1 Tiny Code Generator - Fabrice Bellard.
5 TCG (Tiny Code Generator) began as a generic backend for a C
6 compiler. It was simplified to be used in QEMU. It also has its roots
7 in the QOP code generator written by Paul Brook.
11 The TCG "target" is the architecture for which we generate the
12 code. It is of course not the same as the "target" of QEMU which is
13 the emulated architecture. As TCG started as a generic C backend used
14 for cross compiling, it is assumed that the TCG target is different
15 from the host, although it is never the case for QEMU.
17 A TCG "function" corresponds to a QEMU Translated Block (TB).
19 A TCG "temporary" is a variable only live in a given
20 function. Temporaries are allocated explicitly in each function.
22 A TCG "global" is a variable which is live in all the functions. They
23 are defined before the functions defined. A TCG global can be a memory
24 location (e.g. a QEMU CPU register), a fixed host register (e.g. the
25 QEMU CPU state pointer) or a memory location which is stored in a
26 register outside QEMU TBs (not implemented yet).
28 A TCG "basic block" corresponds to a list of instructions terminated
29 by a branch instruction.
31 3) Intermediate representation
35 TCG instructions operate on variables which are temporaries or
36 globals. TCG instructions and variables are strongly typed. Two types
37 are supported: 32 bit integers and 64 bit integers. Pointers are
38 defined as an alias to 32 bit or 64 bit integers depending on the TCG
41 Each instruction has a fixed number of output variable operands, input
42 variable operands and always constant operands.
44 The notable exception is the call instruction which has a variable
45 number of outputs and inputs.
47 In the textual form, output operands come first, followed by input
48 operands, followed by constant operands. The output type is included
49 in the instruction name. Constants are prefixed with a '$'.
51 add_i32 t0, t1, t2 (t0 <- t1 + t2)
53 sub_i64 t2, t3, $4 (t2 <- t3 - 4)
59 - Basic blocks end after branches (e.g. brcond_i32 instruction),
60 goto_tb and exit_tb instructions.
61 - Basic blocks end before legacy dyngen operations.
62 - Basic blocks start after the end of a previous basic block, at a
63 set_label instruction or after a legacy dyngen operation.
65 After the end of a basic block, temporaries at destroyed and globals
66 are stored at their initial storage (register or memory place
67 depending on their declarations).
69 * Floating point types are not supported yet
71 * Pointers: depending on the TCG target, pointer size is 32 bit or 64
72 bit. The type TCG_TYPE_PTR is an alias to TCG_TYPE_I32 or
77 Using the tcg_gen_helper_x_y it is possible to call any function
78 taking i32, i64 or pointer types types. Before calling an helper, all
79 globals are stored at their canonical location and it is assumed that
80 the function can modify them. In the future, function modifiers will
81 be allowed to tell that the helper does not read or write some globals.
83 On some TCG targets (e.g. x86), several calling conventions are
88 Use the instruction 'br' to jump to a label. Use 'jmp' to jump to an
89 explicit address. Conditional branches can only jump to labels.
91 3.3) Code Optimizations
93 When generating instructions, you can count on at least the following
96 - Single instructions are simplified, e.g.
98 and_i32 t0, t0, $0xffffffff
102 - A liveness analysis is done at the basic block level. The
103 information is used to suppress moves from a dead temporary to
104 another one. It is also used to remove instructions which compute
105 dead results. The later is especially useful for condition code
106 optimization in QEMU.
108 In the following example:
114 only the last instruction is kept.
116 - A macro system is supported (may get closer to function inlining
117 some day). It is useful if the liveness analysis is likely to prove
118 that some results of a computation are indeed not useful. With the
119 macro system, the user can provide several alternative
120 implementations which are used depending on the used results. It is
121 especially useful for condition code optimization in QEMU.
128 The macro identified by the ID "$1" normally returns the values t0
129 and t1. Suppose its implementation is:
132 brcond_i32 t2, $0, $TCG_COND_EQ, $1
141 If t0 is not used after the macro, the user can provide a simpler
148 TCG automatically chooses the right implementation depending on
149 which macro outputs are used after it.
