[PATCH 2/2] drm/amdgpu/vpe: enable vpe dpm
Peyton Lee
peytolee at amd.com
Wed Dec 13 02:49:09 UTC 2023
enable vpe dpm
Signed-off-by: Peyton Lee <peytolee at amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.c | 250 ++++++++++++++++++++++++
drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.h | 12 ++
drivers/gpu/drm/amd/amdgpu/vpe_v6_1.c | 15 ++
3 files changed, 277 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.c
index e81579708e96..2020ddb4182a 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.c
@@ -26,6 +26,7 @@
#include "amdgpu.h"
#include "amdgpu_ucode.h"
#include "amdgpu_vpe.h"
+#include "amdgpu_smu.h"
#include "soc15_common.h"
#include "vpe_v6_1.h"
@@ -33,8 +34,186 @@
/* VPE CSA resides in the 4th page of CSA */
#define AMDGPU_CSA_VPE_OFFSET (4096 * 3)
+/* 1 second timeout */
+#define VPE_IDLE_TIMEOUT msecs_to_jiffies(1000)
+
+#define VPE_MAX_DPM_LEVEL 4
+#define FIXED1_8_BITS_PER_FRACTIONAL_PART 8
+#define GET_PRATIO_INTEGER_PART(x) ((x) >> FIXED1_8_BITS_PER_FRACTIONAL_PART)
+
static void vpe_set_ring_funcs(struct amdgpu_device *adev);
+static inline uint16_t div16_u16_rem(uint16_t dividend, uint16_t divisor, uint16_t *remainder)
+{
+ *remainder = dividend % divisor;
+ return dividend / divisor;
+}
+
+static inline uint16_t complete_integer_division_u16(
+ uint16_t dividend,
+ uint16_t divisor,
+ uint16_t *remainder)
+{
+ return div16_u16_rem(dividend, divisor, (uint16_t *)remainder);
+}
+
+static uint16_t vpe_u1_8_from_fraction(uint16_t numerator, uint16_t denominator)
+{
+ bool arg1_negative = numerator < 0;
+ bool arg2_negative = denominator < 0;
+
+ uint16_t arg1_value = (uint16_t)(arg1_negative ? -numerator : numerator);
+ uint16_t arg2_value = (uint16_t)(arg2_negative ? -denominator : denominator);
+
+ uint16_t remainder;
+
+ /* determine integer part */
+ uint16_t res_value = complete_integer_division_u16(
+ arg1_value, arg2_value, &remainder);
+
+ if (res_value > 127 /* CHAR_MAX */)
+ return 0;
+
+ /* determine fractional part */
+ {
+ unsigned int i = FIXED1_8_BITS_PER_FRACTIONAL_PART;
+
+ do {
+ remainder <<= 1;
+
+ res_value <<= 1;
+
+ if (remainder >= arg2_value) {
+ res_value |= 1;
+ remainder -= arg2_value;
+ }
+ } while (--i != 0);
+ }
+
+ /* round up LSB */
+ {
+ uint16_t summand = (remainder << 1) >= arg2_value;
+
+ if ((res_value + summand) > 32767 /* SHRT_MAX */)
+ return 0;
+
+ res_value += summand;
+ }
+
+ if (arg1_negative ^ arg2_negative)
+ res_value = -res_value;
+
+ return res_value;
+}
+
+static uint16_t vpe_internal_get_pratio(uint16_t from_frequency, uint16_t to_frequency)
+{
+ uint16_t pratio = vpe_u1_8_from_fraction(from_frequency, to_frequency);
+
+ if (GET_PRATIO_INTEGER_PART(pratio) > 1)
+ pratio = 0;
+
+ return pratio;
+}
+
+/*
+ * VPE has 4 DPM levels from level 0 (lowerest) to 3 (highest),
+ * VPE FW will dynamically decide which level should be used according to current loading.
+ *
+ * Get VPE and SOC clocks from PM, and select the appropriate four clock values,
+ * calculate the ratios of adjusting from one clock to another.
+ * The VPE FW can then request the appropriate frequency from the PMFW.
