mpv/video/out/gpu/user_shaders.c

464 lines
14 KiB
C
Raw Normal View History

/*
* This file is part of mpv.
*
* mpv is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* mpv is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with mpv. If not, see <http://www.gnu.org/licenses/>.
*/
#include <assert.h>
#include <math.h>
vo_opengl: refactor into vo_gpu This is done in several steps: 1. refactor MPGLContext -> struct ra_ctx 2. move GL-specific stuff in vo_opengl into opengl/context.c 3. generalize context creation to support other APIs, and add --gpu-api 4. rename all of the --opengl- options that are no longer opengl-specific 5. move all of the stuff from opengl/* that isn't GL-specific into gpu/ (note: opengl/gl_utils.h became opengl/utils.h) 6. rename vo_opengl to vo_gpu 7. to handle window screenshots, the short-term approach was to just add it to ra_swchain_fns. Long term (and for vulkan) this has to be moved to ra itself (and vo_gpu altered to compensate), but this was a stop-gap measure to prevent this commit from getting too big 8. move ra->fns->flush to ra_gl_ctx instead 9. some other minor changes that I've probably already forgotten Note: This is one half of a major refactor, the other half of which is provided by rossy's following commit. This commit enables support for all linux platforms, while his version enables support for all non-linux platforms. Note 2: vo_opengl_cb.c also re-uses ra_gl_ctx so it benefits from the --opengl- options like --opengl-early-flush, --opengl-finish etc. Should be a strict superset of the old functionality. Disclaimer: Since I have no way of compiling mpv on all platforms, some of these ports were done blindly. Specifically, the blind ports included context_mali_fbdev.c and context_rpi.c. Since they're both based on egl_helpers, the port should have gone smoothly without any major changes required. But if somebody complains about a compile error on those platforms (assuming anybody actually uses them), you know where to complain.
2017-09-14 06:04:55 +00:00
#include "common/msg.h"
#include "misc/ctype.h"
#include "user_shaders.h"
static bool parse_rpn_szexpr(struct bstr line, struct szexp out[MAX_SZEXP_SIZE])
{
int pos = 0;
while (line.len > 0) {
struct bstr word = bstr_strip(bstr_splitchar(line, &line, ' '));
if (word.len == 0)
continue;
if (pos >= MAX_SZEXP_SIZE)
return false;
struct szexp *exp = &out[pos++];
if (bstr_eatend0(&word, ".w") || bstr_eatend0(&word, ".width")) {
exp->tag = SZEXP_VAR_W;
exp->val.varname = word;
continue;
}
if (bstr_eatend0(&word, ".h") || bstr_eatend0(&word, ".height")) {
exp->tag = SZEXP_VAR_H;
exp->val.varname = word;
continue;
}
switch (word.start[0]) {
case '+': exp->tag = SZEXP_OP2; exp->val.op = SZEXP_OP_ADD; continue;
case '-': exp->tag = SZEXP_OP2; exp->val.op = SZEXP_OP_SUB; continue;
case '*': exp->tag = SZEXP_OP2; exp->val.op = SZEXP_OP_MUL; continue;
case '/': exp->tag = SZEXP_OP2; exp->val.op = SZEXP_OP_DIV; continue;
case '%': exp->tag = SZEXP_OP2; exp->val.op = SZEXP_OP_MOD; continue;
case '!': exp->tag = SZEXP_OP1; exp->val.op = SZEXP_OP_NOT; continue;
case '>': exp->tag = SZEXP_OP2; exp->val.op = SZEXP_OP_GT; continue;
case '<': exp->tag = SZEXP_OP2; exp->val.op = SZEXP_OP_LT; continue;
case '=': exp->tag = SZEXP_OP2; exp->val.op = SZEXP_OP_EQ; continue;
}
if (mp_isdigit(word.start[0])) {
exp->tag = SZEXP_CONST;
if (bstr_sscanf(word, "%f", &exp->val.cval) != 1)
return false;
continue;
}
// Some sort of illegal expression
return false;
}
return true;
}
// Returns whether successful. 'result' is left untouched on failure
bool eval_szexpr(struct mp_log *log, void *priv,
bool (*lookup)(void *priv, struct bstr var, float size[2]),
struct szexp expr[MAX_SZEXP_SIZE], float *result)
{
float stack[MAX_SZEXP_SIZE] = {0};
int idx = 0; // points to next element to push
for (int i = 0; i < MAX_SZEXP_SIZE; i++) {
switch (expr[i].tag) {
case SZEXP_END:
goto done;
case SZEXP_CONST:
// Since our SZEXPs are bound by MAX_SZEXP_SIZE, it should be
// impossible to overflow the stack
assert(idx < MAX_SZEXP_SIZE);
stack[idx++] = expr[i].val.cval;
continue;
case SZEXP_OP1:
if (idx < 1) {
mp_warn(log, "Stack underflow in RPN expression!\n");
return false;
}
switch (expr[i].val.op) {
case SZEXP_OP_NOT: stack[idx-1] = !stack[idx-1]; break;
default: MP_ASSERT_UNREACHABLE();
}
continue;
