about summary refs log tree commit diff
path: root/sysdeps/x86_64/multiarch/ifunc-memset.h
diff options
context:
space:
mode:
authorH.J. Lu <hjl.tools@gmail.com>2021-03-07 09:44:18 -0800
committerH.J. Lu <hjl.tools@gmail.com>2022-01-27 12:47:19 -0800
commita47bf3df2f5eec32c19e12c0ede3050837e372d6 (patch)
tree8a38c268b8fe2309ab783b5220d241dc8a88ef1c /sysdeps/x86_64/multiarch/ifunc-memset.h
parent5d5ef0cd7fcee11c793b6994723845e467f56b9e (diff)
downloadglibc-a47bf3df2f5eec32c19e12c0ede3050837e372d6.tar.gz
glibc-a47bf3df2f5eec32c19e12c0ede3050837e372d6.tar.xz
glibc-a47bf3df2f5eec32c19e12c0ede3050837e372d6.zip
x86-64: Use ZMM16-ZMM31 in AVX512 memset family functions
Update ifunc-memset.h/ifunc-wmemset.h to select the function optimized
with AVX512 instructions using ZMM16-ZMM31 registers to avoid RTM abort
with usable AVX512VL and AVX512BW since VZEROUPPER isn't needed at
function exit.

(cherry picked from commit 4e2d8f352774b56078c34648b14a2412c38384f4)
Diffstat (limited to 'sysdeps/x86_64/multiarch/ifunc-memset.h')
-rw-r--r--sysdeps/x86_64/multiarch/ifunc-memset.h13
1 files changed, 8 insertions, 5 deletions
diff --git a/sysdeps/x86_64/multiarch/ifunc-memset.h b/sysdeps/x86_64/multiarch/ifunc-memset.h
index 6fdf53ec1c..231cd75897 100644
--- a/sysdeps/x86_64/multiarch/ifunc-memset.h
+++ b/sysdeps/x86_64/multiarch/ifunc-memset.h
@@ -53,13 +53,16 @@ IFUNC_SELECTOR (void)
   if (CPU_FEATURES_ARCH_P (cpu_features, AVX512F_Usable)
       && !CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_AVX512))
     {
-      if (CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER))
-	return OPTIMIZE (avx512_no_vzeroupper);
+      if (CPU_FEATURES_ARCH_P (cpu_features, AVX512VL_Usable)
+	  && CPU_FEATURES_ARCH_P (cpu_features, AVX512BW_Usable))
+	{
+	  if (CPU_FEATURES_CPU_P (cpu_features, ERMS))
+	    return OPTIMIZE (avx512_unaligned_erms);
 
-      if (CPU_FEATURES_CPU_P (cpu_features, ERMS))
-	return OPTIMIZE (avx512_unaligned_erms);
+	  return OPTIMIZE (avx512_unaligned);
+	}
 
-      return OPTIMIZE (avx512_unaligned);
+      return OPTIMIZE (avx512_no_vzeroupper);
     }
 
   if (CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable))