It finishes the work started by 0c2eed0edec, quoting:
The main idea is to avoid setting CB_COLORi_INFO = 0 for i>0
repeatedly
when those colorbuffers aren't used. This is mainly for glamor.
After 0c2eed0edec the r600g was passing null as the dirty_cbufs
pointer,
and the current patch makes the actual use of the argument. In
particular, the check for null dirty_cbufs pointer in
evergreen_do_fast_color_clear() is removed.
I didn't change r600_state.c because evergreen_do_fast_color_clear() is
only used for eg and higher, thus checks in there are useless.
In addition to radeonsi-like modifications I also do reset in
r600_bind_blend_state_internal(), otherwise there're blend
arb_blend_func_extended regressions. Besides, I keep dirty_cbufs in
cb_misc_state, because it rather seem belonging there.
Some statistics: skips ≈70 updates of CB registers per frame in
Kane&Lynch2.
Signed-off-by: Constantine Kharlamov <hi-an...@yandex.ru>
---
src/gallium/drivers/r600/evergreen_state.c | 15 +++++++++++++--
src/gallium/drivers/r600/r600_blit.c | 3 ++-
src/gallium/drivers/r600/r600_hw_context.c | 1 +
src/gallium/drivers/r600/r600_pipe.h | 1 +
src/gallium/drivers/r600/r600_state_common.c | 1 +
src/gallium/drivers/radeon/r600_texture.c | 3 +--
6 files changed, 19 insertions(+), 5 deletions(-)
diff --git a/src/gallium/drivers/r600/evergreen_state.c
b/src/gallium/drivers/r600/evergreen_state.c
index 9a6bb4e4c9..916270ec4b 100644
--- a/src/gallium/drivers/r600/evergreen_state.c
+++ b/src/gallium/drivers/r600/evergreen_state.c
@@ -1426,6 +1426,11 @@ static void
evergreen_set_framebuffer_state(struct pipe_context *ctx,
R600_CONTEXT_FLUSH_AND_INV_DB_META |
R600_CONTEXT_INV_TEX_CACHE;
+ /* Take the maximum of the old and new count. If the new count is
lower,
+ * dirtying is needed to disable the unbound colorbuffers.
+ */
+ rctx->cb_misc_state.dirty_cbufs |=
+ (1 << MAX2(rctx->framebuffer.state.nr_cbufs, state->nr_cbufs))
- 1;
rctx->framebuffer.dirty_zsbuf |= rctx->framebuffer.state.zsbuf !=
state->zsbuf;
util_copy_framebuffer_state(&rctx->framebuffer.state, state);
@@ -1684,6 +1689,9 @@ static void
evergreen_emit_framebuffer_state(struct r600_context *rctx, struct r
for (i = 0; i < nr_cbufs; i++) {
unsigned reloc, cmask_reloc;
+ if (!(rctx->cb_misc_state.dirty_cbufs & (1 << i)))
+ continue;
+
cb = (struct r600_surface*)state->cbufs[i];
if (!cb) {
radeon_set_context_reg(cs, R_028C70_CB_COLOR0_INFO + i
* 0x3C,
@@ -1736,13 +1744,15 @@ static void
evergreen_emit_framebuffer_state(struct r600_context *rctx, struct r
radeon_emit(cs, reloc);
}
/* set CB_COLOR1_INFO for possible dual-src blending */
- if (rctx->framebuffer.dual_src_blend && i == 1 && state->cbufs[0]) {
+ if (rctx->framebuffer.dual_src_blend && i == 1 && state->cbufs[0] &&
+ rctx->cb_misc_state.dirty_cbufs & (1 << 0)) {
radeon_set_context_reg(cs, R_028C70_CB_COLOR0_INFO + 1 * 0x3C,
cb->cb_color_info | tex->cb_color_info);
i++;
}
for (; i < 8 ; i++)
- radeon_set_context_reg(cs, R_028C70_CB_COLOR0_INFO + i * 0x3C,
0);
+ if (rctx->cb_misc_state.dirty_cbufs & (1 << i))
+ radeon_set_context_reg(cs, R_028C70_CB_COLOR0_INFO + i
