diff options
Diffstat (limited to 'sljit/sljitLir.h')
| -rw-r--r-- | sljit/sljitLir.h | 276 |
1 files changed, 187 insertions, 89 deletions
diff --git a/sljit/sljitLir.h b/sljit/sljitLir.h index 8cc6c30..1ca4599 100644 --- a/sljit/sljitLir.h +++ b/sljit/sljitLir.h @@ -56,8 +56,6 @@ Disadvantages: - No automatic register allocation, and temporary results are not stored on the stack. (hence the name comes) - - Limited number of registers (only 6+4 integer registers, max 3+2 - scratch, max 3+2 saved and 6 floating point registers) In practice: - This approach is very effective for interpreters - One of the saved registers typically points to a stack interface @@ -104,71 +102,155 @@ of sljitConfigInternal.h */ /* Registers */ /* --------------------------------------------------------------------- */ +/* + Scratch (R) registers: registers whose may not preserve their values + across function calls. + + Saved (S) registers: registers whose preserve their values across + function calls. + + The scratch and saved register sets are overlap. The last scratch register + is the first saved register, the one before the last is the second saved + register, and so on. + + If an architecture provides two scratch and three saved registers, + its scratch and saved register sets are the following: + + R0 | [S4] | R0 and S4 represent the same physical register + R1 | [S3] | R1 and S3 represent the same physical register + [R2] | S2 | R2 and S2 represent the same physical register + [R3] | S1 | R3 and S1 represent the same physical register + [R4] | S0 | R4 and S0 represent the same physical register + + Note: SLJIT_NUMBER_OF_SCRATCH_REGISTERS would be 2 and + SLJIT_NUMBER_OF_SAVED_REGISTERS would be 3 for this architecture. + + Note: On all supported architectures SLJIT_NUMBER_OF_REGISTERS >= 10 + and SLJIT_NUMBER_OF_SAVED_REGISTERS >= 5. However, 4 registers + are virtual on x86-32. See below. + + The purpose of this definition is convenience. Although a register + is either scratch register or saved register, SLJIT allows accessing + them from the other set. For example, four registers can be used as + scratch registers and the fifth one as saved register on the architecture + above. Of course the last two scratch registers (R2 and R3) from this + four will be saved on the stack, because they are defined as saved + registers in the application binary interface. Still R2 and R3 can be + used for referencing to these registers instead of S2 and S1, which + makes easier to write platform independent code. Scratch registers + can be saved registers in a similar way, but these extra saved + registers will not be preserved across function calls! Hence the + application must save them on those platforms, where the number of + saved registers is too low. This can be done by copy them onto + the stack and restore them after a function call. + + Note: To emphasize that registers assigned to R2-R4 are saved + registers, they are enclosed by square brackets. S3-S4 + are marked in a similar way. + + Note: sljit_emit_enter and sljit_set_context defines whether a register + is S or R register. E.g: when 3 scratches and 1 saved is mapped + by sljit_emit_enter, the allowed register set will be: R0-R2 and + S0. Although S2 is mapped to the same position as R2, it does not + available in the current configuration. Furthermore the R3 (S1) + register does not available as well. +*/ + +/* When SLJIT_UNUSED is specified as destination, the result is