[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Xen-devel] [PATCH v10 01/13] x86emul: support of AVX512* population count insns


  • To: "xen-devel@xxxxxxxxxxxxxxxxxxxx" <xen-devel@xxxxxxxxxxxxxxxxxxxx>
  • From: Jan Beulich <JBeulich@xxxxxxxx>
  • Date: Wed, 17 Jul 2019 06:33:26 +0000
  • Accept-language: en-US
  • Arc-authentication-results: i=1; mx.microsoft.com 1;spf=pass smtp.mailfrom=suse.com;dmarc=pass action=none header.from=suse.com;dkim=pass header.d=suse.com;arc=none
  • Arc-message-signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com; s=arcselector9901; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=i1ge1hDOirKPtMClNsSc5vMAWE5oA+VYTkV6mW+gJJs=; b=GXeuqfh5y5UXspcQuR3HvvUZOlOwLCbXRL76LaXsY2yBth1R7M3HJwwtFTnAz9POsWzt0LZQZemBci2aafDMwK9HcvjvsjEhBE/uce1fPZYzDYCG/t/HBZZMDdof39yMGIICPtnipNlMNdjRhmItXiSaYUTmk9fq/YI904CrNXXSAY5uJ775R+39WZaA0/gf6uqF3CZ76L4soB7bpn40/MlSDPTODUnAURyBkMWiUUbsM3nijZV1PkZrvnAbY7xSyEEew11AOQenkXfzGBEYAFC6FiBgYTJfx3QMMDqC/AepqbQiCf4PbbQ9gpms9r9z5iYXbKyT1BCMicbEwpjmDw==
  • Arc-seal: i=1; a=rsa-sha256; s=arcselector9901; d=microsoft.com; cv=none; b=bgg4glMIodBH7f9mfF0uDeM3ap1AReyMeCTYdr2ljS2avfgTjbN1EOZOSqjiC5BErTsMn7j0H39BN7q24HH6XEtEuySJZbd8fRS8vMQI82R8JeljmKL3NB/mQSg0i5K5sGfKO0OeWdbtqFn9EI1axj0HqxIVBNxnRcjPjJKMUN/Tz3Bppj7GhU5TnZmT3ZNwUoIKNjuJVCmNeWhY2frHOmaLD6MTDNfR/ZzpbKL2a1YtIVGm1GupPtZ/bhEheygGzsLIDZmPRPpw7x2yg/cRLQVv9d6EKTMHjtYD5XpGJyvtW3lXBEzADvuvPsAwP2dq7qm82t2z/tRojxExi2Kvbg==
  • Authentication-results: spf=none (sender IP is ) smtp.mailfrom=JBeulich@xxxxxxxx;
  • Cc: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>, Wei Liu <wl@xxxxxxx>, RogerPau Monne <roger.pau@xxxxxxxxxx>
  • Delivery-date: Wed, 17 Jul 2019 06:33:45 +0000
  • List-id: Xen developer discussion <xen-devel.lists.xenproject.org>
  • Thread-index: AQHVPGmKcM68tOeumkmsB2+Q+F1E2A==
  • Thread-topic: [PATCH v10 01/13] x86emul: support of AVX512* population count insns

Plus the only other AVX512_BITALG one.

As in a few cases before, since the insns here and in particular their
memory access patterns follow the usual scheme, I didn't think it was
necessary to add a contrived test specifically for them, beyond the
Disp8 scaling one.

Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
---
v10: Sort AVX512BW deps by number instead of alphabetically.
v9: Re-base.
v7: Re-base.
v6: New.

--- a/tools/tests/x86_emulator/evex-disp8.c
+++ b/tools/tests/x86_emulator/evex-disp8.c
@@ -538,6 +538,11 @@ static const struct test avx512pf_512[]
      INSNX(scatterpf1q, 66, 0f38, c7, 6, vl, sd, el),
  };
  
+static const struct test avx512_bitalg_all[] = {
+    INSN(popcnt,      66, 0f38, 54, vl, bw, vl),
+    INSN(pshufbitqmb, 66, 0f38, 8f, vl,  b, vl),
+};
+
  static const struct test avx512_vbmi_all[] = {
      INSN(permb,         66, 0f38, 8d, vl, b, vl),
      INSN(permi2b,       66, 0f38, 75, vl, b, vl),
@@ -550,6 +555,10 @@ static const struct test avx512_vbmi2_al
      INSN(pexpand,   66, 0f38, 62, vl, bw, el),
  };
  
