diff mbox series

[v3,1/3] target/arm: Split out recompute_hflags et al

Message ID 20190222024106.9167-2-richard.henderson@linaro.org
State New
Headers show
Series target/arm: Reduce overhead of cpu_get_tb_cpu_state | expand

Commit Message

Richard Henderson Feb. 22, 2019, 2:41 a.m. UTC
We will use these to minimize the computation for every call to
cpu_get_tb_cpu_state.  For now, the env->hflags variable is not used.

Reviewed-by: Alex Bennée <alex.bennee@linaro.org>

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

---
v3: Do not cache VECLEN, VECSTRIDE, VFPEN.
    Move HANDLER and STACKCHECK to rebuild_hflags_a32.
---
 target/arm/cpu.h       |  28 +++--
 target/arm/helper.h    |   3 +
 target/arm/internals.h |   3 +
 target/arm/helper.c    | 254 ++++++++++++++++++++++++-----------------
 4 files changed, 175 insertions(+), 113 deletions(-)

-- 
2.17.2
diff mbox series

Patch

diff --git a/target/arm/cpu.h b/target/arm/cpu.h
index 84ae6849c2..30532bf53e 100644
--- a/target/arm/cpu.h
+++ b/target/arm/cpu.h
@@ -240,6 +240,9 @@  typedef struct CPUARMState {
     uint32_t pstate;
     uint32_t aarch64; /* 1 if CPU is in aarch64 state; inverse of PSTATE.nRW */
 
+    /* Cached TBFLAGS state.  See below for which bits are included.  */
+    uint32_t hflags;
+
     /* Frequently accessed CPSR bits are stored separately for efficiency.
        This contains all the other bits.  Use cpsr_{read,write} to access
        the whole CPSR.  */
@@ -3065,25 +3068,28 @@  static inline bool arm_cpu_data_is_big_endian(CPUARMState *env)
 
 #include "exec/cpu-all.h"
 
-/* Bit usage in the TB flags field: bit 31 indicates whether we are
+/*
+ * Bit usage in the TB flags field: bit 31 indicates whether we are
  * in 32 or 64 bit mode. The meaning of the other bits depends on that.
  * We put flags which are shared between 32 and 64 bit mode at the top
  * of the word, and flags which apply to only one mode at the bottom.
+ *
+ * Unless otherwise noted, these bits are cached in env->hflags.
  */
 FIELD(TBFLAG_ANY, AARCH64_STATE, 31, 1)
 FIELD(TBFLAG_ANY, MMUIDX, 28, 3)
 FIELD(TBFLAG_ANY, SS_ACTIVE, 27, 1)
-FIELD(TBFLAG_ANY, PSTATE_SS, 26, 1)
+FIELD(TBFLAG_ANY, PSTATE_SS, 26, 1)     /* Not cached. */
 /* Target EL if we take a floating-point-disabled exception */
 FIELD(TBFLAG_ANY, FPEXC_EL, 24, 2)
 FIELD(TBFLAG_ANY, BE_DATA, 23, 1)
 
 /* Bit usage when in AArch32 state: */
-FIELD(TBFLAG_A32, THUMB, 0, 1)
-FIELD(TBFLAG_A32, VECLEN, 1, 3)
-FIELD(TBFLAG_A32, VECSTRIDE, 4, 2)
-FIELD(TBFLAG_A32, VFPEN, 7, 1)
-FIELD(TBFLAG_A32, CONDEXEC, 8, 8)
+FIELD(TBFLAG_A32, THUMB, 0, 1)          /* Not cached. */
+FIELD(TBFLAG_A32, VECLEN, 1, 3)         /* Not cached. */
+FIELD(TBFLAG_A32, VECSTRIDE, 4, 2)      /* Not cached. */
+FIELD(TBFLAG_A32, VFPEN, 7, 1)          /* Not cached. */
+FIELD(TBFLAG_A32, CONDEXEC, 8, 8)       /* Not cached. */
 FIELD(TBFLAG_A32, SCTLR_B, 16, 1)
 /* We store the bottom two bits of the CPAR as TB flags and handle
  * checks on the other bits at runtime
@@ -3105,7 +3111,7 @@  FIELD(TBFLAG_A64, SVEEXC_EL, 2, 2)
 FIELD(TBFLAG_A64, ZCR_LEN, 4, 4)
 FIELD(TBFLAG_A64, PAUTH_ACTIVE, 8, 1)
 FIELD(TBFLAG_A64, BT, 9, 1)
-FIELD(TBFLAG_A64, BTYPE, 10, 2)
+FIELD(TBFLAG_A64, BTYPE, 10, 2)         /* Not cached. */
 FIELD(TBFLAG_A64, TBID, 12, 2)
 
