This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
[PATCH] Check if SSE is available with HAS_CPU_FEATURE
- From: "H.J. Lu" <hongjiu dot lu at intel dot com>
- To: GNU C Library <libc-alpha at sourceware dot org>
- Date: Thu, 6 Apr 2017 08:20:34 -0700
- Subject: [PATCH] Check if SSE is available with HAS_CPU_FEATURE
- Authentication-results: sourceware.org; auth=none
- Reply-to: "H.J. Lu" <hjl dot tools at gmail dot com>
Similar to other CPU feature checks, check if SSE is available with
HAS_CPU_FEATURE.
Any comments?
H.J.
---
* sysdeps/i386/fpu/fclrexcpt.c (__feclearexcept): Use
HAS_CPU_FEATURE to check for SSE.
* sysdeps/i386/fpu/fedisblxcpt.c (fedisableexcept): Likewise.
* sysdeps/i386/fpu/feenablxcpt.c (feenableexcept): Likewise.
* sysdeps/i386/fpu/fegetenv.c (__fegetenv): Likewise.
* sysdeps/i386/fpu/fegetmode.c (fegetmode): Likewise.
* sysdeps/i386/fpu/feholdexcpt.c (__feholdexcept): Likewise.
* sysdeps/i386/fpu/fesetenv.c (__fesetenv): Likewise.
* sysdeps/i386/fpu/fesetmode.c (fesetmode): Likewise.
* sysdeps/i386/fpu/fesetround.c (__fesetround): Likewise.
* sysdeps/i386/fpu/feupdateenv.c (__feupdateenv): Likewise.
* sysdeps/i386/fpu/fgetexcptflg.c (__fegetexceptflag): Likewise.
* sysdeps/i386/fpu/fsetexcptflg.c (__fesetexceptflag): Likewise.
* sysdeps/i386/fpu/ftestexcept.c (fetestexcept): Likewise.
* sysdeps/i386/setfpucw.c (__setfpucw): Likewise.
* sysdeps/x86/cpu-features.h (bit_cpu_SSE): New.
(index_cpu_SSE): Likewise.
(reg_SSE): Likewise.
---
sysdeps/i386/fpu/fclrexcpt.c | 2 +-
sysdeps/i386/fpu/fedisblxcpt.c | 2 +-
sysdeps/i386/fpu/feenablxcpt.c | 2 +-
sysdeps/i386/fpu/fegetenv.c | 2 +-
sysdeps/i386/fpu/fegetmode.c | 2 +-
sysdeps/i386/fpu/feholdexcpt.c | 2 +-
sysdeps/i386/fpu/fesetenv.c | 2 +-
sysdeps/i386/fpu/fesetmode.c | 2 +-
sysdeps/i386/fpu/fesetround.c | 2 +-
sysdeps/i386/fpu/feupdateenv.c | 2 +-
sysdeps/i386/fpu/fgetexcptflg.c | 2 +-
sysdeps/i386/fpu/fsetexcptflg.c | 2 +-
sysdeps/i386/fpu/ftestexcept.c | 2 +-
sysdeps/i386/setfpucw.c | 2 +-
sysdeps/x86/cpu-features.h | 4 ++++
15 files changed, 18 insertions(+), 14 deletions(-)
diff --git a/sysdeps/i386/fpu/fclrexcpt.c b/sysdeps/i386/fpu/fclrexcpt.c
index c89fe5b..5d85969 100644
--- a/sysdeps/i386/fpu/fclrexcpt.c
+++ b/sysdeps/i386/fpu/fclrexcpt.c
@@ -41,7 +41,7 @@ __feclearexcept (int excepts)
__asm__ ("fldenv %0" : : "m" (*&temp));
/* If the CPU supports SSE, we clear the MXCSR as well. */
- if ((GLRO(dl_hwcap) & HWCAP_I386_XMM) != 0)
+ if (HAS_CPU_FEATURE (SSE))
{
unsigned int xnew_exc;
diff --git a/sysdeps/i386/fpu/fedisblxcpt.c b/sysdeps/i386/fpu/fedisblxcpt.c
index a12fcaf..f8db665 100644
--- a/sysdeps/i386/fpu/fedisblxcpt.c
+++ b/sysdeps/i386/fpu/fedisblxcpt.c
@@ -38,7 +38,7 @@ fedisableexcept (int excepts)
__asm__ ("fldcw %0" : : "m" (*&new_exc));
/* If the CPU supports SSE we set the MXCSR as well. */
- if ((GLRO(dl_hwcap) & HWCAP_I386_XMM) != 0)
+ if (HAS_CPU_FEATURE (SSE))
{
unsigned int xnew_exc;
diff --git a/sysdeps/i386/fpu/feenablxcpt.c b/sysdeps/i386/fpu/feenablxcpt.c
index 8336c1f..f1c42d7 100644
--- a/sysdeps/i386/fpu/feenablxcpt.c
+++ b/sysdeps/i386/fpu/feenablxcpt.c
@@ -38,7 +38,7 @@ feenableexcept (int excepts)
