drm/i915/gvt: Enable async flip on plane surface mmio writes

According to Intel GFX PRM on 01.org, plane surface address can be updated
synchronously or asynchronously. Synchronous flip will hold plane surface
address update to start of next vsync, which is current implementation.
Asynchronous flip will update the address as soon as possible. Without
async flip, some 3D application could not reach better performance and
the maximum performance is no higher than vsync frequency.

The patch enables the async flip on plane surface address mmio update,
and increment flip count correctly.

With async flip enabled, some 3D applications have significant performance
improvement. i.e. 3DMark Ice Storm has a 300%~400% increment on score.

v2:
Use bit operation definition for flip mode. (zhenyu)

Reviewed-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Colin Xu <colin.xu@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

2 files changed, 55 insertions, 19 deletions

diff --git a/drivers/gpu/drm/i915/gvt/display.c b/drivers/gpu/drm/i915/gvt/display.c
index 035479e273be..d7536e41aff9 100644
--- a/drivers/gpu/drm/i915/gvt/display.c
+++ b/drivers/gpu/drm/i915/gvt/display.c
@@ -407,7 +407,6 @@ static void emulate_vblank_on_pipe(struct intel_vgpu *vgpu, int pipe)
 		if (!pipe_is_enabled(vgpu, pipe))
 			continue;
 
-		vgpu_vreg_t(vgpu, PIPE_FLIPCOUNT_G4X(pipe))++;
 		intel_vgpu_trigger_virtual_event(vgpu, event);
 	}
 
diff --git a/drivers/gpu/drm/i915/gvt/handlers.c b/drivers/gpu/drm/i915/gvt/handlers.c
index de7fb67b5fa7..1a343f99b3b4 100644
--- a/drivers/gpu/drm/i915/gvt/handlers.c
+++ b/drivers/gpu/drm/i915/gvt/handlers.c
@@ -750,18 +750,19 @@ static int pri_surf_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
 		void *p_data, unsigned int bytes)
 {
 	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
-	unsigned int index = DSPSURF_TO_PIPE(offset);
-	i915_reg_t surflive_reg = DSPSURFLIVE(index);
-	int flip_event[] = {
-		[PIPE_A] = PRIMARY_A_FLIP_DONE,
-		[PIPE_B] = PRIMARY_B_FLIP_DONE,
-		[PIPE_C] = PRIMARY_C_FLIP_DONE,
-	};
+	u32 pipe = DSPSURF_TO_PIPE(offset);
+	int event = SKL_FLIP_EVENT(pipe, PLANE_PRIMARY);
 
 	write_vreg(vgpu, offset, p_data, bytes);
-	vgpu_vreg_t(vgpu, surflive_reg) = vgpu_vreg(vgpu, offset);
+	vgpu_vreg_t(vgpu, DSPSURFLIVE(pipe)) = vgpu_vreg(vgpu, offset);
+
+	vgpu_vreg_t(vgpu, PIPE_FLIPCOUNT_G4X(pipe))++;
+
+	if (vgpu_vreg_t(vgpu, DSPCNTR(pipe)) & PLANE_CTL_ASYNC_FLIP)
+		intel_vgpu_trigger_virtual_event(vgpu, event);
+	else
+		set_bit(event, vgpu->irq.flip_done_event[pipe]);
 
-	set_bit(flip_event[index], vgpu->irq.flip_done_event[index]);
 	return 0;
 }
 
@@ -771,18 +772,42 @@ static int pri_surf_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
 static int spr_surf_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
 		void *p_data, unsigned int bytes)
 {
-	unsigned int index = SPRSURF_TO_PIPE(offset);
-	i915_reg_t surflive_reg = SPRSURFLIVE(index);
-	int flip_event[] = {
-		[PIPE_A] = SPRITE_A_FLIP_DONE,
-		[PIPE_B] = SPRITE_B_FLIP_DONE,
-		[PIPE_C] = SPRITE_C_FLIP_DONE,
-	};
+	u32 pipe = SPRSURF_TO_PIPE(offset);
+	int event = SKL_FLIP_EVENT(pipe, PLANE_SPRITE0);
 
 	write_vreg(vgpu, offset, p_data, bytes);
-	vgpu_vreg_t(vgpu, surflive_reg) = vgpu_vreg(vgpu, offset);
+	vgpu_vreg_t(vgpu, SPRSURFLIVE(pipe)) = vgpu_vreg(vgpu, offset);
+
+	if (vgpu_vreg_t(vgpu, SPRCTL(pipe)) & PLANE_CTL_ASYNC_FLIP)
+		intel_vgpu_trigger_virtual_event(vgpu, event);
+	else
+		set_bit(event, vgpu->irq.flip_done_event[pipe]);
+
+	return 0;
+}
+
+static int reg50080_mmio_write(struct intel_vgpu *vgpu,
+			       unsigned int offset, void *p_data,
+			       unsigned int bytes)
+{
+	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
+	enum pipe pipe = REG_50080_TO_PIPE(offset);
+	enum plane_id plane = REG_50080_TO_PLANE(offset);
+	int event = SKL_FLIP_EVENT(pipe, plane);
+
+	write_vreg(vgpu, offset, p_data, bytes);
+	if (plane == PLANE_PRIMARY) {
+		vgpu_vreg_t(vgpu, DSPSURFLIVE(pipe)) = vgpu_vreg(vgpu, offset);
+		vgpu_vreg_t(vgpu, PIPE_FLIPCOUNT_G4X(pipe))++;
+	} else {
+		vgpu_vreg_t(vgpu, SPRSURFLIVE(pipe)) = vgpu_vreg(vgpu, offset);
+	}
+
+	if ((vgpu_vreg(vgpu, offset) & REG50080_FLIP_TYPE_MASK) == REG50080_FLIP_TYPE_ASYNC)
+		intel_vgpu_trigger_virtual_event(vgpu, event);
+	else
+		set_bit(event, vgpu->irq.flip_done_event[pipe]);
 