 static inline bool bswap_code(bool sctlr_b)
@@ -3190,6 +3196,12 @@  void arm_register_pre_el_change_hook(ARMCPU *cpu, ARMELChangeHookFn *hook,
 void arm_register_el_change_hook(ARMCPU *cpu, ARMELChangeHookFn *hook, void
         *opaque);
 
+/**
+ * arm_rebuild_hflags:
+ * Rebuild the cached TBFLAGS for arbitrary changed processor state.
+ */
+void arm_rebuild_hflags(CPUARMState *env);
+
 /**
  * aa32_vfp_dreg:
  * Return a pointer to the Dn register within env in 32-bit mode.
diff --git a/target/arm/helper.h b/target/arm/helper.h
index 923e8e1525..bbc1a48089 100644
--- a/target/arm/helper.h
+++ b/target/arm/helper.h
@@ -89,6 +89,9 @@  DEF_HELPER_4(msr_banked, void, env, i32, i32, i32)
 DEF_HELPER_2(get_user_reg, i32, env, i32)
 DEF_HELPER_3(set_user_reg, void, env, i32, i32)
 
+DEF_HELPER_FLAGS_2(rebuild_hflags_a32, TCG_CALL_NO_RWG, void, env, i32)
+DEF_HELPER_FLAGS_2(rebuild_hflags_a64, TCG_CALL_NO_RWG, void, env, i32)
+
 DEF_HELPER_1(vfp_get_fpscr, i32, env)
 DEF_HELPER_2(vfp_set_fpscr, void, env, i32)
 
diff --git a/target/arm/internals.h b/target/arm/internals.h
index a4bd1becb7..8c1b813364 100644
--- a/target/arm/internals.h
+++ b/target/arm/internals.h
@@ -968,4 +968,7 @@  ARMVAParameters aa64_va_parameters_both(CPUARMState *env, uint64_t va,
 ARMVAParameters aa64_va_parameters(CPUARMState *env, uint64_t va,
                                    ARMMMUIdx mmu_idx, bool data);
 
+uint32_t rebuild_hflags_a32(CPUARMState *env, int el);
+uint32_t rebuild_hflags_a64(CPUARMState *env, int el);
+
 #endif
diff --git a/target/arm/helper.c b/target/arm/helper.c
index a018eb23fe..29486a09f6 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -13886,139 +13886,183 @@  ARMMMUIdx arm_stage1_mmu_idx(CPUARMState *env)
 }
 #endif
 
