[Mesa-dev] [PATCH 6/8] radeonsi: use WRITE_DATA for small glBufferSubData sizes
Marek Olšák
maraeo at gmail.com
Sat Jan 19 01:04:21 UTC 2019
On Fri, Jan 18, 2019 at 6:05 PM Bas Nieuwenhuizen <bas at basnieuwenhuizen.nl>
wrote:
> On Fri, Jan 18, 2019 at 5:44 PM Marek Olšák <maraeo at gmail.com> wrote:
> >
> > From: Marek Olšák <marek.olsak at amd.com>
> >
> > ---
> > src/gallium/drivers/radeonsi/si_buffer.c | 27 ++++++++++++++++++++++++
> > src/gallium/drivers/radeonsi/si_pipe.h | 1 +
> > 2 files changed, 28 insertions(+)
> >
> > diff --git a/src/gallium/drivers/radeonsi/si_buffer.c
> b/src/gallium/drivers/radeonsi/si_buffer.c
> > index 4766cf4bdfa..a1e421b8b0d 100644
> > --- a/src/gallium/drivers/radeonsi/si_buffer.c
> > +++ b/src/gallium/drivers/radeonsi/si_buffer.c
> > @@ -16,20 +16,22 @@
> > * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
> EXPRESS OR
> > * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
> MERCHANTABILITY,
> > * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT
> SHALL
> > * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
> > * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
> > * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE
> OR THE
> > * USE OR OTHER DEALINGS IN THE SOFTWARE.
> > */
> >
> > #include "radeonsi/si_pipe.h"
> > +#include "sid.h"
> > +
> > #include "util/u_memory.h"
> > #include "util/u_upload_mgr.h"
> > #include "util/u_transfer.h"
> > #include <inttypes.h>
> > #include <stdio.h>
> >
> > bool si_rings_is_buffer_referenced(struct si_context *sctx,
> > struct pb_buffer *buf,
> > enum radeon_bo_usage usage)
> > {
> > @@ -506,20 +508,38 @@ static void *si_buffer_transfer_map(struct
> pipe_context *ctx,
> > data = si_buffer_map_sync_with_rings(sctx, rbuffer, usage);
> > if (!data) {
> > return NULL;
> > }
> > data += box->x;
> >
> > return si_buffer_get_transfer(ctx, resource, usage, box,
> > ptransfer, data, NULL, 0);
> > }
> >
> > +static void si_buffer_write_data(struct si_context *sctx, struct
> r600_resource *buf,
> > + unsigned offset, unsigned size, const
> void *data)
> > +{
> > + struct radeon_cmdbuf *cs = sctx->gfx_cs;
> > +
> > + si_need_gfx_cs_space(sctx);
> > +
> > + sctx->flags |= SI_CONTEXT_PS_PARTIAL_FLUSH |
> > + SI_CONTEXT_CS_PARTIAL_FLUSH |
> > + si_get_flush_flags(sctx, SI_COHERENCY_SHADER,
> L2_LRU);
> > + si_emit_cache_flush(sctx);
>
> Maybe only do the cache flush if the buffer is referenced by the
> current cmd buffer?
>
> (I'm kinda surprised reading this that we don't do
> DISCARD_WHOLE_RESOURCE if the offset is 0 and the size equal to the
> buffer size. )
>
Thanks, that's a good point. We promote to DISCARD_WHOLE_RESOURCE in
si_buffer_transfer_map. I'll remove the hunk in si_buffer_subdata, because
it doesn't do optimizations that si_buffer_transfer_map does.
