drm/freedreno/freedreno_ringbuffer.c

183 lines
5.2 KiB
C
Raw Permalink Normal View History

freedreno: add freedreno DRM The libdrm_freedreno helper layer for use by xf86-video-freedreno, fdre (freedreno r/e library and tests for driving gpu), and eventual gallium driver for the Adreno GPU. This uses the msm gpu driver from QCOM's android kernel tree. Note that current msm kernel driver is a bit strange. It provides a DRM interface for GEM, which is basically sufficient to have DRI2 working. But it does not provide KMS. And interface to 2d and 3d cores is via different other devices (/dev/kgsl-*). This is not quite how I'd write a DRM driver, but at this stage it is useful for xf86-video-freedreno and fdre (and eventual gallium driver) to be able to work on existing kernel driver from QCOM, to allow to capture cmdstream dumps from the binary blob drivers without having to reboot. So libdrm_freedreno attempts to hide most of the crazy. The intention is that when there is a proper kernel driver, it will be mostly just changes in libdrm_freedreno to adapt the gallium driver and xf86-video-freedreno (ignoring the fbdev->KMS changes). So don't look at freedreno as an example of how to write a libdrm module or a DRM driver.. it is just an attempt to paper over a non- standard kernel driver architecture. v1: original v2: hold ref's to pending bo's (because qcom's kernel driver doesn't), various bug fixes, add ringbuffer markers so we can emit IB's to portion of ringbuffer (so that gallium driver can use a single ringbuffer for both tile cmds and draw cmds. Signed-off-by: Rob Clark <robclark@freedesktop.org>
2012-10-07 17:57:31 -06:00
/* -*- mode: C; c-file-style: "k&r"; tab-width 4; indent-tabs-mode: t; -*- */
/*
* Copyright (C) 2012 Rob Clark <robclark@freedesktop.org>
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice (including the next
* paragraph) shall be included in all copies or substantial portions of the
* Software.
*
* 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 NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS 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.
*
* Authors:
* Rob Clark <robclark@freedesktop.org>
*/
#include <assert.h>
freedreno: add freedreno DRM The libdrm_freedreno helper layer for use by xf86-video-freedreno, fdre (freedreno r/e library and tests for driving gpu), and eventual gallium driver for the Adreno GPU. This uses the msm gpu driver from QCOM's android kernel tree. Note that current msm kernel driver is a bit strange. It provides a DRM interface for GEM, which is basically sufficient to have DRI2 working. But it does not provide KMS. And interface to 2d and 3d cores is via different other devices (/dev/kgsl-*). This is not quite how I'd write a DRM driver, but at this stage it is useful for xf86-video-freedreno and fdre (and eventual gallium driver) to be able to work on existing kernel driver from QCOM, to allow to capture cmdstream dumps from the binary blob drivers without having to reboot. So libdrm_freedreno attempts to hide most of the crazy. The intention is that when there is a proper kernel driver, it will be mostly just changes in libdrm_freedreno to adapt the gallium driver and xf86-video-freedreno (ignoring the fbdev->KMS changes). So don't look at freedreno as an example of how to write a libdrm module or a DRM driver.. it is just an attempt to paper over a non- standard kernel driver architecture. v1: original v2: hold ref's to pending bo's (because qcom's kernel driver doesn't), various bug fixes, add ringbuffer markers so we can emit IB's to portion of ringbuffer (so that gallium driver can use a single ringbuffer for both tile cmds and draw cmds. Signed-off-by: Rob Clark <robclark@freedesktop.org>
2012-10-07 17:57:31 -06:00
#include "freedreno_drmif.h"
#include "freedreno_priv.h"
#include "freedreno_ringbuffer.h"
drm_public struct fd_ringbuffer *
fd_ringbuffer_new_flags(struct fd_pipe *pipe, uint32_t size,
enum fd_ringbuffer_flags flags)
