aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorNoah Goldstein <goldstein.w.n@gmail.com>2022-06-29 16:07:06 -0700
committerNoah Goldstein <goldstein.w.n@gmail.com>2022-07-05 16:42:42 -0700
commit37ecc657b22c3367a56df5ebd53908e34e65fa16 (patch)
tree6df7e476a885ee4b9e0f89229a2d7bf6a63c9af0
parentb6a02c360655cbb84772bb577fcd8ae54d7f6d82 (diff)
downloadglibc-37ecc657b22c3367a56df5ebd53908e34e65fa16.tar.xz
glibc-37ecc657b22c3367a56df5ebd53908e34e65fa16.zip
x86: Add support for building {w}memset{_chk} with explicit ISA level
1. Refactor files so that all implementations are in the multiarch directory - Moved the implementation portion of memset sse2 from memset.S to multiarch/memset-sse2.S - The non-multiarch file now only includes one of the implementations in the multiarch directory based on the compiled ISA level (only used for non-multiarch builds. Otherwise we go through the ifunc selector). 2. Add ISA level build guards to different implementations. - I.e memset-avx2-unaligned-erms.S which is ISA level 3 will only build if compiled ISA level <= 3. Otherwise there is no reason to include it as we will always use one of the ISA level 4 implementations (memset-evex-unaligned-erms.S). 3. Add new multiarch/rtld-memset.S that just include the non-multiarch memset.S which will in turn select the best implementation based on the compiled ISA level. 4. Refactor the ifunc selector and ifunc implementation list to use the ISA level aware wrapper macros that allow functions below the compiled ISA level (with a guranteed replacement) to be skipped. Tested with and without multiarch on x86_64 for ISA levels: {generic, x86-64-v2, x86-64-v3, x86-64-v4} And m32 with and without multiarch.
-rw-r--r--sysdeps/x86_64/memset.S45
-rw-r--r--sysdeps/x86_64/multiarch/ifunc-impl-list.c249
-rw-r--r--sysdeps/x86_64/multiarch/ifunc-memset.h45
-rw-r--r--sysdeps/x86_64/multiarch/ifunc-wmemset.h21
-rw-r--r--sysdeps/x86_64/multiarch/memset-avx2-unaligned-erms.S5
-rw-r--r--sysdeps/x86_64/multiarch/memset-avx512-no-vzeroupper.S4
-rw-r--r--sysdeps/x86_64/multiarch/memset-avx512-unaligned-erms.S12
-rw-r--r--sysdeps/x86_64/multiarch/memset-evex-unaligned-erms.S12
-rw-r--r--sysdeps/x86_64/multiarch/memset-sse2-unaligned-erms.S57
-rw-r--r--sysdeps/x86_64/multiarch/rtld-memset.S18
10 files changed, 265 insertions, 203 deletions
diff --git a/sysdeps/x86_64/memset.S b/sysdeps/x86_64/memset.S
index a6eea61a4d..f4e1bab601 100644
--- a/sysdeps/x86_64/memset.S
+++ b/sysdeps/x86_64/memset.S
@@ -18,47 +18,18 @@
<https://www.gnu.org/licenses/>. */
#include <sysdep.h>
-#define USE_WITH_SSE2 1
-#define VEC_SIZE 16
-#define MOV_SIZE 3
-#define RET_SIZE 1
+#define MEMSET_SYMBOL(p,s) memset
+#define MEMSET_CHK_SYMBOL(p,s) p
-#define VEC(i) xmm##i
-#define VMOVU movups
-#define VMOVA movaps
+#define WMEMSET_SYMBOL(p,s) __wmemset
+#define WMEMSET_CHK_SYMBOL(p,s) p
-# define MEMSET_SET_VEC0_AND_SET_RETURN(d, r) \
