{"payload":{"feedbackUrl":"https://github.com/orgs/community/discussions/53140","repo":{"id":24047050,"defaultBranch":"call0-4.8.2","name":"gcc-xtensa","ownerLogin":"jcmvbkbc","currentUserCanPush":false,"isFork":false,"isEmpty":false,"createdAt":"2014-09-15T07:27:48.000Z","ownerAvatar":"https://avatars.githubusercontent.com/u/166731?v=4","public":true,"private":false,"isOrgOwned":false},"refInfo":{"name":"","listCacheKey":"v0:1717233717.0","currentOid":""},"activityList":{"items":[{"before":"17ee4e65dbd88b17f1c17fccca3d043bccf211f4","after":"ed96d0a44ac686032634737a323989f055dd8c46","ref":"refs/heads/xtensa-15-updates","pushedAt":"2024-07-30T00:24:44.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"xtensa: Add missing speed cost for TYPE_FARITH in TARGET_INSN_COST\n\nAccording to the implemented pipeline model, this cost can be assumed to be\n1 clock cycle.\n\ngcc/ChangeLog:\n\n\t* config/xtensa/xtensa.cc (xtensa_insn_cost):\n\tAdd a case statement for TYPE_FARITH.","shortMessageHtmlLink":"xtensa: Add missing speed cost for TYPE_FARITH in TARGET_INSN_COST"}},{"before":"29e230baeba397ec924e59cfea8eccc9e34b88d5","after":"17ee4e65dbd88b17f1c17fccca3d043bccf211f4","ref":"refs/heads/xtensa-15-updates","pushedAt":"2024-07-29T07:10:02.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"xtensa: Make use of scaled [U]FLOAT/TRUNC.S instructions\n\n[U]FLOAT.S machine instruction in Xtensa ISA, which converts an integer to\na hardware single-precision FP register, has the ability to divide the\nresult by power of two (0 to 15th).\n\nSimilarly, [U]TRUNC.S instruction, which truncates single-precision FP to\ninteger, can multiply the source value by power of two in advance, but\nneither of these currently uses this function (always specified with 0th\npower of two, i.e. a scaling factor of 1).\n\nThis patch unleashes the scaling ability of the above instructions.\n\n     /* example */\n     float test0(int a) {\n       return a / 2.f;\n     }\n     float test1(unsigned int a) {\n       return a / 32768.f;\n     }\n     int test2(float a) {\n       return a * 2;\n     }\n     unsigned int test3(float a) {\n       return a * 32768;\n     }\n\n     ;; before\n     test0:\n     \tmovi.n\ta9, 0x3f\n     \tfloat.s\tf0, a2, 0\n     \tslli\ta9, a9, 24\n     \twfr\tf1, a9\n     \tmul.s\tf0, f0, f1\n     \trfr\ta2, f0\n     \tret.n\n     test1:\n     \tmovi.n\ta9, 7\n     \tufloat.s\tf0, a2, 0\n     \tslli\ta9, a9, 27\n     \twfr\tf1, a9\n     \tmul.s\tf0, f0, f1\n     \trfr\ta2, f0\n     \tret.n\n     test2:\n     \twfr\tf1, a2\n     \tadd.s\tf0, f1, f1\n     \ttrunc.s\ta2, f0, 0\n     \tret.n\n     test3:\n     \tmovi.n\ta9, 0x47\n     \tslli\ta9, a9, 24\n     \twfr\tf1, a2\n     \twfr\tf2, a9\n     \tmul.s\tf0, f1, f2\n     \tutrunc.s\ta2, f0, 0\n     \tret.n\n\n     ;; after\n     test0:\n     \tfloat.s\tf0, a2, 1\n     \trfr\ta2, f0\n     \tret.n\n     test1:\n     \tufloat.s\tf0, a2, 15\n     \trfr\ta2, f0\n     \tret.n\n     test2:\n     \twfr\tf0, a2\n     \ttrunc.s\ta2, f0, 1\n     \tret.n\n     test3:\n     \twfr\tf0, a2\n     \tutrunc.s\ta2, f0, 15\n     \tret.n\n\ngcc/ChangeLog:\n\n\t* config/xtensa/predicates.md\n\t(fix_scaling_operand, float_scaling_operand): New predicates.\n\t* config/xtensa/xtensa.md\n\t(any_fix/m_fix/s_fix, any_float/m_float/s_float):\n\tNew code iterators and their attributes.