freedreno: add freedreno DRM The libdrm_freedreno helper layer for use by xf86-video-freedreno, fdre (freedreno r/e library and tests for driving gpu), and eventual gallium driver for the Adreno GPU. This uses the msm gpu driver from QCOM's android kernel tree. Note that current msm kernel driver is a bit strange. It provides a DRM interface for GEM, which is basically sufficient to have DRI2 working. But it does not provide KMS. And interface to 2d and 3d cores is via different other devices (/dev/kgsl-*). This is not quite how I'd write a DRM driver, but at this stage it is useful for xf86-video-freedreno and fdre (and eventual gallium driver) to be able to work on existing kernel driver from QCOM, to allow to capture cmdstream dumps from the binary blob drivers without having to reboot. So libdrm_freedreno attempts to hide most of the crazy. The intention is that when there is a proper kernel driver, it will be mostly just changes in libdrm_freedreno to adapt the gallium driver and xf86-video-freedreno (ignoring the fbdev->KMS changes). So don't look at freedreno as an example of how to write a libdrm module or a DRM driver.. it is just an attempt to paper over a non- standard kernel driver architecture. v1: original v2: hold ref's to pending bo's (because qcom's kernel driver doesn't), various bug fixes, add ringbuffer markers so we can emit IB's to portion of ringbuffer (so that gallium driver can use a single ringbuffer for both tile cmds and draw cmds. Signed-off-by: Rob Clark <robclark@freedesktop.org>
2012-10-07 17:57:31 -06:00
{
struct fd_ringbuffer *ring;
freedreno: add freedreno DRM The libdrm_freedreno helper layer for use by xf86-video-freedreno, fdre (freedreno r/e library and tests for driving gpu), and eventual gallium driver for the Adreno GPU. This uses the msm gpu driver from QCOM's android kernel tree. Note that current msm kernel driver is a bit strange. It provides a DRM interface for GEM, which is basically sufficient to have DRI2 working. But it does not provide KMS. And interface to 2d and 3d cores is via different other devices (/dev/kgsl-*). This is not quite how I'd write a DRM driver, but at this stage it is useful for xf86-video-freedreno and fdre (and eventual gallium driver) to be able to work on existing kernel driver from QCOM, to allow to capture cmdstream dumps from the binary blob drivers without having to reboot. So libdrm_freedreno attempts to hide most of the crazy. The intention is that when there is a proper kernel driver, it will be mostly just changes in libdrm_freedreno to adapt the gallium driver and xf86-video-freedreno (ignoring the fbdev->KMS changes). So don't look at freedreno as an example of how to write a libdrm module or a DRM driver.. it is just an attempt to paper over a non- standard kernel driver architecture. v1: original v2: hold ref's to pending bo's (because qcom's kernel driver doesn't), various bug fixes, add ringbuffer markers so we can emit IB's to portion of ringbuffer (so that gallium driver can use a single ringbuffer for both tile cmds and draw cmds. Signed-off-by: Rob Clark <robclark@freedesktop.org>
2012-10-07 17:57:31 -06:00
/* we can't really support "growable" rb's in general for
* stateobj's since we need a single gpu addr (ie. can't
* do the trick of a chain of IB packets):
*/
if (flags & FD_RINGBUFFER_OBJECT)
assert(size);
ring = pipe->funcs->ringbuffer_new(pipe, size, flags);
if (!ring)
return NULL;
freedreno: add freedreno DRM The libdrm_freedreno helper layer for use by xf86-video-freedreno, fdre (freedreno r/e library and tests for driving gpu), and eventual gallium driver for the Adreno GPU. This uses the msm gpu driver from QCOM's android kernel tree. Note that current msm kernel driver is a bit strange. It provides a DRM interface for GEM, which is basically sufficient to have DRI2 working. But it does not provide KMS. And interface to 2d and 3d cores is via different other devices (/dev/kgsl-*). This is not quite how I'd write a DRM driver, but at this stage it is useful for xf86-video-freedreno and fdre (and eventual gallium driver) to be able to work on existing kernel driver from QCOM, to allow to capture cmdstream dumps from the binary blob drivers without having to reboot. So libdrm_freedreno attempts to hide most of the crazy. The intention is that when there is a proper kernel driver, it will be mostly just changes in libdrm_freedreno to adapt the gallium driver and xf86-video-freedreno (ignoring the fbdev->KMS changes). So don't look at freedreno as an example of how to write a libdrm module or a DRM driver.. it is just an attempt to paper over a non- standard kernel driver architecture. v1: original v2: hold ref's to pending bo's (because qcom's kernel driver doesn't), various bug fixes, add ringbuffer markers so we can emit IB's to portion of ringbuffer (so that gallium driver can use a single ringbuffer for both tile cmds and draw cmds. Signed-off-by: Rob Clark <robclark@freedesktop.org>
2012-10-07 17:57:31 -06:00
ring->flags = flags;