- movd d, %xmm0; \
- movq r, %rax; \
- punpcklbw %xmm0, %xmm0; \
- punpcklwd %xmm0, %xmm0; \
- pshufd $0, %xmm0, %xmm0
+#define DEFAULT_IMPL_V1 "multiarch/memset-sse2-unaligned-erms.S"
+#define DEFAULT_IMPL_V3 "multiarch/memset-avx2-unaligned-erms.S"
+#define DEFAULT_IMPL_V4 "multiarch/memset-evex-unaligned-erms.S"
-# define WMEMSET_SET_VEC0_AND_SET_RETURN(d, r) \
- movd d, %xmm0; \
- pshufd $0, %xmm0, %xmm0; \
- movq r, %rax
-
-# define MEMSET_VDUP_TO_VEC0_HIGH()
-# define MEMSET_VDUP_TO_VEC0_LOW()
-
-# define WMEMSET_VDUP_TO_VEC0_HIGH()
-# define WMEMSET_VDUP_TO_VEC0_LOW()
-
-#define SECTION(p) p
-
-#ifndef MEMSET_SYMBOL
-# define MEMSET_CHK_SYMBOL(p,s) p
-# define MEMSET_SYMBOL(p,s) memset
-#endif
-
-#ifndef WMEMSET_SYMBOL
-# define WMEMSET_CHK_SYMBOL(p,s) p
-# define WMEMSET_SYMBOL(p,s) __wmemset
-#endif
-
-#include "multiarch/memset-vec-unaligned-erms.S"
+#include "isa-default-impl.h"
libc_hidden_builtin_def (memset)
diff --git a/sysdeps/x86_64/multiarch/ifunc-impl-list.c b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
index 7858aa316f..21008c72b4 100644
--- a/sysdeps/x86_64/multiarch/ifunc-impl-list.c
+++ b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
@@ -213,94 +213,99 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
IFUNC_IMPL (i, name, __memset_chk,
IFUNC_IMPL_ADD (array, i, __memset_chk, 1,
__memset_chk_erms)
- IFUNC_IMPL_ADD (array, i, __memset_chk, 1,
- __memset_chk_sse2_unaligned)
- IFUNC_IMPL_ADD (array, i, __memset_chk, 1,
- __memset_chk_sse2_unaligned_erms)
- IFUNC_IMPL_ADD (array, i, __memset_chk,
- CPU_FEATURE_USABLE (AVX2),
- __memset_chk_avx2_unaligned)
- IFUNC_IMPL_ADD (array, i, __memset_chk,
- CPU_FEATURE_USABLE (AVX2),
- __memset_chk_avx2_unaligned_erms)
- IFUNC_IMPL_ADD (array, i, __memset_chk,
- (CPU_FEATURE_USABLE (AVX2)
- && CPU_FEATURE_USABLE (RTM)),
- __memset_chk_avx2_unaligned_rtm)
- IFUNC_IMPL_ADD (array, i, __memset_chk,
- (CPU_FEATURE_USABLE (AVX2)
- && CPU_FEATURE_USABLE (RTM)),
- __memset_chk_avx2_unaligned_erms_rtm)
- IFUNC_IMPL_ADD (array, i, __memset_chk,
- (CPU_FEATURE_USABLE (AVX512VL)
- && CPU_FEATURE_USABLE (AVX512BW)
- && CPU_FEATURE_USABLE (BMI2)),
- __memset_chk_evex_unaligned)
- IFUNC_IMPL_ADD (array, i, __memset_chk,
- (CPU_FEATURE_USABLE (AVX512VL)
- && CPU_FEATURE_USABLE (AVX512BW)
- && CPU_FEATURE_USABLE (BMI2)),
- __memset_chk_evex_unaligned_erms)
- IFUNC_IMPL_ADD (array, i, __memset_chk,
- (CPU_FEATURE_USABLE (AVX512VL)
- && CPU_FEATURE_USABLE (AVX512BW)
- && CPU_FEATURE_USABLE (BMI2)),
- __memset_chk_avx512_unaligned_erms)
- IFUNC_IMPL_ADD (array, i, __memset_chk,
- (CPU_FEATURE_USABLE (AVX512VL)
- && CPU_FEATURE_USABLE (AVX512BW)
- && CPU_FEATURE_USABLE (BMI2)),
- __memset_chk_avx512_unaligned)
- IFUNC_IMPL_ADD (array, i, __memset_chk,
- CPU_FEATURE_USABLE (AVX512F),
- __memset_chk_avx512_no_vzeroupper)
+ X86_IFUNC_IMPL_ADD_V4 (array, i, __memset_chk,
+ (CPU_FEATURE_USABLE (AVX512VL)