discarded. */ #define SLJIT_UNUSED 0 -/* Scratch (temporary) registers whose may not preserve their values - across function calls. */ -#define SLJIT_SCRATCH_REG1 1 -#define SLJIT_SCRATCH_REG2 2 -#define SLJIT_SCRATCH_REG3 3 -/* Note: extra registers cannot be used for memory addressing. */ -/* Note: on x86-32, these registers are emulated (using stack - loads & stores). */ -#define SLJIT_SCRATCH_EREG1 4 -#define SLJIT_SCRATCH_EREG2 5 - -/* Saved registers whose preserve their values across function calls. */ -#define SLJIT_SAVED_REG1 6 -#define SLJIT_SAVED_REG2 7 -#define SLJIT_SAVED_REG3 8 -/* Note: extra registers cannot be used for memory addressing. */ -/* Note: on x86-32, these registers are emulated (using stack - loads & stores). */ -#define SLJIT_SAVED_EREG1 9 -#define SLJIT_SAVED_EREG2 10 - -/* Read-only register (cannot be the destination of an operation). - Only SLJIT_MEM1(SLJIT_LOCALS_REG) addressing mode is allowed since - several ABIs has certain limitations about the stack layout. However - sljit_get_local_base() can be used to obtain the offset of a value - on the stack. */ -#define SLJIT_LOCALS_REG 11 - -/* Number of registers. */ -#define SLJIT_NO_TMP_REGISTERS 5 -#define SLJIT_NO_GEN_REGISTERS 5 -#define SLJIT_NO_REGISTERS 11 +/* Scratch registers. */ +#define SLJIT_R0 1 +#define SLJIT_R1 2 +#define SLJIT_R2 3 +/* Note: on x86-32, R3 - R6 are emulated (using stack loads & stores), + so they cannot be used for memory addressing. There is no such + limitation on other CPUs. */ +#define SLJIT_R3 4 +#define SLJIT_R4 5 +#define SLJIT_R5 6 +#define SLJIT_R6 7 +#define SLJIT_R7 8 +#define SLJIT_R8 9 +#define SLJIT_R9 10 +/* All R registers provided by the architecture can be accessed by SLJIT_R(i) + The i parameter must be >= 0 and < SLJIT_NUMBER_OF_REGISTERS. */ +#define SLJIT_R(i) (1 + (i)) + +/* Saved registers. */ +#define SLJIT_S0 (SLJIT_NUMBER_OF_REGISTERS) +#define SLJIT_S1 (SLJIT_NUMBER_OF_REGISTERS - 1) +#define SLJIT_S2 (SLJIT_NUMBER_OF_REGISTERS - 2) +/* Note: on x86-32, S3 - S6 are emulated (using stack loads & stores), + so they cannot be used for memory addressing. There is no such + limitation on other CPUs. */ +#define SLJIT_S3 (SLJIT_NUMBER_OF_REGISTERS - 3) +#define SLJIT_S4 (SLJIT_NUMBER_OF_REGISTERS - 4) +#define SLJIT_S5 (SLJIT_NUMBER_OF_REGISTERS - 5) +#define SLJIT_S6 (SLJIT_NUMBER_OF_REGISTERS - 6) +#define SLJIT_S7 (SLJIT_NUMBER_OF_REGISTERS - 7) +#define SLJIT_S8 (SLJIT_NUMBER_OF_REGISTERS - 8) +#define SLJIT_S9 (SLJIT_NUMBER_OF_REGISTERS - 9) +/* All S registers provided by the architecture can be accessed by SLJIT_S(i) + The i parameter must be >= 0 and < SLJIT_NUMBER_OF_SAVED_REGISTERS. */ +#define SLJIT_S(i) (SLJIT_NUMBER_OF_REGISTERS - (i)) + +/* Registers >= SLJIT_FIRST_SAVED_REG are saved registers. */ +#define SLJIT_FIRST_SAVED_REG (SLJIT_S0 - SLJIT_NUMBER_OF_SAVED_REGISTERS + 1) + +/* The SLJIT_SP provides direct access to the linear stack space allocated by + sljit_emit_enter. It can only be used in the following form: SLJIT_MEM1(SLJIT_SP). + The immediate offset is extended by the relative stack offset automatically. + The sljit_get_local_base can be used to obtain the absolute offset. */ +#define SLJIT_SP (SLJIT_NUMBER_OF_REGISTERS + 1) /* Return with machine word. */ -#define SLJIT_RETURN_REG SLJIT_SCRATCH_REG1 +#define SLJIT_RETURN_REG SLJIT_R0 /* x86 prefers specific registers for special purposes. In case of shift - by register it supports only SLJIT_SCRATCH_REG3 