+static const struct test avx512_vpopcntdq_all[] = {
+    INSN(popcnt, 66, 0f38, 55, vl, dq, vl)
+};
+
  static const unsigned char vl_all[] = { VL_512, VL_128, VL_256 };
  static const unsigned char vl_128[] = { VL_128 };
  static const unsigned char vl_no128[] = { VL_512, VL_256 };
@@ -919,6 +928,8 @@ void evex_disp8_test(void *instr, struct
      RUN(avx512er, 512);
  #define cpu_has_avx512pf cpu_has_avx512f
      RUN(avx512pf, 512);
+    RUN(avx512_bitalg, all);
      RUN(avx512_vbmi, all);
      RUN(avx512_vbmi2, all);
+    RUN(avx512_vpopcntdq, all);
  }
--- a/tools/tests/x86_emulator/x86-emulate.h
+++ b/tools/tests/x86_emulator/x86-emulate.h
@@ -143,6 +143,8 @@ static inline bool xcr0_mask(uint64_t ma
  #define cpu_has_avx512vl  (cp.feat.avx512vl && xcr0_mask(0xe6))
  #define cpu_has_avx512_vbmi (cp.feat.avx512_vbmi && xcr0_mask(0xe6))
  #define cpu_has_avx512_vbmi2 (cp.feat.avx512_vbmi2 && xcr0_mask(0xe6))
+#define cpu_has_avx512_bitalg (cp.feat.avx512_bitalg && xcr0_mask(0xe6))
+#define cpu_has_avx512_vpopcntdq (cp.feat.avx512_vpopcntdq && xcr0_mask(0xe6))
  
  #define cpu_has_xgetbv1   (cpu_has_xsave && cp.xstate.xgetbv1)
  
--- a/xen/arch/x86/x86_emulate/x86_emulate.c
+++ b/xen/arch/x86/x86_emulate/x86_emulate.c
@@ -479,6 +479,7 @@ static const struct ext0f38_table {
      [0x4d] = { .simd_size = simd_scalar_vexw, .d8s = d8s_dq },
      [0x4e] = { .simd_size = simd_packed_fp, .two_op = 1, .d8s = d8s_vl },
      [0x4f] = { .simd_size = simd_scalar_vexw, .d8s = d8s_dq },
+    [0x54 ... 0x55] = { .simd_size = simd_packed_int, .two_op = 1, .d8s = 
d8s_vl },
      [0x58] = { .simd_size = simd_other, .two_op = 1, .d8s = 2 },
      [0x59] = { .simd_size = simd_other, .two_op = 1, .d8s = 3 },
      [0x5a] = { .simd_size = simd_128, .two_op = 1, .d8s = 4 },
@@ -501,6 +502,7 @@ static const struct ext0f38_table {
      [0x8c] = { .simd_size = simd_packed_int },
      [0x8d] = { .simd_size = simd_packed_int, .d8s = d8s_vl },
      [0x8e] = { .simd_size = simd_packed_int, .to_mem = 1 },
+    [0x8f] = { .simd_size = simd_packed_int, .d8s = d8s_vl },
      [0x90 ... 0x93] = { .simd_size = simd_other, .vsib = 1, .d8s = d8s_dq },
      [0x96 ... 0x98] = { .simd_size = simd_packed_fp, .d8s = d8s_vl },
      [0x99] = { .simd_size = simd_scalar_vexw, .d8s = d8s_dq },
@@ -1883,6 +1885,8 @@ in_protmode(
  #define vcpu_has_avx512vl()    (ctxt->cpuid->feat.avx512vl)
  #define vcpu_has_avx512_vbmi() (ctxt->cpuid->feat.avx512_vbmi)
  #define vcpu_has_avx512_vbmi2() (ctxt->cpuid->feat.avx512_vbmi2)
+#define vcpu_has_avx512_bitalg() (ctxt->cpuid->feat.avx512_bitalg)
+#define vcpu_has_avx512_vpopcntdq() (ctxt->cpuid->feat.avx512_vpopcntdq)
  #define vcpu_has_rdpid()       (ctxt->cpuid->feat.rdpid)
  
