[Intel-gfx] [PATCH 2/3] drm/i915: Decode system memory bandwidth
Mahesh Kumar
mahesh1.kumar at intel.com
Wed Feb 15 14:48:39 UTC 2017
This patch adds support to decode system memory bandwidth
which will be used for arbitrated display memory percentage
calculation in GEN9 based system.
Changes from v1:
- Address comments from Paulo
- implement decode function for SKL/KBL also
Changes from v2:
- Rewrite the code as per HW team inputs
- Addresses review comments
Changes from v3:
- Fix compilation warning
Changes from v4:
- Address review comments from Paulo
- Round-off the frequency & bandwidth results (Paulo)
Signed-off-by: Mahesh Kumar <mahesh1.kumar at intel.com>
---
drivers/gpu/drm/i915/i915_drv.c | 179 ++++++++++++++++++++++++++++++++++++++++
drivers/gpu/drm/i915/i915_drv.h | 12 +++
drivers/gpu/drm/i915/i915_reg.h | 31 +++++++
3 files changed, 222 insertions(+)
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index 4c1cb067ca33..9203dad1db07 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -997,6 +997,179 @@ static void intel_sanitize_options(struct drm_i915_private *dev_priv)
DRM_DEBUG_DRIVER("use GPU sempahores? %s\n", yesno(i915.semaphores));
}
+static enum memdev_rank
+skl_memdev_get_channel_rank(uint32_t val)
+{
+ uint8_t l_rank, s_rank;
+ uint8_t l_size, s_size;
+
+ if (!val)
+ return I915_DRAM_RANK_INVALID;
+
+ l_size = (val >> SKL_DRAM_SIZE_L_SHIFT) & SKL_DRAM_SIZE_MASK;
+ s_size = (val >> SKL_DRAM_SIZE_S_SHIFT) & SKL_DRAM_SIZE_MASK;
+ l_rank = (val >> SKL_DRAM_RANK_L_SHIFT) & SKL_DRAM_RANK_MASK;
+ s_rank = (val >> SKL_DRAM_RANK_S_SHIFT) & SKL_DRAM_RANK_MASK;
+
+ if ((l_size == 0) && (s_size == 0))
+ return I915_DRAM_RANK_INVALID;
+
+ if (l_rank == SKL_DRAM_RANK_DUAL || s_rank == SKL_DRAM_RANK_DUAL)
+ return I915_DRAM_RANK_DUAL;
+
+ if ((l_size && l_rank == SKL_DRAM_RANK_SINGLE) &&
+ (s_size && s_rank == SKL_DRAM_RANK_SINGLE))
+ return I915_DRAM_RANK_DUAL;
+ return I915_DRAM_RANK_SINGLE;
+}
+
+static int
+skl_get_memdev_info(struct drm_i915_private *dev_priv)
+{
+ struct memdev_info *memdev_info = &dev_priv->memdev_info;
+ uint32_t mem_freq_khz;
+ uint32_t val;
+ enum memdev_rank ch0_rank, ch1_rank;
+
+ val = I915_READ(SKL_MC_BIOS_DATA_0_0_0_MCHBAR_PCU);
+ mem_freq_khz = DIV_ROUND_UP((val & SKL_REQ_DATA_MASK) *
+ SKL_MEMORY_FREQ_MULTIPLIER_HZ, 1000);
+
+ val = I915_READ(SKL_MAD_DIMM_CH0_0_0_0_MCHBAR_MCMAIN);
+ ch0_rank = skl_memdev_get_channel_rank(val);
+ if (ch0_rank != I915_DRAM_RANK_INVALID)
+ memdev_info->num_channels++;
+
+ val = I915_READ(SKL_MAD_DIMM_CH1_0_0_0_MCHBAR_MCMAIN);
+ ch1_rank = skl_memdev_get_channel_rank(val);
+ if (ch1_rank != I915_DRAM_RANK_INVALID)
+ memdev_info->num_channels++;
+
+ if (memdev_info->num_channels == 0) {
+ DRM_ERROR("Number of memory channels is zero\n");
+ return -EINVAL;
+ }
+
+ memdev_info->bandwidth_kbps = memdev_info->num_channels *
+ mem_freq_khz * 8;
+
+ if (memdev_info->bandwidth_kbps == 0) {
+ DRM_ERROR("Couldn't get system memory bandwidth\n");
+ return -EINVAL;
+ }
+
+ /*
+ * If any of the channel is single rank channel, worst case output
+ * will be same as if single rank memory, so consider single rank
+ * memory.