case SZEXP_OP2:
if (idx < 2) {
mp_warn(log, "Stack underflow in RPN expression!\n");
return false;
}
// Pop the operands in reverse order
float op2 = stack[--idx];
float op1 = stack[--idx];
float res = 0.0;
switch (expr[i].val.op) {
case SZEXP_OP_ADD: res = op1 + op2; break;
case SZEXP_OP_SUB: res = op1 - op2; break;
case SZEXP_OP_MUL: res = op1 * op2; break;
case SZEXP_OP_DIV: res = op1 / op2; break;
case SZEXP_OP_MOD: res = fmodf(op1, op2); break;
case SZEXP_OP_GT: res = op1 > op2; break;
case SZEXP_OP_LT: res = op1 < op2; break;
case SZEXP_OP_EQ: res = op1 == op2; break;
default: MP_ASSERT_UNREACHABLE();
}
if (!isfinite(res)) {
mp_warn(log, "Illegal operation in RPN expression!\n");
return false;
}
stack[idx++] = res;
continue;
case SZEXP_VAR_W:
case SZEXP_VAR_H: {
struct bstr name = expr[i].val.varname;
float size[2];
if (!lookup(priv, name, size)) {
mp_warn(log, "Variable %.*s not found in RPN expression!\n",
BSTR_P(name));
return false;
}
stack[idx++] = (expr[i].tag == SZEXP_VAR_W) ? size[0] : size[1];
continue;
}
}
}
done:
// Return the single stack element
if (idx != 1) {
mp_warn(log, "Malformed stack after RPN expression!\n");
return false;
}
*result = stack[0];
return true;
}
static bool parse_hook(struct mp_log *log, struct bstr *body,
struct gl_user_shader_hook *out)
{
*out = (struct gl_user_shader_hook){
.pass_desc = bstr0("(unknown)"),
.offset = identity_trans,
.align_offset = false,
.width = {{ SZEXP_VAR_W, { .varname = bstr0("HOOKED") }}},
.height = {{ SZEXP_VAR_H, { .varname = bstr0("HOOKED") }}},
.cond = {{ SZEXP_CONST, { .cval = 1.0 }}},
};
int hook_idx = 0;
int bind_idx = 0;
// Parse all headers
while (true) {
struct bstr rest;
struct bstr line = bstr_strip(bstr_getline(*body, &rest));
// Check for the presence of the magic line beginning
if (!bstr_eatstart0(&line, "//!"))
break;
*body = rest;
// Parse the supported commands
if (bstr_eatstart0(&line, "HOOK")) {
if (hook_idx == SHADER_MAX_HOOKS) {
mp_err(log, "Passes may only hook up to %d textures!\n",
SHADER_MAX_HOOKS);
return false;
}
out->hook_tex[hook_idx++] = bstr_strip(line);
continue;
}
if (bstr_eatstart0(&line, "BIND")) {
if (bind_idx == SHADER_MAX_BINDS) {
mp_err(log, "Passes may only bind up to %d textures!\n",
SHADER_MAX_BINDS);
return false;
}
out->bind_tex[bind_idx++] = bstr_strip(line);
continue;
}
if (bstr_eatstart0(&line, "SAVE")) {
out->save_tex = bstr_strip(line);
continue;
}
vo_opengl: refactor vo performance subsystem This replaces `vo-performance` by `vo-passes`, bringing with it a number of changes and improvements: 1. mpv users can now introspect the vo_opengl passes, which is something that has been requested multiple times. 2. performance data is now measured per-pass, which helps both development and debugging. 3. since adding more passes is cheap, we can now report information for more passes (e.g. the blit pass, and the osd pass). Note: we also switch to nanosecond scale, to be able to measure these passes better. 4. `--user-shaders` authors can now describe their own passes, helping users both identify which user shaders are active at any given time as well as helping shader authors identify performance issues. 5. the timing data per pass is now exported as a full list of samples, so projects like Argon-/mpv-stats can immediately read out all of the samples and render a graph without having to manually poll this option constantly. Due to gl_timer's design being complicated (directly reading performance data would block, so we delay the actual read-back until the next _start command), it's vital not to conflate different passes that might be doing different things from one frame to another. To accomplish this, the actual timers are stored as part of the gl_shader_cache's sc_entry, which makes them unique for that exact shader. Starting and stopping the time measurement is easy to unify with the gl_sc architecture, because the existing API already relies on a "generate, render, reset" flow, so we can just put timer_start and timer_stop in sc_generate and sc_reset, respectively. The ugliest thing about this code is that due to the need to keep pass information relatively stable in between frames, we need to distinguish between "new" and "redrawn" frames, which bloats the code somewhat and also feels hacky and vo_opengl-specific. (But then again, this entire thing is vo_opengl-specific)
