Hi Uros: This patch extend pass rpad to handle AVX512F vcvtusi2ss/vcvtusi2sd. 538.image_r would be improved by 4% with single copy run on skylake workstation.
Bootstrap ok. regression test for i386/x86 backend ok. Ok for trunk? Changelog gcc/ * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512): Add avx_partial_xmm_update. gcc/testsuie * gcc.target/i386/pr87007-3.c: New test. -- BR, Hongtao
From 6c759b61c6fd317627791ac7e773465b0b644641 Mon Sep 17 00:00:00 2001 From: liuhongt <hongtao....@intel.com> Date: Thu, 5 Sep 2019 14:00:13 +0800 Subject: [PATCH] Extend pass rpad to handle avx512f vcvtusi2ss vcvtusi2ss 538.imagick_r improved by 4% with single copy run on SKYLAKE workstation. gcc/ * config/i386/i386.md ("*floatuns<SWI48:mode><MODEF:mode>2_avx512"): Add avx_partial_xmm_update. gcc/testsuie * gcc.target/i386/pr87007-3.c: New test. --- gcc/config/i386/i386.md | 1 + gcc/testsuite/gcc.target/i386/pr87007-3.c | 18 ++++++++++++++++++ 2 files changed, 19 insertions(+) create mode 100644 gcc/testsuite/gcc.target/i386/pr87007-3.c diff --git a/gcc/config/i386/i386.md b/gcc/config/i386/i386.md index 7ad97882419..b7e7d126da2 100644 --- a/gcc/config/i386/i386.md +++ b/gcc/config/i386/i386.md @@ -5196,6 +5196,7 @@ "TARGET_AVX512F && TARGET_SSE_MATH" "vcvtusi2<MODEF:ssemodesuffix><SWI48:rex64suffix>\t{%1, %0, %0|%0, %0, %1}" [(set_attr "type" "sseicvt") + (set_attr "avx_partial_xmm_update" "true") (set_attr "prefix" "evex") (set_attr "mode" "<MODEF:MODE>")]) diff --git a/gcc/testsuite/gcc.target/i386/pr87007-3.c b/gcc/testsuite/gcc.target/i386/pr87007-3.c new file mode 100644 index 00000000000..59324fd1a45 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/pr87007-3.c @@ -0,0 +1,18 @@ +/* { dg-do compile } */ +/* { dg-options "-O2 -march=skylake-avx512 -mfpmath=sse" } */ + +extern float f; +extern double d; +extern unsigned char c; + +void +foo (int n, int k) +{ + for (int i = 0; i != n; i++) + if(i < k) + d = c; + else + f = c; +} + +/* { dg-final { scan-assembler-times "vxorps\[^\n\r\]*xmm\[0-9\]" 1 } } */ -- 2.19.1