+ && CPU_FEATURE_USABLE (AVX512BW)
+ && CPU_FEATURE_USABLE (BMI2)),
+ __memset_chk_avx512_unaligned_erms)
+ X86_IFUNC_IMPL_ADD_V4 (array, i, __memset_chk,
+ (CPU_FEATURE_USABLE (AVX512VL)
+ && CPU_FEATURE_USABLE (AVX512BW)
+ && CPU_FEATURE_USABLE (BMI2)),
+ __memset_chk_avx512_unaligned)
+ X86_IFUNC_IMPL_ADD_V4 (array, i, __memset_chk,
+ CPU_FEATURE_USABLE (AVX512F),
+ __memset_chk_avx512_no_vzeroupper)
+ X86_IFUNC_IMPL_ADD_V4 (array, i, __memset_chk,
+ (CPU_FEATURE_USABLE (AVX512VL)
+ && CPU_FEATURE_USABLE (AVX512BW)
+ && CPU_FEATURE_USABLE (BMI2)),
+ __memset_chk_evex_unaligned)
+ X86_IFUNC_IMPL_ADD_V4 (array, i, __memset_chk,
+ (CPU_FEATURE_USABLE (AVX512VL)
+ && CPU_FEATURE_USABLE (AVX512BW)
+ && CPU_FEATURE_USABLE (BMI2)),
+ __memset_chk_evex_unaligned_erms)
+ X86_IFUNC_IMPL_ADD_V3 (array, i, __memset_chk,
+ CPU_FEATURE_USABLE (AVX2),
+ __memset_chk_avx2_unaligned)
+ X86_IFUNC_IMPL_ADD_V3 (array, i, __memset_chk,
+ CPU_FEATURE_USABLE (AVX2),
+ __memset_chk_avx2_unaligned_erms)
+ X86_IFUNC_IMPL_ADD_V3 (array, i, __memset_chk,
+ (CPU_FEATURE_USABLE (AVX2)
+ && CPU_FEATURE_USABLE (RTM)),
+ __memset_chk_avx2_unaligned_rtm)
+ X86_IFUNC_IMPL_ADD_V3 (array, i, __memset_chk,
+ (CPU_FEATURE_USABLE (AVX2)
+ && CPU_FEATURE_USABLE (RTM)),
+ __memset_chk_avx2_unaligned_erms_rtm)
+ /* ISA V2 wrapper for SSE2 implementation because the SSE2
+ implementation is also used at ISA level 2. */
+ X86_IFUNC_IMPL_ADD_V2 (array, i, __memset_chk, 1,
+ __memset_chk_sse2_unaligned)
+ X86_IFUNC_IMPL_ADD_V2 (array, i, __memset_chk, 1,
+ __memset_chk_sse2_unaligned_erms)
)
#endif
/* Support sysdeps/x86_64/multiarch/memset.c. */
IFUNC_IMPL (i, name, memset,
IFUNC_IMPL_ADD (array, i, memset, 1,
- __memset_sse2_unaligned)
- IFUNC_IMPL_ADD (array, i, memset, 1,
- __memset_sse2_unaligned_erms)
- IFUNC_IMPL_ADD (array, i, memset, 1, __memset_erms)
- IFUNC_IMPL_ADD (array, i, memset,
- CPU_FEATURE_USABLE (AVX2),
- __memset_avx2_unaligned)
- IFUNC_IMPL_ADD (array, i, memset,
- CPU_FEATURE_USABLE (AVX2),
- __memset_avx2_unaligned_erms)
- IFUNC_IMPL_ADD (array, i, memset,
- (CPU_FEATURE_USABLE (AVX2)
- && CPU_FEATURE_USABLE (RTM)),
- __memset_avx2_unaligned_rtm)
- IFUNC_IMPL_ADD (array, i, memset,
- (CPU_FEATURE_USABLE (AVX2)
- && CPU_FEATURE_USABLE (RTM)),
- __memset_avx2_unaligned_erms_rtm)
- IFUNC_IMPL_ADD (array, i, memset,
- (CPU_FEATURE_USABLE (AVX512VL)
- && CPU_FEATURE_USABLE (AVX512BW)
- && CPU_FEATURE_USABLE (BMI2)),
- __memset_evex_unaligned)
- IFUNC_IMPL_ADD (array, i, memset,
- (CPU_FEATURE_USABLE (AVX512VL)
- && CPU_FEATURE_USABLE (AVX512BW)
- && CPU_FEATURE_USABLE (BMI2)),
- __memset_evex_unaligned_erms)
- IFUNC_IMPL_ADD (array, i, memset,
- (CPU_FEATURE_USABLE (AVX512VL)
- && CPU_FEATURE_USABLE (AVX512BW)
- && CPU_FEATURE_USABLE (BMI2)),
- __memset_avx512_unaligned_erms)
- IFUNC_IMPL_ADD (array, i, memset,
- (CPU_FEATURE_USABLE (AVX512VL)
- && CPU_FEATURE_USABLE (AVX512BW)