__asm__ ("fldcw %0" : : "m" (*&new_exc));
/* If the CPU supports SSE we set the MXCSR as well. */
- if ((GLRO(dl_hwcap) & HWCAP_I386_XMM) != 0)
+ if (HAS_CPU_FEATURE (SSE))
{
unsigned int xnew_exc;
diff --git a/sysdeps/i386/fpu/fegetenv.c b/sysdeps/i386/fpu/fegetenv.c
index 9ce890e..983f6af 100644
--- a/sysdeps/i386/fpu/fegetenv.c
+++ b/sysdeps/i386/fpu/fegetenv.c
@@ -31,7 +31,7 @@ __fegetenv (fenv_t *envp)
would block all exceptions. */
__asm__ ("fldenv %0" : : "m" (*envp));
- if ((GLRO(dl_hwcap) & HWCAP_I386_XMM) != 0)
+ if (HAS_CPU_FEATURE (SSE))
__asm__ ("stmxcsr %0" : "=m" (envp->__eip));
/* Success. */
diff --git a/sysdeps/i386/fpu/fegetmode.c b/sysdeps/i386/fpu/fegetmode.c
index 3785851..abbce30 100644
--- a/sysdeps/i386/fpu/fegetmode.c
+++ b/sysdeps/i386/fpu/fegetmode.c
@@ -26,7 +26,7 @@ int
fegetmode (femode_t *modep)
{
_FPU_GETCW (modep->__control_word);
- if ((GLRO(dl_hwcap) & HWCAP_I386_XMM) != 0)
+ if (HAS_CPU_FEATURE (SSE))
__asm__ ("stmxcsr %0" : "=m" (modep->__mxcsr));
return 0;
}
diff --git a/sysdeps/i386/fpu/feholdexcpt.c b/sysdeps/i386/fpu/feholdexcpt.c
index cce883c..d327358 100644
--- a/sysdeps/i386/fpu/feholdexcpt.c
+++ b/sysdeps/i386/fpu/feholdexcpt.c
@@ -30,7 +30,7 @@ __feholdexcept (fenv_t *envp)
__asm__ volatile ("fnstenv %0; fnclex" : "=m" (*envp));
/* If the CPU supports SSE we set the MXCSR as well. */
- if ((GLRO(dl_hwcap) & HWCAP_I386_XMM) != 0)
+ if (HAS_CPU_FEATURE (SSE))
{
unsigned int xwork;
diff --git a/sysdeps/i386/fpu/fesetenv.c b/sysdeps/i386/fpu/fesetenv.c
index 18fca30..a338e5d 100644
--- a/sysdeps/i386/fpu/fesetenv.c
+++ b/sysdeps/i386/fpu/fesetenv.c
@@ -79,7 +79,7 @@ __fesetenv (const fenv_t *envp)
__asm__ ("fldenv %0" : : "m" (temp));
- if ((GLRO(dl_hwcap) & HWCAP_I386_XMM) != 0)
+ if (HAS_CPU_FEATURE (SSE))
{
unsigned int mxcsr;
__asm__ ("stmxcsr %0" : "=m" (mxcsr));
diff --git a/sysdeps/i386/fpu/fesetmode.c b/sysdeps/i386/fpu/fesetmode.c
index 2d03846..bd9f74c 100644
--- a/sysdeps/i386/fpu/fesetmode.c
+++ b/sysdeps/i386/fpu/fesetmode.c
@@ -35,7 +35,7 @@ fesetmode (const femode_t *modep)
else
cw = modep->__control_word;
_FPU_SETCW (cw);
- if ((GLRO(dl_hwcap) & HWCAP_I386_XMM) != 0)
+ if (HAS_CPU_FEATURE (SSE))
{
unsigned int mxcsr;
__asm__ ("stmxcsr %0" : "=m" (mxcsr));
diff --git a/sysdeps/i386/fpu/fesetround.c b/sysdeps/i386/fpu/fesetround.c
index fabf0b1..a3fa623 100644
--- a/sysdeps/i386/fpu/fesetround.c
+++ b/sysdeps/i386/fpu/fesetround.c
@@ -37,7 +37,7 @@ __fesetround (int round)
__asm__ ("fldcw %0" : : "m" (*&cw));
/* If the CPU supports SSE we set the MXCSR as well. */
- if ((GLRO(dl_hwcap) & HWCAP_I386_XMM) != 0)
+ if (HAS_CPU_FEATURE (SSE))
{
unsigned int xcw;
diff --git a/sysdeps/i386/fpu/feupdateenv.c b/sysdeps/i386/fpu/feupdateenv.c
index 4b46868..b610289 100644
--- a/sysdeps/i386/fpu/feupdateenv.c
+++ b/sysdeps/i386/fpu/feupdateenv.c
@@ -32,7 +32,7 @@ __feupdateenv (const fenv_t *envp)
__asm__ ("fnstsw %0" : "=m" (*&temp));
/* If the CPU supports SSE we test the MXCSR as well. */
- if ((GLRO(dl_hwcap) & HWCAP_I386_XMM) != 0)
+ if (HAS_CPU_FEATURE (SSE))
__asm__ ("stmxcsr %0" : "=m" (*&xtemp));
temp = (temp | xtemp) & FE_ALL_EXCEPT;
diff --git a/sysdeps/i386/fpu/fgetexcptflg.c b/sysdeps/i386/fpu/fgetexcptflg.c
index bf3c73c..954e5f6 100644
