x86/cpuid: Add AVX512_4VNNIW and AVX512_4FMAPS support
PATCH v2 1/1 x86/cpuid: Add AVX512_4VNNIW and AVX512_4FMAPS support
Add two new AVX512 subfeatures support for guest.
AVX512_4VNNIW:
Vector instructions for deep learning enhanced word variable precision.
AVX512_4FMAPS:
Vector instructions for deep learning floating-point single precision.
---
Changes from v1:
Add new leaf in xen-cpuid.c
--- tools/libxc/xc_cpuid_x86.c | 8 ++++++-- tools/misc/xen-cpuid.c | 10 ++++++++++ xen/arch/x86/cpu/common.c | 2 +- xen/arch/x86/cpuid.c | 2 +- xen/arch/x86/hvm/hvm.c | 1 + xen/arch/x86/traps.c | 5 +++-- xen/include/asm-x86/cpuid.h | 1 + xen/include/public/arch-x86/cpufeatureset.h | 4 ++++ xen/tools/gen-cpuid.py | 2 +- 9 files changed, 28 insertions(+), 7 deletions(-)
tools/libxc/xc_cpuid_x86.c +6 -2
@@ -547,13 +547,15 @@ static void xc_cpuid_hvm_policy(xc_interface *xch, { regs[1 ] = info->featureset[featureword_of (X86_FEATURE_FSGSBASE )]; regs[2 ] = info->featureset[featureword_of (X86_FEATURE_PREFETCHWT1 )]; + regs[3 ] = info->featureset[featureword_of (X86_FEATURE_AVX512_4VNNIW )]; } else { regs[1 ] =0 ; regs[2 ] =0 ; + regs[3 ] =0 ; } - regs[0 ] = regs[3 ] =0 ; + regs[0 ] =0 ;break ;case 0x0000000d :
@@ -638,13 +640,15 @@ static void xc_cpuid_pv_policy(xc_interface *xch, { regs[1 ] = info->featureset[featureword_of (X86_FEATURE_FSGSBASE )]; regs[2 ] = info->featureset[featureword_of (X86_FEATURE_PREFETCHWT1 )]; + regs[3 ] = info->featureset[featureword_of (X86_FEATURE_AVX512_4VNNIW )]; } else { regs[1 ] =0 ; regs[2 ] =0 ; + regs[3 ] =0 ; } - regs[0 ] = regs[3 ] =0 ; + regs[0 ] =0 ;break ;case 0x0000000d :
tools/misc/xen-cpuid.c +10 -0
@@ -143,6 +143,15 @@ static const char *str_e8b[32] = [1 ... 31] = "REZ", }; +static const char *str_7d0[32 ] = +{ + [0 ...1 ] ="REZ" , + + [2 ] ="avx512_4vnniw" , [3 ] ="avx512_4fmaps" , + + [4 ...31 ] ="REZ" , +}; + static struct { const char *name; const char *abbr;
@@ -158,6 +167,7 @@ static struct { { "0x00000007:0.ecx", "7c0", str_7c0 }, { "0x80000007.edx", "e7d", str_e7d }, { "0x80000008.ebx", "e8b", str_e8b }, + { "0x00000007:0.edx", "7d0", str_7d0 }, };#define COL_ALIGN "18"
xen/arch/x86/cpu/common.c +1 -1
@@ -325,7 +325,7 @@ static void generic_identify(struct cpuinfo_x86 *c)cpuid_count (0x00000007 ,0 , &tmp, &c->x86_capability[cpufeat_word (X86_FEATURE_FSGSBASE )], &c->x86_capability[cpufeat_word (X86_FEATURE_PKU )], - &tmp); + &c->x86_capability[cpufeat_word (X86_FEATURE_AVX512_4VNNIW )]); }/*
xen/arch/x86/cpuid.c +1 -1
@@ -78,7 +78,7 @@ static void __init calculate_raw_featureset(void)cpuid_count (0x7 ,0 , &tmp, &raw_featureset[FEATURESET_7b0], &raw_featureset[FEATURESET_7c0], - &tmp); + &raw_featureset[FEATURESET_7d0]);if ( max >=0xd ) cpuid_count(0xd, 1, &raw_featureset[FEATURESET_Da1],
xen/arch/x86/hvm/hvm.c +1 -0
@@ -3503,6 +3503,7 @@ void hvm_cpuid(unsigned int input, unsigned int *eax, unsigned int *ebx, special_features[FEATURESET_7b0]); *ecx &= hvm_featureset[FEATURESET_7c0]; + *edx &= hvm_featureset[FEATURESET_7d0];/* Don't expose HAP-only features to non-hap guests. */ if ( !hap_enabled (d) )
xen/arch/x86/traps.c +3 -2
@@ -1133,6 +1133,7 @@ void pv_cpuid(struct cpu_user_regs *regs) special_features[FEATURESET_7b0]); c &= pv_featureset[FEATURESET_7c0]; + d &= pv_featureset[FEATURESET_7d0];if ( !is_pvh_domain (currd) ) {
@@ -1147,8 +1148,8 @@ void pv_cpuid(struct cpu_user_regs *regs) } }else - b = c =0 ; - a = d =0 ; + b = c = d =0 ; + a =0 ;break ;case XSTATE_CPUID :
xen/include/asm-x86/cpuid.h +1 -0
@@ -17,6 +17,7 @@#define FEATURESET_7c0 6/* 0x00000007:0.ecx */ #define FEATURESET_e7d 7/* 0x80000007.edx */ #define FEATURESET_e8b 8/* 0x80000008.ebx */ +#define FEATURESET_7d0 9/* 0x00000007:0.edx */ #ifndef __ASSEMBLY__#include <xen/types.h>
xen/include/public/arch-x86/cpufeatureset.h +4 -0