- && CPU_FEATURE_USABLE (BMI2)),
- __memset_avx512_unaligned)
- IFUNC_IMPL_ADD (array, i, memset,
- CPU_FEATURE_USABLE (AVX512F),
- __memset_avx512_no_vzeroupper)
+ __memset_erms)
+ X86_IFUNC_IMPL_ADD_V4 (array, i, memset,
+ (CPU_FEATURE_USABLE (AVX512VL)
+ && CPU_FEATURE_USABLE (AVX512BW)
+ && CPU_FEATURE_USABLE (BMI2)),
+ __memset_avx512_unaligned_erms)
+ X86_IFUNC_IMPL_ADD_V4 (array, i, memset,
+ (CPU_FEATURE_USABLE (AVX512VL)
+ && CPU_FEATURE_USABLE (AVX512BW)
+ && CPU_FEATURE_USABLE (BMI2)),
+ __memset_avx512_unaligned)
+ X86_IFUNC_IMPL_ADD_V4 (array, i, memset,
+ CPU_FEATURE_USABLE (AVX512F),
+ __memset_avx512_no_vzeroupper)
+ X86_IFUNC_IMPL_ADD_V4 (array, i, memset,
+ (CPU_FEATURE_USABLE (AVX512VL)
+ && CPU_FEATURE_USABLE (AVX512BW)
+ && CPU_FEATURE_USABLE (BMI2)),
+ __memset_evex_unaligned)
+ X86_IFUNC_IMPL_ADD_V4 (array, i, memset,
+ (CPU_FEATURE_USABLE (AVX512VL)
+ && CPU_FEATURE_USABLE (AVX512BW)
+ && CPU_FEATURE_USABLE (BMI2)),
+ __memset_evex_unaligned_erms)
+ X86_IFUNC_IMPL_ADD_V3 (array, i, memset,
+ CPU_FEATURE_USABLE (AVX2),
+ __memset_avx2_unaligned)
+ X86_IFUNC_IMPL_ADD_V3 (array, i, memset,
+ CPU_FEATURE_USABLE (AVX2),
+ __memset_avx2_unaligned_erms)
+ X86_IFUNC_IMPL_ADD_V3 (array, i, memset,
+ (CPU_FEATURE_USABLE (AVX2)
+ && CPU_FEATURE_USABLE (RTM)),
+ __memset_avx2_unaligned_rtm)
+ X86_IFUNC_IMPL_ADD_V3 (array, i, memset,
+ (CPU_FEATURE_USABLE (AVX2)
+ && CPU_FEATURE_USABLE (RTM)),
+ __memset_avx2_unaligned_erms_rtm)
+ /* ISA V2 wrapper for SSE2 implementation because the SSE2
+ implementation is also used at ISA level 2. */
+ X86_IFUNC_IMPL_ADD_V2 (array, i, memset, 1,
+ __memset_sse2_unaligned)
+ X86_IFUNC_IMPL_ADD_V2 (array, i, memset, 1,
+ __memset_sse2_unaligned_erms)
)
/* Support sysdeps/x86_64/multiarch/rawmemchr.c. */
@@ -821,25 +826,27 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
/* Support sysdeps/x86_64/multiarch/wmemset.c. */
IFUNC_IMPL (i, name, wmemset,
- IFUNC_IMPL_ADD (array, i, wmemset, 1,
- __wmemset_sse2_unaligned)
- IFUNC_IMPL_ADD (array, i, wmemset,
- CPU_FEATURE_USABLE (AVX2),
- __wmemset_avx2_unaligned)
- IFUNC_IMPL_ADD (array, i, wmemset,
- (CPU_FEATURE_USABLE (AVX2)
- && CPU_FEATURE_USABLE (RTM)),
- __wmemset_avx2_unaligned_rtm)
- IFUNC_IMPL_ADD (array, i, wmemset,
- (CPU_FEATURE_USABLE (AVX512VL)
- && CPU_FEATURE_USABLE (AVX512BW)
- && CPU_FEATURE_USABLE (BMI2)),
- __wmemset_evex_unaligned)
- IFUNC_IMPL_ADD (array, i, wmemset,
- (CPU_FEATURE_USABLE (AVX512VL)
- && CPU_FEATURE_USABLE (AVX512BW)
- && CPU_FEATURE_USABLE (BMI2)),
- __wmemset_avx512_unaligned))
+ X86_IFUNC_IMPL_ADD_V4 (array, i, wmemset,
+ (CPU_FEATURE_USABLE (AVX512VL)
+ && CPU_FEATURE_USABLE (AVX512BW)
+ && CPU_FEATURE_USABLE (BMI2)),
+ __wmemset_evex_unaligned)
+ X86_IFUNC_IMPL_ADD_V4 (array, i, wmemset,
+ (CPU_FEATURE_USABLE (AVX512VL)
+ && CPU_FEATURE_USABLE (AVX512BW)
+ && CPU_FEATURE_USABLE (BMI2)),
+ __wmemset_avx512_unaligned)
