summaryrefslogtreecommitdiffstats
path: root/src/mesa/drivers
diff options
context:
space:
mode:
authorEric Anholt <eric@anholt.net>2014-04-25 16:39:18 -0700
committerEric Anholt <eric@anholt.net>2014-04-30 14:33:21 -0700
commit60c5f9716c90090f41e0796e93cc60a297fa883b (patch)
tree28e36c42c0eea0089e09a00e8a186e287470bea3 /src/mesa/drivers
parent973345fc2389e7f74b8732052e781672c4e135c2 (diff)
downloadexternal_mesa3d-60c5f9716c90090f41e0796e93cc60a297fa883b.zip
external_mesa3d-60c5f9716c90090f41e0796e93cc60a297fa883b.tar.gz
external_mesa3d-60c5f9716c90090f41e0796e93cc60a297fa883b.tar.bz2
i965: Track the number of samples in the drawbuffer.
This keeps us from having to emit the nonpipelined state packet on every FBO binding. -4.42003% +/- 1.09961% effect on cairo-perf-trace runtime on glamor (n=110). Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Diffstat (limited to 'src/mesa/drivers')
-rw-r--r--src/mesa/drivers/dri/i965/brw_context.h8
-rw-r--r--src/mesa/drivers/dri/i965/brw_state_upload.c6
-rw-r--r--src/mesa/drivers/dri/i965/gen6_multisample_state.c17
-rw-r--r--src/mesa/drivers/dri/i965/gen8_multisample_state.c12
-rw-r--r--src/mesa/drivers/dri/i965/gen8_ps_state.c8
5 files changed, 29 insertions, 22 deletions
diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h
index 82b38fc..1f7108f 100644
--- a/src/mesa/drivers/dri/i965/brw_context.h
+++ b/src/mesa/drivers/dri/i965/brw_context.h
@@ -181,6 +181,7 @@ enum brw_state_id {
BRW_STATE_META_IN_PROGRESS,
BRW_STATE_INTERPOLATION_MAP,
BRW_STATE_PUSH_CONSTANT_ALLOCATION,
+ BRW_STATE_NUM_SAMPLES,
BRW_NUM_STATE_BITS
};
@@ -220,6 +221,7 @@ enum brw_state_id {
#define BRW_NEW_META_IN_PROGRESS (1 << BRW_STATE_META_IN_PROGRESS)
#define BRW_NEW_INTERPOLATION_MAP (1 << BRW_STATE_INTERPOLATION_MAP)
#define BRW_NEW_PUSH_CONSTANT_ALLOCATION (1 << BRW_STATE_PUSH_CONSTANT_ALLOCATION)
+#define BRW_NEW_NUM_SAMPLES (1 << BRW_STATE_NUM_SAMPLES)
struct brw_state_flags {
/** State update flags signalled by mesa internals */
@@ -1164,6 +1166,12 @@ struct brw_context
const struct gl_geometry_program *geometry_program;
const struct gl_fragment_program *fragment_program;
+ /**
+ * Number of samples in ctx->DrawBuffer, updated by BRW_NEW_NUM_SAMPLES so
+ * that we don't have to reemit that state every time we change FBOs.
+ */
+ int num_samples;
+
/* hw-dependent 3DSTATE_VF_STATISTICS opcode */
uint32_t CMD_VF_STATISTICS;
/* hw-dependent 3DSTATE_PIPELINE_SELECT opcode */
diff --git a/src/mesa/drivers/dri/i965/brw_state_upload.c b/src/mesa/drivers/dri/i965/brw_state_upload.c
index 7ae2e1e..3a452c3 100644
--- a/src/mesa/drivers/dri/i965/brw_state_upload.c
+++ b/src/mesa/drivers/dri/i965/brw_state_upload.c
@@ -507,6 +507,7 @@ static struct dirty_bit_map brw_bits[] = {
DEFINE_BIT(BRW_NEW_META_IN_PROGRESS),
DEFINE_BIT(BRW_NEW_INTERPOLATION_MAP),
DEFINE_BIT(BRW_NEW_PUSH_CONSTANT_ALLOCATION),
+ DEFINE_BIT(BRW_NEW_NUM_SAMPLES),
{0, 0, 0}
};
@@ -604,6 +605,11 @@ void brw_upload_state(struct brw_context *brw)
brw->state.dirty.brw |= BRW_NEW_META_IN_PROGRESS;
}
+ if (brw->num_samples != ctx->DrawBuffer->Visual.samples) {
+ brw->num_samples = ctx->DrawBuffer->Visual.samples;
+ brw->state.dirty.brw |= BRW_NEW_NUM_SAMPLES;
+ }
+
if ((state->mesa | state->cache | state->brw) == 0)
return;
diff --git a/src/mesa/drivers/dri/i965/gen6_multisample_state.c b/src/mesa/drivers/dri/i965/gen6_multisample_state.c