151 Note that if TCG did more expensive optimizations, macros would be
152 less useful. In the previous example a macro is useful because the
153 liveness analysis is done on each basic block separately. Hence TCG
154 cannot remove the code computing 't0' even if it is not used after
155 the first macro implementation.
157 3.4) Instruction Reference
159 ********* Function call
161 * call <ret> <params> ptr
163 call function 'ptr' (pointer type)
165 <ret> optional 32 bit or 64 bit return value
166 <params> optional 32 bit or 64 bit parameters
168 ********* Jumps/Labels
172 Absolute jump to address t0 (pointer type).
176 Define label 'label' at the current program point.
182 * brcond_i32/i64 cond, t0, t1, label
184 Conditional jump if t0 cond t1 is true. cond can be:
187 TCG_COND_LT /* signed */
188 TCG_COND_GE /* signed */
189 TCG_COND_LE /* signed */
190 TCG_COND_GT /* signed */
191 TCG_COND_LTU /* unsigned */
192 TCG_COND_GEU /* unsigned */
193 TCG_COND_LEU /* unsigned */
194 TCG_COND_GTU /* unsigned */
198 * add_i32/i64 t0, t1, t2
202 * sub_i32/i64 t0, t1, t2
208 t0=-t1 (two's complement)
210 * mul_i32/i64 t0, t1, t2
214 * div_i32/i64 t0, t1, t2
216 t0=t1/t2 (signed). Undefined behavior if division by zero or overflow.
218 * divu_i32/i64 t0, t1, t2
220 t0=t1/t2 (unsigned). Undefined behavior if division by zero.
222 * rem_i32/i64 t0, t1, t2
224 t0=t1%t2 (signed). Undefined behavior if division by zero or overflow.
226 * remu_i32/i64 t0, t1, t2
228 t0=t1%t2 (unsigned). Undefined behavior if division by zero.
232 * and_i32/i64 t0, t1, t2
236 * or_i32/i64 t0, t1, t2
240 * xor_i32/i64 t0, t1, t2
246 * shl_i32/i64 t0, t1, t2
248 t0=t1 << t2. Undefined behavior if t2 < 0 or t2 >= 32 (resp 64)
250 * shr_i32/i64 t0, t1, t2
252 t0=t1 >> t2 (unsigned). Undefined behavior if t2 < 0 or t2 >= 32 (resp 64)
254 * sar_i32/i64 t0, t1, t2
256 t0=t1 >> t2 (signed). Undefined behavior if t2 < 0 or t2 >= 32 (resp 64)
264 Move t1 to t0 (both operands must have the same type).
266 * ext8s_i32/i64 t0, t1
268 ext16s_i32/i64 t0, t1
269 ext16u_i32/i64 t0, t1
273 8, 16 or 32 bit sign/zero extension (both operands must have the same type)
277 16 bit byte swap on a 32 bit value. The two high order bytes must be set
290 Indicate that the value of t0 won't be used later. It is useful to
291 force dead code elimination.
293 ********* Type conversions
296 Convert t1 (32 bit) to t0 (64 bit) and does sign extension
298 * extu_i32_i64 t0, t1
299 Convert t1 (32 bit) to t0 (64 bit) and does zero extension
301 * trunc_i64_i32 t0, t1
302 Truncate t1 (64 bit) to t0 (32 bit)
306 * ld_i32/i64 t0, t1, offset
307 ld8s_i32/i64 t0, t1, offset
308 ld8u_i32/i64 t0, t1, offset
309 ld16s_i32/i64 t0, t1, offset
310 ld16u_i32/i64 t0, t1, offset
311 ld32s_i64 t0, t1, offset
312 ld32u_i64 t0, t1, offset
314 t0 = read(t1 + offset)
315 Load 8, 16, 32 or 64 bits with or without sign extension from host memory.
316 offset must be a constant.