freedreno: add freedreno DRM The libdrm_freedreno helper layer for use by xf86-video-freedreno, fdre (freedreno r/e library and tests for driving gpu), and eventual gallium driver for the Adreno GPU. This uses the msm gpu driver from QCOM's android kernel tree. Note that current msm kernel driver is a bit strange. It provides a DRM interface for GEM, which is basically sufficient to have DRI2 working. But it does not provide KMS. And interface to 2d and 3d cores is via different other devices (/dev/kgsl-*). This is not quite how I'd write a DRM driver, but at this stage it is useful for xf86-video-freedreno and fdre (and eventual gallium driver) to be able to work on existing kernel driver from QCOM, to allow to capture cmdstream dumps from the binary blob drivers without having to reboot. So libdrm_freedreno attempts to hide most of the crazy. The intention is that when there is a proper kernel driver, it will be mostly just changes in libdrm_freedreno to adapt the gallium driver and xf86-video-freedreno (ignoring the fbdev->KMS changes). So don't look at freedreno as an example of how to write a libdrm module or a DRM driver.. it is just an attempt to paper over a non- standard kernel driver architecture. v1: original v2: hold ref's to pending bo's (because qcom's kernel driver doesn't), various bug fixes, add ringbuffer markers so we can emit IB's to portion of ringbuffer (so that gallium driver can use a single ringbuffer for both tile cmds and draw cmds. Signed-off-by: Rob Clark <robclark@freedesktop.org>
2012-10-07 17:57:31 -06:00
ring->pipe = pipe;
ring->start = ring->funcs->hostptr(ring);
freedreno: support growable cmdstream buffers The issue that userspace needed to solve is that there is ~two orders of magnitude size difference in cmdstream buffers (both for gmem commands and for draw commands), and that the previous practice of allocating worst-case sizes is quite wasteful. Previously a submit would be constructed (for example) like: CMD TARGET DESCRIPTION g0 N gmem/tiling commands b0 Y binning commands d0 Y draw commands Which, after the one non-IB-target cmd buffer is inserted into the kernel controlled ringbuffer, looks like (not to scale): b0: d0: +-----+ +-----+ IB1 | ... | | ... | +-----+ +-----+ ^ ^ | | +-----+ +-+---------+ g0: | | | +----+----+----+----+----+----+---- IB0 | .. | IB | .. | IB | .. | IB | ... +----+----+----+----+----+----+---- ^ tile0 tile1 | +-----------+ userspace | ~~~~~~~~~~~~~~~~~~~|~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ kernel | ----+----+---- ringbuffer ... | IB | ... ----+----+---- Now, multiple physical cmdstream buffers per fd_ringbuffer are supported, so this becomes: CMD TARGET DESCRIPTION g0 N ... N gmem/tiling commands gN N b0 Y ... Y binning commands bN Y d0 Y ... Y draw commands dN Y Which, after the non-IB-target cmd buffers (g0..gN) are inserted into the kernel controlled ringbuffer, looks like: b0: b1 d0: d1 +-----+ +-----+ +-----+ +-----+ IB1 | ... | | ... | ... | ... | | ... | ... +-----+ +-----+ +-----+ +-----+ ^ ^ ^ ^ | | | | | +-+ | +-----+------+ +-----+ | | | | | | +--+----------+ | g0: | | | | | | +----+----+----+----+----+----+---+----+----+---- IB0 | .. | IB | IB | .. | IB | IB |.. | IB | IB |... +----+----+----+----+----+----+---+----+----+---- ^ tile0 tile1 | to b0 to b1 | | | to|d0 to|d1 | | +----+ | +-+-----------+ | | | | | | | +------+ | +-+-------------+ | | g1: | | | | | | | +----+----+----+----+----+----+---+----+----+---- IB0 | | .. | IB | IB | .. | IB | IB |.. | IB | IB |... | +----+----+----+----+----+----+---+----+----+---- | ^ tileX tileY | | | +-----------+ +-----------+ | userspace | | ~~~~~~~~~~~~~~~~~~~|~~~~|~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ kernel | | ----+----+----+---- ringbuffer ... | IB | IB | ... ----+----+----+---- Signed-off-by: Rob Clark <robclark@freedesktop.org>
2016-06-20 12:06:24 -06:00
ring->end = &(ring->start[ring->size/4]);
freedreno: add freedreno DRM The libdrm_freedreno helper layer for use by xf86-video-freedreno, fdre (freedreno r/e library and tests for driving gpu), and eventual gallium driver for the Adreno GPU. This uses the msm gpu driver from QCOM's android kernel tree. Note that current msm kernel driver is a bit strange. It provides a DRM interface for GEM, which is basically sufficient to have DRI2 working. But it does not provide KMS. And interface to 2d and 3d cores is via different other devices (/dev/kgsl-*). This is not quite how I'd write a DRM driver, but at this stage it is useful for xf86-video-freedreno and fdre (and eventual gallium driver) to be able to work on existing kernel driver from QCOM, to allow to capture cmdstream dumps from the binary blob drivers without having to reboot. So libdrm_freedreno attempts to hide most of the crazy. The intention is that when there is a proper kernel driver, it will be mostly just changes in libdrm_freedreno to adapt the gallium driver and xf86-video-freedreno (ignoring the fbdev->KMS changes). So don't look at freedreno as an example of how to write a libdrm module or a DRM driver.. it is just an attempt to paper over a non- standard kernel driver architecture. v1: original v2: hold ref's to pending bo's (because qcom's kernel driver doesn't), various bug fixes, add ringbuffer markers so we can emit IB's to portion of ringbuffer (so that gallium driver can use a single ringbuffer for both tile cmds and draw cmds. Signed-off-by: Rob Clark <robclark@freedesktop.org>