index 5d65453..701fda2 100644
--- a/src/mesa/drivers/dri/i965/gen6_multisample_state.c
+++ b/src/mesa/drivers/dri/i965/gen6_multisample_state.c
@@ -111,7 +111,8 @@ gen6_determine_sample_mask(struct brw_context *brw)
float coverage_invert = false;
unsigned sample_mask = ~0u;
- unsigned num_samples = ctx->DrawBuffer->Visual.samples;
+ /* BRW_NEW_NUM_SAMPLES */
+ unsigned num_samples = brw->num_samples;
if (ctx->Multisample._Enabled) {
if (ctx->Multisample.SampleCoverage) {
@@ -150,21 +151,17 @@ gen6_emit_3dstate_sample_mask(struct brw_context *brw, unsigned mask)
static void upload_multisample_state(struct brw_context *brw)
{
- struct gl_context *ctx = &brw->ctx;
-
- /* _NEW_BUFFERS, _NEW_MULTISAMPLE */
- unsigned num_samples = ctx->DrawBuffer->Visual.samples;
-
- gen6_emit_3dstate_multisample(brw, num_samples);
+ /* BRW_NEW_NUM_SAMPLES */
+ gen6_emit_3dstate_multisample(brw, brw->num_samples);
gen6_emit_3dstate_sample_mask(brw, gen6_determine_sample_mask(brw));
}
const struct brw_tracked_state gen6_multisample_state = {
.dirty = {
- .mesa = _NEW_BUFFERS |
- _NEW_MULTISAMPLE,
- .brw = BRW_NEW_CONTEXT,
+ .mesa = _NEW_MULTISAMPLE,
+ .brw = (BRW_NEW_CONTEXT |
+ BRW_NEW_NUM_SAMPLES),
.cache = 0
},
.emit = upload_multisample_state
diff --git a/src/mesa/drivers/dri/i965/gen8_multisample_state.c b/src/mesa/drivers/dri/i965/gen8_multisample_state.c
index bfe0d5b..daf3e69 100644
--- a/src/mesa/drivers/dri/i965/gen8_multisample_state.c
+++ b/src/mesa/drivers/dri/i965/gen8_multisample_state.c
@@ -76,19 +76,15 @@ gen8_emit_3dstate_sample_pattern(struct brw_context *brw)
static void
upload_multisample_state(struct brw_context *brw)
{
- struct gl_context *ctx = &brw->ctx;
-
- /* _NEW_BUFFERS, _NEW_MULTISAMPLE */
- unsigned num_samples = ctx->DrawBuffer->Visual.samples;
-
- gen8_emit_3dstate_multisample(brw, num_samples);
+ gen8_emit_3dstate_multisample(brw, brw->num_samples);
gen6_emit_3dstate_sample_mask(brw, gen6_determine_sample_mask(brw));
}
const struct brw_tracked_state gen8_multisample_state = {
.dirty = {
- .mesa = _NEW_BUFFERS | _NEW_MULTISAMPLE,
- .brw = BRW_NEW_CONTEXT,
+ .mesa = _NEW_MULTISAMPLE,
+ .brw = (BRW_NEW_CONTEXT |
+ BRW_NEW_NUM_SAMPLES),
.cache = 0
},
.emit = upload_multisample_state
diff --git a/src/mesa/drivers/dri/i965/gen8_ps_state.c b/src/mesa/drivers/dri/i965/gen8_ps_state.c
index 7d8f954..8d9ad64 100644
--- a/src/mesa/drivers/dri/i965/gen8_ps_state.c
+++ b/src/mesa/drivers/dri/i965/gen8_ps_state.c
@@ -65,8 +65,8 @@ upload_ps_extra(struct brw_context *brw)
if (fp->program.Base.InputsRead & VARYING_BIT_POS)
dw1 |= GEN8_PSX_USES_SOURCE_DEPTH | GEN8_PSX_USES_SOURCE_W;
- /* _NEW_BUFFERS | _NEW_MULTISAMPLE */
- bool multisampled_fbo = ctx->DrawBuffer->Visual.samples > 1;
+ /* BRW_NEW_NUM_SAMPLES | _NEW_MULTISAMPLE */
+ bool multisampled_fbo = brw->num_samples > 1;
if (multisampled_fbo &&
_mesa_get_min_invocations_per_fragment(ctx, &fp->program, false) > 1)
dw1 |= GEN8_PSX_SHADER_IS_PER_SAMPLE;
@@ -85,8 +85,8 @@ upload_ps_extra(struct brw_context *brw)
const struct brw_tracked_state gen8_ps_extra = {
.dirty = {
- .mesa = _NEW_BUFFERS | _NEW_MULTISAMPLE,
- .brw = BRW_NEW_CONTEXT | BRW_NEW_FRAGMENT_PROGRAM,
+ .mesa = _NEW_MULTISAMPLE,
+ .brw = BRW_NEW_CONTEXT | BRW_NEW_FRAGMENT_PROGRAM | BRW_NEW_NUM_SAMPLES,
.cache = 0,
},
.emit = upload_ps_extra,