\n\t(fix<s_fix>_truncsfsi2): Change from \"fix_truncsfsi2\".\n\t(*fix<s_fix>_truncsfsi2_2x, *fix<s_fix>_truncsfsi2_scaled):\n\tNew insn definitions.\n\t(float<s_float>sisf2): Change from \"floatsisf2\".\n\t(*float<s_float>sisf2_scaled): New insn definition.","shortMessageHtmlLink":"xtensa: Make use of scaled [U]FLOAT/TRUNC.S instructions"}},{"before":"9c697d888d5d15f88f72d909471d971ba1f1a45e","after":"29e230baeba397ec924e59cfea8eccc9e34b88d5","ref":"refs/heads/xtensa-15-updates","pushedAt":"2024-07-20T04:55:38.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"xtensa: Make use of scaled [U]FLOAT/TRUNC.S instructions\n\n[U]FLOAT.S machine instruction in Xtensa ISA, which converts an integer to\na hardware single-precision FP register, has the ability to divide the\nresult by power of two (0 to 15th).\n\nSimilarly, [U]TRUNC.S instruction, which truncates single-precision FP to\ninteger, can multiply the source value by power of two in advance, but\nneither of these currently uses this function (always specified with 0th\npower of two, i.e. a scaling factor of 1).\n\nThis patch unleashes the scaling ability of the above instructions.\n\n     /* example */\n     float test0(int a) {\n       return a / 2.f;\n     }\n     float test1(unsigned int a) {\n       return a / 32768.f;\n     }\n     int test2(float a) {\n       return a * 2;\n     }\n     unsigned int test3(float a) {\n       return a * 32768;\n     }\n\n     ;; before\n     test0:\n     \tmovi.n\ta9, 0x3f\n     \tfloat.s\tf0, a2, 0\n     \tslli\ta9, a9, 24\n     \twfr\tf1, a9\n     \tmul.s\tf0, f0, f1\n     \trfr\ta2, f0\n     \tret.n\n     test1:\n     \tmovi.n\ta9, 7\n     \tufloat.s\tf0, a2, 0\n     \tslli\ta9, a9, 27\n     \twfr\tf1, a9\n     \tmul.s\tf0, f0, f1\n     \trfr\ta2, f0\n     \tret.n\n     test2:\n     \twfr\tf1, a2\n     \tadd.s\tf0, f1, f1\n     \ttrunc.s\ta2, f0, 0\n     \tret.n\n     test3:\n     \tmovi.n\ta9, 0x47\n     \tslli\ta9, a9, 24\n     \twfr\tf1, a2\n     \twfr\tf2, a9\n     \tmul.s\tf0, f1, f2\n     \tutrunc.s\ta2, f0, 0\n     \tret.n\n\n     ;; after\n     test0:\n     \tfloat.s\tf0, a2, 1\n     \trfr\ta2, f0\n     \tret.n\n     test1:\n     \tufloat.s\tf0, a2, 15\n     \trfr\ta2, f0\n     \tret.n\n     test2:\n     \twfr\tf0, a2\n     \ttrunc.s\ta2, f0, 1\n     \tret.n\n     test3:\n     \twfr\tf0, a2\n     \tutrunc.s\ta2, f0, 15\n     \tret.n\n\ngcc/ChangeLog:\n\n\t* config/xtensa/predicates.md\n\t(fix_scaling_operand, float_scaling_operand): New predicates.\n\t* config/xtensa/xtensa.md\n\t(any_fix/m_fix/s_fix, any_float/m_float/s_float):\n\tNew code iterators and their attributes.\n\t(fix<s_fix>_truncsfsi2): Change from \"fix_truncsfsi2\".\n\t(*fix<s_fix>_truncsfsi2_2x, *fix<s_fix>_truncsfsi2_scaled):\n\tNew insn definitions.\n\t(float<s_float>sisf2): Change from \"floatsisf2\".\n\t(*float<s_float>sisf2_scaled): New insn definition.","shortMessageHtmlLink":"xtensa: Make use of scaled [U]FLOAT/TRUNC.S instructions"}},{"before":"a5e14a113d6fb6e7d16b409ecac7e86e5488f5b5","after":"9c697d888d5d15f88f72d909471d971ba1f1a45e","ref":"refs/heads/xtensa-15-updates","pushedAt":"2024-07-15T20:22:03.