318 * st_i32/i64 t0, t1, offset
319 st8_i32/i64 t0, t1, offset
320 st16_i32/i64 t0, t1, offset
321 st32_i64 t0, t1, offset
323 write(t0, t1 + offset)
324 Write 8, 16, 32 or 64 bits to host memory.
326 ********* QEMU specific operations
330 Exit the current TB and return the value t0 (word type).
334 Exit the current TB and jump to the TB index 'index' (constant) if the
335 current TB was linked to this TB. Otherwise execute the next
338 * qemu_ld_i32/i64 t0, t1, flags
339 qemu_ld8u_i32/i64 t0, t1, flags
340 qemu_ld8s_i32/i64 t0, t1, flags
341 qemu_ld16u_i32/i64 t0, t1, flags
342 qemu_ld16s_i32/i64 t0, t1, flags
343 qemu_ld32u_i64 t0, t1, flags
344 qemu_ld32s_i64 t0, t1, flags
346 Load data at the QEMU CPU address t1 into t0. t1 has the QEMU CPU
347 address type. 'flags' contains the QEMU memory index (selects user or
348 kernel access) for example.
350 * qemu_st_i32/i64 t0, t1, flags
351 qemu_st8_i32/i64 t0, t1, flags
352 qemu_st16_i32/i64 t0, t1, flags
353 qemu_st32_i64 t0, t1, flags
355 Store the data t0 at the QEMU CPU Address t1. t1 has the QEMU CPU
356 address type. 'flags' contains the QEMU memory index (selects user or
357 kernel access) for example.
359 Note 1: Some shortcuts are defined when the last operand is known to be
360 a constant (e.g. addi for add, movi for mov).
362 Note 2: When using TCG, the opcodes must never be generated directly
363 as some of them may not be available as "real" opcodes. Always use the
364 function tcg_gen_xxx(args).
368 tcg-target.h contains the target specific definitions. tcg-target.c
369 contains the target specific code.
373 The target word size (TCG_TARGET_REG_BITS) is expected to be 32 bit or
374 64 bit. It is expected that the pointer has the same size as the word.
376 On a 32 bit target, all 64 bit operations are converted to 32 bits. A
377 few specific operations must be implemented to allow it (see add2_i32,
378 sub2_i32, brcond2_i32).
380 Floating point operations are not supported in this version. A
381 previous incarnation of the code generator had full support of them,
382 but it is better to concentrate on integer operations first.
384 On a 64 bit target, no assumption is made in TCG about the storage of
385 the 32 bit values in 64 bit registers.
389 GCC like constraints are used to define the constraints of every
390 instruction. Memory constraints are not supported in this
391 version. Aliases are specified in the input operands as for GCC.
393 A target can define specific register or constant constraints. If an
394 operation uses a constant input constraint which does not allow all
395 constants, it must also accept registers in order to have a fallback.
397 The movi_i32 and movi_i64 operations must accept any constants.
399 The mov_i32 and mov_i64 operations must accept any registers of the
402 The ld/st instructions must accept signed 32 bit constant offsets. It
403 can be implemented by reserving a specific register to compute the
404 address if the offset is too big.
406 The ld/st instructions must accept any destination (ld) or source (st)
409 4.3) Function call assumptions
411 - The only supported types for parameters and return value are: 32 and
412 64 bit integers and pointer.
413 - The stack grows downwards.
414 - The first N parameters are passed in registers.
415 - The next parameters are passed on the stack by storing them as words.
416 - Some registers are clobbered during the call.
417 - The function can return 0 or 1 value in registers. On a 32 bit
418 target, functions must be able to return 2 values in registers for
421 5) Migration from dyngen to TCG
423 TCG is backward compatible with QEMU "dyngen" operations. It means
424 that TCG instructions can be freely mixed with dyngen operations. It
425 is expected that QEMU targets will be progressively fully converted to
426 TCG. Once a target is fully converted to TCG, it will be possible
427 to apply more optimizations because more registers will be free for
430 The exception model is the same as the dyngen one.