You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
virglrenderer/src/virglrenderer.c

289 lines
8.5 KiB

/**************************************************************************
*
* Copyright (C) 2014 Red Hat Inc.
*
* 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 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.
*
**************************************************************************/
#include <stdio.h>
#include <time.h>
#include <epoxy/gl.h>
#include <sys/mman.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <errno.h>
#include <unistd.h>
#include <sys/socket.h>
#include <sys/un.h>
#include <sys/eventfd.h>
#include "pipe/p_state.h"
#include "util/u_format.h"
#include "util/u_math.h"
#include "vrend_renderer.h"
#include "virglrenderer.h"
#include "virgl_egl.h"
static struct virgl_egl *egl_info;
static int use_egl_context;
/* new API - just wrap internal API for now */
int virgl_renderer_resource_create(struct virgl_renderer_resource_create_args *args, struct iovec *iov, uint32_t num_iovs)
{
return vrend_renderer_resource_create((struct vrend_renderer_resource_create_args *)args, iov, num_iovs);
}
void virgl_renderer_resource_unref(uint32_t res_handle)
{
vrend_renderer_resource_unref(res_handle);
}
void virgl_renderer_fill_caps(uint32_t set, uint32_t version,
void *caps)
{
vrend_renderer_fill_caps(set, version, (union virgl_caps *)caps);
}
int virgl_renderer_context_create(uint32_t handle, uint32_t nlen, const char *name)
{
return vrend_renderer_context_create(handle, nlen, name);
}
void virgl_renderer_context_destroy(uint32_t handle)
{
vrend_renderer_context_destroy(handle);
}
int virgl_renderer_submit_cmd(void *buffer,
int ctx_id,
int ndw)
{
return vrend_decode_block(ctx_id, buffer, ndw);
}
int virgl_renderer_transfer_write_iov(uint32_t handle,
uint32_t ctx_id,
int level,
uint32_t stride,
uint32_t layer_stride,
struct virgl_box *box,
uint64_t offset,
struct iovec *iovec,
unsigned int iovec_cnt)
{
struct vrend_transfer_info transfer_info;
transfer_info.handle = handle;
transfer_info.ctx_id = ctx_id;
transfer_info.level = level;
transfer_info.stride = stride;
transfer_info.layer_stride = layer_stride;
transfer_info.box = (struct pipe_box *)box;
transfer_info.offset = offset;
transfer_info.iovec = iovec;
transfer_info.iovec_cnt = iovec_cnt;
return vrend_renderer_transfer_iov(&transfer_info, VREND_TRANSFER_WRITE);
}
int virgl_renderer_transfer_read_iov(uint32_t handle, uint32_t ctx_id,
uint32_t level, uint32_t stride,
uint32_t layer_stride,
struct virgl_box *box,
uint64_t offset, struct iovec *iovec,
int iovec_cnt)
{
struct vrend_transfer_info transfer_info;
transfer_info.handle = handle;
transfer_info.ctx_id = ctx_id;
transfer_info.level = level;
transfer_info.stride = stride;
transfer_info.layer_stride = layer_stride;
transfer_info.box = (struct pipe_box *)box;
transfer_info.offset = offset;
transfer_info.iovec = iovec;
transfer_info.iovec_cnt = iovec_cnt;
return vrend_renderer_transfer_iov(&transfer_info, VREND_TRANSFER_READ);
}
int virgl_renderer_resource_attach_iov(int res_handle, struct iovec *iov,
int num_iovs)
{
return vrend_renderer_resource_attach_iov(res_handle, iov, num_iovs);
}
void virgl_renderer_resource_detach_iov(int res_handle, struct iovec **iov_p, int *num_iovs_p)
{
return vrend_renderer_resource_detach_iov(res_handle, iov_p, num_iovs_p);
}
int virgl_renderer_create_fence(int client_fence_id, uint32_t ctx_id)
{
return vrend_renderer_create_fence(client_fence_id, ctx_id);
}
void virgl_renderer_force_ctx_0(void)
{
vrend_renderer_force_ctx_0();
}
void virgl_renderer_ctx_attach_resource(int ctx_id, int res_handle)
{
vrend_renderer_attach_res_ctx(ctx_id, res_handle);
}
void virgl_renderer_ctx_detach_resource(int ctx_id, int res_handle)
{
vrend_renderer_detach_res_ctx(ctx_id, res_handle);
}
int virgl_renderer_resource_get_info(int res_handle,
struct virgl_renderer_resource_info *info)
{
int ret;
ret = vrend_renderer_resource_get_info(res_handle, (struct vrend_renderer_resource_info *)info);
if (ret == 0 && use_egl_context)