for shift argument + by register it supports only SLJIT_R2 for shift argument (which is the src2 argument of sljit_emit_op2). If another register is used, sljit must exchange data between registers which cause a minor slowdown. Other architectures has no such limitation. */ -#define SLJIT_PREF_SHIFT_REG SLJIT_SCRATCH_REG3 +#define SLJIT_PREF_SHIFT_REG SLJIT_R2 /* --------------------------------------------------------------------- */ /* Floating point registers */ /* --------------------------------------------------------------------- */ -/* Note: SLJIT_UNUSED as destination is not valid for floating point - operations, since they cannot be used for setting flags. */ - -/* Floating point operations are performed on double or - single precision values. */ +/* Each floating point register can store a double or single precision + value. The FR and FS register sets are overlap in the same way as R + and S register sets. See above. */ -#define SLJIT_FLOAT_REG1 1 -#define SLJIT_FLOAT_REG2 2 -#define SLJIT_FLOAT_REG3 3 -#define SLJIT_FLOAT_REG4 4 -#define SLJIT_FLOAT_REG5 5 -#define SLJIT_FLOAT_REG6 6 - -#define SLJIT_NO_FLOAT_REGISTERS 6 +/* Note: SLJIT_UNUSED as destination is not valid for floating point + operations, since they cannot be used for setting flags. */ + +/* Floating point scratch registers. */ +#define SLJIT_FR0 1 +#define SLJIT_FR1 2 +#define SLJIT_FR2 3 +#define SLJIT_FR3 4 +#define SLJIT_FR4 5 +#define SLJIT_FR5 6 +/* All FR registers provided by the architecture can be accessed by SLJIT_FR(i) + The i parameter must be >= 0 and < SLJIT_NUMBER_OF_FLOAT_REGISTERS. */ +#define SLJIT_FR(i) (1 + (i)) + +/* Floating point saved registers. */ +#define SLJIT_FS0 (SLJIT_NUMBER_OF_FLOAT_REGISTERS) +#define SLJIT_FS1 (SLJIT_NUMBER_OF_FLOAT_REGISTERS - 1) +#define SLJIT_FS2 (SLJIT_NUMBER_OF_FLOAT_REGISTERS - 2) +#define SLJIT_FS3 (SLJIT_NUMBER_OF_FLOAT_REGISTERS - 3) +#define SLJIT_FS4 (SLJIT_NUMBER_OF_FLOAT_REGISTERS - 4) +#define SLJIT_FS5 (SLJIT_NUMBER_OF_FLOAT_REGISTERS - 5) +/* All S registers provided by the architecture can be accessed by SLJIT_FS(i) + The i parameter must be >= 0 and < SLJIT_NUMBER_OF_SAVED_FLOAT_REGISTERS. */ +#define SLJIT_FS(i) (SLJIT_NUMBER_OF_FLOAT_REGISTERS - (i)) + +/* Float registers >= SLJIT_FIRST_SAVED_FLOAT_REG are saved registers. */ +#define SLJIT_FIRST_SAVED_FLOAT_REG (SLJIT_FS0 - SLJIT_NUMBER_OF_SAVED_FLOAT_REGISTERS + 1) /* --------------------------------------------------------------------- */ /* Main structures and functions */ @@ -216,10 +298,14 @@ struct sljit_compiler { struct sljit_memory_fragment *buf; struct sljit_memory_fragment *abuf; - /* Used local registers. */ + /* Used scratch registers. */ sljit_si scratches; /* Used saved registers. */ sljit_si saveds; + /* Used float scratch registers. */ + sljit_si fscratches; + /* Used float saved registers. */ + sljit_si fsaveds; /* Local stack size. */ sljit_si local_size; /* Code size. */ @@ -229,9 +315,6 @@ struct sljit_compiler { #if (defined SLJIT_CONFIG_X86_32 && SLJIT_CONFIG_X86_32) sljit_si args; - sljit_si locals_offset; - sljit_si scratches_start; - sljit_si saveds_start; #endif #if (defined SLJIT_CONFIG_X86_64 && SLJIT_CONFIG_X86_64) @@ -361,46 +444,61 @@ static SLJIT_INLINE sljit_uw sljit_get_generated_code_size(struct sljit_compiler error, they return with SLJIT_SUCCESS. */ /* - The executable code is basically a function call from the viewpoint of - the C language. The function calls must