  #define vcpu_must_have(feat) \
@@ -8899,6 +8903,19 @@ x86_emulate(
          generate_exception_if(vex.l, EXC_UD);
          goto simd_0f_avx;
  
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x8f): /* vpshufbitqmb 
[xyz]mm/mem,[xyz]mm,k{k} */
+        generate_exception_if(evex.w || !evex.r || !evex.R || evex.z, EXC_UD);
+        /* fall through */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x54): /* vpopcnt{b,w} 
[xyz]mm/mem,[xyz]mm{k} */
+        host_and_vcpu_must_have(avx512_bitalg);
+        generate_exception_if(evex.brs, EXC_UD);
+        elem_bytes = 1 << evex.w;
+        goto avx512f_no_sae;
+
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x55): /* vpopcnt{d,q} 
[xyz]mm/mem,[xyz]mm{k} */
+        host_and_vcpu_must_have(avx512_vpopcntdq);
+        goto avx512f_no_sae;
+
      case X86EMUL_OPC_VEX_66(0x0f38, 0x58): /* vpbroadcastd xmm/m32,{x,y}mm */
      case X86EMUL_OPC_VEX_66(0x0f38, 0x59): /* vpbroadcastq xmm/m64,{x,y}mm */
      case X86EMUL_OPC_VEX_66(0x0f38, 0x78): /* vpbroadcastb xmm/m8,{x,y}mm */
--- a/xen/include/asm-x86/cpufeature.h
+++ b/xen/include/asm-x86/cpufeature.h
@@ -110,6 +110,8 @@
  /* CPUID level 0x00000007:0.ecx */
  #define cpu_has_avx512_vbmi     boot_cpu_has(X86_FEATURE_AVX512_VBMI)
  #define cpu_has_avx512_vbmi2    boot_cpu_has(X86_FEATURE_AVX512_VBMI2)
+#define cpu_has_avx512_bitalg   boot_cpu_has(X86_FEATURE_AVX512_BITALG)
+#define cpu_has_avx512_vpopcntdq boot_cpu_has(X86_FEATURE_AVX512_VPOPCNTDQ)
  #define cpu_has_rdpid           boot_cpu_has(X86_FEATURE_RDPID)
  
  /* CPUID level 0x80000007.edx */
--- a/xen/include/public/arch-x86/cpufeatureset.h
+++ b/xen/include/public/arch-x86/cpufeatureset.h
@@ -229,6 +229,7 @@ XEN_CPUFEATURE(UMIP,          6*32+ 2) /
  XEN_CPUFEATURE(PKU,           6*32+ 3) /*H  Protection Keys for Userspace */
  XEN_CPUFEATURE(OSPKE,         6*32+ 4) /*!  OS Protection Keys Enable */
  XEN_CPUFEATURE(AVX512_VBMI2,  6*32+ 6) /*A  Additional AVX-512 Vector Byte 
Manipulation Instrs */
+XEN_CPUFEATURE(AVX512_BITALG, 6*32+12) /*A  Support for VPOPCNT[B,W] and 
VPSHUFBITQMB */
  XEN_CPUFEATURE(AVX512_VPOPCNTDQ, 6*32+14) /*A  POPCNT for vectors of DW/QW */
  XEN_CPUFEATURE(RDPID,         6*32+22) /*A  RDPID instruction */
  
--- a/xen/tools/gen-cpuid.py
+++ b/xen/tools/gen-cpuid.py
@@ -268,7 +268,7 @@ def crunch_numbers(state):
          # AVX512 extensions acting on vectors of bytes/words are made
          # dependents of AVX512BW (as to requiring wider than 16-bit mask
          # registers), despite the SDM not formally making this connection.
-        AVX512BW: [AVX512_BF16, AVX512_VBMI, AVX512_VBMI2],
+        AVX512BW: [AVX512_VBMI, AVX512_VBMI2, AVX512_BITALG, AVX512_BF16],
  
          # The features:
          #   * Single Thread Indirect Branch Predictors

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.