@@ -234,6 +234,10 @@ XEN_CPUFEATURE(EFRO, 7*32+10) /* APERF/MPERF Read Only interface *//* AMD-defined CPU features, CPUID level 0x80000008.ebx, word 8 */ XEN_CPUFEATURE (CLZERO ,8 *32 +0 )/*A CLZERO instruction */ +/* Intel-defined CPU features, CPUID level 0x00000007:0.edx, word 9 */ +XEN_CPUFEATURE(AVX512_4VNNIW, 9*32+ 2) /*A AVX512 Neural Network Instructions */ +XEN_CPUFEATURE(AVX512_4FMAPS, 9*32+ 3) /*A AVX512 Multiply Accumulation Single Precision */ +#endif /* XEN_CPUFEATURE */ /* Clean up from a default include. Close the enum (for C). */
xen/tools/gen-cpuid.py +1 -1
@@ -253,7 +253,7 @@ def crunch_numbers(state): # 512bit registers, and the instructions themselves. All further AVX512 features # are built on top of AVX512F AVX512F: [AVX512DQ, AVX512IFMA, AVX512PF, AVX512ER, AVX512CD, - AVX512BW, AVX512VL, AVX512VBMI], + AVX512BW, AVX512VL, AVX512VBMI, AVX512_4VNNIW, AVX512_4FMAPS], } deep_features = tuple(sorted(deps.keys()))
Signature · 7 lines
2.7.4 _______________________________________________ Xen-devel mailing list [email protected] https://lists.xen.org/xen-devel
Quoted 11 lines
Quoted 1 line
Quoted 1 line · On 21.11.16 at 07:01, <[email protected]> wrote:
Add two new AVX512 subfeatures support for guest.
AVX512_4VNNIW:
Vector instructions for deep learning enhanced word variable precision.
AVX512_4FMAPS:
Vector instructions for deep learning floating-point single precision.Signed-off-by Luwei Kang <[email protected]>Signed-off-by He Chen <[email protected]>
As before, hypervisor parts
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xen.org/xen-devel
Quoted 10 lines · He Chen
Add two new AVX512 subfeatures support for guest.
AVX512_4VNNIW:
Vector instructions for deep learning enhanced word variable precision.
AVX512_4FMAPS:
Vector instructions for deep learning floating-point single precision.Signed-off-by Luwei Kang <[email protected]>Signed-off-by He Chen <[email protected]>
Thanks. Reviewed-by: Andrew Cooper <[email protected]>
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xen.org/xen-devel
Quoted 10 lines · He Chen
Add two new AVX512 subfeatures support for guest.
AVX512_4VNNIW:
Vector instructions for deep learning enhanced word variable precision.
AVX512_4FMAPS:
Vector instructions for deep learning floating-point single precision.Signed-off-by Luwei Kang <[email protected]>Signed-off-by He Chen <[email protected]>
Quoted 8 lines
---
Changes from v1:
Add new leaf inxen-cpuid.c--- tools/libxc/xc_cpuid_x86.c | 8 ++++++-- tools/misc/xen-cpuid.c | 10 ++++++++++
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xen.org/xen-devel
Quoted 13 lines · Wei Liu
Quoted 12 lines · He Chen
Add two new AVX512 subfeatures support for guest.
AVX512_4VNNIW:
Vector instructions for deep learning enhanced word variable precision.
AVX512_4FMAPS:
Vector instructions for deep learning floating-point single precision.Signed-off-by Luwei Kang <[email protected]>Signed-off-by He Chen <[email protected]>---
Quoted 7 lines
Quoted 6 lines
Changes from v1:
Add new leaf inxen-cpuid.c--- tools/libxc/xc_cpuid_x86.c | 8 ++++++-- tools/misc/xen-cpuid.c | 10 ++++++++++Acked-by Wei Liu <[email protected]>
Ok - queued in my for-4.9 branch
~Andrew
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xen.org/xen-devel
Quoted 14 lines · Andrew Cooper
Quoted 13 lines · Wei Liu
Quoted 12 lines · He Chen
Add two new AVX512 subfeatures support for guest.
AVX512_4VNNIW:
Vector instructions for deep learning enhanced word variable precision.
AVX512_4FMAPS:
Vector instructions for deep learning floating-point single precision.Signed-off-by Luwei Kang <[email protected]>Signed-off-by He Chen <[email protected]>---
Quoted 8 lines
Quoted 7 lines
Quoted 6 lines
Changes from v1:
Add new leaf inxen-cpuid.c--- tools/libxc/xc_cpuid_x86.c | 8 ++++++-- tools/misc/xen-cpuid.c | 10 ++++++++++Acked-by Wei Liu <[email protected]>Ok - queued in my for-4.9 branch
Committed, now that 4.9 is open.
~Andrew
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xen.org/xen-devel