+ X86_IFUNC_IMPL_ADD_V3 (array, i, wmemset,
+ CPU_FEATURE_USABLE (AVX2),
+ __wmemset_avx2_unaligned)
+ X86_IFUNC_IMPL_ADD_V3 (array, i, wmemset,
+ (CPU_FEATURE_USABLE (AVX2)
+ && CPU_FEATURE_USABLE (RTM)),
+ __wmemset_avx2_unaligned_rtm)
+ /* ISA V2 wrapper for SSE2 implementation because the SSE2
+ implementation is also used at ISA level 2. */
+ X86_IFUNC_IMPL_ADD_V2 (array, i, wmemset, 1,
+ __wmemset_sse2_unaligned))
#ifdef SHARED
/* Support sysdeps/x86_64/multiarch/memcpy_chk.c. */
@@ -1049,25 +1056,27 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
#ifdef SHARED
/* Support sysdeps/x86_64/multiarch/wmemset_chk.c. */
IFUNC_IMPL (i, name, __wmemset_chk,
- IFUNC_IMPL_ADD (array, i, __wmemset_chk, 1,
- __wmemset_chk_sse2_unaligned)
- IFUNC_IMPL_ADD (array, i, __wmemset_chk,
- CPU_FEATURE_USABLE (AVX2),
- __wmemset_chk_avx2_unaligned)
- IFUNC_IMPL_ADD (array, i, __wmemset_chk,
- (CPU_FEATURE_USABLE (AVX2)
- && CPU_FEATURE_USABLE (RTM)),
- __wmemset_chk_avx2_unaligned_rtm)
- IFUNC_IMPL_ADD (array, i, __wmemset_chk,
- (CPU_FEATURE_USABLE (AVX512VL)
- && CPU_FEATURE_USABLE (AVX512BW)
- && CPU_FEATURE_USABLE (BMI2)),
- __wmemset_chk_evex_unaligned)
- IFUNC_IMPL_ADD (array, i, __wmemset_chk,
- (CPU_FEATURE_USABLE (AVX512VL)
- && CPU_FEATURE_USABLE (AVX512BW)
- && CPU_FEATURE_USABLE (BMI2)),
- __wmemset_chk_avx512_unaligned))
+ X86_IFUNC_IMPL_ADD_V4 (array, i, __wmemset_chk,
+ (CPU_FEATURE_USABLE (AVX512VL)
+ && CPU_FEATURE_USABLE (AVX512BW)
+ && CPU_FEATURE_USABLE (BMI2)),
+ __wmemset_chk_evex_unaligned)
+ X86_IFUNC_IMPL_ADD_V4 (array, i, __wmemset_chk,
+ (CPU_FEATURE_USABLE (AVX512VL)
+ && CPU_FEATURE_USABLE (AVX512BW)
+ && CPU_FEATURE_USABLE (BMI2)),
+ __wmemset_chk_avx512_unaligned)
+ X86_IFUNC_IMPL_ADD_V3 (array, i, __wmemset_chk,
+ CPU_FEATURE_USABLE (AVX2),
+ __wmemset_chk_avx2_unaligned)
+ X86_IFUNC_IMPL_ADD_V3 (array, i, __wmemset_chk,
+ (CPU_FEATURE_USABLE (AVX2)
+ && CPU_FEATURE_USABLE (RTM)),
+ __wmemset_chk_avx2_unaligned_rtm)
+ /* ISA V2 wrapper for SSE2 implementation because the SSE2
+ implementation is also used at ISA level 2. */
+ X86_IFUNC_IMPL_ADD_V2 (array, i, __wmemset_chk, 1,
+ __wmemset_chk_sse2_unaligned))
#endif
return 0;
diff --git a/sysdeps/x86_64/multiarch/ifunc-memset.h b/sysdeps/x86_64/multiarch/ifunc-memset.h
index 64d179913c..ed514976aa 100644
--- a/sysdeps/x86_64/multiarch/ifunc-memset.h
+++ b/sysdeps/x86_64/multiarch/ifunc-memset.h
@@ -20,10 +20,19 @@
#include <init-arch.h>
extern __typeof (REDIRECT_NAME) OPTIMIZE (erms) attribute_hidden;
-extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2_unaligned)
+
+extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_unaligned)
attribute_hidden;
-extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2_unaligned_erms)
+extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_unaligned_erms)
+ attribute_hidden;
+extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_no_vzeroupper)
+ attribute_hidden;