obey to the ABI (Application - Binary Interface) of the platform, which specify the purpose of machine - registers and stack handling among other things. The sljit_emit_enter - function emits the necessary instructions for setting up a new context - for the executable code and moves function arguments to the saved - registers. The number of arguments are specified in the "args" - parameter and the first argument goes to SLJIT_SAVED_REG1, the second - goes to SLJIT_SAVED_REG2 and so on. The number of scratch and - saved registers are passed in "scratches" and "saveds" arguments - respectively. Since the saved registers contains the arguments, - "args" must be less or equal than "saveds". The sljit_emit_enter - is also capable of allocating a stack space for local variables. The - "local_size" argument contains the size in bytes of this local area - and its staring address is stored in SLJIT_LOCALS_REG. However - the SLJIT_LOCALS_REG is not necessary the machine stack pointer. - The memory bytes between SLJIT_LOCALS_REG (inclusive) and - SLJIT_LOCALS_REG + local_size (exclusive) can be modified freely - until the function returns. The stack space is uninitialized. + The executable code is a function call from the viewpoint of the C + language. The function calls must obey to the ABI (Application + Binary Interface) of the platform, which specify the purpose of + all machine registers and stack handling among other things. The + sljit_emit_enter function emits the necessary instructions for + setting up a new context for the executable code and moves function + arguments to the saved registers. The number of sljit_sw arguments + passed to the function are specified in the "args" parameter. The + number of arguments must be less than or equal to 3. The first + argument goes to SLJIT_S0, the second goes to SLJIT_S1 and so on. + The register set used by the function must be declared as well. + The number of scratch and saved registers used by the function must + be passed to sljit_emit_enter. Only R registers between R0 and + "scratches" argument can be used later. E.g. if "scratches" is set + to 2, the register set will be limited to R0 and R1. The S registers + and the floating point registers ("fscratches" and "fsaveds") + are specified in a similar way. The sljit_emit_enter is also capable + of allocating a stack space for local variables. The "local_size" + argument contains the size in bytes of this local area and its + staring address is stored in SLJIT_SP. The memory area between + SLJIT_SP (inclusive) and SLJIT_SP + local_size (exclusive) can be + modified freely until the function returns. The stack space is not + initialized. + + Note: the following conditions must met: + 0 <= scratches <= SLJIT_NUMBER_OF_REGISTERS + 0 <= saveds <= SLJIT_NUMBER_OF_REGISTERS + scratches + saveds <= SLJIT_NUMBER_OF_REGISTERS + 0 <= fscratches <= SLJIT_NUMBER_OF_FLOAT_REGISTERS + 0 <= fsaveds <= SLJIT_NUMBER_OF_FLOAT_REGISTERS + fscratches + fsaveds <= SLJIT_NUMBER_OF_FLOAT_REGISTERS Note: every call of sljit_emit_enter and sljit_set_context - overwrites the previous context. */ + overwrites the previous context. +*/ #define SLJIT_MAX_LOCAL_SIZE 65536 SLJIT_API_FUNC_ATTRIBUTE sljit_si sljit_emit_enter(struct sljit_compiler *compiler, - sljit_si args, sljit_si scratches, sljit_si saveds, sljit_si local_size); + sljit_si args, sljit_si scratches, sljit_si saveds, + sljit_si fscratches, sljit_si fsaveds, sljit_si local_size); /* The