X-Git-Url: https://notaz.gp2x.de/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=deps%2Flightning%2FChangeLog;h=2cd527357fa627618c6bcc3ecabd84ec73323489;hb=79bfeef6160be4b228a7998ac2b43cd83d882532;hp=76cac91351ae418f714ebf68c5e21a125aba82ae;hpb=a34093eb63d1645fd2de9b412efe2587df9fdb3f;p=pcsx_rearmed.git diff --git a/deps/lightning/ChangeLog b/deps/lightning/ChangeLog index 76cac913..2cd52735 100644 --- a/deps/lightning/ChangeLog +++ b/deps/lightning/ChangeLog @@ -1,3 +1,333 @@ +2023-02-23 Paulo Andrade + + * include/lightning/jit_private.h: Add new 'inst' field to + jit_compiler_t, if __mips__ is defined. This field is a simple + helper for a pending instruction to be emitted, and that can + be emitted out of order. + * lib/jit_fallback.c: Update for changes in internal mips patching + and jumping macros and function calls. + * lib/jit_mips-cpu.c: Core of changes to attempt to fill delay + slots with instructions that can be emitted out of order. + * lib/jit_mips-fpu.c: Update to use delay slot in branches. + * lib/jit_mips.c: Update for new delay slot use logic. + +2023-02-20 Paulo Andrade + + * check/float.tst: Add conditionals for mips release for expected + NaN truncated to an integer. + * check/lightning.c: Add extra preprocessor for mips release. + * include/lightning/jit_mips.h: Make the NEW_ABI preprocessor + defined to zero if using the n32 or n64 abis. This makes it + easier to create runtime checks with an always true or false + condition. + * lib/jit_mips-cpu.c, lib/jit_mips-fpu.c: Implement mips release + 6 support. + * lib/jit_mips.c: Add more reliable mips release detection code. + +2023-02-09 Paulo Andrade + + * check/Makefile.am: Update for new bit.tst test, to check the + new clor, clzr, ctor and ctzr instructions. + * check/all.tst: Update to verify encoding of new instructions. + * check/lightning.c: Update to have the lightning "assembler" + understanding the new instructions. + * include/lightning.h.in: Define new codes for new instructions. + * lib/jit_aarch64.c, lib/jit_alpha.c, lib/jit_arm.c, lib/jit_hppa.c, + lib/jit_ia64.c, lib/jit_loongarch.c, lib/jit_mips.c, lib/jit_ppc.c, + lib/jit_riscv.c, lib/jit_s390.c, lib/jit_sparc.c, lib/jit_x86.c: + Implement fallback version of new instructions. + * lib/jit_fallback.c: Actual implementation of the fallbacks of + the new instructions. + * lib/jit_names.c: Update to print debug information of new + instructions. + +2023-01-26 Paulo Andrade + + * check/riprel.c, check/riprel.ok: New check files. + * check/Makefile.am: Support for new riprel test. + * lib/jit_x86-cpu.c, lib/jit_x86-sse.c, lib/jit_x86.c: Implement + %rip relative addressing when reliable. Currently disabled for + x32 and _WIN32; could be added for positive relative addresses + only where it should work. + * lib/lightning.c: Correct problem added in previous patch due + to not testing on a 32 bit environment. + +2023-01-23 Paulo Andrade + + * lib/jit_mips-cpu.c, lib/jit_mips-cpu.c: Use pseudo instructions + "b" (BEQ(0,0,disp)) and "bal" (BGEZAL(0,disp)) for mips2, when an + unconditional branch or function call is known to be in range of a + relative jump. This should significantly reduce jit size generation. + +2023-01-20 Paulo Andrade + + * lib/jit_mips-cpu.c, lib/jit_mips.c, lib/jit_rewind.c: Adapt + code to implement a variable framesize and optimize frame pointer + for simple leaf functions. + +2023-01-19 Paulo Andrade + + * lib/jit_riscv.c, lib/jit_riscv-cpu.c: Adapt code to use a + variable framesize. Previously it was aligning the stack at + 8 bytes, not 16. Now functions are called with a 16 byte aligned + stack. + +2023-01-18 Paulo Andrade + + * include/lightning/jit_private.h: Include new framesize field + of jit_compiler_t; add