[v10,07/13] x86emul: support VPCLMULQDQ insns
diff mbox series

Message ID 3c414662-98a0-7e58-fbc4-6a49133f68c6@suse.com
State New, archived
Headers show
Series
  • x86emul: remaining AVX512 support
Related show

Commit Message

Jan Beulich July 17, 2019, 6:35 a.m. UTC
As to the feature dependency adjustment, while strictly speaking AVX is
a sufficient prereq (to have YMM registers), 256-bit vectors of integers
have got fully introduced with AVX2 only. Sadly gcc can't be used as a
reference here: They don't provide any AVX512-independent built-in at
all.

Along the lines of PCLMULQDQ, 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@suse.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
---
v10: Re-base.
v9: Re-base. Make VPCLMULQDQ also depend on PCLMULQDQ.
v8: No need to set fault_suppression to false.
v7: New.

Patch
diff mbox series

--- a/tools/tests/x86_emulator/evex-disp8.c
+++ b/tools/tests/x86_emulator/evex-disp8.c
@@ -591,6 +591,10 @@  static const struct test avx512_vpopcntd
      INSN(popcnt, 66, 0f38, 55, vl, dq, vl)
  };
  
+static const struct test vpclmulqdq_all[] = {
+    INSN(pclmulqdq, 66, 0f3a, 44, vl, q_nb, 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 };
@@ -968,4 +972,9 @@  void evex_disp8_test(void *instr, struct
      RUN(avx512_vbmi2, all);
      RUN(avx512_vnni, all);
      RUN(avx512_vpopcntdq, all);
+
+    if ( cpu_has_avx512f )
+    {
+        RUN(vpclmulqdq, all);
+    }
  }
--- a/tools/tests/x86_emulator/x86-emulate.h
+++ b/tools/tests/x86_emulator/x86-emulate.h
@@ -144,6 +144,7 @@  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_vpclmulqdq (cp.feat.vpclmulqdq && xcr0_mask(6))
  #define cpu_has_avx512_vnni (cp.feat.avx512_vnni && 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))
--- a/xen/arch/x86/x86_emulate/x86_emulate.c
+++ b/xen/arch/x86/x86_emulate/x86_emulate.c
@@ -594,7 +594,7 @@  static const struct ext0f3a_table {
      [0x3e ... 0x3f] = { .simd_size = simd_packed_int, .d8s = d8s_vl },
      [0x40 ... 0x41] = { .simd_size = simd_packed_fp },
      [0x42 ... 0x43] = { .simd_size = simd_packed_int, .d8s = d8s_vl },
-    [0x44] = { .simd_size = simd_packed_int },
+    [0x44] = { .simd_size = simd_packed_int, .d8s = d8s_vl },
      [0x46] = { .simd_size = simd_packed_int },
      [0x48 ... 0x49] = { .simd_size = simd_packed_fp, .four_op = 1 },
      [0x4a ... 0x4b] = { .simd_size = simd_packed_fp, .four_op = 1 },
@@ -1890,6 +1890,7 @@  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_vpclmulqdq()  (ctxt->cpuid->feat.vpclmulqdq)
  #define vcpu_has_avx512_vnni() (ctxt->cpuid->feat.avx512_vnni)
  #define vcpu_has_avx512_bitalg() (ctxt->cpuid->feat.avx512_bitalg)
  #define vcpu_has_avx512_vpopcntdq() (ctxt->cpuid->feat.avx512_vpopcntdq)
@@ -10207,13 +10208,19 @@  x86_emulate(
          goto opmask_shift_imm;
  
      case X86EMUL_OPC_66(0x0f3a, 0x44):     /* pclmulqdq $imm8,xmm/m128,xmm */
-    case X86EMUL_OPC_VEX_66(0x0f3a, 0x44): /* vpclmulqdq $imm8,xmm/m128,xmm,xmm */
+    case X86EMUL_OPC_VEX_66(0x0f3a, 0x44): /* vpclmulqdq $imm8,{x,y}mm/mem,{x,y}mm,{x,y}mm */
          host_and_vcpu_must_have(pclmulqdq);
          if ( vex.opcx == vex_none )
              goto simd_0f3a_common;
-        generate_exception_if(vex.l, EXC_UD);
+        if ( vex.l )
+            host_and_vcpu_must_have(vpclmulqdq);
          goto simd_0f_imm8_avx;
  