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"xtensa: Make use of scaled [U]FLOAT/TRUNC.S instructions\n\n[U]FLOAT.S machine instruction in Xtensa ISA, which converts an integer to\na hardware single-precision FP register, has the ability to divide the\nresult by power of two (0 to 15th).\n\nSimilarly, [U]TRUNC.S instruction, which truncates single-precision FP to\ninteger, can multiply the source value by power of two in advance, but\nneither of these currently uses this function (always specified with 0th\npower of two, i.e. a scaling factor of 1).\n\nThis patch unleashes the scaling ability of the above instructions.\n\n     /* example */\n     float test0(int a) {\n       return a / 2.f;\n     }\n     float test1(unsigned int a) {\n       return a / 32768.f;\n     }\n     int test2(float a) {\n       return a * 2;\n     }\n     unsigned int test3(float a) {\n       return a * 32768;\n     }\n\n     ;; before\n     test0:\n     \tmovi.n\ta9, 0x3f\n     \tfloat.s\tf0, a2, 0\n     \tslli\ta9, a9, 24\n     \twfr\tf1, a9\n     \tmul.s\tf0, f0, f1\n     \trfr\ta2, f0\n     \tret.n\n     test1:\n     \tmovi.n\ta9, 7\n     \tufloat.s\tf0, a2, 0\n     \tslli\ta9, a9, 27\n     \twfr\tf1, a9\n     \tmul.s\tf0, f0, f1\n     \trfr\ta2, f0\n     \tret.n\n     test2:\n     \twfr\tf1, a2\n     \tadd.s\tf0, f1, f1\n     \ttrunc.s\ta2, f0, 0\n     \tret.n\n     test3:\n     \tmovi.n\ta9, 0x47\n     \tslli\ta9, a9, 24\n     \twfr\tf1, a2\n     \twfr\tf2, a9\n     \tmul.s\tf0, f1, f2\n     \tutrunc.s\ta2, f0, 0\n     \tret.n\n\n     ;; after\n     test0:\n     \tfloat.s\tf0, a2, 1\n     \trfr\ta2, f0\n     \tret.n\n     test1:\n     \tufloat.s\tf0, a2, 15\n     \trfr\ta2, f0\n     \tret.n\n     test2:\n     \twfr\tf0, a2\n     \ttrunc.s\ta2, f0, 1\n     \tret.n\n     test3:\n     \twfr\tf0, a2\n     \tutrunc.s\ta2, f0, 15\n     \tret.n\n\ngcc/ChangeLog:\n\n\t* config/xtensa/predicates.md\n\t(fix_scaling_operand, float_scaling_operand): New predicates.\n\t* config/xtensa/xtensa.md\n\t(any_fix/m_fix/s_fix, any_float/m_float/s_float):\n\tNew code iterators and their attributes.\n\t(fix<s_fix>_truncsfsi2): Change from \"fix_truncsfsi2\".\n\t(*fix<s_fix>_truncsfsi2_2x, *fix<s_fix>_truncsfsi2_scaled):\n\tNew insn definitions.\n\t(float<s_float>sisf2): Change from \"floatsisf2\".\n\t(*float<s_float>sisf2_scaled): New insn definition.","shortMessageHtmlLink":"xtensa: Make use of scaled [U]FLOAT/TRUNC.S instructions"}},{"before":"d9ea5b0676b3c17e051b3695918f72dddd584c63","after":"a5e14a113d6fb6e7d16b409ecac7e86e5488f5b5","ref":"refs/heads/xtensa-15-updates","pushedAt":"2024-06-19T08:56:39.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"xtensa: Eliminate double MEMW insertions for volatile memory\n\nThis patch makes avoid inserting a MEMW instruction before a load/store\nnstruction with volatile memory reference if there is already a MEMW\nimmediately before it.\n\ngcc/ChangeLog:\n\n\t* config/xtensa/xtensa.cc (print_operand):\n\tWhen outputting MEMW before the instruction, check if the previous\n\tinstruction is already that.","shortMessageHtmlLink":"xtensa: Eliminate double MEMW insertions for volatile memory"}},{"before":"448482d3d5c285a0fe2e2b824797cb78f3dc5418","after":"d9ea5b0676b3c17e051b3695918f72dddd584c63","ref":"refs/heads/xtensa-15-updates","pushedAt":"2024-06-19T03:29:33.