+
+extern __typeof (REDIRECT_NAME) OPTIMIZE (evex_unaligned)
+ attribute_hidden;
+extern __typeof (REDIRECT_NAME) OPTIMIZE (evex_unaligned_erms)
attribute_hidden;
+
extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2_unaligned) attribute_hidden;
extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2_unaligned_erms)
attribute_hidden;
@@ -31,31 +40,26 @@ extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2_unaligned_rtm)
attribute_hidden;
extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2_unaligned_erms_rtm)
attribute_hidden;
-extern __typeof (REDIRECT_NAME) OPTIMIZE (evex_unaligned)
- attribute_hidden;
-extern __typeof (REDIRECT_NAME) OPTIMIZE (evex_unaligned_erms)
- attribute_hidden;
-extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_unaligned)
- attribute_hidden;
-extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_unaligned_erms)
+
+extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2_unaligned)
attribute_hidden;
-extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_no_vzeroupper)
+extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2_unaligned_erms)
attribute_hidden;
static inline void *
IFUNC_SELECTOR (void)
{
- const struct cpu_features* cpu_features = __get_cpu_features ();
+ const struct cpu_features *cpu_features = __get_cpu_features ();
if (CPU_FEATURES_ARCH_P (cpu_features, Prefer_ERMS))
return OPTIMIZE (erms);
- if (CPU_FEATURE_USABLE_P (cpu_features, AVX512F)
+ if (X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, AVX512F)
&& !CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_AVX512))
{
- if (CPU_FEATURE_USABLE_P (cpu_features, AVX512VL)
- && CPU_FEATURE_USABLE_P (cpu_features, AVX512BW)
- && CPU_FEATURE_USABLE_P (cpu_features, BMI2))
+ if (X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, AVX512VL)
+ && X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, AVX512BW)
+ && X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, BMI2))
{
if (CPU_FEATURE_USABLE_P (cpu_features, ERMS))
return OPTIMIZE (avx512_unaligned_erms);
@@ -66,11 +70,11 @@ IFUNC_SELECTOR (void)
return OPTIMIZE (avx512_no_vzeroupper);
}
- if (CPU_FEATURE_USABLE_P (cpu_features, AVX2))
+ if (X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, AVX2))
{
- if (CPU_FEATURE_USABLE_P (cpu_features, AVX512VL)
- && CPU_FEATURE_USABLE_P (cpu_features, AVX512BW)
- && CPU_FEATURE_USABLE_P (cpu_features, BMI2))
+ if (X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, AVX512VL)
+ && X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, AVX512BW)
+ && X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, BMI2))
{
if (CPU_FEATURE_USABLE_P (cpu_features, ERMS))
return OPTIMIZE (evex_unaligned_erms);
@@ -86,7 +90,8 @@ IFUNC_SELECTOR (void)
return OPTIMIZE (avx2_unaligned_rtm);
}
- if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER))
+ if (X86_ISA_CPU_FEATURES_ARCH_P (cpu_features,
+ Prefer_No_VZEROUPPER, !))
