|
| 1 | +2024-02-09 Jakub Jelinek < [email protected]> |
| 2 | + |
| 3 | + PR tree-optimization/113783 |
| 4 | + * gimple-lower-bitint.cc (bitint_large_huge::lower_stmt): Look |
| 5 | + through VIEW_CONVERT_EXPR for final cast checks. Handle |
| 6 | + VIEW_CONVERT_EXPRs from large/huge _BitInt to > MAX_FIXED_MODE_SIZE |
| 7 | + INTEGER_TYPEs. |
| 8 | + (gimple_lower_bitint): Don't merge mergeable operations or other |
| 9 | + casts with VIEW_CONVERT_EXPRs to > MAX_FIXED_MODE_SIZE INTEGER_TYPEs. |
| 10 | + * expr.cc (expand_expr_real_1): Don't use convert_modes if either |
| 11 | + mode is BLKmode. |
| 12 | + |
| 13 | +2024-02-09 Jakub Jelinek < [email protected]> |
| 14 | + |
| 15 | + * hwint.h (GCC_PRISZ, fmt_size_t, HOST_SIZE_T_PRINT_DEC, |
| 16 | + HOST_SIZE_T_PRINT_UNSIGNED, HOST_SIZE_T_PRINT_HEX, |
| 17 | + HOST_SIZE_T_PRINT_HEX_PURE): Define. |
| 18 | + * ira-conflicts.cc (build_conflict_bit_table): Use it. Formatting |
| 19 | + fixes. |
| 20 | + |
| 21 | +2024-02-09 Jakub Jelinek < [email protected]> |
| 22 | + |
| 23 | + PR middle-end/113415 |
| 24 | + * cfgexpand.cc (expand_asm_stmt): For asm goto, use |
| 25 | + duplicate_insn_chain to duplicate after_rtl_seq sequence instead |
| 26 | + of hand written loop with emit_insn of copy_insn and emit original |
| 27 | + after_rtl_seq on the last edge. |
| 28 | + |
| 29 | +2024-02-09 Jakub Jelinek < [email protected]> |
| 30 | + |
| 31 | + PR tree-optimization/113818 |
| 32 | + * gimple-lower-bitint.cc (add_eh_edge): New function. |
| 33 | + (bitint_large_huge::handle_load, |
| 34 | + bitint_large_huge::lower_mergeable_stmt, |
| 35 | + bitint_large_huge::lower_muldiv_stmt): Use it. |
| 36 | + |
| 37 | +2024-02-09 Jakub Jelinek < [email protected]> |
| 38 | + |
| 39 | + PR tree-optimization/113774 |
| 40 | + * gimple-lower-bitint.cc (bitint_large_huge::handle_cast): Don't |
| 41 | + emit any comparison if m_first and low + 1 is equal to |
| 42 | + m_upwards_2limb, simplify condition for that. If not |
| 43 | + single_comparison, not m_first and we can prove that the idx <= low |
| 44 | + comparison will be always true, emit instead of idx <= low |
| 45 | + comparison low <= low such that cfg cleanup will optimize it at |
| 46 | + the end of the pass. |
| 47 | + |
1 | 48 | 2024-02-08 Aldy Hernandez < [email protected]>
|
2 | 49 |
|
3 | 50 | PR tree-optimization/113735
|
|
0 commit comments