From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: by sourceware.org (Postfix, from userid 2153) id 8F62E3856DE9; Wed, 11 May 2022 06:24:36 +0000 (GMT) DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 8F62E3856DE9 MIME-Version: 1.0 Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset="utf-8" From: Jakub Jelinek To: gcc-cvs@gcc.gnu.org Subject: [gcc r9-10126] i386: Fix up cvtsd2ss splitter [PR104502] X-Act-Checkin: gcc X-Git-Author: Jakub Jelinek X-Git-Refname: refs/heads/releases/gcc-9 X-Git-Oldrev: c7e7ca915dc4c526b1f58a4808c7ec4ceaa30348 X-Git-Newrev: cb412e0e881adcc5440ed37a8a415a77fe3df980 Message-Id: <20220511062436.8F62E3856DE9@sourceware.org> Date: Wed, 11 May 2022 06:24:36 +0000 (GMT) X-BeenThere: gcc-cvs@gcc.gnu.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Gcc-cvs mailing list List-Unsubscribe: , List-Archive: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 11 May 2022 06:24:36 -0000 https://gcc.gnu.org/g:cb412e0e881adcc5440ed37a8a415a77fe3df980 commit r9-10126-gcb412e0e881adcc5440ed37a8a415a77fe3df980 Author: Jakub Jelinek Date: Sat Feb 12 11:17:41 2022 +0100 i386: Fix up cvtsd2ss splitter [PR104502] The following testcase ICEs, because AVX512F is enabled, AVX512VL is not, and the cvtsd2ss insn has %xmm0-15 as output operand and %xmm16-31 as input operand. For output operand %xmm16+ the splitter just gives up in such case, but for such input it just emits vmovddup which requires AVX512VL if either operand is EXT_REX_SSE_REG_P (when it is 128-bit). The following patch fixes it by treating that case like the pre-SSE3 output != input case - move the input to output and do everything on the output reg which is known to be < %xmm16. 2022-02-12 Jakub Jelinek PR target/104502 * config/i386/i386.md (cvtsd2ss splitter): If operands[1] is xmm16+ and AVX512VL isn't available, move operands[1] to operands[0] first. * gcc.target/i386/pr104502.c: New test. (cherry picked from commit 0538d42cdd68f6b65d72ed7768f1d00ba44f8631) Diff: --- gcc/config/i386/i386.md | 4 ++-- gcc/testsuite/gcc.target/i386/pr104502.c | 31 +++++++++++++++++++++++++++++++ 2 files changed, 33 insertions(+), 2 deletions(-) diff --git a/gcc/config/i386/i386.md b/gcc/config/i386/i386.md index 8144d7ba9c3..00a1db3ec0c 100644 --- a/gcc/config/i386/i386.md +++ b/gcc/config/i386/i386.md @@ -4613,8 +4613,8 @@ movddup is available. */ if (REG_P (operands[1])) { - if (!TARGET_SSE3 - && REGNO (operands[0]) != REGNO (operands[1])) + if ((!TARGET_SSE3 && REGNO (operands[0]) != REGNO (operands[1])) + || (EXT_REX_SSE_REG_P (operands[1]) && !TARGET_AVX512VL)) { rtx tmp = lowpart_subreg (DFmode, operands[0], SFmode); emit_move_insn (tmp, operands[1]); diff --git a/gcc/testsuite/gcc.target/i386/pr104502.c b/gcc/testsuite/gcc.target/i386/pr104502.c new file mode 100644 index 00000000000..7a6eb260e41 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/pr104502.c @@ -0,0 +1,31 @@ +/* PR target/104502 */ +/* { dg-do compile { target fstack_protector } } */ +/* { dg-options "-O -flive-range-shrinkage -march=barcelona -fstack-protector-all -mavx512f" } */ + +typedef char __attribute__((__vector_size__ (8))) U; +typedef int __attribute__((__vector_size__ (8))) A; +typedef int __attribute__((__vector_size__ (16))) B; +typedef int __attribute__((__vector_size__ (32))) C; +typedef int __attribute__((__vector_size__ (64))) D; +typedef __float128 __attribute__((__vector_size__ (32))) F; + +char s; +U u; +A a; +int i; +C c; +double d; + +U +foo (U u0, A a0, B b0, B b1, C c0, C c1, C c2, C c3, A a1, A a2, F f0) +{ + C ca = c |= (short) (float) d; + C cb = c0 + c1 + c2 + c3 + ca + (C) f0; + U ua = s << (u & 4); + B ba = ((union {C a; B b;}) cb).b + b0 + b1; + U ub = ((union {B a; U b;}) ba).b + + u0 + u + ua + (U) a + (U) a + (U) a0 + (U) a1 + (U) a2; + long long u64_r = i + d; + char u8_r = u64_r; + return ub + u8_r; +}