Marek
> > +
> > + si_cp_write_data(sctx, buf, offset, size, V_370_TC_L2, V_370_ME,
> data);
> > +
> > + radeon_emit(cs, PKT3(PKT3_PFP_SYNC_ME, 0, 0));
> > + radeon_emit(cs, 0);
> > +}
> > +
> > static void si_buffer_do_flush_region(struct pipe_context *ctx,
> > struct pipe_transfer *transfer,
> > const struct pipe_box *box)
> > {
> > struct si_transfer *stransfer = (struct si_transfer*)transfer;
> > struct r600_resource *rbuffer =
> r600_resource(transfer->resource);
> >
> > if (stransfer->u.staging) {
> > /* Copy the staging buffer into the original one. */
> > si_copy_buffer((struct si_context*)ctx,
> transfer->resource,
> > @@ -568,20 +588,27 @@ static void si_buffer_transfer_unmap(struct
> pipe_context *ctx,
> >
> > static void si_buffer_subdata(struct pipe_context *ctx,
> > struct pipe_resource *buffer,
> > unsigned usage, unsigned offset,
> > unsigned size, const void *data)
> > {
> > struct pipe_transfer *transfer = NULL;
> > struct pipe_box box;
> > uint8_t *map = NULL;
> >
> > + if (size <= SI_TRANSFER_WRITE_DATA_THRESHOLD &&
> > + offset % 4 == 0 && size % 4 == 0 && (uintptr_t)data % 4 ==
> 0) {
> > + si_buffer_write_data((struct si_context*)ctx,
> > + r600_resource(buffer), offset,
> size, data);
> > + return;
> > + }
> > +
> > u_box_1d(offset, size, &box);
> > map = si_buffer_transfer_map(ctx, buffer, 0,
> > PIPE_TRANSFER_WRITE |
> > PIPE_TRANSFER_DISCARD_RANGE |
> > usage,
> > &box, &transfer);
> > if (!map)
> > return;
> >
> > memcpy(map, data, size);
> > diff --git a/src/gallium/drivers/radeonsi/si_pipe.h
> b/src/gallium/drivers/radeonsi/si_pipe.h
> > index 5bd3d9641d2..f79828f3438 100644
> > --- a/src/gallium/drivers/radeonsi/si_pipe.h
> > +++ b/src/gallium/drivers/radeonsi/si_pipe.h
> > @@ -94,20 +94,21 @@
> > #define SI_PREFETCH_ES (1 << 3)
> > #define SI_PREFETCH_GS (1 << 4)
> > #define SI_PREFETCH_VS (1 << 5)
> > #define SI_PREFETCH_PS (1 << 6)
> >
> > #define SI_MAX_BORDER_COLORS 4096
> > #define SI_MAX_VIEWPORTS 16
> > #define SIX_BITS 0x3F
> > #define SI_MAP_BUFFER_ALIGNMENT 64
> > #define SI_MAX_VARIABLE_THREADS_PER_BLOCK 1024
> > +#define SI_TRANSFER_WRITE_DATA_THRESHOLD 64
> >
> > #define SI_RESOURCE_FLAG_TRANSFER (PIPE_RESOURCE_FLAG_DRV_PRIV <<
> 0)
> > #define SI_RESOURCE_FLAG_FLUSHED_DEPTH (PIPE_RESOURCE_FLAG_DRV_PRIV <<
> 1)
> > #define SI_RESOURCE_FLAG_FORCE_MSAA_TILING (PIPE_RESOURCE_FLAG_DRV_PRIV
> << 2)
> > #define SI_RESOURCE_FLAG_DISABLE_DCC (PIPE_RESOURCE_FLAG_DRV_PRIV <<
> 3)
> > #define SI_RESOURCE_FLAG_UNMAPPABLE (PIPE_RESOURCE_FLAG_DRV_PRIV <<
> 4)
> > #define SI_RESOURCE_FLAG_READ_ONLY (PIPE_RESOURCE_FLAG_DRV_PRIV <<
> 5)
> > #define SI_RESOURCE_FLAG_32BIT (PIPE_RESOURCE_FLAG_DRV_PRIV <<
> 6)
> > #define SI_RESOURCE_FLAG_SO_FILLED_SIZE
> (PIPE_RESOURCE_FLAG_DRV_PRIV << 7)
> >
> > --
> > 2.17.1
> >
> > _______________________________________________
> > mesa-dev mailing list
> > mesa-dev at lists.freedesktop.org
> > https://lists.freedesktop.org/mailman/listinfo/mesa-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20190118/cbfb23e1/attachment-0001.html>
More information about the mesa-dev
mailing list