aboutsummaryrefslogtreecommitdiffstats
path: root/backend
Commit message (Collapse)AuthorAgeFilesLines
* Merge branch 'master' into dwarfBernhard Schommer2015-01-233-6/+6
|\
| * Merge branch 'named-structs'Xavier Leroy2015-01-233-6/+6
| |\ | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | - Switch CompCert C / Clight AST of composite types (structs and unions) from a structural representation to a nominal representation, closer to concrete syntax. - This avoids algorithmic inefficiencies due to the structural representation. - Closes PR#4. - Smallstep: make small-step semantics more polymorphic in the type of the global environment. - Globalenvs: introduce Senv.t (symbol environments) as a restricted view on Genv.t (full global environments). - Events, Smallstep: use Senv instead of Genv to talk about global names.
| | * Introduce symbol environments (type Senv.t) as a restricted view on global ↵Xavier Leroy2014-11-263-6/+6
| | | | | | | | | | | | environments (type Genv.t). Use symbol environments instead of global environments for external functions (module Events).
* | | Merge branch 'master' into dwarfBernhard Schommer2015-01-205-250/+273
|\| |
| * | Protect against redefinition of the __i64_xxx helper library functions.Xavier Leroy2015-01-204-244/+252
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This is achieved by declaring these functions in C2C.ml, then re-checking their declarations in the global environment during the Selection pass. In passing, the "hf" parameter for SelectLong functions was removed and replaced by fixed identifiers corresponding to the actual names of the helper functions.
| * | Follow-up to [5aecefe]: be conservative also in the case of a "common" ↵Xavier Leroy2015-01-201-6/+21
| | | | | | | | | | | | global const variable.
* | | Merge branch 'master' into dwarfBernhard Schommer2015-01-121-3/+6
|\| |
| * | More prudent analysis of uninitialized const global variables.Xavier Leroy2015-01-091-3/+6
| | | | | | | | | | | | | | | | | | | | | In the presence of separate compilation and linking, an uninitialized const global variable may be initialized elsewhere with a pointer value, falsifying the points-to analysis. Report and fix by Chung-Kil Hur and Jeehoon Kang.
* | | Merge branch 'master' into dwarfBernhard Schommer2015-01-121-0/+67
|\| | | |/ |/| | | | | Conflicts: powerpc/PrintAsm.ml
| * In -g -S mode, annotate the generated asm file with the C source code in ↵Xavier Leroy2015-01-071-0/+67
|/ | | | | | comments. Refactor printing of .loc debug directives in backend/PrintAnnot.ml
* Verification of the Unusedglob pass (removal of unreferenced static global ↵Xavier Leroy2014-11-244-95/+1401
| | | | definitions). Assorted changes to ia32/Op.v. PowerPC and ARM need updating.
* Add Genv.public_symbol operation.Xavier Leroy2014-11-2413-56/+136
| | | | | Restrict pointer event values to public global names. Update proofs accordingly. PowerPC and ARM need updating.
* Record public global definitions via field "prog_public" in AST.program.Xavier Leroy2014-11-242-0/+3
| | | | For the moment, this field is ignored.
* Add flags to control individual optimization passes + flag -O0 for turning ↵Xavier Leroy2014-11-162-8/+5
| | | | them off.
