2025-02-10 16:56:52 +01:00
|
|
|
/* SPDX-FileCopyrightText: 2024 Blender Authors
|
2023-05-31 16:19:06 +02:00
|
|
|
*
|
|
|
|
|
* SPDX-License-Identifier: GPL-2.0-or-later */
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
|
2025-02-10 16:56:52 +01:00
|
|
|
#include "IO_string_utils.hh"
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
|
2022-05-13 09:24:28 +10:00
|
|
|
/* NOTE: we could use C++17 <charconv> from_chars to parse
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
* floats, but even if some compilers claim full support,
|
|
|
|
|
* their standard libraries are not quite there yet.
|
|
|
|
|
* LLVM/libc++ only has a float parser since LLVM 14,
|
2025-03-06 10:49:51 +11:00
|
|
|
* and GCC/libstdc++ since 11.1. So until at least these are
|
2022-04-20 09:16:24 +10:00
|
|
|
* the minimum spec, use an external library. */
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
#include "fast_float.h"
|
|
|
|
|
#include <charconv>
|
|
|
|
|
|
2025-02-10 16:56:52 +01:00
|
|
|
namespace blender::io {
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
|
|
|
|
|
StringRef read_next_line(StringRef &buffer)
|
|
|
|
|
{
|
|
|
|
|
const char *start = buffer.begin();
|
|
|
|
|
const char *end = buffer.end();
|
|
|
|
|
size_t len = 0;
|
|
|
|
|
const char *ptr = start;
|
|
|
|
|
while (ptr < end) {
|
|
|
|
|
char c = *ptr++;
|
2022-07-10 18:27:07 +03:00
|
|
|
if (c == '\n') {
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
++len;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
buffer = StringRef(ptr, end);
|
|
|
|
|
return StringRef(start, len);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static bool is_whitespace(char c)
|
|
|
|
|
{
|
2022-07-10 18:27:07 +03:00
|
|
|
return c <= ' ';
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void fixup_line_continuations(char *p, char *end)
|
|
|
|
|
{
|
|
|
|
|
while (true) {
|
|
|
|
|
/* Find next backslash, if any. */
|
|
|
|
|
char *backslash = std::find(p, end, '\\');
|
2022-07-30 13:10:35 +03:00
|
|
|
if (backslash == end) {
|
2022-07-10 18:27:07 +03:00
|
|
|
break;
|
2022-07-30 13:10:35 +03:00
|
|
|
}
|
2025-03-13 13:41:17 +11:00
|
|
|
/* Skip over possible white-space right after it. */
|
2022-07-10 18:27:07 +03:00
|
|
|
p = backslash + 1;
|
2022-07-30 13:10:35 +03:00
|
|
|
while (p < end && is_whitespace(*p) && *p != '\n') {
|
2022-07-10 18:27:07 +03:00
|
|
|
++p;
|
2022-07-30 13:10:35 +03:00
|
|
|
}
|
2022-07-10 18:27:07 +03:00
|
|
|
/* If then we have a newline, turn both backslash
|
|
|
|
|
* and the newline into regular spaces. */
|
|
|
|
|
if (p < end && *p == '\n') {
|
|
|
|
|
*backslash = ' ';
|
|
|
|
|
*p = ' ';
|
|
|
|
|
}
|
|
|
|
|
}
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
}
|
|
|
|
|
|
OBJ: improve new importer file parsing performance on windows
The OBJ parser was primarily using StringRef for convenience, with
functions like "skip whitespace" or "parse a number" taking an input
stringref, representing an input line, and returning a new stringref,
representing the remainder of the line. This is convenient, but does
more work than strictly needed -- while parsing, only the "beginning"
of the line ever changes by moving forward; the end of the line
always stays the same. We can change the code to take a pair of
pointers (begin of line, end of line) as input, and make the
functions return the new begin of line pointer. This makes the return
value neatly fit into a processor register, which StringRef did not.
On Windows, this does result in non-trivial speedups in the actual
OBJ file parsing part, due to Windows calling convention where return
values larger than 64 bits are returned via memory. Does not
measurably affect performance on Mac/Linux, because the calling
convention there uses a pair of 64-bit registers to return a
StringRef.