2012-10-07 17:57:31 -06:00
ring->cur = ring->last_start = ring->start;
return ring;
}
drm_public struct fd_ringbuffer *
fd_ringbuffer_new(struct fd_pipe *pipe, uint32_t size)
{
return fd_ringbuffer_new_flags(pipe, size, 0);
}
drm_public struct fd_ringbuffer *
fd_ringbuffer_new_object(struct fd_pipe *pipe, uint32_t size)
{
return fd_ringbuffer_new_flags(pipe, size, FD_RINGBUFFER_OBJECT);
}
drm_public void fd_ringbuffer_del(struct fd_ringbuffer *ring)
freedreno: add freedreno DRM The libdrm_freedreno helper layer for use by xf86-video-freedreno, fdre (freedreno r/e library and tests for driving gpu), and eventual gallium driver for the Adreno GPU. This uses the msm gpu driver from QCOM's android kernel tree. Note that current msm kernel driver is a bit strange. It provides a DRM interface for GEM, which is basically sufficient to have DRI2 working. But it does not provide KMS. And interface to 2d and 3d cores is via different other devices (/dev/kgsl-*). This is not quite how I'd write a DRM driver, but at this stage it is useful for xf86-video-freedreno and fdre (and eventual gallium driver) to be able to work on existing kernel driver from QCOM, to allow to capture cmdstream dumps from the binary blob drivers without having to reboot. So libdrm_freedreno attempts to hide most of the crazy. The intention is that when there is a proper kernel driver, it will be mostly just changes in libdrm_freedreno to adapt the gallium driver and xf86-video-freedreno (ignoring the fbdev->KMS changes). So don't look at freedreno as an example of how to write a libdrm module or a DRM driver.. it is just an attempt to paper over a non- standard kernel driver architecture. v1: original v2: hold ref's to pending bo's (because qcom's kernel driver doesn't), various bug fixes, add ringbuffer markers so we can emit IB's to portion of ringbuffer (so that gallium driver can use a single ringbuffer for both tile cmds and draw cmds. Signed-off-by: Rob Clark <robclark@freedesktop.org>
2012-10-07 17:57:31 -06:00
{
if (!atomic_dec_and_test(&ring->refcnt))
return;
fd_ringbuffer_reset(ring);
ring->funcs->destroy(ring);
freedreno: add freedreno DRM The libdrm_freedreno helper layer for use by xf86-video-freedreno, fdre (freedreno r/e library and tests for driving gpu), and eventual gallium driver for the Adreno GPU. This uses the msm gpu driver from QCOM's android kernel tree. Note that current msm kernel driver is a bit strange. It provides a DRM interface for GEM, which is basically sufficient to have DRI2 working. But it does not provide KMS. And interface to 2d and 3d cores is via different other devices (/dev/kgsl-*). This is not quite how I'd write a DRM driver, but at this stage it is useful for xf86-video-freedreno and fdre (and eventual gallium driver) to be able to work on existing kernel driver from QCOM, to allow to capture cmdstream dumps from the binary blob drivers without having to reboot. So libdrm_freedreno attempts to hide most of the crazy. The intention is that when there is a proper kernel driver, it will be mostly just changes in libdrm_freedreno to adapt the gallium driver and xf86-video-freedreno (ignoring the fbdev->KMS changes). So don't look at freedreno as an example of how to write a libdrm module or a DRM driver.. it is just an attempt to paper over a non- standard kernel driver architecture. v1: original v2: hold ref's to pending bo's (because qcom's kernel driver doesn't), various bug fixes, add ringbuffer markers so we can emit IB's to portion of ringbuffer (so that gallium driver can use a single ringbuffer for both tile cmds and draw cmds. Signed-off-by: Rob Clark <robclark@freedesktop.org>
2012-10-07 17:57:31 -06:00
}
drm_public struct fd_ringbuffer *
fd_ringbuffer_ref(struct fd_ringbuffer *ring)
{
STATIC_ASSERT(sizeof(ring->refcnt) <= sizeof(ring->__pad));
atomic_inc(&ring->refcnt);
return ring;
}
/* ringbuffers which are IB targets should set the toplevel rb (ie.