+static uint32_t common_hflags(CPUARMState *env, int el, ARMMMUIdx mmu_idx,
+                              int fp_el, uint32_t flags)
+{
+    flags = FIELD_DP32(flags, TBFLAG_ANY, FPEXC_EL, fp_el);
+    flags = FIELD_DP32(flags, TBFLAG_ANY, MMUIDX,
+                       arm_to_core_mmu_idx(mmu_idx));
+    if (arm_cpu_data_is_big_endian(env)) {
+        flags = FIELD_DP32(flags, TBFLAG_ANY, BE_DATA, 1);
+    }
+    if (arm_singlestep_active(env)) {
+        flags = FIELD_DP32(flags, TBFLAG_ANY, SS_ACTIVE, 1);
+    }
+    return flags;
+}
+
+uint32_t rebuild_hflags_a32(CPUARMState *env, int el)
+{
+    uint32_t flags = 0;
+    ARMMMUIdx mmu_idx;
+    int fp_el;
+
+    flags = FIELD_DP32(flags, TBFLAG_A32, SCTLR_B, arm_sctlr_b(env));
+    flags = FIELD_DP32(flags, TBFLAG_A32, NS, !access_secure_reg(env));
+    flags = FIELD_DP32(flags, TBFLAG_A32, XSCALE_CPAR, env->cp15.c15_cpar);
+
+    if (arm_v7m_is_handler_mode(env)) {
+        flags = FIELD_DP32(flags, TBFLAG_A32, HANDLER, 1);
+    }
+
+    mmu_idx = arm_mmu_idx(env);
+
+    /* v8M always applies stack limit checks unless CCR.STKOFHFNMIGN is
+     * suppressing them because the requested execution priority is less than 0.
+     */
+    if (arm_feature(env, ARM_FEATURE_V8) &&
+        arm_feature(env, ARM_FEATURE_M) &&
+        !((mmu_idx & ARM_MMU_IDX_M_NEGPRI) &&
+          (env->v7m.ccr[env->v7m.secure] & R_V7M_CCR_STKOFHFNMIGN_MASK))) {
+        flags = FIELD_DP32(flags, TBFLAG_A32, STACKCHECK, 1);
+    }
+
+    fp_el = fp_exception_el(env, el);
+    return common_hflags(env, el, mmu_idx, fp_el, flags);
+}
+
+uint32_t rebuild_hflags_a64(CPUARMState *env, int el)
+{
+    ARMCPU *cpu = arm_env_get_cpu(env);
+    ARMMMUIdx mmu_idx = arm_mmu_idx(env);
+    ARMMMUIdx stage1 = stage_1_mmu_idx(mmu_idx);
+    ARMVAParameters p0 = aa64_va_parameters_both(env, 0, stage1);
+    int fp_el = fp_exception_el(env, el);
+    uint32_t flags = 0;
+    uint64_t sctlr;
+    int tbii, tbid;
+
+    flags = FIELD_DP32(flags, TBFLAG_ANY, AARCH64_STATE, 1);
+
+    /* Get control bits for tagged addresses.  */
+    /* FIXME: ARMv8.1-VHE S2 translation regime.  */
+    if (regime_el(env, stage1) < 2) {
+        ARMVAParameters p1 = aa64_va_parameters_both(env, -1, stage1);
+        tbid = (p1.tbi << 1) | p0.tbi;
+        tbii = tbid & ~((p1.tbid << 1) | p0.tbid);
+    } else {
+        tbid = p0.tbi;
+        tbii = tbid & !p0.tbid;
+    }
+
+    flags = FIELD_DP32(flags, TBFLAG_A64, TBII, tbii);
+    flags = FIELD_DP32(flags, TBFLAG_A64, TBID, tbid);
+
+    if (cpu_isar_feature(aa64_sve, cpu)) {
+        int sve_el = sve_exception_el(env, el);
+        uint32_t zcr_len;
+
+        /* If SVE is disabled, but FP is enabled,
+         * then the effective len is 0.
+         */
+        if (sve_el != 0 && fp_el == 0) {
+            zcr_len = 0;
+        } else {
+            zcr_len = sve_zcr_len_for_el(env, el);
+        }
+        flags = FIELD_DP32(flags, TBFLAG_A64, SVEEXC_EL, sve_el);
+        flags = FIELD_DP32(flags, TBFLAG_A64, ZCR_LEN, zcr_len);
+    }
+
+    if (el == 0) {
+        /* FIXME: ARMv8.1-VHE S2 translation regime.  */
+        sctlr = env->cp15.sctlr_el[1];
+    } else {
+        sctlr = env->cp15.sctlr_el[el];
+    }
+    if (cpu_isar_feature(aa64_pauth, cpu)) {
+        /*
+         * In order to save space in flags, we record only whether
+         * pauth is "inactive", meaning all insns are implemented as
+         * a nop, or "active" when some action must be performed.
+         * The decision of which action to take is left to a helper.
+         */
+        if (sctlr & (SCTLR_EnIA | SCTLR_EnIB | SCTLR_EnDA | SCTLR_EnDB)) {
+            flags = FIELD_DP32(flags, TBFLAG_A64, PAUTH_ACTIVE, 1);
+        }
+    }
+
+    if (cpu_isar_feature(aa64_bti, cpu)) {
+        /* Note that SCTLR_EL[23].BT == SCTLR_BT1.  */
+        if (sctlr & (el == 0 ? SCTLR_BT0 : SCTLR_BT1)) {
+            flags = FIELD_DP32(flags, TBFLAG_A64, BT, 1);
+        }
+    }
+
+    return common_hflags(env, el, mmu_idx, fp_el, flags);
+}
+
+void arm_rebuild_hflags(CPUARMState *env)
+{
+    int el = arm_current_el(env);
+    env->hflags = (is_a64(env)
+                   ? rebuild_hflags_a64(env, el)
+                   : rebuild_hflags_a32(env, el));
+}
+
+void HELPER(rebuild_hflags_a32)(CPUARMState *env, uint32_t el)
+{
+    tcg_debug_assert(!is_a64(env));
+    env->hflags = rebuild_hflags_a32(env, el);
+}
+
+void HELPER(rebuild_hflags_a64)(CPUARMState *env, uint32_t el)
+{
+    tcg_debug_assert(is_a64(env));
+    env->hflags = rebuild_hflags_a64(env, el);
+}
+
 void cpu_get_tb_cpu_state(CPUARMState *env, target_ulong *pc,
                           target_ulong *cs_base, uint32_t *pflags)
 {
-    ARMMMUIdx mmu_idx = arm_mmu_idx(env);
     int current_el = arm_current_el(env);
-    int fp_el = fp_exception_el(env, current_el);
-    uint32_t flags = 0;
+    uint32_t flags;
+    uint32_t pstate_for_ss;
 