2017-06-29 15:00:06 +00:00
if (bstr_eatstart0(&line, "DESC")) {
out->pass_desc = bstr_strip(line);
vo_opengl: refactor vo performance subsystem This replaces `vo-performance` by `vo-passes`, bringing with it a number of changes and improvements: 1. mpv users can now introspect the vo_opengl passes, which is something that has been requested multiple times. 2. performance data is now measured per-pass, which helps both development and debugging. 3. since adding more passes is cheap, we can now report information for more passes (e.g. the blit pass, and the osd pass). Note: we also switch to nanosecond scale, to be able to measure these passes better. 4. `--user-shaders` authors can now describe their own passes, helping users both identify which user shaders are active at any given time as well as helping shader authors identify performance issues. 5. the timing data per pass is now exported as a full list of samples, so projects like Argon-/mpv-stats can immediately read out all of the samples and render a graph without having to manually poll this option constantly. Due to gl_timer's design being complicated (directly reading performance data would block, so we delay the actual read-back until the next _start command), it's vital not to conflate different passes that might be doing different things from one frame to another. To accomplish this, the actual timers are stored as part of the gl_shader_cache's sc_entry, which makes them unique for that exact shader. Starting and stopping the time measurement is easy to unify with the gl_sc architecture, because the existing API already relies on a "generate, render, reset" flow, so we can just put timer_start and timer_stop in sc_generate and sc_reset, respectively. The ugliest thing about this code is that due to the need to keep pass information relatively stable in between frames, we need to distinguish between "new" and "redrawn" frames, which bloats the code somewhat and also feels hacky and vo_opengl-specific. (But then again, this entire thing is vo_opengl-specific)
2017-06-29 15:00:06 +00:00
continue;
}
if (bstr_eatstart0(&line, "OFFSET")) {
line = bstr_strip(line);
if (bstr_equals0(line, "ALIGN")) {
out->align_offset = true;
} else {
float ox, oy;
if (bstr_sscanf(line, "%f %f", &ox, &oy) != 2) {
mp_err(log, "Error while parsing OFFSET!\n");
return false;
}
out->offset.t[0] = ox;
out->offset.t[1] = oy;
}
continue;
}
if (bstr_eatstart0(&line, "WIDTH")) {
if (!parse_rpn_szexpr(line, out->width)) {
mp_err(log, "Error while parsing WIDTH!\n");
return false;
}
continue;
}
if (bstr_eatstart0(&line, "HEIGHT")) {
if (!parse_rpn_szexpr(line, out->height)) {
mp_err(log, "Error while parsing HEIGHT!\n");
return false;
}
continue;
}
if (bstr_eatstart0(&line, "WHEN")) {
if (!parse_rpn_szexpr(line, out->cond)) {
mp_err(log, "Error while parsing WHEN!\n");
return false;
}
continue;
}
if (bstr_eatstart0(&line, "COMPONENTS")) {
if (bstr_sscanf(line, "%d", &out->components) != 1) {
mp_err(log, "Error while parsing COMPONENTS!\n");
return false;
}
continue;
}
if (bstr_eatstart0(&line, "COMPUTE")) {
struct compute_info *ci = &out->compute;
int num = bstr_sscanf(line, "%d %d %d %d", &ci->block_w, &ci->block_h,
&ci->threads_w, &ci->threads_h);
if (num == 2 || num == 4) {
ci->active = true;
ci->directly_writes = true;
} else {
mp_err(log, "Error while parsing COMPUTE!\n");
return false;
}
continue;
}
// Unknown command type
mp_err(log, "Unrecognized command '%.*s'!\n", BSTR_P(line));
return false;
}
// The rest of the file up until the next magic line beginning (if any)
// shall be the shader body
if (bstr_split_tok(*body, "//!", &out->pass_body, body)) {
// Make sure the magic line is part of the rest
body->start -= 3;
body->len += 3;
}
// Sanity checking
if (hook_idx == 0)
mp_warn(log, "Pass has no hooked textures (will be ignored)!\n");
return true;
}
static bool parse_tex(struct mp_log *log, struct ra *ra, struct bstr *body,
struct gl_user_shader_tex *out)
{
*out = (struct gl_user_shader_tex){
.name = bstr0("USER_TEX"),
.params = {
.dimensions = 2,
.w = 1, .h = 1, .d = 1,
.render_src = true,
.src_linear = true,
},
};
struct ra_tex_params *p = &out->params;
while (true) {
struct bstr rest;
struct bstr line = bstr_strip(bstr_getline(*body, &rest));
if (!bstr_eatstart0(&line, "//!"))
