG1, 0, src1, src1w); @@ -2044,7 +2010,7 @@ static sljit_s32 emit_shift(struct sljit *inst |= mode; EMIT_MOV(compiler, SLJIT_PREF_SHIFT_REG, 0, TMP_REG1, 0); } - else if (FAST_IS_REG(dst) && dst != src2 && !ADDRESSING_DEPENDS_ON(src2, dst)) { + else if (SLOW_IS_REG(dst) && dst != src2 && !ADDRESSING_DEPENDS_ON(src2, dst)) { if (src1 != dst) EMIT_MOV(compiler, dst, 0, src1, src1w); EMIT_MOV(compiler, TMP_REG1, 0, SLJIT_PREF_SHIFT_REG, 0); @@ -2057,27 +2023,24 @@ static sljit_s32 emit_shift(struct sljit else { /* This case is complex since ecx itself may be used for addressing, and this case must be supported as well. */ -#if (defined SLJIT_CONFIG_X86_32 && SLJIT_CONFIG_X86_32) EMIT_MOV(compiler, TMP_REG1, 0, src1, src1w); +#if (defined SLJIT_CONFIG_X86_32 && SLJIT_CONFIG_X86_32) EMIT_MOV(compiler, SLJIT_MEM1(SLJIT_SP), 0, SLJIT_PREF_SHIFT_REG, 0); EMIT_MOV(compiler, SLJIT_PREF_SHIFT_REG, 0, src2, src2w); inst = emit_x86_instruction(compiler, 1 | EX86_SHIFT_INS, SLJIT_PREF_SHIFT_REG, 0, TMP_REG1, 0); FAIL_IF(!inst); *inst |= mode; EMIT_MOV(compiler, SLJIT_PREF_SHIFT_REG, 0, SLJIT_MEM1(SLJIT_SP), 0); - EMIT_MOV(compiler, dst, dstw, TMP_REG1, 0); #else - EMIT_MOV(compiler, TMP_REG1, 0, src1, src1w); - EMIT_MOV(compiler, TMP_REG2, 0, src2, src2w); - inst = emit_x86_instruction(compiler, 1, TMP_REG2, 0, SLJIT_PREF_SHIFT_REG, 0); - FAIL_IF(!inst); - *inst = XCHG_r_rm; + EMIT_MOV(compiler, TMP_REG2, 0, SLJIT_PREF_SHIFT_REG, 0); + EMIT_MOV(compiler, SLJIT_PREF_SHIFT_REG, 0, src2, src2w); inst = emit_x86_instruction(compiler, 1 | EX86_SHIFT_INS, SLJIT_PREF_SHIFT_REG, 0, TMP_REG1, 0); FAIL_IF(!inst); *inst |= mode; EMIT_MOV(compiler, SLJIT_PREF_SHIFT_REG, 0, TMP_REG2, 0); - EMIT_MOV(compiler, dst, dstw, TMP_REG1, 0); #endif + if (dst != SLJIT_UNUSED) + return emit_mov(compiler, dst, dstw, TMP_REG1, 0); } return SLJIT_SUCCESS; @@ -2101,7 +2064,7 @@ static sljit_s32 emit_shift_with_flags(s if (!set_flags) return emit_mov(compiler, dst, dstw, src1, src1w); /* OR dst, src, 0 */ - return emit_cum_binary(compiler, OR_r_rm, OR_rm_r, OR, OR_EAX_i32, + return emit_cum_binary(compiler, BINARY_OPCODE(OR), dst, dstw, src1, src1w, SLJIT_IMM, 0); } @@ -2111,10 +2074,10 @@ static sljit_s32 emit_shift_with_flags(s if (!FAST_IS_REG(dst)) FAIL_IF(emit_cmp_binary(compiler, src1, src1w, SLJIT_IMM, 0)); - FAIL_IF(emit_shift(compiler,mode, dst, dstw, src1, src1w, src2, src2w)); + FAIL_IF(emit_shift(compiler, mode, dst, dstw, src1, src1w, src2, src2w)); if (FAST_IS_REG(dst)) - return emit_cmp_binary(compiler, dst, dstw, SLJIT_IMM, 0); + return emit_cmp_binary(compiler, (dst == SLJIT_UNUSED) ? TMP_REG1 : dst, dstw, SLJIT_IMM, 0); return SLJIT_SUCCESS; } @@ -2145,10 +2108,10 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit if (emit_lea_binary(compiler, dst, dstw, src1, src1w, src2, src2w) != SLJIT_ERR_UNSUPPORTED) return compiler->error; } - return emit_cum_binary(compiler, ADD_r_rm, ADD_rm_r, ADD, ADD_EAX_i32, + return