+    case X86EMUL_OPC_EVEX_66(0x0f3a, 0x44): /* vpclmulqdq $imm8,[xyz]mm/mem,[xyz]mm,[xyz]mm */
+        host_and_vcpu_must_have(vpclmulqdq);
+        generate_exception_if(evex.brs || evex.opmsk, EXC_UD);
+        goto avx512f_imm8_no_sae;
+
      case X86EMUL_OPC_VEX_66(0x0f3a, 0x4a): /* vblendvps {x,y}mm,{x,y}mm/mem,{x,y}mm,{x,y}mm */
      case X86EMUL_OPC_VEX_66(0x0f3a, 0x4b): /* vblendvpd {x,y}mm,{x,y}mm/mem,{x,y}mm,{x,y}mm */
          generate_exception_if(vex.w, EXC_UD);
--- a/xen/include/asm-x86/cpufeature.h
+++ b/xen/include/asm-x86/cpufeature.h
@@ -111,6 +111,7 @@ 
  /* 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_vpclmulqdq      boot_cpu_has(X86_FEATURE_VPCLMULQDQ)
  #define cpu_has_avx512_vnni     boot_cpu_has(X86_FEATURE_AVX512_VNNI)
  #define cpu_has_avx512_bitalg   boot_cpu_has(X86_FEATURE_AVX512_BITALG)
  #define cpu_has_avx512_vpopcntdq boot_cpu_has(X86_FEATURE_AVX512_VPOPCNTDQ)
--- a/xen/include/public/arch-x86/cpufeatureset.h
+++ b/xen/include/public/arch-x86/cpufeatureset.h
@@ -121,7 +121,7 @@  XEN_CPUFEATURE(PBE,           0*32+31) /
  
  /* Intel-defined CPU features, CPUID level 0x00000001.ecx, word 1 */
  XEN_CPUFEATURE(SSE3,          1*32+ 0) /*A  Streaming SIMD Extensions-3 */
-XEN_CPUFEATURE(PCLMULQDQ,     1*32+ 1) /*A  Carry-less mulitplication */
+XEN_CPUFEATURE(PCLMULQDQ,     1*32+ 1) /*A  Carry-less multiplication */
  XEN_CPUFEATURE(DTES64,        1*32+ 2) /*   64-bit Debug Store */
  XEN_CPUFEATURE(MONITOR,       1*32+ 3) /*   Monitor/Mwait support */
  XEN_CPUFEATURE(DSCPL,         1*32+ 4) /*   CPL Qualified Debug Store */
@@ -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(VPCLMULQDQ,    6*32+10) /*A  Vector Carry-less Multiplication Instrs */
  XEN_CPUFEATURE(AVX512_VNNI,   6*32+11) /*A  Vector Neural Network 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 */
--- a/xen/tools/gen-cpuid.py
+++ b/xen/tools/gen-cpuid.py
@@ -254,8 +254,9 @@  def crunch_numbers(state):
  
          # This is just the dependency between AVX512 and AVX2 of XSTATE
          # feature flags.  If want to use AVX512, AVX2 must be supported and
-        # enabled.
-        AVX2: [AVX512F],
+        # enabled.  Certain later extensions, acting on 256-bit vectors of
+        # integers, better depend on AVX2 than AVX.
+        AVX2: [AVX512F, VPCLMULQDQ],
  
          # AVX512F is taken to mean hardware support for 512bit registers
          # (which in practice depends on the EVEX prefix to encode) as well
@@ -270,6 +271,10 @@  def crunch_numbers(state):
          # registers), despite the SDM not formally making this connection.
          AVX512BW: [AVX512_VBMI, AVX512_VBMI2, AVX512_BITALG, AVX512_BF16],
  
+        # Extensions with VEX/EVEX encodings keyed to a separate feature
+        # flag are made dependents of their respective legacy feature.
+        PCLMULQDQ: [VPCLMULQDQ],
+
          # The features:
          #   * Single Thread Indirect Branch Predictors
          #   * Speculative Store Bypass Disable