From: "Clément Léger" <cleger@rivosinc.com>
To: qemu-riscv@nongnu.org, Palmer Dabbelt <palmer@dabbelt.com>,
Alistair Francis <alistair.francis@wdc.com>,
Bin Meng <bin.meng@windriver.com>,
Weiwei Li <liwei1518@gmail.com>,
Daniel Henrique Barboza <dbarboza@ventanamicro.com>,
Liu Zhiwei <zhiwei_liu@linux.alibaba.com>
Cc: "Clément Léger" <cleger@rivosinc.com>,
qemu-devel@nongnu.org, "Atish Patra" <atishp@rivosinc.com>
Subject: [PATCH] target/riscv: fix instructions count handling in icount mode
Date: Thu, 11 Apr 2024 13:34:05 +0200 [thread overview]
Message-ID: <20240411113406.1301906-1-cleger@rivosinc.com> (raw)
When icount is enabled, rather than returning the virtual CPU time, we
should return the instruction count itself. Add an instructions bool
parameter to get_ticks() to correctly return icount_get_raw() when
icount_enabled() == 1 and instruction count is queried. This will modify
the existing behavior which was returning an instructions count close to
the number of cycles (CPI ~= 1).
Signed-off-by: Clément Léger <cleger@rivosinc.com>
---
target/riscv/csr.c | 29 ++++++++++++++++-------------
1 file changed, 16 insertions(+), 13 deletions(-)
diff --git a/target/riscv/csr.c b/target/riscv/csr.c
index 726096444f..5f1dcee102 100644
--- a/target/riscv/csr.c
+++ b/target/riscv/csr.c
@@ -762,14 +762,17 @@ static RISCVException write_vcsr(CPURISCVState *env, int csrno,
}
/* User Timers and Counters */
-static target_ulong get_ticks(bool shift)
+static target_ulong get_ticks(bool shift, bool instructions)
{
int64_t val;
target_ulong result;
#if !defined(CONFIG_USER_ONLY)
if (icount_enabled()) {
- val = icount_get();
+ if (instructions)
+ val = icount_get_raw();
+ else
+ val = icount_get();
} else {
val = cpu_get_host_ticks();
}
@@ -804,14 +807,14 @@ static RISCVException read_timeh(CPURISCVState *env, int csrno,
static RISCVException read_hpmcounter(CPURISCVState *env, int csrno,
target_ulong *val)
{
- *val = get_ticks(false);
+ *val = get_ticks(false, (csrno == CSR_INSTRET));
return RISCV_EXCP_NONE;
}
static RISCVException read_hpmcounterh(CPURISCVState *env, int csrno,
target_ulong *val)
{
- *val = get_ticks(true);
+ *val = get_ticks(true, (csrno == CSR_INSTRETH));
return RISCV_EXCP_NONE;
}
@@ -875,11 +878,11 @@ static RISCVException write_mhpmcounter(CPURISCVState *env, int csrno,
int ctr_idx = csrno - CSR_MCYCLE;
PMUCTRState *counter = &env->pmu_ctrs[ctr_idx];
uint64_t mhpmctr_val = val;
+ bool instr = riscv_pmu_ctr_monitor_instructions(env, ctr_idx);
counter->mhpmcounter_val = val;
- if (riscv_pmu_ctr_monitor_cycles(env, ctr_idx) ||
- riscv_pmu_ctr_monitor_instructions(env, ctr_idx)) {
- counter->mhpmcounter_prev = get_ticks(false);
+ if (riscv_pmu_ctr_monitor_cycles(env, ctr_idx) || instr) {
+ counter->mhpmcounter_prev = get_ticks(false, instr);
if (ctr_idx > 2) {
if (riscv_cpu_mxl(env) == MXL_RV32) {
mhpmctr_val = mhpmctr_val |
@@ -902,12 +905,12 @@ static RISCVException write_mhpmcounterh(CPURISCVState *env, int csrno,
PMUCTRState *counter = &env->pmu_ctrs[ctr_idx];
uint64_t mhpmctr_val = counter->mhpmcounter_val;
uint64_t mhpmctrh_val = val;
+ bool instr = riscv_pmu_ctr_monitor_instructions(env, ctr_idx);
counter->mhpmcounterh_val = val;
mhpmctr_val = mhpmctr_val | (mhpmctrh_val << 32);
- if (riscv_pmu_ctr_monitor_cycles(env, ctr_idx) ||
- riscv_pmu_ctr_monitor_instructions(env, ctr_idx)) {
- counter->mhpmcounterh_prev = get_ticks(true);
+ if (riscv_pmu_ctr_monitor_cycles(env, ctr_idx) || instr) {
+ counter->mhpmcounterh_prev = get_ticks(true, instr);
if (ctr_idx > 2) {
riscv_pmu_setup_timer(env, mhpmctr_val, ctr_idx);
}
@@ -926,6 +929,7 @@ static RISCVException riscv_pmu_read_ctr(CPURISCVState *env, target_ulong *val,
counter->mhpmcounter_prev;
target_ulong ctr_val = upper_half ? counter->mhpmcounterh_val :
counter->mhpmcounter_val;
+ bool instr = riscv_pmu_ctr_monitor_instructions(env, ctr_idx);
if (get_field(env->mcountinhibit, BIT(ctr_idx))) {
/*
@@ -946,9 +950,8 @@ static RISCVException riscv_pmu_read_ctr(CPURISCVState *env, target_ulong *val,
* The kernel computes the perf delta by subtracting the current value from
* the value it initialized previously (ctr_val).
*/
- if (riscv_pmu_ctr_monitor_cycles(env, ctr_idx) ||
- riscv_pmu_ctr_monitor_instructions(env, ctr_idx)) {
- *val = get_ticks(upper_half) - ctr_prev + ctr_val;
+ if (riscv_pmu_ctr_monitor_cycles(env, ctr_idx) || instr) {
+ *val = get_ticks(upper_half, instr) - ctr_prev + ctr_val;
} else {
*val = ctr_val;
}
--
2.43.0
next reply other threads:[~2024-04-11 11:34 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-04-11 11:34 Clément Léger [this message]
2024-04-11 17:22 ` [PATCH] target/riscv: fix instructions count handling in icount mode Atish Kumar Patra
2024-04-29 2:37 ` Alistair Francis
2024-05-27 23:47 ` Alistair Francis
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20240411113406.1301906-1-cleger@rivosinc.com \
--to=cleger@rivosinc.com \
--cc=alistair.francis@wdc.com \
--cc=atishp@rivosinc.com \
--cc=bin.meng@windriver.com \
--cc=dbarboza@ventanamicro.com \
--cc=liwei1518@gmail.com \
--cc=palmer@dabbelt.com \
--cc=qemu-devel@nongnu.org \
--cc=qemu-riscv@nongnu.org \
--cc=zhiwei_liu@linux.alibaba.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).