+ */
+ if (ch0_rank == I915_DRAM_RANK_SINGLE
+ || ch1_rank == I915_DRAM_RANK_SINGLE)
+ memdev_info->rank = I915_DRAM_RANK_SINGLE;
+ else
+ memdev_info->rank = max(ch0_rank, ch1_rank);
+
+ if (memdev_info->rank == I915_DRAM_RANK_INVALID) {
+ DRM_ERROR("couldn't get memory rank information\n");
+ return -EINVAL;
+ }
+
+ memdev_info->valid = true;
+ return 0;
+}
+
+static int
+bxt_get_memdev_info(struct drm_i915_private *dev_priv)
+{
+ struct memdev_info *memdev_info = &dev_priv->memdev_info;
+ uint32_t dram_channels;
+ uint32_t mem_freq_khz, val;
+ uint8_t num_active_channels;
+ int i;
+
+ val = I915_READ(BXT_P_CR_MC_BIOS_REQ_0_0_0);
+ mem_freq_khz = DIV_ROUND_UP((val & BXT_REQ_DATA_MASK) *
+ BXT_MEMORY_FREQ_MULTIPLIER_HZ, 1000);
+
+ dram_channels = (val >> BXT_DRAM_CHANNEL_ACTIVE_SHIFT) &
+ BXT_DRAM_CHANNEL_ACTIVE_MASK;
+ num_active_channels = hweight32(dram_channels);
+
+ /* Each active bit represents 4-byte channel */
+ memdev_info->bandwidth_kbps = (mem_freq_khz * num_active_channels * 4);
+
+ if (memdev_info->bandwidth_kbps == 0) {
+ DRM_ERROR("Couldn't get system memory bandwidth\n");
+ return -EINVAL;
+ }
+
+ /*
+ * Now read each DUNIT8/9/10/11 to check the rank of each dimms.
+ */
+ for (i = BXT_D_CR_DRP0_DUNIT_START; i <= BXT_D_CR_DRP0_DUNIT_END; i++) {
+ val = I915_READ(BXT_D_CR_DRP0_DUNIT(i));
+ if (val != 0xFFFFFFFF) {
+ uint8_t rank;
+ enum memdev_rank ch_rank;
+
+ memdev_info->num_channels++;
+ rank = val & BXT_DRAM_RANK_MASK;
+ if (rank == BXT_DRAM_RANK_SINGLE)
+ ch_rank = I915_DRAM_RANK_SINGLE;
+ else if (rank == BXT_DRAM_RANK_DUAL)
+ ch_rank = I915_DRAM_RANK_DUAL;
+ else
+ ch_rank = I915_DRAM_RANK_INVALID;
+
+ /*
+ * If any of the channel is single rank channel,
+ * worst case output will be same as if single rank
+ * memory, so consider single rank memory.
+ */
+ if (memdev_info->rank == I915_DRAM_RANK_INVALID)
+ memdev_info->rank = ch_rank;
+ else if (ch_rank == I915_DRAM_RANK_SINGLE)
+ memdev_info->rank = I915_DRAM_RANK_SINGLE;
+ }
+ }
+
+ if (memdev_info->rank == I915_DRAM_RANK_INVALID) {
+ DRM_ERROR("couldn't get memory rank information\n");
+ return -EINVAL;
+ }
+
+ memdev_info->valid = true;
+ return 0;
+}
+
+static void
+intel_get_memdev_info(struct drm_i915_private *dev_priv)
+{
+ struct memdev_info *memdev_info = &dev_priv->memdev_info;
+ int ret;
+
+ memdev_info->valid = false;
+ memdev_info->rank = I915_DRAM_RANK_INVALID;
+ memdev_info->num_channels = 0;
+
+ if (!IS_GEN9(dev_priv))
+ return;
+
+ if (IS_BROXTON(dev_priv))
+ ret = bxt_get_memdev_info(dev_priv);
+ else
+ ret = skl_get_memdev_info(dev_priv);
+ if (ret)
+ return;
+
+ DRM_DEBUG_DRIVER("DRAM bandwidth: %u KBps, total-channels: %u\n",
+ memdev_info->bandwidth_kbps,
+ memdev_info->num_channels);
+ DRM_DEBUG_DRIVER("DRAM rank: %s rank\n",
+ (memdev_info->rank == I915_DRAM_RANK_DUAL) ?