+    *cs_base = 0;
     if (is_a64(env)) {
-        ARMCPU *cpu = arm_env_get_cpu(env);
-        uint64_t sctlr;
-
         *pc = env->pc;
-        flags = FIELD_DP32(flags, TBFLAG_ANY, AARCH64_STATE, 1);
-
-        /* Get control bits for tagged addresses.  */
-        {
-            ARMMMUIdx stage1 = stage_1_mmu_idx(mmu_idx);
-            ARMVAParameters p0 = aa64_va_parameters_both(env, 0, stage1);
-            int tbii, tbid;
-
-            /* FIXME: ARMv8.1-VHE S2 translation regime.  */
-            if (regime_el(env, stage1) < 2) {
-                ARMVAParameters p1 = aa64_va_parameters_both(env, -1, stage1);
-                tbid = (p1.tbi << 1) | p0.tbi;
-                tbii = tbid & ~((p1.tbid << 1) | p0.tbid);
-            } else {
-                tbid = p0.tbi;
-                tbii = tbid & !p0.tbid;
-            }
-
-            flags = FIELD_DP32(flags, TBFLAG_A64, TBII, tbii);
-            flags = FIELD_DP32(flags, TBFLAG_A64, TBID, tbid);
-        }
-
-        if (cpu_isar_feature(aa64_sve, cpu)) {
-            int sve_el = sve_exception_el(env, current_el);
-            uint32_t zcr_len;
-
-            /* If SVE is disabled, but FP is enabled,
-             * then the effective len is 0.
-             */
-            if (sve_el != 0 && fp_el == 0) {
-                zcr_len = 0;
-            } else {
-                zcr_len = sve_zcr_len_for_el(env, current_el);
-            }
-            flags = FIELD_DP32(flags, TBFLAG_A64, SVEEXC_EL, sve_el);
-            flags = FIELD_DP32(flags, TBFLAG_A64, ZCR_LEN, zcr_len);
-        }
-
-        if (current_el == 0) {
-            /* FIXME: ARMv8.1-VHE S2 translation regime.  */
-            sctlr = env->cp15.sctlr_el[1];
-        } else {
-            sctlr = env->cp15.sctlr_el[current_el];
-        }
-        if (cpu_isar_feature(aa64_pauth, cpu)) {
-            /*
-             * In order to save space in flags, we record only whether
-             * pauth is "inactive", meaning all insns are implemented as
-             * a nop, or "active" when some action must be performed.
-             * The decision of which action to take is left to a helper.
-             */
-            if (sctlr & (SCTLR_EnIA | SCTLR_EnIB | SCTLR_EnDA | SCTLR_EnDB)) {
-                flags = FIELD_DP32(flags, TBFLAG_A64, PAUTH_ACTIVE, 1);
-            }
-        }
-
-        if (cpu_isar_feature(aa64_bti, cpu)) {
-            /* Note that SCTLR_EL[23].BT == SCTLR_BT1.  */
-            if (sctlr & (current_el == 0 ? SCTLR_BT0 : SCTLR_BT1)) {
-                flags = FIELD_DP32(flags, TBFLAG_A64, BT, 1);
-            }
-            flags = FIELD_DP32(flags, TBFLAG_A64, BTYPE, env->btype);
-        }
+        flags = rebuild_hflags_a64(env, current_el);
+        flags = FIELD_DP32(flags, TBFLAG_A64, BTYPE, env->btype);
+        pstate_for_ss = env->pstate;
     } else {
         *pc = env->regs[15];
+        flags = rebuild_hflags_a32(env, current_el);
         flags = FIELD_DP32(flags, TBFLAG_A32, THUMB, env->thumb);
+        flags = FIELD_DP32(flags, TBFLAG_A32, CONDEXEC, env->condexec_bits);
         flags = FIELD_DP32(flags, TBFLAG_A32, VECLEN, env->vfp.vec_len);
         flags = FIELD_DP32(flags, TBFLAG_A32, VECSTRIDE, env->vfp.vec_stride);
-        flags = FIELD_DP32(flags, TBFLAG_A32, CONDEXEC, env->condexec_bits);
-        flags = FIELD_DP32(flags, TBFLAG_A32, SCTLR_B, arm_sctlr_b(env));
-        flags = FIELD_DP32(flags, TBFLAG_A32, NS, !access_secure_reg(env));
         if (env->vfp.xregs[ARM_VFP_FPEXC] & (1 << 30)
             || arm_el_is_aa64(env, 1)) {
             flags = FIELD_DP32(flags, TBFLAG_A32, VFPEN, 1);
         }
-        flags = FIELD_DP32(flags, TBFLAG_A32, XSCALE_CPAR, env->cp15.c15_cpar);
+        pstate_for_ss = env->uncached_cpsr;
     }
 