* - Support "switch" statements over 64-bit integersxleroy2014-08-1713-479/+845
| | | | | | | | | | | | | (in CompCert C to Cminor, included) - Translation of "switch" to decision trees or jumptables made generic over the sizes of integers and moved to the Cminor->CminorSel pass instead of CminorSel->RTL as before. - CminorSel: add "exitexpr" to support the above. - ValueDomain: more precise analysis of comparisons against an integer literal. E.g. "x >=u 0" is always true. git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2565 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Merge the various $(ARCH)/$(VARIANT)/xxx.v files into $(ARCH)/xxx.v.xleroy2014-07-231-1/+1
| | | | | | | | | The only platform where we have two variants is ARM, and it's easier to share the callling convention code between the two than to maintain both variants separately. git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2540 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Merge of "newspilling" branch:xleroy2014-07-2327-1134/+1011
| | | | | | | | | | | | | | | - Support single-precision floats as first-class values - Introduce chunks Many32, Many64 and types Tany32, Tany64 to support saving and restoring registers without knowing the exact types (int/single/float) of their contents, just their sizes. - Memory model: generalize the opaque encoding of pointers to apply to any value, not just pointers, if chunks Many32/Many64 are selected. - More properties of FP arithmetic proved. git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2537 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Refactoring: move symbol_offset into Genv.xleroy2014-05-243-11/+8
| | | | git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2503 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Integration of Jacques-Henri Jourdan's verified parser.xleroy2014-04-292-13/+5
| | | | | | | (Merge of branch newparser.) git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2469 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Constprop: use "not" for "xorimm(-1)"; optimize == 1 and != 0 comparisons ↵xleroy2014-04-093-32/+40
| | | | | | | | | | over booleans. Select*: more systematic constant propagation; don't CP shifts by amounts outside of [0..31]. Driver: timer for whole compilation. git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2452 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Reducing compilation times: (by 35% on one example)xleroy2014-04-061-20/+18
| | | | | | | | - Use a hashtable instead of an AVL set for adjSet - Inlining and algebraic simplification of spillCost into selectSpill. git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2450 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Merge of branch linear-typing:xleroy2014-04-066-671/+1214
| | | | | | | | | | | | | | | | | | | | | | | 1) Revised division of labor between RTLtyping and Lineartyping: - RTLtyping no longer keeps track of single-precision floats, switches from subtype-based inference to unification-based inference. - Unityping: new library for unification-based inference. - Locations: don't normalize at assignment in a stack slot - Allocation, Allocproof: simplify accordingly. - Lineartyping: add inference of locations that contain a single-precision float. - Stackingproof: adapted accordingly. This addresses a defect report whereas RTLtyping was rejecting code that used a RTL pseudoreg to hold both double- and single-precision floats (see test/regression/singlefloats.c). This corresponds to commits 2435+2436 plus improvements in Lineartyping. 2) Add -dtimings option to measure compilation times. Moved call to C parser from Elab to Parse, to make it easier to measure parsing time independently of elaboration time. git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2449 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Inlining: preserve all RTL regs mentioned in the function, not justxleroy2014-03-293-107/+164
| | | | | | | | those defined in the function. Semantically, both are correct, but the latter may cause RTLtyping to fail if some regs are uninitialized and a collision occurs between regs of different types. RTL: move the function resource computations there, as it could be useful for other passes some day. git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2440 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Revert commits r2435 and r2436 (coarser RTLtyping / finer Lineartyping):xleroy2014-03-286-1076/+661
| | | | | | | the new Lineartyping can't keep track of single floats that were spilled. git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2438 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Commentsxleroy2014-03-271-0/+7
| | | | git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2437 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Move wt_instr_inv where it belongs.xleroy2014-03-272-10/+10
| | | | | | | Update Makefile and dependencies. git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2436 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Revised division of labor between RTLtyping and Lineartyping:xleroy2014-03-276-661/+1069
| | | | | | | | | | | | | | | | | - RTLtyping no longer keeps track of single-precision floats, switches from subtype-based inference to unification-based inference. - Unityping: new library for unification-based inference. - Locations: don't normalize at assignment in a stack slot - Allocation, Allocproof: simplify accordingly. - Lineartyping: add inference of machine registers that contain a single-precision float. - Stackingproof: adapted accordingly. This addresses a defect report by D. Dickman whereas RTLtyping was rejecting code that used a RTL pseudoreg to hold both double- and single-precision floats. git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2435 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* In Regalloc, dead code elimination, don't eliminate move operationsxleroy2014-02-233-3/+13
| | | | | | | | that pop the x87 FP stack (var <- FP0). Otherwise, (void) f(); where f returns a float eventually produces a FP stack overflow. git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2416 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Add option -Os to optimize for code size rather than for execution speed.xleroy2014-02-198-58/+78