return virgl_egl_get_fourcc_for_texture(egl_info, info->tex_id, info->virgl_format, &info->drm_fourcc);
return ret;
}
void virgl_renderer_get_cap_set(uint32_t cap_set, uint32_t *max_ver,
uint32_t *max_size)
{
vrend_renderer_get_cap_set(cap_set, max_ver, max_size);
}
void virgl_renderer_get_rect(int resource_id, struct iovec *iov, unsigned int num_iovs,
uint32_t offset, int x, int y, int width, int height)
{
vrend_renderer_get_rect(resource_id, iov, num_iovs, offset, x, y, width, height);
}
static struct virgl_renderer_callbacks *rcbs;
static void *dev_cookie;
static struct vrend_if_cbs virgl_cbs;
static void virgl_write_fence(uint32_t fence_id)
{
rcbs->write_fence(dev_cookie, fence_id);
}
static virgl_renderer_gl_context create_gl_context(int scanout_idx, struct virgl_gl_ctx_param *param)
{
struct virgl_renderer_gl_ctx_param vparam;
if (use_egl_context)
return virgl_egl_create_context(egl_info, param);
vparam.version = 1;
vparam.shared = param->shared;
vparam.major_ver = param->major_ver;
vparam.minor_ver = param->minor_ver;
return rcbs->create_gl_context(dev_cookie, scanout_idx, &vparam);
}
static void destroy_gl_context(virgl_renderer_gl_context ctx)
{
if (use_egl_context)
return virgl_egl_destroy_context(egl_info, ctx);
return rcbs->destroy_gl_context(dev_cookie, ctx);
}
static int make_current(int scanout_idx, virgl_renderer_gl_context ctx)
{
if (use_egl_context)
return virgl_egl_make_context_current(egl_info, ctx);
return rcbs->make_current(dev_cookie, scanout_idx, ctx);
}
static struct vrend_if_cbs virgl_cbs = {
virgl_write_fence,
create_gl_context,
destroy_gl_context,
make_current,
};
void *virgl_renderer_get_cursor_data(uint32_t resource_id, uint32_t *width, uint32_t *height)
{
return vrend_renderer_get_cursor_contents(resource_id, width, height);
}
void virgl_renderer_poll(void)
{
vrend_renderer_check_queries();
vrend_renderer_check_fences();
}
void virgl_renderer_cleanup(void *cookie)
{
vrend_renderer_fini();
if (use_egl_context) {
virgl_egl_destroy(egl_info);
egl_info = NULL;
use_egl_context = 0;
}
}
int virgl_renderer_init(void *cookie, int flags, struct virgl_renderer_callbacks *cbs)
{
renderer: use a thread to block for fences. Instead of polling the fences regularly, have a thread that blocks for a single fence using a separate shared context, then uses eventfd to wake up the main thread when something happens. Inside the guest, glmark2 typicially runs twice as fast with the thread sync. Although in general, the performances seems to be about +30%. The benefits is mostly for CPU-bounds tasks (when main the thread hits 100%) A naive perf stat of the vtest renderer with glmark2 "build" test with a fixed number of frames (500) results in the following stats data: (do not value timing related informations, since the renderer is ran and stopped manually) without thread: 3032.282265 task-clock (msec) # 0.420 CPUs utilized 4,277 context-switches # 0.001 M/sec 102 cpu-migrations # 0.034 K/sec 9,020 page-faults # 0.003 M/sec 7,884,098,254 cycles # 2.600 GHz 4,440,126,451 stalled-cycles-frontend # 56.32% frontend cycles idle <not supported> stalled-cycles-backend 11,024,091,578 instructions # 1.40 insns per cycle # 0.40 stalled # cycles per insn 1,091,831,588 branches # 360.069 M/sec 5,426,846 branch-misses # 0.50% of all branches with thread: 3403.592921 task-clock (msec) # 0.452 CPUs utilized 7,145 context-switches # 0.002 M/sec 410 cpu-migrations # 0.120 K/sec 6,191 page-faults # 0.002 M/sec 7,475,038,064 cycles # 2.196 GHz 4,487,043,071 stalled-cycles-frontend # 60.03% frontend cycles idle <not supported> stalled-cycles-backend 9,925,205,494 instructions # 1.33 insns per cycle # 0.45 stalled # cycles per insn 834,375,503 branches # 245.146 M/sec 4,919,995 branch-misses # 0.59% of all branches Signed-off-by: Marc-André Lureau <marcandre.lureau@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years ago
uint32_t renderer_flags = 0;
if (!cookie || !cbs)
return -1;
if (cbs->version != 1)
return -1;
dev_cookie = cookie;
rcbs = cbs;