-	set_bit(flip_event[index], vgpu->irq.flip_done_event[index]);
 	return 0;
 }
 
@@ -1969,6 +1994,8 @@ static int init_generic_mmio_info(struct intel_gvt *gvt)
 	MMIO_DH(DSPSURF(PIPE_A), D_ALL, NULL, pri_surf_mmio_write);
 	MMIO_D(DSPOFFSET(PIPE_A), D_ALL);
 	MMIO_D(DSPSURFLIVE(PIPE_A), D_ALL);
+	MMIO_DH(REG_50080(PIPE_A, PLANE_PRIMARY), D_ALL, NULL,
+		reg50080_mmio_write);
 
 	MMIO_D(DSPCNTR(PIPE_B), D_ALL);
 	MMIO_D(DSPADDR(PIPE_B), D_ALL);
@@ -1978,6 +2005,8 @@ static int init_generic_mmio_info(struct intel_gvt *gvt)
 	MMIO_DH(DSPSURF(PIPE_B), D_ALL, NULL, pri_surf_mmio_write);
 	MMIO_D(DSPOFFSET(PIPE_B), D_ALL);
 	MMIO_D(DSPSURFLIVE(PIPE_B), D_ALL);
+	MMIO_DH(REG_50080(PIPE_B, PLANE_PRIMARY), D_ALL, NULL,
+		reg50080_mmio_write);
 
 	MMIO_D(DSPCNTR(PIPE_C), D_ALL);
 	MMIO_D(DSPADDR(PIPE_C), D_ALL);
@@ -1987,6 +2016,8 @@ static int init_generic_mmio_info(struct intel_gvt *gvt)
 	MMIO_DH(DSPSURF(PIPE_C), D_ALL, NULL, pri_surf_mmio_write);
 	MMIO_D(DSPOFFSET(PIPE_C), D_ALL);
 	MMIO_D(DSPSURFLIVE(PIPE_C), D_ALL);
+	MMIO_DH(REG_50080(PIPE_C, PLANE_PRIMARY), D_ALL, NULL,
+		reg50080_mmio_write);
 
 	MMIO_D(SPRCTL(PIPE_A), D_ALL);
 	MMIO_D(SPRLINOFF(PIPE_A), D_ALL);
@@ -2000,6 +2031,8 @@ static int init_generic_mmio_info(struct intel_gvt *gvt)
 	MMIO_D(SPROFFSET(PIPE_A), D_ALL);
 	MMIO_D(SPRSCALE(PIPE_A), D_ALL);
 	MMIO_D(SPRSURFLIVE(PIPE_A), D_ALL);
+	MMIO_DH(REG_50080(PIPE_A, PLANE_SPRITE0), D_ALL, NULL,
+		reg50080_mmio_write);
 
 	MMIO_D(SPRCTL(PIPE_B), D_ALL);
 	MMIO_D(SPRLINOFF(PIPE_B), D_ALL);
@@ -2013,6 +2046,8 @@ static int init_generic_mmio_info(struct intel_gvt *gvt)
 	MMIO_D(SPROFFSET(PIPE_B), D_ALL);
 	MMIO_D(SPRSCALE(PIPE_B), D_ALL);
 	MMIO_D(SPRSURFLIVE(PIPE_B), D_ALL);
+	MMIO_DH(REG_50080(PIPE_B, PLANE_SPRITE0), D_ALL, NULL,
+		reg50080_mmio_write);
 
 	MMIO_D(SPRCTL(PIPE_C), D_ALL);
 	MMIO_D(SPRLINOFF(PIPE_C), D_ALL);
@@ -2026,6 +2061,8 @@ static int init_generic_mmio_info(struct intel_gvt *gvt)
 	MMIO_D(SPROFFSET(PIPE_C), D_ALL);
 	MMIO_D(SPRSCALE(PIPE_C), D_ALL);
 	MMIO_D(SPRSURFLIVE(PIPE_C), D_ALL);
+	MMIO_DH(REG_50080(PIPE_C, PLANE_SPRITE0), D_ALL, NULL,
+		reg50080_mmio_write);
 
 	MMIO_D(HTOTAL(TRANSCODER_A), D_ALL);
 	MMIO_D(HBLANK(TRANSCODER_A), D_ALL);