2012-05-11 Andreas Jaeger Carlos O'Donell [BZ #14059] * sysdeps/x86_64/multiarch/init-arch.h (bit_AVX_Usable): New macro. (bit_FMA4_Usable): New macro. (bit_XMM_state): New macro. (bit_YMM_state): New macro. [__ASSEMBLER__] (index_YMM_Usable): Rename to... [__ASSEMBLER__] (index_AVX_Usable): ... this. [__ASSEMBLER__] (index_FMA4_Usable): New macro. (CPUID_OSXSAVE): New macro. (CPUID_AVX): New macro. (CPUID_FMA4): New macro. (index_YMM_Usable): Rename to... (index_AVX_Usable): ... this. (HAS_AVX): Use HAS_ARCH_FEATURE. (HAS_FMA4): Likewise. (HAS_YMM_USABLE): Remove. * sysdeps/x86_64/multiarch/init-arch.c (__init_cpu_features): Fix check for AVX, enable FMA4 only if it exists and if AVX is usable. * sysdeps/x86_64/multiarch/strcmp.S: Use bit_AVX_Usable. diff --git a/sysdeps/x86_64/multiarch/init-arch.c b/sysdeps/x86_64/multiarch/init-arch.c index 80527ec..e68e5fa 100644 --- a/sysdeps/x86_64/multiarch/init-arch.c +++ b/sysdeps/x86_64/multiarch/init-arch.c @@ -1,6 +1,6 @@ /* Initialize CPU feature data. This file is part of the GNU C Library. - Copyright (C) 2008, 2009, 2010, 2011, 2012 Free Software Foundation, Inc. + Copyright (C) 2008-2012 Free Software Foundation, Inc. Contributed by Ulrich Drepper . The GNU C Library is free software; you can redistribute it and/or @@ -143,18 +143,23 @@ __init_cpu_features (void) else kind = arch_kind_other; - if (__cpu_features.cpuid[COMMON_CPUID_INDEX_1].ecx & bit_AVX) + if (CPUID_AVX) { - /* Reset the AVX bit in case OSXSAVE is disabled. */ - if ((__cpu_features.cpuid[COMMON_CPUID_INDEX_1].ecx & bit_OSXSAVE) != 0 + /* Determine if AVX is usable. */ + if (CPUID_OSXSAVE && ({ unsigned int xcrlow; unsigned int xcrhigh; asm ("xgetbv" : "=a" (xcrlow), "=d" (xcrhigh) : "c" (0)); - (xcrlow & 6) == 6; })) - __cpu_features.feature[index_YMM_Usable] |= bit_YMM_Usable; + (xcrlow & (bit_YMM_state | bit_XMM_state)) == + (bit_YMM_state | bit_XMM_state); })) + __cpu_features.feature[index_AVX_Usable] = 1; } + /* FMA4 depends on AVX support. */ + if (CPUID_FMA4 && HAS_AVX) + __cpu_features.feature[index_FMA4_Usable] = 1; + __cpu_features.family = family; __cpu_features.model = model; atomic_write_barrier (); diff --git a/sysdeps/x86_64/multiarch/init-arch.h b/sysdeps/x86_64/multiarch/init-arch.h index 5054e46..e6a10bb 100644 --- a/sysdeps/x86_64/multiarch/init-arch.h +++ b/sysdeps/x86_64/multiarch/init-arch.h @@ -1,5 +1,5 @@ /* This file is part of the GNU C Library. - Copyright (C) 2008, 2009, 2010, 2011, 2012 Free Software Foundation, Inc. + Copyright (C) 2008-2012 Free Software Foundation, Inc. The GNU C Library is free software; you can redistribute it and/or modify it under the terms of the GNU Lesser General Public @@ -21,8 +21,10 @@ #define bit_Prefer_SSE_for_memop (1 << 3) #define bit_Fast_Unaligned_Load (1 << 4) #define bit_Prefer_PMINUB_for_stringop (1 << 5) -#define bit_YMM_Usable (1 << 6) +#define bit_AVX_Usable (1 << 6) +#define bit_FMA4_Usable (1 << 7) +/* CPUID Feature flags. */ #define bit_SSE2 (1 << 26) #define bit_SSSE3 (1 << 9) #define bit_SSE4_1 (1 << 19) @@ -33,6 +35,10 @@ #define bit_FMA (1 << 12) #define bit_FMA4 (1 << 16) +/* XCR0 Feature flags. */ +#define bit_XMM_state (1 << 1) +#define bit_YMM_state (2 << 1) + #ifdef __ASSEMBLER__ # include @@ -49,7 +55,8 @@ # define index_Prefer_SSE_for_memop FEATURE_INDEX_1*FEATURE_SIZE # define index_Fast_Unaligned_Load FEATURE_INDEX_1*FEATURE_SIZE # define index_Prefer_PMINUB_for_stringop FEATURE_INDEX_1*FEATURE_SIZE -# define index_YMM_Usable FEATURE_INDEX_1*FEATURE_SIZE +# define index_AVX_Usable FEATURE_INDEX_1*FEATURE_SIZE +# define index_FMA4_Usable