break;
*body = rest;
if (bstr_eatstart0(&line, "TEXTURE")) {
out->name = bstr_strip(line);
continue;
}
if (bstr_eatstart0(&line, "SIZE")) {
p->dimensions = bstr_sscanf(line, "%d %d %d", &p->w, &p->h, &p->d);
if (p->dimensions < 1 || p->dimensions > 3 ||
p->w < 1 || p->h < 1 || p->d < 1)
{
mp_err(log, "Error while parsing SIZE!\n");
return false;
}
continue;
}
if (bstr_eatstart0(&line, "FORMAT ")) {
p->format = NULL;
for (int n = 0; n < ra->num_formats; n++) {
const struct ra_format *fmt = ra->formats[n];
if (bstr_equals0(line, fmt->name)) {
p->format = fmt;
break;
}
}
// (pixel_size==0 is for opaque formats)
if (!p->format || !p->format->pixel_size) {
mp_err(log, "Unrecognized/unavailable FORMAT name: '%.*s'!\n",
BSTR_P(line));
return false;
}
continue;
}
if (bstr_eatstart0(&line, "FILTER")) {
line = bstr_strip(line);
if (bstr_equals0(line, "LINEAR")) {
p->src_linear = true;
} else if (bstr_equals0(line, "NEAREST")) {
p->src_linear = false;
} else {
mp_err(log, "Unrecognized FILTER: '%.*s'!\n", BSTR_P(line));
return false;
}
continue;
}
if (bstr_eatstart0(&line, "BORDER")) {
line = bstr_strip(line);
if (bstr_equals0(line, "CLAMP")) {
p->src_repeat = false;
} else if (bstr_equals0(line, "REPEAT")) {
p->src_repeat = true;
} else {
mp_err(log, "Unrecognized BORDER: '%.*s'!\n", BSTR_P(line));
return false;
}
continue;
}
mp_err(log, "Unrecognized command '%.*s'!\n", BSTR_P(line));
return false;
}
if (!p->format) {
mp_err(log, "No FORMAT specified.\n");
return false;
}
if (p->src_linear && !p->format->linear_filter) {
mp_err(log, "The specified texture format cannot be filtered!\n");
return false;
}
// Decode the rest of the section (up to the next //! marker) as raw hex
// data for the texture
struct bstr hexdata;
if (bstr_split_tok(*body, "//!", &hexdata, body)) {
// Make sure the magic line is part of the rest
body->start -= 3;
body->len += 3;
}
struct bstr tex;
if (!bstr_decode_hex(NULL, bstr_strip(hexdata), &tex)) {
mp_err(log, "Error while parsing TEXTURE body: must be a valid "
"hexadecimal sequence, on a single line!\n");
return false;
}
int expected_len = p->w * p->h * p->d * p->format->pixel_size;
if (tex.len != expected_len) {
mp_err(log, "Shader TEXTURE size mismatch: got %zd bytes, expected %d!\n",
tex.len, expected_len);
talloc_free(tex.start);
return false;
}
p->initial_data = tex.start;
return true;
}
void parse_user_shader(struct mp_log *log, struct ra *ra, struct bstr shader,
void *priv,
bool (*dohook)(void *p, const struct gl_user_shader_hook *hook),
bool (*dotex)(void *p, struct gl_user_shader_tex tex))
{
if (!dohook || !dotex || !shader.len)
return;
// Skip all garbage (e.g. comments) before the first header
int pos = bstr_find(shader, bstr0("//!"));
if (pos < 0) {
mp_warn(log, "Shader appears to contain no headers!\n");
return;
}
shader = bstr_cut(shader, pos);
// Loop over the file
while (shader.len > 0)
{
// Peek at the first header to dispatch the right type
if (bstr_startswith0(shader, "//!TEXTURE")) {
struct gl_user_shader_tex t;
if (!parse_tex(log, ra, &shader, &t) || !dotex(priv, t))
return;
continue;
}
struct gl_user_shader_hook h;
if (!parse_hook(log, &shader, &h) || !dohook(priv, &h))
return;
}
}