+ */
+int amdgpu_vpe_configure_dpm(struct amdgpu_vpe *vpe)
+{
+ struct amdgpu_device *adev = vpe->ring.adev;
+ uint32_t dpm_ctl;
+
+ if (adev->pm.dpm_enabled) {
+ struct dpm_clocks clock_table = { 0 };
+ struct dpm_clock *VPEClks;
+ struct dpm_clock *SOCClks;
+ uint32_t idx;
+ uint32_t pratio_vmax_vnorm = 0, pratio_vnorm_vmid = 0, pratio_vmid_vmin = 0;
+ uint16_t pratio_vmin_freq = 0, pratio_vmid_freq = 0, pratio_vnorm_freq = 0, pratio_vmax_freq = 0;
+
+ dpm_ctl = RREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_enable));
+ dpm_ctl |= 1; /* DPM enablement */
+ WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_enable), dpm_ctl);
+
+ /* Get VPECLK and SOCCLK */
+ if (amdgpu_dpm_get_dpm_clock_table(adev, &clock_table)) {
+ dev_dbg(adev->dev, "%s: get clock failed!\n", __func__);
+ goto disable_dpm;
+ }
+
+ SOCClks = clock_table.SocClocks;
+ VPEClks = clock_table.VPEClocks;
+
+ /* vpe dpm only cares 4 levels. */
+ for (idx = 0; idx < VPE_MAX_DPM_LEVEL; idx++) {
+ uint32_t soc_dpm_level;
+ uint32_t min_freq;
+
+ if (idx == 0)
+ soc_dpm_level = 0;
+ else
+ soc_dpm_level = (idx * 2) + 1;
+
+ /* clamp the max level */
+ if (soc_dpm_level > PP_SMU_NUM_VPECLK_DPM_LEVELS - 1)
+ soc_dpm_level = PP_SMU_NUM_VPECLK_DPM_LEVELS - 1;
+
+ min_freq = (SOCClks[soc_dpm_level].Freq < VPEClks[soc_dpm_level].Freq) ?
+ SOCClks[soc_dpm_level].Freq : VPEClks[soc_dpm_level].Freq;
+
+ switch (idx) {
+ case 0:
+ pratio_vmin_freq = min_freq;
+ break;
+ case 1:
+ pratio_vmid_freq = min_freq;
+ break;
+ case 2:
+ pratio_vnorm_freq = min_freq;
+ break;
+ case 3:
+ pratio_vmax_freq = min_freq;
+ break;
+ default:
+ break;
+ }
+ }
+
+ if (pratio_vmin_freq && pratio_vmid_freq && pratio_vnorm_freq && pratio_vmax_freq) {
+ uint32_t pratio_ctl;
+
+ pratio_vmax_vnorm = (uint32_t)vpe_internal_get_pratio(pratio_vmax_freq, pratio_vnorm_freq);
+ pratio_vnorm_vmid = (uint32_t)vpe_internal_get_pratio(pratio_vnorm_freq, pratio_vmid_freq);
+ pratio_vmid_vmin = (uint32_t)vpe_internal_get_pratio(pratio_vmid_freq, pratio_vmin_freq);
+
+ pratio_ctl = pratio_vmax_vnorm | (pratio_vnorm_vmid << 9) | (pratio_vmid_vmin << 18);
+ WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_pratio), pratio_ctl); /* PRatio */
+ WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_request_interval), 24000); /* 1ms, unit=1/24MHz */
+ WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_decision_threshold), 1200000); /* 50ms */
+ WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_busy_clamp_threshold), 1200000);/* 50ms */
+ WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_idle_clamp_threshold), 1200000);/* 50ms */
+ dev_dbg(adev->dev, "%s: configure vpe dpm pratio done!\n", __func__);
+ } else {
+ dev_dbg(adev->dev, "%s: invalid pratio parameters!\n", __func__);
+ goto disable_dpm;
+ }
+ }
+ return 0;
+
+disable_dpm:
+ dpm_ctl = RREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_enable));
+ dpm_ctl &= 0xfffffffe; /* Disable DPM */
+ WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_enable), dpm_ctl);