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"xtensa: constantsynth: Reforge to fix some non-fatal issues\n\nThe previous constant synthesis logic had some issues that were non-fatal\nbut worth considering:\n\n- It didn't work with DFmode literals, because those were cast to SImode\n   rather SFmode when splitting into two natural-width words by\n   split_double().\n\n- It didn't work with large literals when TARGET_AUTO_LITPOOLS was enabled,\n   because those were relaxed MOVI immediates rather references to literal\n   pool entries,\n\n- It didn't take into account that when literals with the same RTL\n   representation are pooled multiple times within a function, those entries\n   are shared (especially important when optimizing for size).\n\nThis patch addresses the above issues by making appropriate tweaks to the\nconstant synthesis logic.\n\ngcc/ChangeLog:\n\n\t* config/xtensa/xtensa-protos.h (xtensa_constantsynth):\n\tChange the second argument from HOST_WIDE_INT to rtx.\n\t* config/xtensa/xtensa.cc (#include):\n\tAdd \"context.h\" and \"pass_manager.h\".\n\t(machine_function): Add a new hash_map field \"litpool_usage\".\n\t(xtensa_constantsynth): Make \"src\" (the second operand) accept\n\tRTX literal instead of its value, and treat both bare and pooled\n\tSI/SFmode literals equally by bit-exact canonicalization into\n\tCONST_INT RTX internally.  And then, make avoid synthesis if\n\tsuch multiple identical canonicalized literals are found in same\n\tfunction when optimizing for size.  Finally, for literals where\n\tsynthesis is not possible or has been avoided, re-emit \"move\"\n\tRTXes with canonicalized ones to increase the chances of sharing\n\tliteral pool entries.\n\t* config/xtensa/xtensa.md (split patterns for constant synthesis):\n\tChange to simply invoke xtensa_constantsynth() as mentioned above,\n\tand add new patterns for when TARGET_AUTO_LITPOOLS is enabled.","shortMessageHtmlLink":"xtensa: constantsynth: Reforge to fix some non-fatal issues"}},{"before":"4fc3487aa82750f6e3fc586f8c79d7e26ef710bc","after":"448482d3d5c285a0fe2e2b824797cb78f3dc5418","ref":"refs/heads/xtensa-15-updates","pushedAt":"2024-06-17T18:00:40.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"xtensa: constantsynth: Reforge to fix some non-fatal issues\n\nThe previous constant synthesis logic had some issues that were non-fatal\nbut worth considering:\n\n- It didn't work with DFmode literals, because those were cast to SImode\n  rather SFmode when splitting into two natural-width words by\n  split_double().\n\n- It didn't work with large literals when TARGET_AUTO_LITPOOLS was enabled,\n  because those were relaxed MOVI immediates rather references to literal\n  pool entries,\n\n- It didn't take into account that when literals with the same RTL\n  representation are pooled multiple times within a function, those entries\n  are shared (especially important when optimizing for size).\n\nThis patch addresses the above issues by making appropriate tweaks to the\nconstant synthesis logic.\n\ngcc/ChangeLog:\n\n\t* config/xtensa/xtensa-protos.h (xtensa_constantsynth):\n\tChange the second argument from HOST_WIDE_INT to rtx.\n\t* config/xtensa/xtensa.cc (#include):\n\tAdd \"context.h\" and \"pass_manager.h\".\n\t(machine_function): Add a new hash_map field \"litpool_usage\".