machine code has a context (which contains the local stack space size, number of used registers, etc.) which initialized by sljit_emit_enter. Several functions (like sljit_emit_return) requres this context to be able to generate the appropriate code. However, some code fragments (like inline cache) may have - no normal entry point so their context is unknown for the compiler. Using the - function below we can specify their context. + no normal entry point so their context is unknown for the compiler. Their context + can be provided to the compiler by the sljit_set_context function. Note: every call of sljit_emit_enter and sljit_set_context overwrites the previous context. */ SLJIT_API_FUNC_ATTRIBUTE void sljit_set_context(struct sljit_compiler *compiler, - sljit_si args, sljit_si scratches, sljit_si saveds, sljit_si local_size); + sljit_si args, sljit_si scratches, sljit_si saveds, + sljit_si fscratches, sljit_si fsaveds, sljit_si local_size); /* Return from machine code. The op argument can be SLJIT_UNUSED which means the function does not return with anything or any opcode between SLJIT_MOV and @@ -558,23 +656,23 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_si sljit_emit_fast_return(struct sljit_compiler * it can even decrease the runtime in a few cases. */ #define SLJIT_NOP 1 /* Flags: - (may destroy flags) - Unsigned multiplication of SLJIT_SCRATCH_REG1 and SLJIT_SCRATCH_REG2. - Result goes to SLJIT_SCRATCH_REG2:SLJIT_SCRATCH_REG1 (high:low) word */ + Unsigned multiplication of SLJIT_R0 and SLJIT_R1. + Result goes to SLJIT_R1:SLJIT_R0 (high:low) word */ #define SLJIT_UMUL 2 /* Flags: - (may destroy flags) - Signed multiplication of SLJIT_SCRATCH_REG1 and SLJIT_SCRATCH_REG2. - Result goes to SLJIT_SCRATCH_REG2:SLJIT_SCRATCH_REG1 (high:low) word */ + Signed multiplication of SLJIT_R0 and SLJIT_R1. + Result goes to SLJIT_R1:SLJIT_R0 (high:low) word */ #define SLJIT_SMUL 3 /* Flags: I - (may destroy flags) - Unsigned divide of the value in SLJIT_SCRATCH_REG1 by the value in SLJIT_SCRATCH_REG2. - The result is placed in SLJIT_SCRATCH_REG1 and the remainder goes to SLJIT_SCRATCH_REG2. - Note: if SLJIT_SCRATCH_REG2 contains 0, the behaviour is undefined. */ + Unsigned divide of the value in SLJIT_R0 by the value in SLJIT_R1. + The result is placed in SLJIT_R0 and the remainder goes to SLJIT_R1. + Note: if SLJIT_R1 contains 0, the behaviour is undefined. */ #define SLJIT_UDIV 4 #define SLJIT_IUDIV (SLJIT_UDIV | SLJIT_INT_OP) /* Flags: I - (may destroy flags) - Signed divide of the value in SLJIT_SCRATCH_REG1 by the value in SLJIT_SCRATCH_REG2. - The result is placed in SLJIT_SCRATCH_REG1 and the remainder goes to SLJIT_SCRATCH_REG2. - Note: if SLJIT_SCRATCH_REG2 contains 0, the behaviour is undefined. */ + Signed divide of the value in SLJIT_R0 by the value in SLJIT_R1. + The result is placed in SLJIT_R0 and the remainder goes to SLJIT_R1. + Note: if SLJIT_R1 contains 0, the behaviour is undefined. */ #define SLJIT_SDIV 5 #define SLJIT_ISDIV (SLJIT_SDIV | SLJIT_INT_OP) @@ -906,7 +1004,7 @@ SLJIT_API_FUNC_ATTRIBUTE sljit_si sljit_emit_op_flags(struct sljit_compiler *com sljit_si src, sljit_sw srcw, sljit_si type); -/* Copies the base address of SLJIT_LOCALS_REG+offset to dst. +/* Copies the base address of SLJIT_SP + offset to dst. Flags: - (never set any flags) */ SLJIT_API_FUNC_ATTRIBUTE sljit_si sljit_get_local_base(struct sljit_compiler *compiler, sljit_si dst, sljit_sw dstw, sljit_sw offset); |