* the IB source) as it's parent before emitting reloc's, to ensure
* the bookkeeping works out properly.
*/
drm_public void fd_ringbuffer_set_parent(struct fd_ringbuffer *ring,
struct fd_ringbuffer *parent)
{
/* state objects should not be parented! */
assert(!(ring->flags & FD_RINGBUFFER_OBJECT));
ring->parent = parent;
}
drm_public void fd_ringbuffer_reset(struct fd_ringbuffer *ring)
freedreno: add freedreno DRM The libdrm_freedreno helper layer for use by xf86-video-freedreno, fdre (freedreno r/e library and tests for driving gpu), and eventual gallium driver for the Adreno GPU. This uses the msm gpu driver from QCOM's android kernel tree. Note that current msm kernel driver is a bit strange. It provides a DRM interface for GEM, which is basically sufficient to have DRI2 working. But it does not provide KMS. And interface to 2d and 3d cores is via different other devices (/dev/kgsl-*). This is not quite how I'd write a DRM driver, but at this stage it is useful for xf86-video-freedreno and fdre (and eventual gallium driver) to be able to work on existing kernel driver from QCOM, to allow to capture cmdstream dumps from the binary blob drivers without having to reboot. So libdrm_freedreno attempts to hide most of the crazy. The intention is that when there is a proper kernel driver, it will be mostly just changes in libdrm_freedreno to adapt the gallium driver and xf86-video-freedreno (ignoring the fbdev->KMS changes). So don't look at freedreno as an example of how to write a libdrm module or a DRM driver.. it is just an attempt to paper over a non- standard kernel driver architecture. v1: original v2: hold ref's to pending bo's (because qcom's kernel driver doesn't), various bug fixes, add ringbuffer markers so we can emit IB's to portion of ringbuffer (so that gallium driver can use a single ringbuffer for both tile cmds and draw cmds. Signed-off-by: Rob Clark <robclark@freedesktop.org>
2012-10-07 17:57:31 -06:00
{
uint32_t *start = ring->start;
if (ring->pipe->id == FD_PIPE_2D)
start = &ring->start[0x140];
ring->cur = ring->last_start = start;
if (ring->funcs->reset)
ring->funcs->reset(ring);
freedreno: add freedreno DRM The libdrm_freedreno helper layer for use by xf86-video-freedreno, fdre (freedreno r/e library and tests for driving gpu), and eventual gallium driver for the Adreno GPU. This uses the msm gpu driver from QCOM's android kernel tree. Note that current msm kernel driver is a bit strange. It provides a DRM interface for GEM, which is basically sufficient to have DRI2 working. But it does not provide KMS. And interface to 2d and 3d cores is via different other devices (/dev/kgsl-*). This is not quite how I'd write a DRM driver, but at this stage it is useful for xf86-video-freedreno and fdre (and eventual gallium driver) to be able to work on existing kernel driver from QCOM, to allow to capture cmdstream dumps from the binary blob drivers without having to reboot. So libdrm_freedreno attempts to hide most of the crazy. The intention is that when there is a proper kernel driver, it will be mostly just changes in libdrm_freedreno to adapt the gallium driver and xf86-video-freedreno (ignoring the fbdev->KMS changes). So don't look at freedreno as an example of how to write a libdrm module or a DRM driver.. it is just an attempt to paper over a non- standard kernel driver architecture. v1: original v2: hold ref's to pending bo's (because qcom's kernel driver doesn't), various bug fixes, add ringbuffer markers so we can emit IB's to portion of ringbuffer (so that gallium driver can use a single ringbuffer for both tile cmds and draw cmds. Signed-off-by: Rob Clark <robclark@freedesktop.org>
2012-10-07 17:57:31 -06:00
}
drm_public int fd_ringbuffer_flush(struct fd_ringbuffer *ring)