\n\t(xtensa_constantsynth): Make \"src\" (the second operand) accept\n\tRTX literal instead of its value, and treat both bare and pooled\n\tSI/SFmode literals equally by bit-exact canonicalization into\n\tCONST_INT RTX internally.  And then, make avoid synthesis if\n\tsuch multiple identical canonicalized literals are found in same\n\tfunction when optimizing for size.  Finally, for literals where\n\tsynthesis is not possible or has been avoided, re-emit \"move\"\n\tRTXes with canonicalized ones to increase the chances of sharing\n\tliteral pool entries.\n\t* config/xtensa/xtensa.md (split patterns for constant synthesis):\n\tChange to simply invoke xtensa_constantsynth() as mentioned above,\n\tand add new patterns for when TARGET_AUTO_LITPOOLS is enabled.","shortMessageHtmlLink":"xtensa: constantsynth: Reforge to fix some non-fatal issues"}},{"before":"0ca5885f069c2827ced971f980f985fa04c525c6","after":"4fc3487aa82750f6e3fc586f8c79d7e26ef710bc","ref":"refs/heads/xtensa-15-updates","pushedAt":"2024-05-31T16:44:56.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"xtensa: Prepend \"(use A0_REG)\" to sibling call CALL_INSN_FUNCTION_USAGE instead of emitting it as insn at the end of epilogue\n\nNo functional changes.\n\ngcc/ChangeLog:\n\n\t* config/xtensa/xtensa-protos.h (xtensa_expand_call):\n\tAdd the third argument as boolean.\n\t(xtensa_expand_epilogue): Remove the first argument.\n\t* config/xtensa/xtensa.cc (xtensa_expand_call):\n\tAdd the third argument \"sibcall_p\", and modify in order to prepend\n\t\"(use A0_REG)\" to CALL_INSN_FUNCTION_USAGE if the argument is true.\n\t(xtensa_expand_epilogue): Remove the first argument \"sibcall_p\" and\n\tits conditional clause.\n\t* config/xtensa/xtensa.md (call, call_value, sibcall, sibcall_value):\n\tAppend a boolean value to the argument of xtensa_expand_call()\n\tindicating whether it is sibling call or not.\n\t(epilogue): Remove the boolean argument from xtensa_expand_epilogue(),\n\tand then append emitting \"(return)\".\n\t(sibcall_epilogue): Remove the boolean argument from\n\txtensa_expand_epilogue().","shortMessageHtmlLink":"xtensa: Prepend \"(use A0_REG)\" to sibling call CALL_INSN_FUNCTION_USA…"}},{"before":null,"after":"0ca5885f069c2827ced971f980f985fa04c525c6","ref":"refs/heads/xtensa-15-updates","pushedAt":"2024-05-30T17:59:22.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"xtensa: Use epilogue_completed rather than cfun->machine->epilogue_done\n\nIn commit ad89d820bf, an \"epilogue_done\" member was added to the\nmachine_function structure, but it is sufficient to use the existing\n\"epilogue_completed\" global variable.\n\ngcc/ChangeLog:\n\n\t* config/xtensa/xtensa-protos.h\n\t(xtensa_use_return_instruction_p): Remove.\n\t* config/xtensa/xtensa.cc\n\t(machine_function): Remove \"epilogue_done\" field.\n\t(xtensa_expand_epilogue): Remove \"cfun->machine->epilogue_done\" usage.\n\t(xtensa_use_return_instruction_p): Remove.\n\t* config/xtensa/xtensa.md (\"return\"):\n\tReplace calling \"xtensa_use_return_instruction_p()\" with inline code.","shortMessageHtmlLink":"xtensa: Use epilogue_completed rather than cfun-&gt;machine-&gt;epilogue_done"}},{"before":"8885e288ddf0dbf4849709f07e73e90c53be4318","after":"ef4a1117b74266068dbc8036922bcd86e3f53a8e","ref":"refs/heads/xtensa-14.1-fdpic","pushedAt":"2024-05-28T08:34:30.