new alist field for jit_function_t; add + new cvt_offset and need_stack fields specific to x86. + * lib/jit_x86.c, lib/jit_x86-cpu: Rewrite code to create stack + frames, so that less stack space can be used if no, or very few + callee save registers are modified in a function. + * jit_x86-sse.c, jit_x86-x87.c: Make CVT_OFFSET variable, and + dynamically allocated; this is required to avoid needing to + modify twice %rsp at function prologs, even if no stack space + is used. + +2022-11-09 Paulo Andrade + + * configure.ac: Add new --enable-devel-strong-type-checking + option. + * include/lightning.h.in: Rework to not need to know if + PACKED_STACK is defined, and add a new argument to _jit_arg, + _jit_putarg{r,i}, _jit_pusharg{r,i} and _jit_ret{r,i} to have + the same code path if PACKED_STACK is defined or not, and also + to implement STRONG_TYPE_CHECK enabled with the new + --enable-devel-strong-type-checking. + * include/lightning/jit_private.h: Add new macros to add assertions + for STRONG_TYPE_CHECK and avoid pasting tokens in jit_inc_synth* + when the token is not a static known value. + * lib/jit_aarch64.c: The first implementation of the new code, + working correctly in Apple M1 and with and without STRONG_TYPE_CHECK + in Linux. + +2022-11-08 Paulo Andrade + + Add support for packed stack arguments as used by Apple M1 + aarch64 cpus. This requires a major redesign in how Lightning + works, because contrary to all other supported ports, in this + case arguments must be truncated and sign/zero extended if + passed in registers, but when receiving the argument, there + is no need to truncate and sign/zero extend. + Return values are also treated this way. The callee must + truncate sign/zero extend, not the caller. + check/Makefile.am: Add LIGHTNING_CFLAGS to AM_CFLAGS. + check/all.tst: Implement paired arg/getarg/pusharg/putarg/ret + codes to validate they do not generate assertions. + * check/allocar.tst, check/call.tst, check/fib.tst, check/put.tst, + check/stack.tst: Update to pass in all build types. + check/lightning.c: Add new codes for extra codes to handle + packed stack. + * configure.ac: Add a preprocessor define to know if packed stack + need is required. This is not really used, as it was moved to + jit_aarch64.h. + * doc/Makefile.am: Add LIGHTNING_CFLAGS to AM_CFLAGS. + * doc/rpn.c: Update to pass in all build types. + include/lightning.h.in: Add new codes and reorder enum. + * include/lightning/jit_aarch64.h: Detect condition of needing + a packed stack. + * lib/jit_aarch64-sz.c: Regenerate. + * lib/jit_aarch64.c: Major updates for packed stack. + * lib/jit_names.c: Updates for debug output. + * lib/lightning.c: Update for new codes. + +2022-10-31 Marc Nieper-Wißkirchen + + Add new skip instruction. + * .gitignore: Update from Gnulib. + * check/Makefile.am: Add tests. + * check/lightning.c: Handle skip instructions. + * check/protect.c: Rewrite with skip. + * check/skip.ok: New test. + * check/skip.tst: New test. + * doc/body.texi: Document the skip instruction. + * include/lightning.h.in: Add the skip instruction. + * lib/jit_aarch64-sz.c: Update for skip instruction. + * lib/jit_aarch64.c: Implement skip instruction. + * lib/jit_alpha-sz.c: Update for skip instruction. + * lib/jit_alpha.c: Implement skip instruction. + * lib/jit_arm-sz.c: Update for skip instruction. + * lib/jit_arm.c: Implement skip instruction. + * lib/jit_hppa-sz.c: Update for skip instruction. + * lib/jit_hppa.c: Implement skip instruction. + * lib/jit_ia64-sz.c: Update for skip instruction. + * lib/jit_ia64.c: Implement skip instruction. + * lib/jit_loongarch-sz.c: Update for skip instruction. + * lib/jit_loongarch.c: Implement skip instruction. + * lib/jit_mips-sz.c: Update