{
if (CPU_FEATURE_USABLE_P (cpu_features, ERMS))
return OPTIMIZE (avx2_unaligned_erms);
diff --git a/sysdeps/x86_64/multiarch/ifunc-wmemset.h b/sysdeps/x86_64/multiarch/ifunc-wmemset.h
index 87c48e2387..3810c719c6 100644
--- a/sysdeps/x86_64/multiarch/ifunc-wmemset.h
+++ b/sysdeps/x86_64/multiarch/ifunc-wmemset.h
@@ -18,22 +18,26 @@
#include <init-arch.h>
-extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2_unaligned) attribute_hidden;
+extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_unaligned) attribute_hidden;
+
+extern __typeof (REDIRECT_NAME) OPTIMIZE (evex_unaligned) attribute_hidden;
+
extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2_unaligned) attribute_hidden;
extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2_unaligned_rtm)
attribute_hidden;
-extern __typeof (REDIRECT_NAME) OPTIMIZE (evex_unaligned) attribute_hidden;
-extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_unaligned) attribute_hidden;
+
+extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2_unaligned) attribute_hidden;
static inline void *
IFUNC_SELECTOR (void)
{
- const struct cpu_features* cpu_features = __get_cpu_features ();
+ const struct cpu_features *cpu_features = __get_cpu_features ();
- if (CPU_FEATURE_USABLE_P (cpu_features, AVX2)
- && CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load))
+ if (X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, AVX2)
+ && X86_ISA_CPU_FEATURES_ARCH_P (cpu_features,
+ AVX_Fast_Unaligned_Load, !))
{
- if (CPU_FEATURE_USABLE_P (cpu_features, AVX512VL))
+ if (X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, AVX512VL))
{
if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_AVX512))
return OPTIMIZE (avx512_unaligned);
@@ -44,7 +48,8 @@ IFUNC_SELECTOR (void)
if (CPU_FEATURE_USABLE_P (cpu_features, RTM))
return OPTIMIZE (avx2_unaligned_rtm);
- if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER))
+ if (X86_ISA_CPU_FEATURES_ARCH_P (cpu_features,
+ Prefer_No_VZEROUPPER, !))
return OPTIMIZE (avx2_unaligned);
}
diff --git a/sysdeps/x86_64/multiarch/memset-avx2-unaligned-erms.S b/sysdeps/x86_64/multiarch/memset-avx2-unaligned-erms.S
index c0bf2875d0..a9054a9122 100644
--- a/sysdeps/x86_64/multiarch/memset-avx2-unaligned-erms.S
+++ b/sysdeps/x86_64/multiarch/memset-avx2-unaligned-erms.S
@@ -1,4 +1,7 @@
-#if IS_IN (libc)
+#include <isa-level.h>
+
+#if ISA_SHOULD_BUILD (3)
+
# define USE_WITH_AVX2 1
# define VEC_SIZE 32
diff --git a/sysdeps/x86_64/multiarch/memset-avx512-no-vzeroupper.S b/sysdeps/x86_64/multiarch/memset-avx512-no-vzeroupper.S
index c5be8f57ef..8cc9c16d73 100644
--- a/sysdeps/x86_64/multiarch/memset-avx512-no-vzeroupper.S
+++ b/sysdeps/x86_64/multiarch/memset-avx512-no-vzeroupper.S
@@ -17,8 +17,10 @@
<https://www.gnu.org/licenses/>. */
#include <sysdep.h>
+#include <isa-level.h>
+
+#if ISA_SHOULD_BUILD (4)
-#if IS_IN (libc)
#include "asm-syntax.h"
#ifndef MEMSET
diff --git a/sysdeps/x86_64/multiarch/memset-avx512-unaligned-erms.S b/sysdeps/x86_64/multiarch/memset-avx512-unaligned-erms.S
index 5241216a77..47623b8ee8 100644
--- a/sysdeps/x86_64/multiarch/memset-avx512-unaligned-erms.S
+++ b/sysdeps/x86_64/multiarch/memset-avx512-unaligned-erms.S
@@ -1,4 +1,7 @@
-#if IS_IN (libc)
+#include <isa-level.h>
+
+#if ISA_SHOULD_BUILD (4)
+
# define USE_WITH_AVX512 1
# define VEC_SIZE 64
@@ -30,8 +33,15 @@
# define WMEMSET_VDUP_TO_VEC0_LOW()
# define SECTION(p) p##.evex512