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"libstdc++-v3/src/libbacktrace: add -funwind-tables\n\nlibstdc++-v3/\n\n\t* src/libbacktrace/Makefile.am (AM_CFLAGS, AM_CXXFLAGS): Add\n\t-funwind-tables\n\t* src/libbacktrace/Makefile.in: Regenerate.","shortMessageHtmlLink":"libstdc++-v3/src/libbacktrace: add -funwind-tables"}},{"before":"aa59c8c018f6fd01b4ebc97532fa91e9d68face3","after":"8885e288ddf0dbf4849709f07e73e90c53be4318","ref":"refs/heads/xtensa-14.1-fdpic","pushedAt":"2024-05-26T21:34:30.000Z","pushType":"push","commitsCount":2,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"WIP: libstdc++-v3/src/libbacktrace: add -funwind-tables","shortMessageHtmlLink":"WIP: libstdc++-v3/src/libbacktrace: add -funwind-tables"}},{"before":"175612de38be5beb85c0808e46a99743c4123cd4","after":"aa59c8c018f6fd01b4ebc97532fa91e9d68face3","ref":"refs/heads/xtensa-14.1-fdpic","pushedAt":"2024-05-19T08:11:18.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"WIP: libbacktrace: add FDPIC support","shortMessageHtmlLink":"WIP: libbacktrace: add FDPIC support"}},{"before":null,"after":"175612de38be5beb85c0808e46a99743c4123cd4","ref":"refs/heads/xtensa-14.1-fdpic","pushedAt":"2024-05-11T02:18:27.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"Revert \"libstdc++: use -D_GNU_SOURCE when building libbacktrace\"\n\nThis reverts commit 2b64e4a54042fb8f75f1c1429eb1c13afb9fa118.","shortMessageHtmlLink":"Revert \"libstdc++: use -D_GNU_SOURCE when building libbacktrace\""}},{"before":"8c58465fa6a54e9abb45f298b0e593cb37c3c983","after":"17f9c9e98fc6214fd87549b47cf2dec14f1a3bc1","ref":"refs/heads/xtensa-14-9655-fdpic","pushedAt":"2024-05-02T18:21:24.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"WIP disable target libbacktrace (compilation issue due to fdpic types) and libgomp (linking issue)\n\nSigned-off-by: Max Filippov <jcmvbkbc@gmail.com>","shortMessageHtmlLink":"WIP disable target libbacktrace (compilation issue due to fdpic types…"}},{"before":"8619cd2db4b91bcf5edc444bb366cd92df5da1eb","after":"8c58465fa6a54e9abb45f298b0e593cb37c3c983","ref":"refs/heads/xtensa-14-9655-fdpic","pushedAt":"2024-05-02T11:21:59.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"WIP disable target libbacktrace (compilation issue due to fdpic types) and libgomp (linking issue)\n\nSigned-off-by: Max Filippov <jcmvbkbc@gmail.com>","shortMessageHtmlLink":"WIP disable target libbacktrace (compilation issue due to fdpic types…"}},{"before":"7733f6c6811340b2b69e2e0071cb2bbe0d0cae79","after":"15fbea4e485e3591327a774dbb9c302de1230a12","ref":"refs/heads/xtensa-14-9655-fdpic-musl","pushedAt":"2024-05-02T07:05:45.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"WIP: xtensa*-*-{musl,uclibc}fdpic support","shortMessageHtmlLink":"WIP: xtensa*-*-{musl,uclibc}fdpic support"}},{"before":null,"after":"8619cd2db4b91bcf5edc444bb366cd92df5da1eb","ref":"refs/heads/xtensa-14-9655-fdpic","pushedAt":"2024-04-30T12:39:53.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"WIP disable target libbacktrace (compilation issue due to fdpic types) and libgomp (linking issue)\n\nSigned-off-by: Max Filippov <jcmvbkbc@gmail.com>","shortMessageHtmlLink":"WIP disable target libbacktrace (compilation issue due to fdpic types…"}},{"before":"baafbda55558beae8bb22a984f49da307ceef5e8","after":"7733f6c6811340b2b69e2e0071cb2bbe0d0cae79","ref":"refs/heads/xtensa-14-9655-fdpic-musl","pushedAt":"2024-04-30T12:36:09.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"gcc: xtensa: use GOTOFFFUNCDESC for function calls\n\nNow that the assembler allows GOTOFFFUNCDESC for dynamic symbols use it\nfor function calls. This saves one instruction and replaces\ndynamically-allocated function descriptor with the descriptor in the\ncaller's GOT.","shortMessageHtmlLink":"gcc: xtensa: use GOTOFFFUNCDESC for function calls"}},{"before":"c627fade0a75654221bbfabe6d5846749618bea6","after":"baafbda55558beae8bb22a984f49da307ceef5e8","ref":"refs/heads/xtensa-14-9655-fdpic-musl","pushedAt":"2024-04-02T02:01:20.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"gcc: xtensa: add SECREL and GOTSECBASE for local references","shortMessageHtmlLink":"gcc: xtensa: add SECREL and GOTSECBASE for local references"}},{"before":"3f3c47daf8cca47f09b66237361489e1072cca30","after":"c627fade0a75654221bbfabe6d5846749618bea6","ref":"refs/heads/xtensa-14-9655-fdpic-musl","pushedAt":"2024-04-01T07:39:13.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"gcc: xtensa: add SECREL and GOTSECBASE for local references","shortMessageHtmlLink":"gcc: xtensa: add SECREL and GOTSECBASE for local references"}},{"before":null,"after":"3f3c47daf8cca47f09b66237361489e1072cca30","ref":"refs/heads/xtensa-14-9655-fdpic-musl","pushedAt":"2024-03-28T19:22:52.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"WIP: basic compatibility for musl","shortMessageHtmlLink":"WIP: basic compatibility for musl"}},{"before":"8c24b6d40e997b7ce3a5162d264fad46b73d4baa","after":"3cede49aacd7f2b40d575fb71c56cfd40c897e14","ref":"refs/heads/xtensa-14-updates","pushedAt":"2024-03-26T15:34:39.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"xtensa: Make use of std::swap where appropriate\n\nNo functional changes.\n\ngcc/ChangeLog:\n\n\t* config/xtensa/xtensa.cc\n\t(gen_int_relational, gen_float_relational): Replace tempvar-based\n\tvalue-swapping codes with std::swap.\n\t* config/xtensa/xtensa.md (movdi_internal, movdf_internal):\n\tDitto.","shortMessageHtmlLink":"xtensa: Make use of std::swap where appropriate"}},{"before":"3e3ba6c2d9bed3c5183be59f515260a8cda3402a","after":"8c24b6d40e997b7ce3a5162d264fad46b73d4baa","ref":"refs/heads/xtensa-14-updates","pushedAt":"2024-03-23T04:39:14.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"WIP: gcc: xtensa: add movcc pattern","shortMessageHtmlLink":"WIP: gcc: xtensa: add movcc pattern"}},{"before":"f78cd44b053486c9fe38ef13b959d10d6348adb1","after":"3e3ba6c2d9bed3c5183be59f515260a8cda3402a","ref":"refs/heads/xtensa-14-updates","pushedAt":"2024-03-22T20:18:08.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"xtensa: Add supplementary split pattern for \"*addsubx\"\n\nint test(int a) {\n   return a * 4 + 30000;\n}\n\nIn the example above, since Xtensa has instructions to add register value\nscaled by 2, 4 or 8 (and corresponding define_insns), we would expect them\nto be used but not, because it is transformed before reaching the RTL\ngeneration pass as below:\n\nint test(int a) {\n   return (a + 7500) * 4;\n}\n\nFortunately, the RTL combination pass tries a splitting pattern that matches\nthe first example, so it is easy to solve by defining that pattern.