Mesa (master): nvc0: fix monitoring multiple MP counters queries on Fermi
Samuel Pitoiset
hakzsam at kemper.freedesktop.org
Fri Oct 16 19:59:20 UTC 2015
Module: Mesa
Branch: master
Commit: 7abd707251f29aaf27f83644e47d2dc8b75e10c6
URL: http://cgit.freedesktop.org/mesa/mesa/commit/?id=7abd707251f29aaf27f83644e47d2dc8b75e10c6
Author: Samuel Pitoiset <samuel.pitoiset at gmail.com>
Date: Fri Oct 9 11:18:45 2015 +0200
nvc0: fix monitoring multiple MP counters queries on Fermi
For strange reasons, the signal id depends on the slot selected on Fermi
but not on Kepler. Fortunately, the signal ids are just offseted by the
slot id!
Signed-off-by: Samuel Pitoiset <samuel.pitoiset at gmail.com>
Reviewed-by: Ilia Mirkin <imirkin at alum.mit.edu>
---
.../drivers/nouveau/nvc0/nvc0_query_hw_sm.c | 163 +++++++++++---------
1 file changed, 87 insertions(+), 76 deletions(-)
diff --git a/src/gallium/drivers/nouveau/nvc0/nvc0_query_hw_sm.c b/src/gallium/drivers/nouveau/nvc0/nvc0_query_hw_sm.c
index 99e9073..2f827b0 100644
--- a/src/gallium/drivers/nouveau/nvc0/nvc0_query_hw_sm.c
+++ b/src/gallium/drivers/nouveau/nvc0/nvc0_query_hw_sm.c
@@ -128,9 +128,9 @@ struct nvc0_hw_sm_counter_cfg
{
uint32_t func : 16; /* mask or 4-bit logic op (depending on mode) */
uint32_t mode : 4; /* LOGOP,B6,LOGOP_B6(_PULSE) */
- uint32_t num_src : 3; /* number of sources (1 - 6, only for NVC0:NVE4) */
uint32_t sig_dom : 1; /* if 0, MP_PM_A (per warp-sched), if 1, MP_PM_B */
uint32_t sig_sel : 8; /* signal group */
+ uint32_t src_mask; /* mask for signal selection (only for NVC0:NVE4) */
uint32_t src_sel; /* signal selection for up to 4 sources */
};
@@ -150,19 +150,19 @@ struct nvc0_hw_sm_query_cfg
uint8_t norm[2]; /* normalization num,denom */
};
-#define _Q1A(n, f, m, g, s, nu, dn) [NVE4_HW_SM_QUERY_##n] = { { { f, NVE4_COMPUTE_MP_PM_FUNC_MODE_##m, 0, 0, NVE4_COMPUTE_MP_PM_A_SIGSEL_##g, s }, {}, {}, {} }, 1, NVC0_COUNTER_OPn_SUM, { nu, dn } }
-#define _Q1B(n, f, m, g, s, nu, dn) [NVE4_HW_SM_QUERY_##n] = { { { f, NVE4_COMPUTE_MP_PM_FUNC_MODE_##m, 0, 1, NVE4_COMPUTE_MP_PM_B_SIGSEL_##g, s }, {}, {}, {} }, 1, NVC0_COUNTER_OPn_SUM, { nu, dn } }
+#define _Q1A(n, f, m, g, s, nu, dn) [NVE4_HW_SM_QUERY_##n] = { { { f, NVE4_COMPUTE_MP_PM_FUNC_MODE_##m, 0, NVE4_COMPUTE_MP_PM_A_SIGSEL_##g, 0, s }, {}, {}, {} }, 1, NVC0_COUNTER_OPn_SUM, { nu, dn } }
+#define _Q1B(n, f, m, g, s, nu, dn) [NVE4_HW_SM_QUERY_##n] = { { { f, NVE4_COMPUTE_MP_PM_FUNC_MODE_##m, 1, NVE4_COMPUTE_MP_PM_B_SIGSEL_##g, 0, s }, {}, {}, {} }, 1, NVC0_COUNTER_OPn_SUM, { nu, dn } }
#define _M2A(n, f0, m0, g0, s0, f1, m1, g1, s1, o, nu, dn) [NVE4_HW_SM_QUERY_METRIC_##n] = { { \
- { f0, NVE4_COMPUTE_MP_PM_FUNC_MODE_##m0, 0, 0, NVE4_COMPUTE_MP_PM_A_SIGSEL_##g0, s0 }, \
- { f1, NVE4_COMPUTE_MP_PM_FUNC_MODE_##m1, 0, 0, NVE4_COMPUTE_MP_PM_A_SIGSEL_##g1, s1 }, \
+ { f0, NVE4_COMPUTE_MP_PM_FUNC_MODE_##m0, 0, NVE4_COMPUTE_MP_PM_A_SIGSEL_##g0, 0, s0 }, \
+ { f1, NVE4_COMPUTE_MP_PM_FUNC_MODE_##m1, 0, NVE4_COMPUTE_MP_PM_A_SIGSEL_##g1, 0, s1 }, \
{}, {}, }, 2, NVC0_COUNTER_OP2_##o, { nu, dn } }
#define _M2B(n, f0, m0, g0, s0, f1, m1, g1, s1, o, nu, dn) [NVE4_HW_SM_QUERY_METRIC_##n] = { { \
- { f0, NVE4_COMPUTE_MP_PM_FUNC_MODE_##m0, 0, 1, NVE4_COMPUTE_MP_PM_B_SIGSEL_##g0, s0 }, \
- { f1, NVE4_COMPUTE_MP_PM_FUNC_MODE_##m1, 0, 1, NVE4_COMPUTE_MP_PM_B_SIGSEL_##g1, s1 }, \
+ { f0, NVE4_COMPUTE_MP_PM_FUNC_MODE_##m0, 1, NVE4_COMPUTE_MP_PM_B_SIGSEL_##g0, 0, s0 }, \
+ { f1, NVE4_COMPUTE_MP_PM_FUNC_MODE_##m1, 1, NVE4_COMPUTE_MP_PM_B_SIGSEL_##g1, 0, s1 }, \
{}, {}, }, 2, NVC0_COUNTER_OP2_##o, { nu, dn } }
#define _M2AB(n, f0, m0, g0, s0, f1, m1, g1, s1, o, nu, dn) [NVE4_HW_SM_QUERY_METRIC_##n] = { { \
- { f0, NVE4_COMPUTE_MP_PM_FUNC_MODE_##m0, 0, 0, NVE4_COMPUTE_MP_PM_A_SIGSEL_##g0, s0 }, \
- { f1, NVE4_COMPUTE_MP_PM_FUNC_MODE_##m1, 0, 1, NVE4_COMPUTE_MP_PM_B_SIGSEL_##g1, s1 }, \
+ { f0, NVE4_COMPUTE_MP_PM_FUNC_MODE_##m0, 0, NVE4_COMPUTE_MP_PM_A_SIGSEL_##g0, 0, s0 }, \
+ { f1, NVE4_COMPUTE_MP_PM_FUNC_MODE_##m1, 1, NVE4_COMPUTE_MP_PM_B_SIGSEL_##g1, 0, s1 }, \
{}, {}, }, 2, NVC0_COUNTER_OP2_##o, { nu, dn } }
/* NOTES:
@@ -280,78 +280,78 @@ static const uint64_t nvc0_read_hw_sm_counters_code[] =
0x8000000000001de7ULL
};
-#define _C(f, o, g, s) { f, NVC0_COMPUTE_MP_PM_OP_MODE_##o, 0, 0, g, s }
+#define _C(f, o, g, m, s) { f, NVC0_COMPUTE_MP_PM_OP_MODE_##o, 0, g, m, s }
#define _Q(n, c, ...) [NVC0_HW_SM_QUERY_##n] = { \
{ __VA_ARGS__ }, c, NVC0_COUNTER_OPn_SUM, { 1, 1 }, \
}
static const struct nvc0_hw_sm_query_cfg nvc0_hw_sm_queries[] =
{
- _Q(ACTIVE_CYCLES, 1, _C(0xaaaa, LOGOP, 0x11, 0x00000000)),
- _Q(ACTIVE_WARPS, 6, _C(0xaaaa, LOGOP, 0x24, 0x00000010),
- _C(0xaaaa, LOGOP, 0x24, 0x00000021),
- _C(0xaaaa, LOGOP, 0x24, 0x00000032),
- _C(0xaaaa, LOGOP, 0x24, 0x00000043),
- _C(0xaaaa, LOGOP, 0x24, 0x00000054),
- _C(0xaaaa, LOGOP, 0x24, 0x00000065)),
- _Q(ATOM_COUNT, 1, _C(0xaaaa, LOGOP, 0x63, 0x00000030)),
- _Q(BRANCH, 2, _C(0xaaaa, LOGOP, 0x1a, 0x00000000),
- _C(0xaaaa, LOGOP, 0x1a, 0x00000011)),
- _Q(DIVERGENT_BRANCH, 2, _C(0xaaaa, LOGOP, 0x19, 0x00000020),
- _C(0xaaaa, LOGOP, 0x19, 0x00000031)),
- _Q(GLD_REQUEST, 1, _C(0xaaaa, LOGOP, 0x64, 0x00000030)),
- _Q(GRED_COUNT, 1, _C(0xaaaa, LOGOP, 0x63, 0x00000040)),
- _Q(GST_REQUEST, 1, _C(0xaaaa, LOGOP, 0x64, 0x00000060)),
- _Q(INST_EXECUTED, 3, _C(0xaaaa, LOGOP, 0x2d, 0x00000000),
- _C(0xaaaa, LOGOP, 0x2d, 0x00000011),
- _C(0xaaaa, LOGOP, 0x2d, 0x00000022)),
- _Q(INST_ISSUED1_0, 1, _C(0xaaaa, LOGOP, 0x7e, 0x00000010)),
- _Q(INST_ISSUED1_1, 1, _C(0xaaaa, LOGOP, 0x7e, 0x00000040)),
- _Q(INST_ISSUED2_0, 1, _C(0xaaaa, LOGOP, 0x7e, 0x00000020)),
- _Q(INST_ISSUED2_1, 1, _C(0xaaaa, LOGOP, 0x7e, 0x00000050)),
- _Q(LOCAL_LD, 1, _C(0xaaaa, LOGOP, 0x64, 0x00000020)),
- _Q(LOCAL_ST, 1, _C(0xaaaa, LOGOP, 0x64, 0x00000050)),
- _Q(PROF_TRIGGER_0, 1, _C(0xaaaa, LOGOP, 0x01, 0x00000000)),
- _Q(PROF_TRIGGER_1, 1, _C(0xaaaa, LOGOP, 0x01, 0x00000010)),
- _Q(PROF_TRIGGER_2, 1, _C(0xaaaa, LOGOP, 0x01, 0x00000020)),
- _Q(PROF_TRIGGER_3, 1, _C(0xaaaa, LOGOP, 0x01, 0x00000030)),
- _Q(PROF_TRIGGER_4, 1, _C(0xaaaa, LOGOP, 0x01, 0x00000040)),
- _Q(PROF_TRIGGER_5, 1, _C(0xaaaa, LOGOP, 0x01, 0x00000050)),
- _Q(PROF_TRIGGER_6, 1, _C(0xaaaa, LOGOP, 0x01, 0x00000060)),
- _Q(PROF_TRIGGER_7, 1, _C(0xaaaa, LOGOP, 0x01, 0x00000070)),
- _Q(SHARED_LD, 1, _C(0xaaaa, LOGOP, 0x64, 0x00000010)),
- _Q(SHARED_ST, 1, _C(0xaaaa, LOGOP, 0x64, 0x00000040)),
- _Q(THREADS_LAUNCHED, 6, _C(0xaaaa, LOGOP, 0x26, 0x00000010),
- _C(0xaaaa, LOGOP, 0x26, 0x00000021),
- _C(0xaaaa, LOGOP, 0x26, 0x00000032),
- _C(0xaaaa, LOGOP, 0x26, 0x00000043),
- _C(0xaaaa, LOGOP, 0x26, 0x00000054),
- _C(0xaaaa, LOGOP, 0x26, 0x00000065)),
- _Q(TH_INST_EXECUTED_0, 6, _C(0xaaaa, LOGOP, 0xa3, 0x00000000),
- _C(0xaaaa, LOGOP, 0xa3, 0x00000011),
- _C(0xaaaa, LOGOP, 0xa3, 0x00000022),
- _C(0xaaaa, LOGOP, 0xa3, 0x00000033),
- _C(0xaaaa, LOGOP, 0xa3, 0x00000044),
- _C(0xaaaa, LOGOP, 0xa3, 0x00000055)),
- _Q(TH_INST_EXECUTED_1, 6, _C(0xaaaa, LOGOP, 0xa5, 0x00000000),
- _C(0xaaaa, LOGOP, 0xa5, 0x00000011),
- _C(0xaaaa, LOGOP, 0xa5, 0x00000022),
- _C(0xaaaa, LOGOP, 0xa5, 0x00000033),
- _C(0xaaaa, LOGOP, 0xa5, 0x00000044),
- _C(0xaaaa, LOGOP, 0xa5, 0x00000055)),
- _Q(TH_INST_EXECUTED_2, 6, _C(0xaaaa, LOGOP, 0xa4, 0x00000000),
- _C(0xaaaa, LOGOP, 0xa4, 0x00000011),
- _C(0xaaaa, LOGOP, 0xa4, 0x00000022),
- _C(0xaaaa, LOGOP, 0xa4, 0x00000033),
- _C(0xaaaa, LOGOP, 0xa4, 0x00000044),
- _C(0xaaaa, LOGOP, 0xa4, 0x00000055)),
- _Q(TH_INST_EXECUTED_3, 6, _C(0xaaaa, LOGOP, 0xa6, 0x00000000),
- _C(0xaaaa, LOGOP, 0xa6, 0x00000011),
- _C(0xaaaa, LOGOP, 0xa6, 0x00000022),
- _C(0xaaaa, LOGOP, 0xa6, 0x00000033),
- _C(0xaaaa, LOGOP, 0xa6, 0x00000044),
- _C(0xaaaa, LOGOP, 0xa6, 0x00000055)),
- _Q(WARPS_LAUNCHED, 1, _C(0xaaaa, LOGOP, 0x26, 0x00000000)),
+ _Q(ACTIVE_CYCLES, 1, _C(0xaaaa, LOGOP, 0x11, 0x000000ff, 0x00000000)),
+ _Q(ACTIVE_WARPS, 6, _C(0xaaaa, LOGOP, 0x24, 0x000000ff, 0x00000010),
+ _C(0xaaaa, LOGOP, 0x24, 0x000000ff, 0x00000020),
+ _C(0xaaaa, LOGOP, 0x24, 0x000000ff, 0x00000030),
+ _C(0xaaaa, LOGOP, 0x24, 0x000000ff, 0x00000040),
+ _C(0xaaaa, LOGOP, 0x24, 0x000000ff, 0x00000050),
+ _C(0xaaaa, LOGOP, 0x24, 0x000000ff, 0x00000060)),
+ _Q(ATOM_COUNT, 1, _C(0xaaaa, LOGOP, 0x63, 0x000000ff, 0x00000030)),
+ _Q(BRANCH, 2, _C(0xaaaa, LOGOP, 0x1a, 0x000000ff, 0x00000000),
+ _C(0xaaaa, LOGOP, 0x1a, 0x000000ff, 0x00000010)),
+ _Q(DIVERGENT_BRANCH, 2, _C(0xaaaa, LOGOP, 0x19, 0x000000ff, 0x00000020),
+ _C(0xaaaa, LOGOP, 0x19, 0x000000ff, 0x00000030)),
+ _Q(GLD_REQUEST, 1, _C(0xaaaa, LOGOP, 0x64, 0x000000ff, 0x00000030)),
+ _Q(GRED_COUNT, 1, _C(0xaaaa, LOGOP, 0x63, 0x000000ff, 0x00000040)),
+ _Q(GST_REQUEST, 1, _C(0xaaaa, LOGOP, 0x64, 0x000000ff, 0x00000060)),
+ _Q(INST_EXECUTED, 3, _C(0xaaaa, LOGOP, 0x2d, 0x000000ff, 0x00000000),
+ _C(0xaaaa, LOGOP, 0x2d, 0x000000ff, 0x00000010),
+ _C(0xaaaa, LOGOP, 0x2d, 0x000000ff, 0x00000020)),
+ _Q(INST_ISSUED1_0, 1, _C(0xaaaa, LOGOP, 0x7e, 0x000000ff, 0x00000010)),
+ _Q(INST_ISSUED1_1, 1, _C(0xaaaa, LOGOP, 0x7e, 0x000000ff, 0x00000040)),
+ _Q(INST_ISSUED2_0, 1, _C(0xaaaa, LOGOP, 0x7e, 0x000000ff, 0x00000020)),
+ _Q(INST_ISSUED2_1, 1, _C(0xaaaa, LOGOP, 0x7e, 0x000000ff, 0x00000050)),
+ _Q(LOCAL_LD, 1, _C(0xaaaa, LOGOP, 0x64, 0x000000ff, 0x00000020)),
+ _Q(LOCAL_ST, 1, _C(0xaaaa, LOGOP, 0x64, 0x000000ff, 0x00000050)),
+ _Q(PROF_TRIGGER_0, 1, _C(0xaaaa, LOGOP, 0x01, 0x000000ff, 0x00000000)),
+ _Q(PROF_TRIGGER_1, 1, _C(0xaaaa, LOGOP, 0x01, 0x000000ff, 0x00000010)),
+ _Q(PROF_TRIGGER_2, 1, _C(0xaaaa, LOGOP, 0x01, 0x000000ff, 0x00000020)),
+ _Q(PROF_TRIGGER_3, 1, _C(0xaaaa, LOGOP, 0x01, 0x000000ff, 0x00000030)),
+ _Q(PROF_TRIGGER_4, 1, _C(0xaaaa, LOGOP, 0x01, 0x000000ff, 0x00000040)),
+ _Q(PROF_TRIGGER_5, 1, _C(0xaaaa, LOGOP, 0x01, 0x000000ff, 0x00000050)),
+ _Q(PROF_TRIGGER_6, 1, _C(0xaaaa, LOGOP, 0x01, 0x000000ff, 0x00000060)),
+ _Q(PROF_TRIGGER_7, 1, _C(0xaaaa, LOGOP, 0x01, 0x000000ff, 0x00000070)),
+ _Q(SHARED_LD, 1, _C(0xaaaa, LOGOP, 0x64, 0x000000ff, 0x00000010)),
+ _Q(SHARED_ST, 1, _C(0xaaaa, LOGOP, 0x64, 0x000000ff, 0x00000040)),
+ _Q(THREADS_LAUNCHED, 6, _C(0xaaaa, LOGOP, 0x26, 0x000000ff, 0x00000010),
+ _C(0xaaaa, LOGOP, 0x26, 0x000000ff, 0x00000020),
+ _C(0xaaaa, LOGOP, 0x26, 0x000000ff, 0x00000030),
+ _C(0xaaaa, LOGOP, 0x26, 0x000000ff, 0x00000040),
+ _C(0xaaaa, LOGOP, 0x26, 0x000000ff, 0x00000050),
+ _C(0xaaaa, LOGOP, 0x26, 0x000000ff, 0x00000060)),
+ _Q(TH_INST_EXECUTED_0, 6, _C(0xaaaa, LOGOP, 0xa3, 0x000000ff, 0x00000000),
+ _C(0xaaaa, LOGOP, 0xa3, 0x000000ff, 0x00000010),
+ _C(0xaaaa, LOGOP, 0xa3, 0x000000ff, 0x00000020),
+ _C(0xaaaa, LOGOP, 0xa3, 0x000000ff, 0x00000030),
+ _C(0xaaaa, LOGOP, 0xa3, 0x000000ff, 0x00000040),
+ _C(0xaaaa, LOGOP, 0xa3, 0x000000ff, 0x00000050)),
+ _Q(TH_INST_EXECUTED_1, 6, _C(0xaaaa, LOGOP, 0xa5, 0x000000ff, 0x00000000),
+ _C(0xaaaa, LOGOP, 0xa5, 0x000000ff, 0x00000010),