End-to-end times of importing several test files, on Windows
(VS2022 build, Ryzen 5950X):
- Monkey subdivided to level 6, no normals (220MB file): 1.25s -> 0.85s
- Rungholt minecraft level (270MB file): 7.0s -> 5.8s
- Blender 3 splash scene (2.4GB file): 49.1s -> 45.5s
The full import process has a lot of other overhead besides actual
OBJ file parsing (mostly creating actual blender objects out of
parsed data). In pure parsing, in the monkey test scene above, the
parsing part goes 1.0s -> 0.6s.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14936
2022-05-12 13:48:55 +03:00
|
|
|
const char *drop_whitespace(const char *p, const char *end)
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
{
|
OBJ: improve new importer file parsing performance on windows
The OBJ parser was primarily using StringRef for convenience, with
functions like "skip whitespace" or "parse a number" taking an input
stringref, representing an input line, and returning a new stringref,
representing the remainder of the line. This is convenient, but does
more work than strictly needed -- while parsing, only the "beginning"
of the line ever changes by moving forward; the end of the line
always stays the same. We can change the code to take a pair of
pointers (begin of line, end of line) as input, and make the
functions return the new begin of line pointer. This makes the return
value neatly fit into a processor register, which StringRef did not.
On Windows, this does result in non-trivial speedups in the actual
OBJ file parsing part, due to Windows calling convention where return
values larger than 64 bits are returned via memory. Does not
measurably affect performance on Mac/Linux, because the calling
convention there uses a pair of 64-bit registers to return a
StringRef.
End-to-end times of importing several test files, on Windows
(VS2022 build, Ryzen 5950X):
- Monkey subdivided to level 6, no normals (220MB file): 1.25s -> 0.85s
- Rungholt minecraft level (270MB file): 7.0s -> 5.8s
- Blender 3 splash scene (2.4GB file): 49.1s -> 45.5s
The full import process has a lot of other overhead besides actual
OBJ file parsing (mostly creating actual blender objects out of
parsed data). In pure parsing, in the monkey test scene above, the
parsing part goes 1.0s -> 0.6s.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14936
2022-05-12 13:48:55 +03:00
|
|
|
while (p < end && is_whitespace(*p)) {
|
|
|
|
|
++p;
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
}
|
OBJ: improve new importer file parsing performance on windows
The OBJ parser was primarily using StringRef for convenience, with
functions like "skip whitespace" or "parse a number" taking an input
stringref, representing an input line, and returning a new stringref,
representing the remainder of the line. This is convenient, but does
more work than strictly needed -- while parsing, only the "beginning"
of the line ever changes by moving forward; the end of the line
always stays the same. We can change the code to take a pair of
pointers (begin of line, end of line) as input, and make the
functions return the new begin of line pointer. This makes the return
value neatly fit into a processor register, which StringRef did not.
On Windows, this does result in non-trivial speedups in the actual
OBJ file parsing part, due to Windows calling convention where return
values larger than 64 bits are returned via memory. Does not
measurably affect performance on Mac/Linux, because the calling
convention there uses a pair of 64-bit registers to return a
StringRef.
End-to-end times of importing several test files, on Windows
(VS2022 build, Ryzen 5950X):
- Monkey subdivided to level 6, no normals (220MB file): 1.25s -> 0.85s
- Rungholt minecraft level (270MB file): 7.0s -> 5.8s
- Blender 3 splash scene (2.4GB file): 49.1s -> 45.5s
The full import process has a lot of other overhead besides actual
OBJ file parsing (mostly creating actual blender objects out of
parsed data). In pure parsing, in the monkey test scene above, the
parsing part goes 1.0s -> 0.6s.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14936
2022-05-12 13:48:55 +03:00
|
|
|
return p;
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
}
|
|
|
|
|
|
OBJ: improve new importer file parsing performance on windows
The OBJ parser was primarily using StringRef for convenience, with
functions like "skip whitespace" or "parse a number" taking an input
stringref, representing an input line, and returning a new stringref,
representing the remainder of the line. This is convenient, but does
more work than strictly needed -- while parsing, only the "beginning"
of the line ever changes by moving forward; the end of the line
always stays the same. We can change the code to take a pair of
pointers (begin of line, end of line) as input, and make the
functions return the new begin of line pointer. This makes the return
value neatly fit into a processor register, which StringRef did not.
