On 10/8/24 12:25 AM, Tsung Chun Lin wrote:
0001-RISC-V-Enable-builtin-__riscv_mul-with-Zmmul-extensi.patch
From d5b254e19d1f37fe27c7e98a0160e5c22446cfea Mon Sep 17 00:00:00 2001
From: Jim Lin<j...@andestech.com>
Date: Tue, 8 Oct 2024 13:14:32 +0800
Subject: [PATCH] RISC-V: Enable builtin __riscv_mul with Zmmul extension.
gcc/ChangeLog:
* config/riscv/riscv-c.cc:
Enable builtin __riscv_mul with Zmmul extension.
Tsung Chun,
Both of your patches caused various regressions in the GCC testsuite.
As an example on riscv32-elf:
Tests that now fail, but worked before (49 tests):
unix/-march=rv32gcv: gcc: gcc.target/riscv/predef-14.c -O0 (test for excess
errors)
unix/-march=rv32gcv: gcc: gcc.target/riscv/predef-14.c -O1 (test for excess
errors)
unix/-march=rv32gcv: gcc: gcc.target/riscv/predef-14.c -O2 (test for excess
errors)
unix/-march=rv32gcv: gcc: gcc.target/riscv/predef-14.c -O2 -flto
-fno-use-linker-plugin -flto-partition=none (test for excess errors)
unix/-march=rv32gcv: gcc: gcc.target/riscv/predef-14.c -O2 -flto
-fuse-linker-plugin -fno-fat-lto-objects (test for excess errors)
unix/-march=rv32gcv: gcc: gcc.target/riscv/predef-14.c -O3 -g (test for
excess errors)
unix/-march=rv32gcv: gcc: gcc.target/riscv/predef-14.c -Os (test for excess
errors)
[ More similar failures follow. ]
I should have asked if you had run the regression testsuite to be sure
your patches weren't going to regress something unexpectedly. GIven all
the regressions where in the risc-v specific subdirectory, you can use
make check-gcc RUNTESTFLAGS="riscv.exp"
To run just the risc-v specific tests, which will be much faster than
running the full testsuite. This shortcut isn't usually appropriate,
but I think it is for the two patches you've submitted. There is a web
page with further information on the testsuite, but the basic idea would
be to build the compiler without your patch, run the testsuite, save the
resulting .sum files, then apply your patch and run the testsuite again
and compare the resulting .sum files. Ideally there should be no
differences.
Anyway, I'm going to be reverting both patches. Can you please
re-submit them after fixing the testsuite failures?
Also note that if you use the tag "[RISC-V]" in your subject line your
patch will be automatically picked up by a pre-commit tester that can be
subsequently examined to verify behavior.
Thanks,
Jeff