+ _C(0xaaaa, LOGOP, 0xa5, 0x000000ff, 0x00000020),
+ _C(0xaaaa, LOGOP, 0xa5, 0x000000ff, 0x00000030),
+ _C(0xaaaa, LOGOP, 0xa5, 0x000000ff, 0x00000040),
+ _C(0xaaaa, LOGOP, 0xa5, 0x000000ff, 0x00000050)),
+ _Q(TH_INST_EXECUTED_2, 6, _C(0xaaaa, LOGOP, 0xa4, 0x000000ff, 0x00000000),
+ _C(0xaaaa, LOGOP, 0xa4, 0x000000ff, 0x00000010),
+ _C(0xaaaa, LOGOP, 0xa4, 0x000000ff, 0x00000020),
+ _C(0xaaaa, LOGOP, 0xa4, 0x000000ff, 0x00000030),
+ _C(0xaaaa, LOGOP, 0xa4, 0x000000ff, 0x00000040),
+ _C(0xaaaa, LOGOP, 0xa4, 0x000000ff, 0x00000050)),
+ _Q(TH_INST_EXECUTED_3, 6, _C(0xaaaa, LOGOP, 0xa6, 0x000000ff, 0x00000000),
+ _C(0xaaaa, LOGOP, 0xa6, 0x000000ff, 0x00000010),
+ _C(0xaaaa, LOGOP, 0xa6, 0x000000ff, 0x00000020),
+ _C(0xaaaa, LOGOP, 0xa6, 0x000000ff, 0x00000030),
+ _C(0xaaaa, LOGOP, 0xa6, 0x000000ff, 0x00000040),
+ _C(0xaaaa, LOGOP, 0xa6, 0x000000ff, 0x00000050)),
+ _Q(WARPS_LAUNCHED, 1, _C(0xaaaa, LOGOP, 0x26, 0x000000ff, 0x00000000)),
};
#undef _Q
@@ -479,6 +479,8 @@ nvc0_hw_sm_begin_query(struct nvc0_context *nvc0, struct nvc0_hw_query *hq)
hq->sequence++;
for (i = 0; i < cfg->num_counters; ++i) {
+ uint32_t mask_sel = 0x00000000;
+
if (!screen->pm.num_hw_sm_active[0]) {
BEGIN_NVC0(push, SUBC_SW(0x0600), 1);
PUSH_DATA (push, 0x80000000);
@@ -493,11 +495,20 @@ nvc0_hw_sm_begin_query(struct nvc0_context *nvc0, struct nvc0_hw_query *hq)
}
}
+ /* Oddly-enough, the signal id depends on the slot selected on Fermi but
+ * not on Kepler. Fortunately, the signal ids are just offseted by the
+ * slot id! */
+ mask_sel |= c;
+ mask_sel |= (c << 8);
+ mask_sel |= (c << 16);
+ mask_sel |= (c << 24);
+ mask_sel &= cfg->ctr[i].src_mask;
+
/* configure and reset the counter(s) */
BEGIN_NVC0(push, NVC0_COMPUTE(MP_PM_SIGSEL(c)), 1);
PUSH_DATA (push, cfg->ctr[i].sig_sel);
BEGIN_NVC0(push, NVC0_COMPUTE(MP_PM_SRCSEL(c)), 1);
- PUSH_DATA (push, cfg->ctr[i].src_sel);
+ PUSH_DATA (push, cfg->ctr[i].src_sel | mask_sel);
BEGIN_NVC0(push, NVC0_COMPUTE(MP_PM_OP(c)), 1);
PUSH_DATA (push, (cfg->ctr[i].func << 4) | cfg->ctr[i].mode);
BEGIN_NVC0(push, NVC0_COMPUTE(MP_PM_SET(c)), 1);
More information about the mesa-commit
mailing list