On Windows, this does result in non-trivial speedups in the actual
OBJ file parsing part, due to Windows calling convention where return
values larger than 64 bits are returned via memory. Does not
measurably affect performance on Mac/Linux, because the calling
convention there uses a pair of 64-bit registers to return a
StringRef.
End-to-end times of importing several test files, on Windows
(VS2022 build, Ryzen 5950X):
- Monkey subdivided to level 6, no normals (220MB file): 1.25s -> 0.85s
- Rungholt minecraft level (270MB file): 7.0s -> 5.8s
- Blender 3 splash scene (2.4GB file): 49.1s -> 45.5s
The full import process has a lot of other overhead besides actual
OBJ file parsing (mostly creating actual blender objects out of
parsed data). In pure parsing, in the monkey test scene above, the
parsing part goes 1.0s -> 0.6s.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14936
2022-05-12 13:48:55 +03:00
|
|
|
const char *drop_non_whitespace(const char *p, const char *end)
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
{
|
OBJ: improve new importer file parsing performance on windows
The OBJ parser was primarily using StringRef for convenience, with
functions like "skip whitespace" or "parse a number" taking an input
stringref, representing an input line, and returning a new stringref,
representing the remainder of the line. This is convenient, but does
more work than strictly needed -- while parsing, only the "beginning"
of the line ever changes by moving forward; the end of the line
always stays the same. We can change the code to take a pair of
pointers (begin of line, end of line) as input, and make the
functions return the new begin of line pointer. This makes the return
value neatly fit into a processor register, which StringRef did not.
On Windows, this does result in non-trivial speedups in the actual
OBJ file parsing part, due to Windows calling convention where return
values larger than 64 bits are returned via memory. Does not
measurably affect performance on Mac/Linux, because the calling
convention there uses a pair of 64-bit registers to return a
StringRef.
End-to-end times of importing several test files, on Windows
(VS2022 build, Ryzen 5950X):
- Monkey subdivided to level 6, no normals (220MB file): 1.25s -> 0.85s
- Rungholt minecraft level (270MB file): 7.0s -> 5.8s
- Blender 3 splash scene (2.4GB file): 49.1s -> 45.5s
The full import process has a lot of other overhead besides actual
OBJ file parsing (mostly creating actual blender objects out of
parsed data). In pure parsing, in the monkey test scene above, the
parsing part goes 1.0s -> 0.6s.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14936
2022-05-12 13:48:55 +03:00
|
|
|
while (p < end && !is_whitespace(*p)) {
|
|
|
|
|
++p;
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
}
|
OBJ: improve new importer file parsing performance on windows
The OBJ parser was primarily using StringRef for convenience, with
functions like "skip whitespace" or "parse a number" taking an input
stringref, representing an input line, and returning a new stringref,
representing the remainder of the line. This is convenient, but does
more work than strictly needed -- while parsing, only the "beginning"
of the line ever changes by moving forward; the end of the line
always stays the same. We can change the code to take a pair of
pointers (begin of line, end of line) as input, and make the
functions return the new begin of line pointer. This makes the return
value neatly fit into a processor register, which StringRef did not.
On Windows, this does result in non-trivial speedups in the actual
OBJ file parsing part, due to Windows calling convention where return
values larger than 64 bits are returned via memory. Does not
measurably affect performance on Mac/Linux, because the calling
convention there uses a pair of 64-bit registers to return a
StringRef.