* 0x3C, 0);
for (; i < 12; i++)
radeon_set_context_reg(cs, R_028E50_CB_COLOR8_INFO + (i - 8) * 0x1C,
0);
@@ -1810,6 +1820,7 @@ static void
evergreen_emit_framebuffer_state(struct r600_context *rctx, struct r
rctx->ps_iter_samples, 0,
sc_mode_cntl_1);
}
rctx->framebuffer.dirty_zsbuf = false;
+ rctx->cb_misc_state.dirty_cbufs = 0;
}
static void evergreen_emit_polygon_offset(struct r600_context *rctx,
struct r600_atom *a)
diff --git a/src/gallium/drivers/r600/r600_blit.c
b/src/gallium/drivers/r600/r600_blit.c
index 62035a7bee..a4434b77c2 100644
--- a/src/gallium/drivers/r600/r600_blit.c
+++ b/src/gallium/drivers/r600/r600_blit.c
@@ -407,7 +407,8 @@ static void r600_clear(struct pipe_context *ctx,
unsigned buffers,
if (buffers & PIPE_CLEAR_COLOR && rctx->b.chip_class >= EVERGREEN) {
evergreen_do_fast_color_clear(&rctx->b, fb,
&rctx->framebuffer.atom,
- &buffers, NULL, color);
+ &buffers,
+ &rctx->cb_misc_state.dirty_cbufs,
color);
if (!buffers)
return; /* all buffers have been fast cleared */
}
diff --git a/src/gallium/drivers/r600/r600_hw_context.c
b/src/gallium/drivers/r600/r600_hw_context.c
index c85e346307..4054896393 100644
--- a/src/gallium/drivers/r600/r600_hw_context.c
+++ b/src/gallium/drivers/r600/r600_hw_context.c
@@ -375,6 +375,7 @@ void r600_begin_new_cs(struct r600_context *ctx)
assert(!ctx->b.gfx.cs->prev_dw);
ctx->b.initial_gfx_cs_size = ctx->b.gfx.cs->current.cdw;
ctx->framebuffer.dirty_zsbuf = true;
+ ctx->cb_misc_state.dirty_cbufs = (1 << 8) - 1;
}
void r600_emit_pfp_sync_me(struct r600_context *rctx)
diff --git a/src/gallium/drivers/r600/r600_pipe.h
b/src/gallium/drivers/r600/r600_pipe.h
index aeba1f2635..8f1c1a8095 100644
--- a/src/gallium/drivers/r600/r600_pipe.h
+++ b/src/gallium/drivers/r600/r600_pipe.h
@@ -133,6 +133,7 @@ struct r600_cb_misc_state {
unsigned blend_colormask; /* 8*4 bits for 8 RGBA colorbuffers */
unsigned nr_cbufs;
unsigned nr_ps_color_outputs;
+ unsigned dirty_cbufs;
bool multiwrite;
bool dual_src_blend;
};
diff --git a/src/gallium/drivers/r600/r600_state_common.c
b/src/gallium/drivers/r600/r600_state_common.c
index ffb0f5a811..7716c83f84 100644
--- a/src/gallium/drivers/r600/r600_state_common.c
+++ b/src/gallium/drivers/r600/r600_state_common.c
@@ -187,6 +187,7 @@ static void r600_bind_blend_state_internal(struct
r600_context *rctx,
}
if (update_cb) {
r600_mark_atom_dirty(rctx, &rctx->cb_misc_state.atom);
+ rctx->cb_misc_state.dirty_cbufs |= (1 <<
rctx->framebuffer.state.nr_cbufs) - 1;
}
if (rctx->framebuffer.dual_src_blend != blend->dual_src_blend) {
rctx->framebuffer.dual_src_blend = blend->dual_src_blend;
diff --git a/src/gallium/drivers/radeon/r600_texture.c
b/src/gallium/drivers/radeon/r600_texture.c
index 8110ec1484..ca8c75639a 100644
--- a/src/gallium/drivers/radeon/r600_texture.c
+++ b/src/gallium/drivers/radeon/r600_texture.c
@@ -2753,8 +2753,7 @@ void evergreen_do_fast_color_clear(struct
r600_common_context *rctx,
evergreen_set_clear_color(tex, fb->cbufs[i]->format, color);
- if (dirty_cbufs)
- *dirty_cbufs |= 1 << i;
+ *dirty_cbufs |= 1 << i;
rctx->set_atom_dirty(rctx, fb_state, true);
*buffers &= ~clear_bit;
}
