diff options
Diffstat (limited to 'docs')
| -rw-r--r-- | docs/devel/tcg-ops.rst | 230 |
1 files changed, 129 insertions, 101 deletions
diff --git a/docs/devel/tcg-ops.rst b/docs/devel/tcg-ops.rst index 9adc0c9b6c..561c416574 100644 --- a/docs/devel/tcg-ops.rst +++ b/docs/devel/tcg-ops.rst @@ -7,67 +7,51 @@ TCG Intermediate Representation Introduction ============ -TCG (Tiny Code Generator) began as a generic backend for a C -compiler. It was simplified to be used in QEMU. It also has its roots -in the QOP code generator written by Paul Brook. +TCG (Tiny Code Generator) began as a generic backend for a C compiler. +It was simplified to be used in QEMU. It also has its roots in the +QOP code generator written by Paul Brook. Definitions =========== -TCG receives RISC-like *TCG ops* and performs some optimizations on them, -including liveness analysis and trivial constant expression -evaluation. TCG ops are then implemented in the host CPU back end, -also known as the TCG target. - -The TCG *target* is the architecture for which we generate the -code. It is of course not the same as the "target" of QEMU which is -the emulated architecture. As TCG started as a generic C backend used -for cross compiling, it is assumed that the TCG target is different -from the host, although it is never the case for QEMU. +The TCG *target* is the architecture for which we generate the code. +It is of course not the same as the "target" of QEMU which is the +emulated architecture. As TCG started as a generic C backend used +for cross compiling, the assumption was that TCG target might be +different from the host, although this is never the case for QEMU. In this document, we use *guest* to specify what architecture we are emulating; *target* always means the TCG target, the machine on which we are running QEMU. -A TCG *function* corresponds to a QEMU Translated Block (TB). - -A TCG *temporary* is a variable only live in a basic block. Temporaries are allocated explicitly in each function. - -A TCG *local temporary* is a variable only live in a function. Local temporaries are allocated explicitly in each function. - -A TCG *global* is a variable which is live in all the functions -(equivalent of a C global variable). They are defined before the -functions defined. A TCG global can be a memory location (e.g. a QEMU -CPU register), a fixed host register (e.g. the QEMU CPU state pointer) -or a memory location which is stored in a register outside QEMU TBs -(not implemented yet). - -A TCG *basic block* corresponds to a list of instructions terminated -by a branch instruction. - An operation with *undefined behavior* may result in a crash. An operation with *unspecified behavior* shall not crash. However, the result may be one of several possibilities so may be considered an *undefined result*. -Intermediate representation -=========================== +Basic Blocks +============ -Introduction ------------- +A TCG *basic block* is a single entry, multiple exit region which +corresponds to a list of instructions terminated by a label, or +any branch instruction. -TCG instructions operate on variables which are temporaries, local -temporaries or globals. TCG instructions and variables are strongly -typed. Two types are supported: 32 bit integers and 64 bit -integers. Pointers are defined as an alias to 32 bit or 64 bit -integers depending on the TCG target word size. +A TCG *extended basic block* is a single entry, multiple exit region +which corresponds to a list of instructions terminated by a label or +an unconditional branch. Specifically, an extended basic block is +a sequence of basic blocks connected by the fall-through paths of +zero or more conditional branch instructions. -Each instruction has a fixed number of output variable operands, input -variable operands and always constant operands. +Operations +========== -The notable exception is the call instruction which has a variable -number of outputs and inputs. +TCG instructions or *ops* operate on TCG *variables*, both of which +are strongly typed. Each instruction has a fixed number of output +variable operands, input variable operands and constant operands. +Vector instructions have a field specifying the element size within +the vector. The notable exception is the call instruction which has +a variable number of outputs and inputs. In the textual form, output operands usually come first, followed by input operands, followed by constant operands. The output type is @@ -77,68 +61,127 @@ included in the instruction name. Constants are prefixed with a '$'. add_i32 t0, t1, t2 /* (t0 <- t1 + t2) */ +Variables +========= -Assumptions ------------ +* ``TEMP_FIXED`` -Basic blocks -^^^^^^^^^^^^ + There is one TCG *fixed global* variable, ``cpu_env``, which is + live in all translation blocks, and holds a pointer to ``CPUArchState``. + This variable is held in a host cpu register at all times in all + translation blocks. -* Basic blocks end after branches (e.g. brcond_i32 instruction), - goto_tb and exit_tb instructions. +* ``TEMP_GLOBAL`` -* Basic blocks start after the end of a previous basic block, or at a - set_label instruction. + A TCG *global* is a variable which is live in all translation blocks, + and corresponds to memory location that is within ``CPUArchState``. + These may be specified as an offset from ``cpu_env``, in which case + they are called *direct globals*, or may be specified as an offset + from a direct global, in which case they are called *indirect globals*. + Even indirect globals should still reference memory within + ``CPUArchState``. All TCG globals are defined during + ``TCGCPUOps.initialize``, before any translation blocks are generated. -After the end of a basic block, the content of temporaries is -destroyed, but local temporaries and globals are preserved. +* ``TEMP_CONST`` -Floating point types -^^^^^^^^^^^^^^^^^^^^ + A TCG *constant* is a variable which is live throughout the entire + translation block, and contains a constant value. These variables + are allocated on demand during translation and are hashed so that + there is exactly one variable holding a given value. -* Floating point types are not supported yet +* ``TEMP_TB`` -Pointers -^^^^^^^^ + A TCG *translation block temporary* is a variable which is live + throughout the entire translation block, but dies on any exit. + These temporaries are allocated explicitly during translation. -* Depending on the TCG target, pointer size is 32 bit or 64 - bit. The type ``TCG_TYPE_PTR`` is an alias to ``TCG_TYPE_I32`` or - ``TCG_TYPE_I64``. +* ``TEMP_EBB`` -Helpers -^^^^^^^ + A TCG *extended basic block temporary* is a variable which is live + throughout an extended basic block, but dies on any exit. + These temporaries are allocated explicitly during translation. + +Types +===== + +* ``TCG_TYPE_I32`` + + A 32-bit integer. + +* ``TCG_TYPE_I64`` + + A 64-bit integer. For 32-bit hosts, such variables are split into a pair + of variables with ``type=TCG_TYPE_I32`` and ``base_type=TCG_TYPE_I64``. + The ``temp_subindex`` for each indicates where it falls within the + host-endian representation. -* Using the tcg_gen_helper_x_y it is possible to call any function - taking i32, i64 or pointer types. By default, before calling a helper, - all globals are stored at their canonical location and it is assumed - that the function can modify them. By default, the helper is allowed to - modify the CPU state or raise an exception. +* ``TCG_TYPE_PTR`` - This can be overridden using the following function modifiers: + An alias for ``TCG_TYPE_I32`` or ``TCG_TYPE_I64``, depending on the size + of a pointer for the host. - - ``TCG_CALL_NO_READ_GLOBALS`` means that the helper does not read globals, - either directly or via an exception. They will not be saved to their - canonical locations before calling the helper. +* ``TCG_TYPE_REG`` - - ``TCG_CALL_NO_WRITE_GLOBALS`` means that the helper does not modify any globals. - They will only be saved to their canonical location before calling helpers, - but they won't be reloaded afterwards. + An alias for ``TCG_TYPE_I32`` or ``TCG_TYPE_I64``, depending on the size + of the integer registers for the host. This may be larger + than ``TCG_TYPE_PTR`` depending on the host ABI. - - ``TCG_CALL_NO_SIDE_EFFECTS`` means that the call to the function is removed if - the return value is not used. +* ``TCG_TYPE_I128`` - Note that ``TCG_CALL_NO_READ_GLOBALS`` implies ``TCG_CALL_NO_WRITE_GLOBALS``. + A 128-bit integer. For all hosts, such variables are split into a number + of variables with ``type=TCG_TYPE_REG`` and ``base_type=TCG_TYPE_I128``. + The ``temp_subindex`` for each indicates where it falls within the + host-endian representation. - On some TCG targets (e.g. x86), several calling conventions are - supported. +* ``TCG_TYPE_V64`` -Branches -^^^^^^^^ + A 64-bit vector. This type is valid only if the TCG target + sets ``TCG_TARGET_HAS_v64``. -* Use the instruction 'br' to jump to a label. +* ``TCG_TYPE_V128`` + + A 128-bit vector. This type is valid only if the TCG target + sets ``TCG_TARGET_HAS_v128``. + +* ``TCG_TYPE_V256`` + + A 256-bit vector. This type is valid only if the TCG target + sets ``TCG_TARGET_HAS_v256``. + +Helpers +======= + +Helpers are registered in a guest-specific ``helper.h``, +which is processed to generate ``tcg_gen_helper_*`` functions. +With these functions it is possible to call a function taking +i32, i64, i128 or pointer types. + +By default, before calling a helper, all globals are stored at their +canonical location. By default, the helper is allowed to modify the +CPU state (including the state represented by tcg globals) +or may raise an exception. This default can be overridden using the +following function modifiers: + +* ``TCG_CALL_NO_WRITE_GLOBALS`` + + The helper does not modify any globals, but may read them. + Globals will be saved to their canonical location before calling helpers, + but need not be reloaded afterwards. + +* ``TCG_CALL_NO_READ_GLOBALS`` + + The helper does not read globals, either directly or via an exception. + They will not be saved to their canonical locations before calling + the helper. This implies ``TCG_CALL_NO_WRITE_GLOBALS``. + +* ``TCG_CALL_NO_SIDE_EFFECTS`` + + The call to the helper function may be removed if the return value is + not used. This means that it may not modify any CPU state nor may it + raise an exception. Code Optimizations ------------------- +================== When generating instructions, you can count on at least the following optimizations: @@ -908,20 +951,9 @@ Recommended coding rules for best performance often modified, e.g. the integer registers and the condition codes. TCG will be able to use host registers to store them. -- Avoid globals stored in fixed registers. They must be used only to - store the pointer to the CPU state and possibly to store a pointer - to a register window. - -- Use temporaries. Use local temporaries only when really needed, - e.g. when you need to use a value after a jump. Local temporaries - introduce a performance hit in the current TCG implementation: their - content is saved to memory at end of each basic block. - -- Free temporaries and local temporaries when they are no longer used - (tcg_temp_free). Since tcg_const_x() also creates a temporary, you - should free it after it is used. Freeing temporaries does not yield - a better generated code, but it reduces the memory usage of TCG and - the speed of the translation. +- Free temporaries when they are no longer used (``tcg_temp_free``). + Since ``tcg_const_x`` also creates a temporary, you should free it + after it is used. - Don't hesitate to use helpers for complicated or seldom used guest instructions. There is little performance advantage in using TCG to @@ -932,10 +964,6 @@ Recommended coding rules for best performance the instruction is mostly doing loads and stores, and in those cases inline TCG may still be faster for longer sequences. -- The hard limit on the number of TCG instructions you can generate - per guest instruction is set by ``MAX_OP_PER_INSTR`` in ``exec-all.h`` -- - you cannot exceed this without risking a buffer overrun. - - Use the 'discard' instruction if you know that TCG won't be able to prove that a given global is "dead" at a given program point. The x86 guest uses it to improve the condition codes optimisation. |