End-to-end times of importing several test files, on Windows
(VS2022 build, Ryzen 5950X):
- Monkey subdivided to level 6, no normals (220MB file): 1.25s -> 0.85s
- Rungholt minecraft level (270MB file): 7.0s -> 5.8s
- Blender 3 splash scene (2.4GB file): 49.1s -> 45.5s
The full import process has a lot of other overhead besides actual
OBJ file parsing (mostly creating actual blender objects out of
parsed data). In pure parsing, in the monkey test scene above, the
parsing part goes 1.0s -> 0.6s.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14936
2022-05-12 13:48:55 +03:00
|
|
|
return p;
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
}
|
|
|
|
|
|
2025-02-10 16:56:52 +01:00
|
|
|
static const char *drop_sign(const char *p, const char *end, int &sign)
|
|
|
|
|
{
|
|
|
|
|
sign = 1;
|
|
|
|
|
if (p < end) {
|
|
|
|
|
if (*p == '+') {
|
|
|
|
|
++p;
|
|
|
|
|
}
|
|
|
|
|
if (*p == '-') {
|
|
|
|
|
sign = -1;
|
|
|
|
|
++p;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
return p;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
const char *try_parse_float(
|
|
|
|
|
const char *p, const char *end, int fallback, bool &success, float &dst, bool skip_space)
|
|
|
|
|
{
|
|
|
|
|
if (skip_space) {
|
|
|
|
|
p = drop_whitespace(p, end);
|
|
|
|
|
}
|
|
|
|
|
int sign = 0;
|
|
|
|
|
p = drop_sign(p, end, sign);
|
|
|
|
|
fast_float::from_chars_result res = fast_float::from_chars(p, end, dst);
|
|
|
|
|
if (ELEM(res.ec, std::errc::invalid_argument, std::errc::result_out_of_range) || res.ptr < end) {
|
|
|
|
|
dst = fallback;
|
|
|
|
|
success = false;
|
|
|
|
|
}
|
|
|
|
|
else {
|
|
|
|
|
dst *= sign;
|
|
|
|
|
success = true;
|
|
|
|
|
}
|
|
|
|
|
return res.ptr;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
const char *try_parse_int(
|
|
|
|
|
const char *p, const char *end, int fallback, bool &success, int &dst, bool skip_space)
|
|
|
|
|
{
|
|
|
|
|
if (skip_space) {
|
|
|
|
|
p = drop_whitespace(p, end);
|
|
|
|
|
}
|
|
|
|
|
int sign = 0;
|
|
|
|
|
p = drop_sign(p, end, sign);
|
|
|
|
|
std::from_chars_result res = std::from_chars(p, end, dst);
|
|
|
|
|
if (ELEM(res.ec, std::errc::invalid_argument, std::errc::result_out_of_range) || res.ptr < end) {
|
|
|
|
|
dst = fallback;
|
|
|
|
|
success = false;
|
|
|
|
|
}
|
|
|
|
|
else {
|
|
|
|
|
dst *= sign;
|
|
|
|
|
success = true;
|
|
|
|
|
}
|
|
|
|
|
return res.ptr;
|
|
|
|
|
}
|
|
|
|
|
|
OBJ: improve new importer file parsing performance on windows
The OBJ parser was primarily using StringRef for convenience, with
functions like "skip whitespace" or "parse a number" taking an input
stringref, representing an input line, and returning a new stringref,
representing the remainder of the line. This is convenient, but does
more work than strictly needed -- while parsing, only the "beginning"
of the line ever changes by moving forward; the end of the line
always stays the same. We can change the code to take a pair of
pointers (begin of line, end of line) as input, and make the
functions return the new begin of line pointer. This makes the return
value neatly fit into a processor register, which StringRef did not.
On Windows, this does result in non-trivial speedups in the actual
OBJ file parsing part, due to Windows calling convention where return
values larger than 64 bits are returned via memory. Does not
measurably affect performance on Mac/Linux, because the calling
convention there uses a pair of 64-bit registers to return a
StringRef.
End-to-end times of importing several test files, on Windows
(VS2022 build, Ryzen 5950X):
- Monkey subdivided to level 6, no normals (220MB file): 1.25s -> 0.85s
- Rungholt minecraft level (270MB file): 7.0s -> 5.8s
- Blender 3 splash scene (2.4GB file): 49.1s -> 45.5s
The full import process has a lot of other overhead besides actual
OBJ file parsing (mostly creating actual blender objects out of
parsed data). In pure parsing, in the monkey test scene above, the
parsing part goes 1.0s -> 0.6s.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14936
2022-05-12 13:48:55 +03:00
|
|
|
static const char *drop_plus(const char *p, const char *end)
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
{
|
OBJ: improve new importer file parsing performance on windows
The OBJ parser was primarily using StringRef for convenience, with
functions like "skip whitespace" or "parse a number" taking an input
stringref, representing an input line, and returning a new stringref,
representing the remainder of the line. This is convenient, but does
more work than strictly needed -- while parsing, only the "beginning"
of the line ever changes by moving forward; the end of the line
always stays the same. We can change the code to take a pair of
pointers (begin of line, end of line) as input, and make the
functions return the new begin of line pointer. This makes the return
value neatly fit into a processor register, which StringRef did not.