| | | | | | | | Refactored compilation flags that affect the Coq part (module Compopts). Added support for C99 for loops with declarations. git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2410 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Recognize .i and .p source files as C sources not to be preprocessed.xleroy2014-02-051-0/+11
| | | | | | | Add CompCert version number and command line as comments in generated asm. git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2407 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* - Back to origins: suppress Mfloat64al32 chunk and align Mfloat64 to 4.xleroy2014-01-127-34/+9
| | | | | | | - Revised printing of intermediate RTL code. git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2403 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Introduce and use the platform-specific Archi module giving:xleroy2014-01-033-20/+22
| | | | | | | | | | - endianness - alignment constraints for 8-byte types (which is 4 for x86 ABI and 8 for other ABIs) - NaN handling options (superceding the Nan module, removed). git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2402 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Updated neededness analysis for IA32.xleroy2014-01-021-1/+14
| | | | git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2400 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Updated ARM backend wrt new static analyses and optimizations.xleroy2014-01-025-128/+176
| | | | | | | | NeedOp, Deadcode: must have distinct needs per argument of an operator. This change remains to be propagated to IA32 and PPC. git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2399 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Ctypes.sizeof ty = 0 for empty types ty (zero-sized array, empty struct/union).xleroy2013-12-303-10/+4
| | | | | | | __builtin_memcpy_aligned now supports the case sz = 0. git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2392 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Simpler, more robust emulation of calls to variadic functions:xleroy2013-12-289-25/+26
| | | | | | | | | | | | | | - C function types and Cminor signatures annotated by calling conventions. esp. vararg / not vararg - Cshmgen: generate correct code for function call where there are more arguments than listed in the function prototype. This is still undefined behavior according to the formal semantics, but correct code is generated. - C2C, */PrintAsm.ml: remove "printf$iif" hack. - powerpc/, checklink/: don't generate stubs for variadic functions. git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2386 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Support "default" cases in the middle of a "switch", not just at the end.xleroy2013-12-211-1/+1
| | | | git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2383 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Merge of branch value-analysis.xleroy2013-12-2019-2289/+10535
| | | | git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2381 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* - Recognize __builtin_fabs as an operator, not just a builtin,xleroy2013-11-061-5/+7
| | | | | | | | | enabling more aggressive optimizations. - Less aggressive CSE for EF_builtin builtins, causes problems for __builtin_write{16,32}_reversed. git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2363 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Revised renumbering of nodes and registers so that main function is not ↵xleroy2013-10-183-48/+114
| | | | | | shifted by 1 git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2346 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Cminor parsing and printing (from Andrew Tolmach)xleroy2013-10-164-19/+158
| | | | git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2345 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Do not use Format for faster printing of RTL, XTL, LTL, Machxleroy2013-09-265-136/+128
| | | | git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2337 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Small improvements in compilation times for the register allocation pass.xleroy2013-09-203-21/+24
| | | | | | | Maps.v: add a PTree.fold1 operation that doesn't maintain the key. git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2329 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Floats.v, Nan.v: hard-wire the general shape of binop_pl, so that no axioms xleroy2013-09-144-6/+46
| | | | | | | | | are necessary, only two parameters (default_pl and choose_binop_pl). SelectDiv: optimize FP division by a power of 2. ConstpropOp: optimize 2.0 * x and x * 2.0 into x + x. git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2326 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Oops, wrong commit of generated files.xleroy2013-08-242-758/+0
| | | | git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2311 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Forgot to add these two files.xleroy2013-08-242-0/+758
| | | | git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2310 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Simplify LPMap by smashing bottoms.xleroy2013-08-121-7/+8
| | | | git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2306 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Change interface of Kildall solvers to avoid precomputing the map pc -> list ↵xleroy2013-08-1219-236/+287
| | | | | | of successors. git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2305 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Optimize integer divisions by positive constants, turning them intoxleroy2013-07-295-1/+707
| | | | | | | multiply-high and shifts. git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2300 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
* Treat casts int64 -> float32 as primitive operations instead of twoxleroy2013-07-037-6/+51
| | | | | | | | casts int64 -> float64 -> float32. The latter causes double rounding errors. git-svn-id: https://yquem.inria.fr/compcert/svn/compcert/trunk@2290 fca1b0fc-160b-0410-b1d3-a4f43f01ea2e
generated by cgit (git 2.34.1) at 2024-06-28 22:08:56 +0000