FEATURE_INDEX_1*FEATURE_SIZE #else /* __ASSEMBLER__ */ @@ -113,35 +120,36 @@ extern const struct cpu_features *__get_cpu_features (void) /* Following are the feature tests used throughout libc. */ +/* CPUID_* evaluates to true if the feature flag is enabled. */ +# define CPUID_OSXSAVE HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, bit_OSXSAVE) +# define CPUID_AVX HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, bit_AVX) +# define CPUID_FMA4 HAS_CPU_FEATURE (COMMON_CPUID_INDEX_80000001, ecx, bit_FMA4) + +/* HAS_* evaluates to true if we may use the feature at runtime. */ # define HAS_SSE2 HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, edx, bit_SSE2) # define HAS_POPCOUNT HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, bit_POPCOUNT) # define HAS_SSSE3 HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, bit_SSSE3) # define HAS_SSE4_1 HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, bit_SSE4_1) # define HAS_SSE4_2 HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, bit_SSE4_2) # define HAS_FMA HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, bit_FMA) -# define HAS_AVX HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, bit_AVX) -# define HAS_FMA4 HAS_CPU_FEATURE (COMMON_CPUID_INDEX_80000001, ecx, bit_FMA4) # define index_Fast_Rep_String FEATURE_INDEX_1 # define index_Fast_Copy_Backward FEATURE_INDEX_1 # define index_Slow_BSF FEATURE_INDEX_1 # define index_Prefer_SSE_for_memop FEATURE_INDEX_1 # define index_Fast_Unaligned_Load FEATURE_INDEX_1 -# define index_YMM_Usable FEATURE_INDEX_1 +# define index_AVX_Usable FEATURE_INDEX_1 +# define index_FMA4_Usable FEATURE_INDEX_1 # define HAS_ARCH_FEATURE(name) \ ((__get_cpu_features ()->feature[index_##name] & (bit_##name)) != 0) -# define HAS_FAST_REP_STRING HAS_ARCH_FEATURE (Fast_Rep_String) - -# define HAS_FAST_COPY_BACKWARD HAS_ARCH_FEATURE (Fast_Copy_Backward) - -# define HAS_SLOW_BSF HAS_ARCH_FEATURE (Slow_BSF) - -# define HAS_PREFER_SSE_FOR_MEMOP HAS_ARCH_FEATURE (Prefer_SSE_for_memop) - -# define HAS_FAST_UNALIGNED_LOAD HAS_ARCH_FEATURE (Fast_Unaligned_Load) - -# define HAS_YMM_USABLE HAS_ARCH_FEATURE (YMM_Usable) +# define HAS_FAST_REP_STRING HAS_ARCH_FEATURE (Fast_Rep_String) +# define HAS_FAST_COPY_BACKWARD HAS_ARCH_FEATURE (Fast_Copy_Backward) +# define HAS_SLOW_BSF HAS_ARCH_FEATURE (Slow_BSF) +# define HAS_PREFER_SSE_FOR_MEMOP HAS_ARCH_FEATURE (Prefer_SSE_for_memop) +# define HAS_FAST_UNALIGNED_LOAD HAS_ARCH_FEATURE (Fast_Unaligned_Load) +# define HAS_AVX HAS_ARCH_FEATURE (AVX_Usable) +# define HAS_FMA4 HAS_ARCH_FEATURE (FMA4_Usable) #endif /* __ASSEMBLER__ */ diff --git a/sysdeps/x86_64/multiarch/strcmp.S b/sysdeps/x86_64/multiarch/strcmp.S index 2b9870b..d76e55b 100644 --- a/sysdeps/x86_64/multiarch/strcmp.S +++ b/sysdeps/x86_64/multiarch/strcmp.S @@ -1,5 +1,5 @@ /* strcmp with SSE4.2 - Copyright (C) 2009, 2010, 2011 Free Software Foundation, Inc. + Copyright (C) 2009-2012 Free Software Foundation, Inc. Contributed by Intel Corporation. This file is part of the GNU C Library. @@ -106,7 +106,7 @@ ENTRY(__strcasecmp) 1: # ifdef HAVE_AVX_SUPPORT leaq __strcasecmp_avx(%rip), %rax - testl $bit_AVX, __cpu_features+CPUID_OFFSET+index_AVX(%rip) + testl $bit_AVX_Usable, __cpu_features+FEATURE_OFFSET+index_AVX_Usable(%rip) jnz 2f # endif leaq __strcasecmp_sse42(%rip), %rax @@ -129,7 +129,7 @@ ENTRY(__strncasecmp) 1: # ifdef HAVE_AVX_SUPPORT leaq __strncasecmp_avx(%rip), %rax - testl $bit_AVX, __cpu_features+CPUID_OFFSET+index_AVX(%rip) + testl $bit_AVX_Usable, __cpu_features+FEATURE_OFFSET+index_AVX_Usable(%rip) jnz 2f # endif leaq __strncasecmp_sse42(%rip), %rax