if (flags & VIRGL_RENDERER_USE_EGL) {
egl_info = virgl_egl_init();
if (!egl_info)
return -1;
use_egl_context = 1;
}
renderer: use a thread to block for fences. Instead of polling the fences regularly, have a thread that blocks for a single fence using a separate shared context, then uses eventfd to wake up the main thread when something happens. Inside the guest, glmark2 typicially runs twice as fast with the thread sync. Although in general, the performances seems to be about +30%. The benefits is mostly for CPU-bounds tasks (when main the thread hits 100%) A naive perf stat of the vtest renderer with glmark2 "build" test with a fixed number of frames (500) results in the following stats data: (do not value timing related informations, since the renderer is ran and stopped manually) without thread: 3032.282265 task-clock (msec) # 0.420 CPUs utilized 4,277 context-switches # 0.001 M/sec 102 cpu-migrations # 0.034 K/sec 9,020 page-faults # 0.003 M/sec 7,884,098,254 cycles # 2.600 GHz 4,440,126,451 stalled-cycles-frontend # 56.32% frontend cycles idle <not supported> stalled-cycles-backend 11,024,091,578 instructions # 1.40 insns per cycle # 0.40 stalled # cycles per insn 1,091,831,588 branches # 360.069 M/sec 5,426,846 branch-misses # 0.50% of all branches with thread: 3403.592921 task-clock (msec) # 0.452 CPUs utilized 7,145 context-switches # 0.002 M/sec 410 cpu-migrations # 0.120 K/sec 6,191 page-faults # 0.002 M/sec 7,475,038,064 cycles # 2.196 GHz 4,487,043,071 stalled-cycles-frontend # 60.03% frontend cycles idle <not supported> stalled-cycles-backend 9,925,205,494 instructions # 1.33 insns per cycle # 0.45 stalled # cycles per insn 834,375,503 branches # 245.146 M/sec 4,919,995 branch-misses # 0.59% of all branches Signed-off-by: Marc-André Lureau <marcandre.lureau@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years ago
if (flags & VIRGL_RENDERER_THREAD_SYNC)
renderer_flags |= VREND_USE_THREAD_SYNC;
return vrend_renderer_init(&virgl_cbs, renderer_flags);
}
int virgl_renderer_get_fd_for_texture(uint32_t tex_id, int *fd)
{
return virgl_egl_get_fd_for_texture(egl_info, tex_id, fd);
}
void virgl_renderer_reset(void)
{
vrend_renderer_reset();
}
renderer: use a thread to block for fences. Instead of polling the fences regularly, have a thread that blocks for a single fence using a separate shared context, then uses eventfd to wake up the main thread when something happens. Inside the guest, glmark2 typicially runs twice as fast with the thread sync. Although in general, the performances seems to be about +30%. The benefits is mostly for CPU-bounds tasks (when main the thread hits 100%) A naive perf stat of the vtest renderer with glmark2 "build" test with a fixed number of frames (500) results in the following stats data: (do not value timing related informations, since the renderer is ran and stopped manually) without thread: 3032.282265 task-clock (msec) # 0.420 CPUs utilized 4,277 context-switches # 0.001 M/sec 102 cpu-migrations # 0.034 K/sec 9,020 page-faults # 0.003 M/sec 7,884,098,254 cycles # 2.600 GHz 4,440,126,451 stalled-cycles-frontend # 56.32% frontend cycles idle <not supported> stalled-cycles-backend 11,024,091,578 instructions # 1.40 insns per cycle # 0.40 stalled # cycles per insn 1,091,831,588 branches # 360.069 M/sec 5,426,846 branch-misses # 0.50% of all branches with thread: 3403.592921 task-clock (msec) # 0.452 CPUs utilized 7,145 context-switches # 0.002 M/sec 410 cpu-migrations # 0.120 K/sec 6,191 page-faults # 0.002 M/sec 7,475,038,064 cycles # 2.196 GHz 4,487,043,071 stalled-cycles-frontend # 60.03% frontend cycles idle <not supported> stalled-cycles-backend 9,925,205,494 instructions # 1.33 insns per cycle # 0.45 stalled # cycles per insn 834,375,503 branches # 245.146 M/sec 4,919,995 branch-misses # 0.59% of all branches Signed-off-by: Marc-André Lureau <marcandre.lureau@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years ago
int virgl_renderer_get_poll_fd(void)
{
return vrend_renderer_get_poll_fd();
}