emit_cum_binary(compiler, BINARY_OPCODE(ADD), dst, dstw, src1, src1w, src2, src2w); case SLJIT_ADDC: - return emit_cum_binary(compiler, ADC_r_rm, ADC_rm_r, ADC, ADC_EAX_i32, + return emit_cum_binary(compiler, BINARY_OPCODE(ADC), dst, dstw, src1, src1w, src2, src2w); case SLJIT_SUB: if (!HAS_FLAGS(op)) { @@ -2158,23 +2121,23 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit if (dst == SLJIT_UNUSED) return emit_cmp_binary(compiler, src1, src1w, src2, src2w); - return emit_non_cum_binary(compiler, SUB_r_rm, SUB_rm_r, SUB, SUB_EAX_i32, + return emit_non_cum_binary(compiler, BINARY_OPCODE(SUB), dst, dstw, src1, src1w, src2, src2w); case SLJIT_SUBC: - return emit_non_cum_binary(compiler, SBB_r_rm, SBB_rm_r, SBB, SBB_EAX_i32, + return emit_non_cum_binary(compiler, BINARY_OPCODE(SBB), dst, dstw, src1, src1w, src2, src2w); case SLJIT_MUL: return emit_mul(compiler, dst, dstw, src1, src1w, src2, src2w); case SLJIT_AND: if (dst == SLJIT_UNUSED) return emit_test_binary(compiler, src1, src1w, src2, src2w); - return emit_cum_binary(compiler, AND_r_rm, AND_rm_r, AND, AND_EAX_i32, + return emit_cum_binary(compiler, BINARY_OPCODE(AND), dst, dstw, src1, src1w, src2, src2w); case SLJIT_OR: - return emit_cum_binary(compiler, OR_r_rm, OR_rm_r, OR, OR_EAX_i32, + return emit_cum_binary(compiler, BINARY_OPCODE(OR), dst, dstw, src1, src1w, src2, src2w); case SLJIT_XOR: - return emit_cum_binary(compiler, XOR_r_rm, XOR_rm_r, XOR, XOR_EAX_i32, + return emit_cum_binary(compiler, BINARY_OPCODE(XOR), dst, dstw, src1, src1w, src2, src2w); case SLJIT_SHL: return emit_shift_with_flags(compiler, SHL, HAS_FLAGS(op), @@ -2203,7 +2166,11 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit_get_float_register_index(sljit_s32 reg) { CHECK_REG_INDEX(check_sljit_get_float_register_index(reg)); +#if (defined SLJIT_CONFIG_X86_32 && SLJIT_CONFIG_X86_32) return reg; +#else + return freg_map[reg]; +#endif } SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit_emit_op_custom(struct sljit_compiler *compiler, @@ -2345,6 +2312,7 @@ static SLJIT_INLINE sljit_s32 sljit_emit FAIL_IF(emit_sse2_load(compiler, op & SLJIT_F32_OP, TMP_FREG, src1, src1w)); src1 = TMP_FREG; } + return emit_sse2_logic(compiler, UCOMISD_x_xm, !(op & SLJIT_F32_OP), src1, src2, src2w); } @@ -2516,9 +2484,6 @@ SLJIT_API_FUNC_ATTRIBUTE struct sljit_ju set_jump(jump, compiler, type & SLJIT_REWRITABLE_JUMP); type &= 0xff; - if (type >= SLJIT_CALL1) - PTR_FAIL_IF(call_with_args(compiler, type)); - /* Worst case size. */ #if (defined SLJIT_CONFIG_X86_32 && SLJIT_CONFIG_X86_32) compiler->size += (type >= SLJIT_JUMP) ? 