for skip instruction. + * lib/jit_mips.c: Implement skip instruction. + * lib/jit_names.c: Update for skip instruction. + * lib/jit_ppc-sz.c: Update for skip instruction. + * lib/jit_ppc.c: Implement skip instruction. + * lib/jit_riscv-sz.c: Update for skip instruction. + * lib/jit_riscv.c: Implement skip instruction. + * lib/jit_s390-sz.c: Update for skip instruction. + * lib/jit_s390.c: Implement skip instruction. + * lib/jit_size.c: Treat align and skip in a special way. + * lib/jit_sparc-sz.c: Update for skip instruction. + * lib/jit_sparc.c: Implement skip instruction. + * lib/jit_x86-sz.c: Update for skip instruction. + * lib/jit_x86.c: Implement skip instruction. + * lib/lightning.c: Classify skip instruction. + +2022-10-30 Marc Nieper-Wißkirchen + + Add user-visible functions jit_protect and jit_unprotect. + * check/Makefile.am: Add test for jit_protect and jit_unprotect. + * check/protect.c: New test. + * doc/body.texi: Add documentation for jit_protect and + jit_unprotect. + * include/lightning.h.in: Add prototypes for jit_protect and + jit_unprotect. + * include/lightning/jit_private.h: Add a field to store the size + of the protected memory. + * lib/lightning.c: Remember the size of the protected memory and + implement the two new functions. + +2022-10-12 Paulo Andrade + + * include/lightning/jit_loongarch.h, lib/jit_loongarch-cpu.c, + lib/jit_loongarch-fpu.c, lib/jit_loongarch-sz.c, lib/jit_loongarch.c: + New files implementing the first version of the new loongarch port. + * check/float.tst: Add preprocessor checks for NaN and +-Inf + values converted to integers for loongarch. + * configure.ac: Add check and conditionals for new architecture. + * include/lightning.h.in, check/lightning.c, + include/lightning/Makefile.am, include/lightning/jit_private.h, + lib/Makefile.am, lib/jit_size.c, lib/lightning.c: Update for new + port. + +2022-10-05 Paulo Andrade + + * check/lightning.c: Remove -Dmacro=value from usage and attempt + to parse it. It was buggy and not properly implemented. Now + it pass any extra options to the generated jit. To pass any + option starting with '-' need to also use '--'. + * check/collatz.e: New sample file showing an example of jit + generation. + +2022-10-04 Paulo Andrade + + * include/lightning/jit_private.h: Add new flag to jit_block_t. + * lib/lightning.c: Rewrite register liveness and state at block + entry code to avoid a very expensive and non scaling code path. + Now it attempts to do as few as possible recomputations when + merging state of adjacent blocks, still doing one extra nop pass + (in the sense that it will not find any changes) to make sure the + logic is correct. + +2022-09-30 Paulo Andrade + + * include/lightning/jit_private.h: Implement new data structures + specific to riscv. + * lib/jit_disasm.c: Extra disassemble code for riscv constant pool. + * lib/jit_riscv-cpu.c: Modify movi to use constant pool if 3 or + more instructions are required to construct constant and modify + movi_p to use a pc relative load from a constant pool. + lib/jit_riscv-sz.c: Update for new constant pool code. Most + instructions that need 64 bit constants are significantly reduced. + * lib/jit_riscv.c: Implement most of the constant pool code. + * lib/jit_size.c: Update for estimate of code generation size. + * lib/lightning.c: Update for riscv specific code, and also make + sure to mprotect the constant pool as executable. + +2022-09-08 Paulo Andrade + + * lib/jit_fallback.c: Implement fallback compare and swap with + pthreads. + * check/Makefile.am: Update for new cas{r,i} simple test. + * check/catomic.c, check/catomic.ok: New test case for + simple compare and swap atomic operation. + * check/lightning.c: Add entries to be able to