On 7/29/2026 4:10 AM, Liqin Weng wrote:
From: wengliqin <[email protected]>
gcc/ChangeLog:
* config.gcc: Add riscv-fold-base-idx.cc to riscv extra_objs.
* config/riscv/bitmanip.md: Update sh1add/sh2add/sh3add/add.uw patterns.
* config/riscv/riscv-passes.def: Add fold-base-idx pass.
* config/riscv/riscv-protos.h: Add fold_base_idx_pass prototype.
* config/riscv/riscv.cc: Add base_idx_add symbol.
* config/riscv/riscv.md: Update add patterns.
* config/riscv/t-riscv: Add riscv-fold-base-idx.cc to build rules.
* config/riscv/riscv-fold-base-idx.cc: Implement fold-base-idx
optimization pass for RISC-V to optimize global array accesses.
gcc/testsuite/ChangeLog:
* gcc.target/riscv/fold-base-idx-array-1.c: New test.
* gcc.target/riscv/fold-base-idx-array-2.c: New test.
* gcc.target/riscv/fold-base-idx-array-3.c: New test.
* gcc.target/riscv/fold-base-idx-array-4.c: New test.
* gcc.target/riscv/fold-base-idx-array-5.c: New test.
* gcc.target/riscv/fold-base-idx-array-6.c: New test.
* gcc.target/riscv/fold-base-idx-array-7.c: New test.
* gcc.target/riscv/fold-base-idx-strcut-1.c: New test.
* gcc.target/riscv/fold-base-idx-strcut-2.c: New test.
As I noted before, I don't see any justification for using an UNSPEC in
here. UNSPECS are to be used when we can not reasonably describe the
semantics of the instruction in RTL.
I really don't want to see a new pass for this. I know you indicated it
was hard, but I'd like to see more details about why it was hard to make
this work in the existing passes.
Given this depends on significantly new assembler behavior, the patterns
will need to be conditional on testing the assembler's behavior.
Otherwise if someone were to use a new compiler with an older assembler
they'll get errors.
Is the new syntax for these instruction variants documented anywhere?
jeff