freedreno: add freedreno DRM The libdrm_freedreno helper layer for use by xf86-video-freedreno, fdre (freedreno r/e library and tests for driving gpu), and eventual gallium driver for the Adreno GPU. This uses the msm gpu driver from QCOM's android kernel tree. Note that current msm kernel driver is a bit strange. It provides a DRM interface for GEM, which is basically sufficient to have DRI2 working. But it does not provide KMS. And interface to 2d and 3d cores is via different other devices (/dev/kgsl-*). This is not quite how I'd write a DRM driver, but at this stage it is useful for xf86-video-freedreno and fdre (and eventual gallium driver) to be able to work on existing kernel driver from QCOM, to allow to capture cmdstream dumps from the binary blob drivers without having to reboot. So libdrm_freedreno attempts to hide most of the crazy. The intention is that when there is a proper kernel driver, it will be mostly just changes in libdrm_freedreno to adapt the gallium driver and xf86-video-freedreno (ignoring the fbdev->KMS changes). So don't look at freedreno as an example of how to write a libdrm module or a DRM driver.. it is just an attempt to paper over a non- standard kernel driver architecture. v1: original v2: hold ref's to pending bo's (because qcom's kernel driver doesn't), various bug fixes, add ringbuffer markers so we can emit IB's to portion of ringbuffer (so that gallium driver can use a single ringbuffer for both tile cmds and draw cmds. Signed-off-by: Rob Clark <robclark@freedesktop.org>
2012-10-07 17:57:31 -06:00
{
return ring->funcs->flush(ring, ring->last_start, -1, NULL);
}
drm_public int fd_ringbuffer_flush2(struct fd_ringbuffer *ring, int in_fence_fd,
int *out_fence_fd)
{
return ring->funcs->flush(ring, ring->last_start, in_fence_fd, out_fence_fd);
freedreno: add freedreno DRM The libdrm_freedreno helper layer for use by xf86-video-freedreno, fdre (freedreno r/e library and tests for driving gpu), and eventual gallium driver for the Adreno GPU. This uses the msm gpu driver from QCOM's android kernel tree. Note that current msm kernel driver is a bit strange. It provides a DRM interface for GEM, which is basically sufficient to have DRI2 working. But it does not provide KMS. And interface to 2d and 3d cores is via different other devices (/dev/kgsl-*). This is not quite how I'd write a DRM driver, but at this stage it is useful for xf86-video-freedreno and fdre (and eventual gallium driver) to be able to work on existing kernel driver from QCOM, to allow to capture cmdstream dumps from the binary blob drivers without having to reboot. So libdrm_freedreno attempts to hide most of the crazy. The intention is that when there is a proper kernel driver, it will be mostly just changes in libdrm_freedreno to adapt the gallium driver and xf86-video-freedreno (ignoring the fbdev->KMS changes). So don't look at freedreno as an example of how to write a libdrm module or a DRM driver.. it is just an attempt to paper over a non- standard kernel driver architecture. v1: original v2: hold ref's to pending bo's (because qcom's kernel driver doesn't), various bug fixes, add ringbuffer markers so we can emit IB's to portion of ringbuffer (so that gallium driver can use a single ringbuffer for both tile cmds and draw cmds. Signed-off-by: Rob Clark <robclark@freedesktop.org>
2012-10-07 17:57:31 -06:00
}
drm_public void fd_ringbuffer_grow(struct fd_ringbuffer *ring, uint32_t ndwords)
freedreno: support growable cmdstream buffers The issue that userspace needed to solve is that there is ~two orders of magnitude size difference in cmdstream buffers (both for gmem commands and for draw commands), and that the previous practice of allocating worst-case sizes is quite wasteful. Previously a submit would be constructed (for example) like: CMD TARGET DESCRIPTION g0 N gmem/tiling commands b0 Y binning commands d0 Y draw commands Which, after the one non-IB-target cmd buffer is inserted into the kernel controlled ringbuffer, looks like (not to scale): b0: d0: +-----+ +-----+ IB1 | ... | | ... | +-----+ +-----+ ^ ^ | | +-----+ +-+---------+ g0: | | | +----+----+----+----+----+----+---- IB0 | .. | IB | .. | IB | .. | IB | ... +----+----+----+----+----+----+---- ^ tile0 tile1 | +-----------+ userspace | ~~~~~~~~~~~~~~~~~~~|~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ kernel | ----+----+---- ringbuffer ... | IB | ... ----+----+---- Now, multiple physical cmdstream buffers per fd_ringbuffer are supported, so this becomes: CMD TARGET DESCRIPTION g0 N ... N gmem/tiling commands gN N b0 Y ... Y binning commands bN Y d0 Y ... Y draw commands dN Y Which, after the non-IB-target cmd buffers (g0..gN) are inserted into the kernel controlled ringbuffer, looks like: b0: b1 d0: d1 +-----+ +-----+ +-----+ +-----+ IB1 | ... | | ... | ... | ... | | ... | ... +-----+ +-----+ +-----+ +-----+ ^ ^ ^ ^ | | | | | +-+ | +-----+------+ +-----+ | | | | | | +--+----------+ | g0: | | | | | | +----+----+----+----+----+----+---+----+----+---- IB0 | .. | IB | IB | .. | IB | IB |.. | IB | IB |... +----+----+----+----+----+----+---+----+----+---- ^ tile0 tile1 | to b0 to b1 | | | to|d0 to|d1 | | +----+ | +-+-----------+ | | | | | | | +------+ | +-+-------------+ | | g1: | | | | | | | +----+----+----+----+----+----+---+----+----+---- IB0 | | .. | IB | IB | .. | IB | IB |.. | IB | IB |... | +----+----+----+----+----+----+---+----+----+---- | ^ tileX tileY | | | +-----------+ +-----------+ | userspace | | ~~~~~~~~~~~~~~~~~~~|~~~~|~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ kernel | | ----+----+----+---- ringbuffer ... | IB | IB | ... ----+----+----+---- Signed-off-by: Rob Clark <robclark@freedesktop.org>
2016-06-20 12:06:24 -06:00
{
assert(ring->funcs->grow); /* unsupported on kgsl */
/* there is an upper bound on IB size, which appears to be 0x100000 */
if (ring->size < 0x100000)
ring->size *= 2;
ring->funcs->grow(ring, ring->size);
ring->start = ring->funcs->hostptr(ring);
ring->end = &(ring->start[ring->size/4]);
ring->cur = ring->last_start = ring->start;
}
drm_public uint32_t fd_ringbuffer_timestamp(struct fd_ringbuffer *ring)
freedreno: add freedreno DRM The libdrm_freedreno helper layer for use by xf86-video-freedreno, fdre (freedreno r/e library and tests for driving gpu), and eventual gallium driver for the Adreno GPU. This uses the msm gpu driver from QCOM's android kernel tree. Note that current msm kernel driver is a bit strange. It provides a DRM interface for GEM, which is basically sufficient to have DRI2 working. But it does not provide KMS. And interface to 2d and 3d cores is via different other devices (/dev/kgsl-*). This is not quite how I'd write a DRM driver, but at this stage it is useful for xf86-video-freedreno and fdre (and eventual gallium driver) to be able to work on existing kernel driver from QCOM, to allow to capture cmdstream dumps from the binary blob drivers without having to reboot. So libdrm_freedreno attempts to hide most of the crazy. The intention is that when there is a proper kernel driver, it will be mostly just changes in libdrm_freedreno to adapt the gallium driver and xf86-video-freedreno (ignoring the fbdev->KMS changes). So don't look at freedreno as an example of how to write a libdrm module or a DRM driver.. it is just an attempt to paper over a non- standard kernel driver architecture. v1: original v2: hold ref's to pending bo's (because qcom's kernel driver doesn't), various bug fixes, add ringbuffer markers so we can emit IB's to portion of ringbuffer (so that gallium driver can use a single ringbuffer for both tile cmds and draw cmds. Signed-off-by: Rob Clark <robclark@freedesktop.org>
2012-10-07 17:57:31 -06:00
{
return ring->last_timestamp;
}
drm_public void fd_ringbuffer_reloc(struct fd_ringbuffer *ring,
const struct fd_reloc *reloc)
{
assert(ring->pipe->gpu_id < 500);
ring->funcs->emit_reloc(ring, reloc);
}
drm_public void fd_ringbuffer_reloc2(struct fd_ringbuffer *ring,
const struct fd_reloc *reloc)
{
ring->funcs->emit_reloc(ring, reloc);
}
drm_public uint32_t fd_ringbuffer_cmd_count(struct fd_ringbuffer *ring)