+ dev_dbg(adev->dev, "%s: disable vpe dpm\n", __func__);
+ return 0;
+}
+
int amdgpu_vpe_psp_update_sram(struct amdgpu_device *adev)
{
struct amdgpu_firmware_info ucode = {
@@ -134,6 +313,20 @@ static int vpe_early_init(void *handle)
return 0;
}
+static void vpe_idle_work_handler(struct work_struct *work)
+{
+ struct amdgpu_device *adev =
+ container_of(work, struct amdgpu_device, vpe.idle_work.work);
+ struct amdgpu_vpe *vpe = &adev->vpe;
+ unsigned int fences = 0;
+
+ fences += amdgpu_fence_count_emitted(&adev->vpe.ring);
+
+ if (fences == 0)
+ amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VPE, AMD_PG_STATE_GATE);
+ else
+ schedule_delayed_work(&adev->vpe.idle_work, VPE_IDLE_TIMEOUT);
+}
static int vpe_common_init(struct amdgpu_vpe *vpe)
{
@@ -150,6 +343,9 @@ static int vpe_common_init(struct amdgpu_vpe *vpe)
return r;
}
+ vpe->context_started = false;
+ INIT_DELAYED_WORK(&adev->vpe.idle_work, vpe_idle_work_handler);
+
return 0;
}
@@ -219,6 +415,9 @@ static int vpe_hw_fini(void *handle)
vpe_ring_stop(vpe);
+ /* Power off VPE */
+ amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VPE, AMD_PG_STATE_GATE);
+
return 0;
}
@@ -226,6 +425,8 @@ static int vpe_suspend(void *handle)
{
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+ cancel_delayed_work_sync(&adev->vpe.idle_work);
+
return vpe_hw_fini(adev);
}
@@ -430,6 +631,21 @@ static int vpe_set_clockgating_state(void *handle,
static int vpe_set_powergating_state(void *handle,
enum amd_powergating_state state)
{
+ struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+ struct amdgpu_vpe *vpe = &adev->vpe;
+
+ if (!adev->pm.dpm_enabled)
+ dev_err(adev->dev, "Without PM, cannot support powergating\n");
+
+ dev_dbg(adev->dev, "%s: %s!\n", __func__, (state == AMD_PG_STATE_GATE) ? "GATE":"UNGATE");
+
+ if (state == AMD_PG_STATE_GATE) {
+ amdgpu_dpm_enable_vpe(adev, false);
+ vpe->context_started = false;
+ } else {
+ amdgpu_dpm_enable_vpe(adev, true);
+ }
+
return 0;
}
@@ -595,6 +811,38 @@ static int vpe_ring_test_ib(struct amdgpu_ring *ring, long timeout)
return ret;
}
+static void vpe_ring_begin_use(struct amdgpu_ring *ring)
+{
+ struct amdgpu_device *adev = ring->adev;
+ struct amdgpu_vpe *vpe = &adev->vpe;
+
+ cancel_delayed_work_sync(&adev->vpe.idle_work);
+
+ /* Power on VPE and notify VPE of new context */
+ if (!vpe->context_started) {
+ uint32_t context_notify;
+
+ /* Power on VPE */
+ amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VPE, AMD_PG_STATE_UNGATE);
+
+ /* Indicates that a job from a new context has been submitted. */
+ context_notify = RREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.context_indicator));
+ if ((context_notify & 0x1) == 0)
+ context_notify |= 0x1;
+ else
+ context_notify &= ~(0x1);
+ WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.context_indicator), context_notify);
+ vpe->context_started = true;
+ }
+}
+
+static void vpe_ring_end_use(struct amdgpu_ring *ring)
+{
+ struct amdgpu_device *adev = ring->adev;