5 : 6; @@ -2534,14 +2499,6 @@ SLJIT_API_FUNC_ATTRIBUTE struct sljit_ju return jump; } -#if (defined SLJIT_CONFIG_X86_64 && SLJIT_CONFIG_X86_64) -#ifndef _WIN64 -#define IS_REG_CHANGED_BY_CALL(src, type) ((src) == SLJIT_R3) -#else -#define IS_REG_CHANGED_BY_CALL(src, type) ((src) == SLJIT_R2) -#endif -#endif - SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit_emit_ijump(struct sljit_compiler *compiler, sljit_s32 type, sljit_s32 src, sljit_sw srcw) { sljit_u8 *inst; @@ -2553,25 +2510,6 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_s32 sljit CHECK_EXTRA_REGS(src, srcw, (void)0); - if (type >= SLJIT_CALL1) { -#if (defined SLJIT_CONFIG_X86_32 && SLJIT_CONFIG_X86_32) -#if (defined SLJIT_X86_32_FASTCALL && SLJIT_X86_32_FASTCALL) - if (src == SLJIT_R2) { - EMIT_MOV(compiler, TMP_REG1, 0, src, 0); - src = TMP_REG1; - } - if (src == SLJIT_MEM1(SLJIT_SP) && type >= SLJIT_CALL3) - srcw += sizeof(sljit_sw); -#endif -#else - if ((src & SLJIT_MEM) || IS_REG_CHANGED_BY_CALL(src, type)) { - EMIT_MOV(compiler, TMP_REG2, 0, src, srcw); - src = TMP_REG2; - } -#endif - FAIL_IF(call_with_args(compiler, type)); - } - if (src == SLJIT_IMM) { jump = (struct sljit_jump*)ensure_abuf(compiler, sizeof(struct sljit_jump)); FAIL_IF_NULL(jump); Modified: tomcat/jk/trunk/native/iis/pcre/sljit/sljitUtils.c URL: http://svn.apache.org/viewvc/tomcat/jk/trunk/native/iis/pcre/sljit/sljitUtils.c?rev=1838867&r1=1838866&r2=1838867&view=diff ============================================================================== --- tomcat/jk/trunk/native/iis/pcre/sljit/sljitUtils.c (original) +++ tomcat/jk/trunk/native/iis/pcre/sljit/sljitUtils.c Fri Aug 24 14:36:12 2018 @@ -48,12 +48,12 @@ static SLJIT_INLINE void allocator_relea #if (defined SLJIT_UTIL_GLOBAL_LOCK && SLJIT_UTIL_GLOBAL_LOCK) -SLJIT_API_FUNC_ATTRIBUTE void SLJIT_CALL sljit_grab_lock(void) +SLJIT_API_FUNC_ATTRIBUTE void SLJIT_FUNC sljit_grab_lock(void) { /* Always successful. */ } -SLJIT_API_FUNC_ATTRIBUTE void SLJIT_CALL sljit_release_lock(void) +SLJIT_API_FUNC_ATTRIBUTE void SLJIT_FUNC sljit_release_lock(void) { /* Always successful. */ } @@ -88,7 +88,7 @@ static SLJIT_INLINE void allocator_relea static HANDLE global_mutex = 0; -SLJIT_API_FUNC_ATTRIBUTE void SLJIT_CALL sljit_grab_lock(void) +SLJIT_API_FUNC_ATTRIBUTE void SLJIT_FUNC sljit_grab_lock(void) { /* No idea what to do if an error occures. Static mutexes should never fail... */ if (!global_mutex) @@ -97,7 +97,7 @@ SLJIT_API_FUNC_ATTRIBUTE void SLJIT_CALL WaitForSingleObject(global_mutex, INFINITE); } -SLJIT_API_FUNC_ATTRIBUTE void SLJIT_CALL sljit_release_lock(void) +SLJIT_API_FUNC_ATTRIBUTE void SLJIT_FUNC sljit_release_lock(void) { ReleaseMutex(global_mutex); } @@ -130,12 +130,12 @@ static SLJIT_INLINE void allocator_relea static pthread_mutex_t global_mutex = PTHREAD_MUTEX_INITIALIZER; -SLJIT_API_FUNC_ATTRIBUTE void SLJIT_CALL sljit_grab_lock(void) +SLJIT_API_FUNC_ATTRIBUTE void SLJIT_FUNC sljit_grab_lock(void) { pthread_mutex_lock(&global_mutex); } -SLJIT_API_FUNC_ATTRIBUTE void SLJIT_CALL sljit_release_lock(void) +SLJIT_API_FUNC_ATTRIBUTE void SLJIT_FUNC sljit_release_lock(void) { pthread_mutex_unlock(&global_mutex); } @@ -203,7 +203,7 @@ static SLJIT_INLINE sljit_s32 open_dev_z /* Planning to make it even more clever in the future. */ static sljit_sw sljit_page_align = 0; -SLJIT_API_FUNC_ATTRIBUTE struct sljit_stack* SLJIT_CALL sljit_allocate_stack(sljit_uw limit, sljit_uw max_limit, void *allocator_data) +SLJIT_API_FUNC_ATTRIBUTE struct sljit_stack* SLJIT_FUNC sljit_allocate_stack(sljit_uw start_size, sljit_uw max_size, void *allocator_data) { struct sljit_stack *stack; void *ptr; @@ -212,7 +212,7 @@ SLJIT_API_FUNC_ATTRIBUTE struct sljit_st #endif SLJIT_UNUSED_ARG(allocator_data); - if (limit > max_limit || limit < 1) + if (start_size > max_size || start_size < 1) return NULL; #ifdef _WIN32 @@ -234,25 +234,27 @@ SLJIT_API_FUNC_ATTRIBUTE struct sljit_st if (!stack) return NULL; - /* Align max_limit. */ - max_limit = (max_limit + sljit_page_align) & ~sljit_page_align; + /* Align max_size. */ + max_size = (max_size + sljit_page_align) & ~sljit_page_align; #ifdef _WIN32 - ptr = VirtualAlloc(NULL, max_limit, MEM_RESERVE, PAGE_READWRITE); + ptr = VirtualAlloc(NULL, max_size, MEM_RESERVE, PAGE_READWRITE); if (!ptr) { SLJIT_FREE(stack, allocator_data); return NULL; } - stack->max_limit = (sljit_u8 *)ptr; - stack->base = stack->max_limit + max_limit; - stack->limit = stack->base; - if (sljit_stack_resize(stack, stack->base - limit)) { + + stack->min_start = (sljit_u8 *)ptr; + stack->end = stack->min_start + max_size; + stack->start = stack->end; + + if (sljit_stack_resize(stack, stack->end - start_size) == NULL) { sljit_free_stack(stack, allocator_data); return NULL; } #else #ifdef MAP_ANON - ptr = mmap(NULL, max_limit, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANON, -1, 0); + ptr = mmap(NULL, max_size, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANON, -1, 0); #else if (dev_zero < 0) { if (open_dev_zero()) { @@ -260,73 +262,70 @@ SLJIT_API_FUNC_ATTRIBUTE struct sljit_st return NULL; } } - ptr = mmap(NULL, max_limit, PROT_READ | PROT_WRITE, MAP_PRIVATE, dev_zero, 0); + ptr = mmap(NULL, max_size, PROT_READ | PROT_WRITE, MAP_PRIVATE, dev_zero, 0); #endif if (ptr == MAP_FAILED) { SLJIT_FREE(stack, allocator_data); return NULL; } - stack->max_limit = (sljit_u8 *)ptr; - stack->base = stack->max_limit + max_limit; - stack->limit = stack->base - limit; + stack->min_start = (sljit_u8 *)ptr; + stack->end = stack->min_start + max_size; + stack->start = stack->end - start_size; #endif - stack->top = stack->base; + stack->top = stack->end; return stack; } #undef PAGE_ALIGN -SLJIT_API_FUNC_ATTRIBUTE void SLJIT_CALL sljit_free_stack(struct sljit_stack *stack, void *allocator_data) +SLJIT_API_FUNC_ATTRIBUTE void SLJIT_FUNC sljit_free_stack(struct sljit_stack *stack, void *allocator_data) { SLJIT_UNUSED_ARG(allocator_data); #ifdef _WIN32 - VirtualFree((void*)stack->max_limit, 0, MEM_RELEASE); + VirtualFree((void*)stack->min_start, 0, MEM_RELEASE); #else - munmap((void*)stack->max_limit, stack->base - stack->max_limit); + munmap((void*)stack->min_start, stack->end - stack->min_start); #endif SLJIT_FREE(stack, allocator_data); } -SLJIT_API_FUNC_ATTRIBUTE sljit_sw SLJIT_CALL sljit_stack_resize(struct sljit_stack *stack, sljit_u8 *new_limit) +SLJIT_API_FUNC_ATTRIBUTE sljit_u8 *SLJIT_FUNC sljit_stack_resize(struct sljit_stack *stack, sljit_u8 *new_start) { - sljit_uw aligned_old_limit; - sljit_uw aligned_new_limit; + sljit_uw aligned_old_start; + sljit_uw aligned_new_start; + + if ((new_start < stack->min_start) || (new_start >= stack->end)) + return NULL; - if ((new_limit < stack->max_limit) || (new_limit >= stack->base)) - return -1; #ifdef _WIN32 - aligned_new_limit = (sljit_uw)new_limit & ~sljit_page_align; - aligned_old_limit = ((sljit_uw)stack->limit) & ~sljit_page_align; - if (aligned_new_limit != aligned_old_limit) { - if (aligned_new_limit < aligned_old_limit) { - if (!VirtualAlloc((void*)aligned_new_limit, aligned_old_limit - aligned_new_limit, MEM_COMMIT, PAGE_READWRITE)) - return -1; + aligned_new_start = (sljit_uw)new_start & ~sljit_page_align; + aligned_old_start = ((sljit_uw)stack->start) & ~sljit_page_align; + if (aligned_new_start != aligned_old_start) { + if (aligned_new_start < aligned_old_start) { + if (!VirtualAlloc((void*)aligned_new_start, aligned_old_start - aligned_new_start, MEM_COMMIT, PAGE_READWRITE)) + return NULL; } else { - if (!VirtualFree((void*)aligned_old_limit, aligned_new_limit - aligned_old_limit, MEM_DECOMMIT)) - return -1; + if (!VirtualFree((void*)aligned_old_start, aligned_new_start - aligned_old_start, MEM_DECOMMIT)) + return NULL; } } - stack->limit = new_limit; - return 0; #else - if (new_limit <= stack->limit) { - stack->limit = new_limit; - return 0; - } - aligned_new_limit = (sljit_uw)new_limit & ~sljit_page_align; - aligned_old_limit = ((sljit_uw)stack->limit) & ~sljit_page_align; - /* If madvise is available, we release the unnecessary space. */ + if (stack->start < new_start) { + aligned_new_start = (sljit_uw)new_start & ~sljit_page_align; + aligned_old_start = ((sljit_uw)stack->start) & ~sljit_page_align; + /* If madvise is available, we release the unnecessary space. */ #if defined(MADV_DONTNEED) - if (aligned_new_limit > aligned_old_limit) - madvise((void*)aligned_old_limit, aligned_new_limit - aligned_old_limit, MADV_DONTNEED); + if (aligned_new_start > aligned_old_start) + madvise((void*)aligned_old_start, aligned_new_start - aligned_old_start, MADV_DONTNEED); #elif defined(POSIX_MADV_DONTNEED) - if (aligned_new_limit > aligned_old_limit) - posix_madvise((void*)aligned_old_limit, aligned_new_limit - aligned_old_limit, POSIX_MADV_DONTNEED); + if (aligned_new_start > aligned_old_start) + posix_madvise((void*)aligned_old_start, aligned_new_start - aligned_old_start, POSIX_MADV_DONTNEED); #endif - stack->limit = new_limit; - return 0; + } #endif + stack->start = new_start; + return new_start; } #endif /* SLJIT_UTIL_STACK */ Modified: tomcat/jk/trunk/native/iis/pcre/test-driver