freedreno: support growable cmdstream buffers The issue that userspace needed to solve is that there is ~two orders of magnitude size difference in cmdstream buffers (both for gmem commands and for draw commands), and that the previous practice of allocating worst-case sizes is quite wasteful. Previously a submit would be constructed (for example) like: CMD TARGET DESCRIPTION g0 N gmem/tiling commands b0 Y binning commands d0 Y draw commands Which, after the one non-IB-target cmd buffer is inserted into the kernel controlled ringbuffer, looks like (not to scale): b0: d0: +-----+ +-----+ IB1 | ... | | ... | +-----+ +-----+ ^ ^ | | +-----+ +-+---------+ g0: | | | +----+----+----+----+----+----+---- IB0 | .. | IB | .. | IB | .. | IB | ... +----+----+----+----+----+----+---- ^ tile0 tile1 | +-----------+ userspace | ~~~~~~~~~~~~~~~~~~~|~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ kernel | ----+----+---- ringbuffer ... | IB | ... ----+----+---- Now, multiple physical cmdstream buffers per fd_ringbuffer are supported, so this becomes: CMD TARGET DESCRIPTION g0 N ... N gmem/tiling commands gN N b0 Y ... Y binning commands bN Y d0 Y ... Y draw commands dN Y Which, after the non-IB-target cmd buffers (g0..gN) are inserted into the kernel controlled ringbuffer, looks like: b0: b1 d0: d1 +-----+ +-----+ +-----+ +-----+ IB1 | ... | | ... | ... | ... | | ... | ... +-----+ +-----+ +-----+ +-----+ ^ ^ ^ ^ | | | | | +-+ | +-----+------+ +-----+ | | | | | | +--+----------+ | g0: | | | | | | +----+----+----+----+----+----+---+----+----+---- IB0 | .. | IB | IB | .. | IB | IB |.. | IB | IB |... +----+----+----+----+----+----+---+----+----+---- ^ tile0 tile1 | to b0 to b1 | | | to|d0 to|d1 | | +----+ | +-+-----------+ | | | | | | | +------+ | +-+-------------+ | | g1: | | | | | | | +----+----+----+----+----+----+---+----+----+---- IB0 | | .. | IB | IB | .. | IB | IB |.. | IB | IB |... | +----+----+----+----+----+----+---+----+----+---- | ^ tileX tileY | | | +-----------+ +-----------+ | userspace | | ~~~~~~~~~~~~~~~~~~~|~~~~|~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ kernel | | ----+----+----+---- ringbuffer ... | IB | IB | ... ----+----+----+---- Signed-off-by: Rob Clark <robclark@freedesktop.org>
2016-06-20 12:06:24 -06:00
{
if (!ring->funcs->cmd_count)
return 1;
return ring->funcs->cmd_count(ring);
freedreno: add freedreno DRM The libdrm_freedreno helper layer for use by xf86-video-freedreno, fdre (freedreno r/e library and tests for driving gpu), and eventual gallium driver for the Adreno GPU. This uses the msm gpu driver from QCOM's android kernel tree. Note that current msm kernel driver is a bit strange. It provides a DRM interface for GEM, which is basically sufficient to have DRI2 working. But it does not provide KMS. And interface to 2d and 3d cores is via different other devices (/dev/kgsl-*). This is not quite how I'd write a DRM driver, but at this stage it is useful for xf86-video-freedreno and fdre (and eventual gallium driver) to be able to work on existing kernel driver from QCOM, to allow to capture cmdstream dumps from the binary blob drivers without having to reboot. So libdrm_freedreno attempts to hide most of the crazy. The intention is that when there is a proper kernel driver, it will be mostly just changes in libdrm_freedreno to adapt the gallium driver and xf86-video-freedreno (ignoring the fbdev->KMS changes). So don't look at freedreno as an example of how to write a libdrm module or a DRM driver.. it is just an attempt to paper over a non- standard kernel driver architecture. v1: original v2: hold ref's to pending bo's (because qcom's kernel driver doesn't), various bug fixes, add ringbuffer markers so we can emit IB's to portion of ringbuffer (so that gallium driver can use a single ringbuffer for both tile cmds and draw cmds. Signed-off-by: Rob Clark <robclark@freedesktop.org>
2012-10-07 17:57:31 -06:00
}
drm_public uint32_t
fd_ringbuffer_emit_reloc_ring_full(struct fd_ringbuffer *ring,
struct fd_ringbuffer *target, uint32_t cmd_idx)
{
return ring->funcs->emit_reloc_ring(ring, target, cmd_idx);
}
drm_public uint32_t
fd_ringbuffer_size(struct fd_ringbuffer *ring)
{
/* only really needed for stateobj ringbuffers, and won't really
* do what you expect for growable rb's.. so lets just restrict
* this to stateobj's for now:
*/
assert(ring->flags & FD_RINGBUFFER_OBJECT);
return offset_bytes(ring->cur, ring->start);
}