ssa_operands: Speed up GIMPLE_SWITCH handling
GCC's SSA operand scanner now processes only the index operand of GIMPLE_SWITCH statements, improving performance.
The operands_scanner::parse_ssa_operands function now only processes the index operand of GIMPLE_SWITCH statements, skipping the case labels. Since the case labels are skipped anyway, this change reduces the amount of work performed by the operand scanner, leading to faster compilation times. This change is also presented as a preliminary step in modifying GIMPLE_SWITCH slightly.
In Details
This commit optimizes the operands_scanner::parse_ssa_operands function in tree-ssa-operands.cc by only processing the index operand of a GIMPLE_SWITCH statement. The cases within the switch statement are represented as CASE_LABEL_EXPR nodes, which are already skipped by the get_expr_operands method. By avoiding redundant scanning of these case labels, the commit improves the performance of the SSA operand scanner. This is a localized optimization that is also a step towards future changes to the representation of GIMPLE_SWITCH.
For Context
GCC's internal representation of code, GIMPLE, uses Single Static Assignment (SSA) form during optimization. The ssa_operands pass analyzes the operands of GIMPLE statements to identify optimization opportunities. GIMPLE_SWITCH statements are used to implement switch statements in C/C++. This commit optimizes the scanning of GIMPLE_SWITCH operands by skipping the case labels, as they are already handled elsewhere. This optimization improves the performance of the SSA analysis, potentially leading to faster compilation times.