+
+ schedule_delayed_work(&adev->vpe.idle_work, VPE_IDLE_TIMEOUT);
+}
+
static const struct amdgpu_ring_funcs vpe_ring_funcs = {
.type = AMDGPU_RING_TYPE_VPE,
.align_mask = 0xf,
@@ -625,6 +873,8 @@ static const struct amdgpu_ring_funcs vpe_ring_funcs = {
.init_cond_exec = vpe_ring_init_cond_exec,
.patch_cond_exec = vpe_ring_patch_cond_exec,
.preempt_ib = vpe_ring_preempt_ib,
+ .begin_use = vpe_ring_begin_use,
+ .end_use = vpe_ring_end_use,
};
static void vpe_set_ring_funcs(struct amdgpu_device *adev)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.h
index 29d56f7ae4a9..1153ddaea64d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.h
@@ -47,6 +47,15 @@ struct vpe_regs {
uint32_t queue0_rb_wptr_lo;
uint32_t queue0_rb_wptr_hi;
uint32_t queue0_preempt;
+
+ uint32_t dpm_enable;
+ uint32_t dpm_pratio;
+ uint32_t dpm_request_interval;
+ uint32_t dpm_decision_threshold;
+ uint32_t dpm_busy_clamp_threshold;
+ uint32_t dpm_idle_clamp_threshold;
+ uint32_t dpm_request_lv;
+ uint32_t context_indicator;
};
struct amdgpu_vpe {
@@ -63,12 +72,15 @@ struct amdgpu_vpe {
struct amdgpu_bo *cmdbuf_obj;
uint64_t cmdbuf_gpu_addr;
uint32_t *cmdbuf_cpu_addr;
+ struct delayed_work idle_work;
+ bool context_started;
};
int amdgpu_vpe_psp_update_sram(struct amdgpu_device *adev);
int amdgpu_vpe_init_microcode(struct amdgpu_vpe *vpe);
int amdgpu_vpe_ring_init(struct amdgpu_vpe *vpe);
int amdgpu_vpe_ring_fini(struct amdgpu_vpe *vpe);
+int amdgpu_vpe_configure_dpm(struct amdgpu_vpe *vpe);
#define vpe_ring_init(vpe) ((vpe)->funcs->ring_init ? (vpe)->funcs->ring_init((vpe)) : 0)
#define vpe_ring_start(vpe) ((vpe)->funcs->ring_start ? (vpe)->funcs->ring_start((vpe)) : 0)
diff --git a/drivers/gpu/drm/amd/amdgpu/vpe_v6_1.c b/drivers/gpu/drm/amd/amdgpu/vpe_v6_1.c
index 174f13eff575..d20060a51e05 100644
--- a/drivers/gpu/drm/amd/amdgpu/vpe_v6_1.c
+++ b/drivers/gpu/drm/amd/amdgpu/vpe_v6_1.c
@@ -96,6 +96,10 @@ static int vpe_v6_1_load_microcode(struct amdgpu_vpe *vpe)
adev->vpe.cmdbuf_cpu_addr[1] = f32_cntl;
amdgpu_vpe_psp_update_sram(adev);
+
+ /* Config DPM */
+ amdgpu_vpe_configure_dpm(vpe);
+
return 0;
}
@@ -128,6 +132,8 @@ static int vpe_v6_1_load_microcode(struct amdgpu_vpe *vpe)
}
vpe_v6_1_halt(vpe, false);
+ /* Config DPM */
+ amdgpu_vpe_configure_dpm(vpe);
return 0;
}
@@ -264,6 +270,15 @@ static int vpe_v6_1_set_regs(struct amdgpu_vpe *vpe)
vpe->regs.queue0_rb_wptr_hi = regVPEC_QUEUE0_RB_WPTR_HI;
vpe->regs.queue0_preempt = regVPEC_QUEUE0_PREEMPT;
+ vpe->regs.dpm_enable = regVPEC_PUB_DUMMY2;
+ vpe->regs.dpm_pratio = regVPEC_QUEUE6_DUMMY4;
+ vpe->regs.dpm_request_interval = regVPEC_QUEUE5_DUMMY3;
+ vpe->regs.dpm_decision_threshold = regVPEC_QUEUE5_DUMMY4;
+ vpe->regs.dpm_busy_clamp_threshold = regVPEC_QUEUE7_DUMMY2;
+ vpe->regs.dpm_idle_clamp_threshold = regVPEC_QUEUE7_DUMMY3;
+ vpe->regs.dpm_request_lv = regVPEC_QUEUE7_DUMMY1;
+ vpe->regs.context_indicator = regVPEC_QUEUE6_DUMMY3;
+
return 0;
}
--
2.34.1
More information about the amd-gfx
mailing list