-    flags = FIELD_DP32(flags, TBFLAG_ANY, MMUIDX, arm_to_core_mmu_idx(mmu_idx));
-
     /* The SS_ACTIVE and PSTATE_SS bits correspond to the state machine
      * states defined in the ARM ARM for software singlestep:
      *  SS_ACTIVE   PSTATE.SS   State
      *     0            x       Inactive (the TB flag for SS is always 0)
      *     1            0       Active-pending
      *     1            1       Active-not-pending
+     * SS_ACTIVE is set in hflags; PSTATE_SS is computed every TB.
      */
-    if (arm_singlestep_active(env)) {
-        flags = FIELD_DP32(flags, TBFLAG_ANY, SS_ACTIVE, 1);
-        if (is_a64(env)) {
-            if (env->pstate & PSTATE_SS) {
-                flags = FIELD_DP32(flags, TBFLAG_ANY, PSTATE_SS, 1);
-            }
-        } else {
-            if (env->uncached_cpsr & PSTATE_SS) {
-                flags = FIELD_DP32(flags, TBFLAG_ANY, PSTATE_SS, 1);
-            }
-        }
-    }
-    if (arm_cpu_data_is_big_endian(env)) {
-        flags = FIELD_DP32(flags, TBFLAG_ANY, BE_DATA, 1);
-    }
-    flags = FIELD_DP32(flags, TBFLAG_ANY, FPEXC_EL, fp_el);
-
-    if (arm_v7m_is_handler_mode(env)) {
-        flags = FIELD_DP32(flags, TBFLAG_A32, HANDLER, 1);
-    }
-
-    /* v8M always applies stack limit checks unless CCR.STKOFHFNMIGN is
-     * suppressing them because the requested execution priority is less than 0.
-     */
-    if (arm_feature(env, ARM_FEATURE_V8) &&
-        arm_feature(env, ARM_FEATURE_M) &&
-        !((mmu_idx  & ARM_MMU_IDX_M_NEGPRI) &&
-          (env->v7m.ccr[env->v7m.secure] & R_V7M_CCR_STKOFHFNMIGN_MASK))) {
-        flags = FIELD_DP32(flags, TBFLAG_A32, STACKCHECK, 1);
+    if (FIELD_EX32(flags, TBFLAG_ANY, SS_ACTIVE)
+        && (pstate_for_ss & PSTATE_SS)) {
+        flags = FIELD_DP32(flags, TBFLAG_ANY, PSTATE_SS, 1);
     }
 
     *pflags = flags;
-    *cs_base = 0;
 }
 
 #ifdef TARGET_AARCH64