use + the new compare and swap atomic operation. Still missing + a general test, only the basic C version. + * include/lightning.h.in: Include pthread.h, even if not + needing a fallback compare and swap. + * include/lightning/jit_private.h: Add support for a register pair + in second argument. Required by the new casr and casi operations. + * lib/jit_aarch64-cpu.c, lib/jit_aarch64-sz.c, lib/jit_aarch64.c, + lib/jit_ppc-cpu.c, lib/jit_ppc-sz.c, lib/jit_ppc.c, lib/jit_x86-cpu.c, + lib/jit_x86-sz.c, lib/jit_x86.c: Implement inline code for compare + and swap. + * lib/jit_arm-cpu.c, lib/jit_arm-sz.c, lib/jit_arm.c: Implement + inline code for compare and swap if cpu is armv7, otherwise, use + a fallback with pthreads. + * lib/jit_alpha-cpu.c, lib/jit_alpha-sz.c, lib/jit_alpha.c, + lib/jit_hppa-cpu.c, lib/jit_hppa-sz.c, lib/jit_hppa.c, + lib/jit_ia64-cpu.c, lib/jit_ia64-sz.c, lib/jit_ia64.c, + lib/jit_mips-cpu.c, lib/jit_mips-sz.c, lib/jit_mips.c, + lib/jit_riscv-cpu.c, lib/jit_riscv-sz.c, lib/jit_riscv.c, + lib/jit_s390-cpu.c, lib/jit_s390-sz.c, lib/jit_s390.c, + lib/jit_sparc-cpu.c, lib/jit_sparc-sz.c, lib/jit_sparc.c: Implement + fallback compare and swap with pthreads. At least some of these + should be updated for inline code generation. + * lib/jit_names.c, lib/jit_print.c: lib/lightning.c: Update for the + new compare and swap operation. + * doc/body.texi: Add simple documentation of the compare and swap + new operation. + +2022-08-12 Marc Nieper-Wißkirchen + + Document jit_align. + * doc/body.texi: Add documentation for jit_align. + +2022-05-14 Paulo Andrade + + * include/lightning.h.in: Reorder jit_mov{n,z}r in instruction list. + * lib/jit_alpha.c, lib/jit_alpha-cpu.c, lib/jit_hppa.c, + lib/jit_hppa-cpu.c, lib/jit_ia64.c, lib/jit_ia64-cpu.c, + lib/jit_riscv.c, lib/jit_riscv-cpu.c, lib/jit_s390.c, + lib/jit_s390-cpu.c, lib/jit_sparc.c, lib/jit_sparc-cpu.c: + Implement fallback jit_mov{n,z}r. These are a somewhat cheap + implementation, but should be reviewed for the arches that already + have a proper conditional move. + * lib/jit_arm-sz.c, lib/jit_mips-sz.c: Add missing maximum size + estimative and reorder. + * lib/jit_aarch64-sz.c, lib/jit_x86-sz.c, lib/jit_ppc-sz.c: + Reorder entry to match definition order. + * lib/jit_aarch64-sz.c, lib/jit_alpha-sz.c, lib/jit_hppa-sz.c, + lib/jit_ia64-sz.c, lib/jit_riscv-sz.c, lib/jit_s390-sz.c, + lib/jit_sparc-sz.c: Add heuristic value, basically the sum of + the cost of a movr + beqr. + * lib/jit_names.c: Add entries for debug output of mov{n,z}r. + * lib/lightning.c: Use proper bitmask in jit_classify. + +2021-04-03 Marc Nieper-Wißkirchen + + * check/Makefile.am: Add test for the live instruction. + * check/live.ok: New file. + * check/live.tst: New file. + * doc/body.texi: Add documentation for the live instruction and + for jit_get_reg/jit_get_unreg. Fix menu entries. + * include/lightning.h.in (jit_get_reg, jit_unget_reg): Expose the + macros in the public header file. + * include/lightning/jit_private.h (jit_get_reg, jit_unget_reg): + Remove the macros from the private header file. + +2021-04-03 Marc Nieper-Wißkirchen + + * Makefile.am, check/Makefile.am, doc/Makefile.am, + lib/Makefile.am: Include $(top_builddir)/include in include paths + for the autoconf-generated header file lightning.h. + +2021-04-03 Marc Nieper-Wißkirchen + + * doc/.gitignore: Add version.texi to list of ignored files. + * doc/version.texi: Remove file from version control. + +2020-18-04 Paulo Andrade + + * lib/jit_x86-cpu.c, lib/jit_x86.c: Implement %rip relative + calls and jumps on x86_64. Currently very conservative, assuming + a jit block can be larger than 2G, so, if a jump or call is in + the same jit generation, but target is unknown, use