+ "dual" : "single");
+}
+
/**
* i915_driver_init_hw - setup state requiring device access
* @dev_priv: device private
@@ -1103,6 +1276,12 @@ static int i915_driver_init_hw(struct drm_i915_private *dev_priv)
if (ret)
goto out_ggtt;
+ /*
+ * Fill the memdev structure to get the system raw bandwidth
+ * This will be used by WM algorithm, to implement GEN9 based WA
+ */
+ intel_get_memdev_info(dev_priv);
+
return 0;
out_ggtt:
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 918118a268b8..a88b2721dbe7 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -2384,6 +2384,18 @@ struct drm_i915_private {
bool distrust_bios_wm;
} wm;
+ struct memdev_info {
+ bool valid;
+ uint32_t bandwidth_kbps;
+ uint8_t num_channels;
+ enum memdev_rank {
+ I915_DRAM_RANK_INVALID = 0,
+ I915_DRAM_RANK_SINGLE,
+ I915_DRAM_RANK_DUAL
+ } rank;
+ } memdev_info;
+
+
struct i915_runtime_pm pm;
struct {
diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
index 2ee26dc67053..8a3898c81a26 100644
--- a/drivers/gpu/drm/i915/i915_reg.h
+++ b/drivers/gpu/drm/i915/i915_reg.h
@@ -8110,6 +8110,37 @@ enum {
#define DC_STATE_DEBUG_MASK_CORES (1<<0)
#define DC_STATE_DEBUG_MASK_MEMORY_UP (1<<1)
+#define BXT_P_CR_MC_BIOS_REQ_0_0_0 _MMIO(MCHBAR_MIRROR_BASE_SNB + 0x7114)
+#define BXT_REQ_DATA_MASK 0x3F
+#define BXT_DRAM_CHANNEL_ACTIVE_SHIFT 12
+#define BXT_DRAM_CHANNEL_ACTIVE_MASK 0xF
+#define BXT_MEMORY_FREQ_MULTIPLIER_HZ 133333333
+
+#define BXT_D_CR_DRP0_DUNIT8 0x1000
+#define BXT_D_CR_DRP0_DUNIT9 0x1200
+#define BXT_D_CR_DRP0_DUNIT_START 8
+#define BXT_D_CR_DRP0_DUNIT_END 11
+#define _MMIO_MCHBAR_DUNIT(x, a, b) _MMIO(MCHBAR_MIRROR_BASE_SNB + (a) + (x - BXT_D_CR_DRP0_DUNIT_START)*((b)-(a)))
+#define BXT_D_CR_DRP0_DUNIT(x) _MMIO_MCHBAR_DUNIT(x, BXT_D_CR_DRP0_DUNIT8, BXT_D_CR_DRP0_DUNIT9)
+#define BXT_DRAM_RANK_MASK 0x3
+#define BXT_DRAM_RANK_SINGLE 0x1
+#define BXT_DRAM_RANK_DUAL 0x3
+
+#define SKL_MEMORY_FREQ_MULTIPLIER_HZ 266666666
+#define SKL_MC_BIOS_DATA_0_0_0_MCHBAR_PCU _MMIO(MCHBAR_MIRROR_BASE_SNB + 0x5E04)
+#define SKL_REQ_DATA_MASK (0xF << 0)
+
+#define SKL_MAD_DIMM_CH0_0_0_0_MCHBAR_MCMAIN _MMIO(MCHBAR_MIRROR_BASE_SNB + 0x500C)
+#define SKL_MAD_DIMM_CH1_0_0_0_MCHBAR_MCMAIN _MMIO(MCHBAR_MIRROR_BASE_SNB + 0x5010)
+#define SKL_DRAM_SIZE_MASK 0x3F
+#define SKL_DRAM_SIZE_L_SHIFT 0
+#define SKL_DRAM_SIZE_S_SHIFT 16
+#define SKL_DRAM_RANK_MASK 0x1
+#define SKL_DRAM_RANK_L_SHIFT 10
+#define SKL_DRAM_RANK_S_SHIFT 26
+#define SKL_DRAM_RANK_SINGLE 0x0
+#define SKL_DRAM_RANK_DUAL 0x1
+
/* Please see hsw_read_dcomp() and hsw_write_dcomp() before using this register,
* since on HSW we can't write to it using I915_WRITE. */
#define D_COMP_HSW _MMIO(MCHBAR_MIRROR_BASE_SNB + 0x5F0C)
--
2.11.0
More information about the Intel-gfx
mailing list