URL: http://svn.apache.org/viewvc/tomcat/jk/trunk/native/iis/pcre/test-driver?rev=1838867&r1=1838866&r2=1838867&view=diff ============================================================================== --- tomcat/jk/trunk/native/iis/pcre/test-driver (original) +++ tomcat/jk/trunk/native/iis/pcre/test-driver Fri Aug 24 14:36:12 2018 @@ -1,9 +1,9 @@ #! /bin/sh # test-driver - basic testsuite driver script. -scriptversion=2013-07-13.22; # UTC +scriptversion=2016-01-11.22; # UTC -# Copyright (C) 2011-2014 Free Software Foundation, Inc. +# Copyright (C) 2011-2017 Free Software Foundation, Inc. # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by @@ -143,6 +143,6 @@ echo ":copy-in-global-log: $gcopy" >> $t # eval: (add-hook 'write-file-hooks 'time-stamp) # time-stamp-start: "scriptversion=" # time-stamp-format: "%:y-%02m-%02d.%02H" -# time-stamp-time-zone: "UTC" +# time-stamp-time-zone: "UTC0" # time-stamp-end: "; # UTC" # End: Modified: tomcat/jk/trunk/native/iis/pcre/testdata/testinput2 URL: http://svn.apache.org/viewvc/tomcat/jk/trunk/native/iis/pcre/testdata/testinput2?rev=1838867&r1=1838866&r2=1838867&view=diff ============================================================================== Binary files - no diff available. Modified: tomcat/jk/trunk/native/iis/pcre/testdata/testinput5 URL: http://svn.apache.org/viewvc/tomcat/jk/trunk/native/iis/pcre/testdata/testinput5?rev=1838867&r1=1838866&r2=1838867&view=diff ============================================================================== Binary files - no diff available. Modified: tomcat/jk/trunk/native/iis/pcre/testdata/testoutput2 URL: http://svn.apache.org/viewvc/tomcat/jk/trunk/native/iis/pcre/testdata/testoutput2?rev=1838867&r1=1838866&r2=1838867&view=diff ============================================================================== Binary files - no diff available. Modified: tomcat/jk/trunk/native/iis/pcre/testdata/testoutput5 URL: http://svn.apache.org/viewvc/tomcat/jk/trunk/native/iis/pcre/testdata/testoutput5?rev=1838867&r1=1838866&r2=1838867&view=diff ============================================================================== Binary files - no diff available. Modified: tomcat/jk/trunk/xdocs/miscellaneous/changelog.xml URL: http://svn.apache.org/viewvc/tomcat/jk/trunk/xdocs/miscellaneous/changelog.xml?rev=1838867&r1=1838866&r2=1838867&view=diff ============================================================================== --- tomcat/jk/trunk/xdocs/miscellaneous/changelog.xml (original) +++ tomcat/jk/trunk/xdocs/miscellaneous/changelog.xml Fri Aug 24 14:36:12 2018 @@ -88,6 +88,9 @@ now effectively hard-coded to CollpaseSlashesAll due to the changes made to align normalization with that implemented in Tomcat. (markt) </add> + <update> + Update PCRE bundled with the ISAPI redirector to 8.42. (rjung) + </update> </changelog> </subsection> </section> --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