an indirect + branch (could have an option to assume a jit code block is + never larger than 2G). Also a deoptimization is that now does + not always generate ip relative jmpi; previously implicitly + assumed jmpi would never be larger than 2G. Overall still an + optimization. + 2020-23-01 Paulo Andrade * lib/lightning.c: Add a proper fix to the condition of considering @@ -26,7 +356,7 @@ to the set of registers to scan for live range, what might consume a lot of cpu time, doing nothing. -2019-09-16 Marc Nieper-WiÃkirchen +2019-09-16 Marc Nieper-Wißkirchen * include/lightning/jit_x86.h, lib/jit_x86.c: Correct x86_64 backend, made %r12 a callee-save register as dictated by the @@ -65,7 +395,7 @@ https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=925129 * THANKS: update. -2019-08-29 Marc Nieper-WiÃkirchen +2019-08-29 Marc Nieper-Wißkirchen * include/lightning/jit_private.h: Move definition of offsetof from the public header file here. @@ -1415,7 +1745,7 @@ * include/lightning/jit_arm.h: Correct wrong jit_f macro definition. - * include/lightning/jit_ia64.h, include/lightning/jit_ppc.h: + * include/lightning/jit_ia64.h, include/lightning/jit_ppc.h: Correct wrong jit_r macro definition. * lib/jit_x86-x87.c, lib/jit_x86.c: Actually use the @@ -1884,7 +2214,7 @@ divr_f and divrd_d implementation. * check/lightning.c: Add __ia64__ preprocessor define - on Itanium. + on Itanium. * check/alu.inc, check/clobber.tst, check/float.tst: Define several macros conditionally to __ia64__. This is required @@ -2991,7 +3321,7 @@ * include/lightning/jit_x86.h: Correct typo in macro name. * lib/jit_arm.c, lib/jit_arm-cpu.c, lib/jit_mips.c, - lib/jit_mips-cpu.c, lib/jit_ppc.c, lib/jit_ppc-cpu.c, + lib/jit_mips-cpu.c, lib/jit_ppc.c, lib/jit_ppc-cpu.c, lib/jit_x86.c, lib/jit_x86-cpu.c: Correct wrong code to get current jit function pointer. @@ -3225,7 +3555,7 @@ * lightning/i386/core.h (jit_ldr_c, jit_ldxr_c, jit_ldr_s, jit_ldxr_s): Move... * lightning/i386/core-32.h: ... here. - * lightning/i386/core-64.h (jit_ldr_c, jit_ldxr_c, jit_ldr_s, + * lightning/i386/core-64.h (jit_ldr_c, jit_ldxr_c, jit_ldr_s): Use movsbq and movswq. 2010-08-10 Paulo César Pereira de Andrade @@ -3306,11 +3636,11 @@ * lightning/ppc/funcs.h (jit_flush_code): modified the computation of start/end. The pointer arithmetic was done without casting. It - prevented compilation with recent gcc versions. + prevented compilation with recent gcc versions. * lightning/ppc/core.h (jit_pushr_i): The offset for the store was incorrect. Should have been 4 bytes below SP (not above). - * lightning/ppc/core.h (jit_popr_i): The offset for the load was - incorrect. Should have been 0 (not +8). + * lightning/ppc/core.h (jit_popr_i): The offset for the load was + incorrect. Should have been 0 (not +8). 2008-06-17 Paolo Bonzini @@ -3589,7 +3919,7 @@ * lightning/i386/fp-32.h: ... here. * lightning/i386/fp-64.h: Write the code. * lightning/sparc/fp.h: Fix jit_extr_{f_d,d_f} register order. - + 2006-11-22 Paolo Bonzini * lightning/i386/asm-i386.h: Move x86-64 instructions... @@ -3803,7 +4133,7 @@ 2006-01-23 Paolo Bonzini * configure.ac: Fix comments in config.h.in. - + 2005-11-25 Paolo Bonzini * lightning/sparc/fp.h: Fix header comment. @@ -4123,7 +4453,7 @@ * lightning/ppc/funcs.h: correctly align stack pointer No changelogs for the assemblers (lightning directory) until 1.0 - + 2003-03-27 Paolo Bonzini * tests/printf2.c: new test @@ -4135,7 +4465,7 @@ No changelogs for the assemblers (lightning directory) until 1.0 2001-01-19 Paolo Bonzini * configure.in: support cross-assembling - + * disass/bfd.h, disass/dis-asm.h, disass/dis-buf.c, disass/i386-dis.c, disass/i386.h, disass/ppc-dis.c, disass/ppc.h, disass/ppc-opc.c, disass/sparc-dis.c,