\n\ngcc/ChangeLog:\n\n\t* config/xtensa/xtensa.md: Add new split pattern described above.","shortMessageHtmlLink":"xtensa: Add supplementary split pattern for \"*addsubx\""}},{"before":null,"after":"6be6cd96a08a49d83d5d9300e4fd270089b6eb5d","ref":"refs/heads/xtensa-14-9480-fdpic","pushedAt":"2024-03-15T08:20:05.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"WIP disable target libbacktrace (compilation issue due to fdpic types) and libgomp (linking issue)\n\nSigned-off-by: Max Filippov <jcmvbkbc@gmail.com>","shortMessageHtmlLink":"WIP disable target libbacktrace (compilation issue due to fdpic types…"}},{"before":"d1dca6acd6b1fd60498e98e41b8d8c52c41283e3","after":"f78cd44b053486c9fe38ef13b959d10d6348adb1","ref":"refs/heads/xtensa-14-updates","pushedAt":"2024-03-14T11:57:49.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"gcc: xtensa: reorder movsi_internal patterns for better code generation during LRA\n\nAfter switching to LRA xtensa backend generates the following code for\nsaving/loading registers:\n\n    movi     a9, 0x190\n    add      a9, a9, sp\n    s32i.n   a3, a9, 0\n\ninstead of the shorter and more efficient\n\n    s32i     a3, a9, 0x190\n\nThis happens in the LRA pass because s32i.n and l32i.n are listed before\nthe s32i and l32i in the movsi_internal pattern and alternative\nconsideration loop stops early.\n\ngcc/\n\t* config/xtensa/xtensa.md (movsi_internal): Move l32i and s32i\n\tpatterns ahead of the l32i.n and s32i.n.","shortMessageHtmlLink":"gcc: xtensa: reorder movsi_internal patterns for better code generati…"}},{"before":"0a7fbd4b72cd4db42e059a5c91122032e3548081","after":"67b0605494f32811364e25328d3522467aaf0638","ref":"refs/heads/xtensa-14-8789-fdpic","pushedAt":"2024-02-07T19:31:31.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"libgcc: libstdc++: restrict ARM-specific __FDPIC__ changes\n\nOther ports use data/text-based pointers notion instead of overloading\npc-relative pointer representation meaning.\nAlso the test in the read_encoded_value_with_base doesn't look right as\nit is satisfied not only by the DW_EH_PE_pcrel encoding, but also by the\nDW_EH_PE_datarel.","shortMessageHtmlLink":"libgcc: libstdc++: restrict ARM-specific __FDPIC__ changes"}},{"before":null,"after":"0a7fbd4b72cd4db42e059a5c91122032e3548081","ref":"refs/heads/xtensa-14-8789-fdpic","pushedAt":"2024-02-07T19:30:17.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"},"commit":{"message":"gcc: xtensa: FDPIC: fix nested function trampoline generation","shortMessageHtmlLink":"gcc: xtensa: FDPIC: fix nested function trampoline generation"}},{"before":"3ed7e1c9322c6c2a8884d198111a6264b8d82179","after":null,"ref":"refs/heads/xtensa-13-updates","pushedAt":"2024-02-05T13:37:40.000Z","pushType":"branch_deletion","commitsCount":0,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"}},{"before":"6475a1a0e05fe1a064aec71c081bc026e434f4d7","after":null,"ref":"refs/heads/xtensa-14-lra","pushedAt":"2024-02-04T10:29:25.000Z","pushType":"branch_deletion","commitsCount":0,"pusher":{"login":"jcmvbkbc","name":"Max Filippov","path":"/jcmvbkbc","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/166731?s=80&v=4"}}],"hasNextPage":true,"hasPreviousPage":false,"activityType":"all","actor":null,"timePeriod":"all","sort":"DESC","perPage":30,"cursor":"djE6ks8AAAAEjG-j9gA","startCursor":null,"endCursor":null}},"title":"Activity · jcmvbkbc/gcc-xtensa"}