--- a/sysdeps/i386/fpu/fgetexcptflg.c
+++ b/sysdeps/i386/fpu/fgetexcptflg.c
@@ -34,7 +34,7 @@ __fegetexceptflag (fexcept_t *flagp, int excepts)
*flagp = temp & excepts & FE_ALL_EXCEPT;
/* If the CPU supports SSE, we clear the MXCSR as well. */
- if ((GLRO(dl_hwcap) & HWCAP_I386_XMM) != 0)
+ if (HAS_CPU_FEATURE (SSE))
{
unsigned int sse_exc;
diff --git a/sysdeps/i386/fpu/fsetexcptflg.c b/sysdeps/i386/fpu/fsetexcptflg.c
index efea610..efa64aa 100644
--- a/sysdeps/i386/fpu/fsetexcptflg.c
+++ b/sysdeps/i386/fpu/fsetexcptflg.c
@@ -41,7 +41,7 @@ __fesetexceptflag (const fexcept_t *flagp, int excepts)
__asm__ ("fldenv %0" : : "m" (*&temp));
/* If the CPU supports SSE, we set the MXCSR as well. */
- if ((GLRO(dl_hwcap) & HWCAP_I386_XMM) != 0)
+ if (HAS_CPU_FEATURE (SSE))
{
unsigned int xnew_exc;
diff --git a/sysdeps/i386/fpu/ftestexcept.c b/sysdeps/i386/fpu/ftestexcept.c
index 476db1e..f523f9e 100644
--- a/sysdeps/i386/fpu/ftestexcept.c
+++ b/sysdeps/i386/fpu/ftestexcept.c
@@ -32,7 +32,7 @@ fetestexcept (int excepts)
__asm__ ("fnstsw %0" : "=a" (temp));
/* If the CPU supports SSE we test the MXCSR as well. */
- if ((GLRO(dl_hwcap) & HWCAP_I386_XMM) != 0)
+ if (HAS_CPU_FEATURE (SSE))
__asm__ ("stmxcsr %0" : "=m" (*&xtemp));
return (temp | xtemp) & excepts & FE_ALL_EXCEPT;
diff --git a/sysdeps/i386/setfpucw.c b/sysdeps/i386/setfpucw.c
index 7c41486..40b995f 100644
--- a/sysdeps/i386/setfpucw.c
+++ b/sysdeps/i386/setfpucw.c
@@ -39,7 +39,7 @@ __setfpucw (fpu_control_t set)
__asm__ ("fldcw %0" : : "m" (*&cw));
/* If the CPU supports SSE, we set the MXCSR as well. */
- if ((GLRO(dl_hwcap) & HWCAP_I386_XMM) != 0)
+ if (HAS_CPU_FEATURE (SSE))
{
unsigned int xnew_exc;
diff --git a/sysdeps/x86/cpu-features.h b/sysdeps/x86/cpu-features.h
index 95f0fcf..8ec1562 100644
--- a/sysdeps/x86/cpu-features.h
+++ b/sysdeps/x86/cpu-features.h
@@ -45,6 +45,7 @@
/* COMMON_CPUID_INDEX_1. */
#define bit_cpu_CX8 (1 << 8)
#define bit_cpu_CMOV (1 << 15)
+#define bit_cpu_SSE (1 << 25)
#define bit_cpu_SSE2 (1 << 26)
#define bit_cpu_SSSE3 (1 << 9)
#define bit_cpu_SSE4_1 (1 << 19)
@@ -82,6 +83,7 @@
# define index_cpu_CX8 COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_EDX_OFFSET
# define index_cpu_CMOV COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_EDX_OFFSET
+# define index_cpu_SSE COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_EDX_OFFSET
# define index_cpu_SSE2 COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_EDX_OFFSET
# define index_cpu_SSSE3 COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_ECX_OFFSET
# define index_cpu_SSE4_1 COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_ECX_OFFSET
@@ -228,6 +230,7 @@ extern const struct cpu_features *__get_cpu_features (void)
# define index_cpu_CX8 COMMON_CPUID_INDEX_1
# define index_cpu_CMOV COMMON_CPUID_INDEX_1
+# define index_cpu_SSE COMMON_CPUID_INDEX_1
# define index_cpu_SSE2 COMMON_CPUID_INDEX_1
# define index_cpu_SSSE3 COMMON_CPUID_INDEX_1
# define index_cpu_SSE4_1 COMMON_CPUID_INDEX_1
@@ -246,6 +249,7 @@ extern const struct cpu_features *__get_cpu_features (void)
# define reg_CX8 edx
# define reg_CMOV edx
+# define reg_SSE edx
# define reg_SSE2 edx
# define reg_SSSE3 ecx
# define reg_SSE4_1 ecx
--
2.9.3