On Windows, this does result in non-trivial speedups in the actual
OBJ file parsing part, due to Windows calling convention where return
values larger than 64 bits are returned via memory. Does not
measurably affect performance on Mac/Linux, because the calling
convention there uses a pair of 64-bit registers to return a
StringRef.
End-to-end times of importing several test files, on Windows
(VS2022 build, Ryzen 5950X):
- Monkey subdivided to level 6, no normals (220MB file): 1.25s -> 0.85s
- Rungholt minecraft level (270MB file): 7.0s -> 5.8s
- Blender 3 splash scene (2.4GB file): 49.1s -> 45.5s
The full import process has a lot of other overhead besides actual
OBJ file parsing (mostly creating actual blender objects out of
parsed data). In pure parsing, in the monkey test scene above, the
parsing part goes 1.0s -> 0.6s.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14936
2022-05-12 13:48:55 +03:00
|
|
|
if (p < end && *p == '+') {
|
|
|
|
|
++p;
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
}
|
OBJ: improve new importer file parsing performance on windows
The OBJ parser was primarily using StringRef for convenience, with
functions like "skip whitespace" or "parse a number" taking an input
stringref, representing an input line, and returning a new stringref,
representing the remainder of the line. This is convenient, but does
more work than strictly needed -- while parsing, only the "beginning"
of the line ever changes by moving forward; the end of the line
always stays the same. We can change the code to take a pair of
pointers (begin of line, end of line) as input, and make the
functions return the new begin of line pointer. This makes the return
value neatly fit into a processor register, which StringRef did not.
On Windows, this does result in non-trivial speedups in the actual
OBJ file parsing part, due to Windows calling convention where return
values larger than 64 bits are returned via memory. Does not
measurably affect performance on Mac/Linux, because the calling
convention there uses a pair of 64-bit registers to return a
StringRef.
End-to-end times of importing several test files, on Windows
(VS2022 build, Ryzen 5950X):
- Monkey subdivided to level 6, no normals (220MB file): 1.25s -> 0.85s
- Rungholt minecraft level (270MB file): 7.0s -> 5.8s
- Blender 3 splash scene (2.4GB file): 49.1s -> 45.5s
The full import process has a lot of other overhead besides actual
OBJ file parsing (mostly creating actual blender objects out of
parsed data). In pure parsing, in the monkey test scene above, the
parsing part goes 1.0s -> 0.6s.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14936
2022-05-12 13:48:55 +03:00
|
|
|
return p;
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
}
|
|
|
|
|
|
2022-07-07 11:34:04 +03:00
|
|
|
const char *parse_float(const char *p,
|
|
|
|
|
const char *end,
|
|
|
|
|
float fallback,
|
|
|
|
|
float &dst,
|
|
|
|
|
bool skip_space,
|
|
|
|
|
bool require_trailing_space)
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
{
|
|
|
|
|
if (skip_space) {
|
OBJ: improve new importer file parsing performance on windows
The OBJ parser was primarily using StringRef for convenience, with
functions like "skip whitespace" or "parse a number" taking an input
stringref, representing an input line, and returning a new stringref,
representing the remainder of the line. This is convenient, but does
more work than strictly needed -- while parsing, only the "beginning"
of the line ever changes by moving forward; the end of the line
always stays the same. We can change the code to take a pair of
pointers (begin of line, end of line) as input, and make the
functions return the new begin of line pointer. This makes the return
value neatly fit into a processor register, which StringRef did not.
On Windows, this does result in non-trivial speedups in the actual
OBJ file parsing part, due to Windows calling convention where return
values larger than 64 bits are returned via memory. Does not
measurably affect performance on Mac/Linux, because the calling
convention there uses a pair of 64-bit registers to return a
StringRef.
End-to-end times of importing several test files, on Windows
(VS2022 build, Ryzen 5950X):
- Monkey subdivided to level 6, no normals (220MB file): 1.25s -> 0.85s
- Rungholt minecraft level (270MB file): 7.0s -> 5.8s
- Blender 3 splash scene (2.4GB file): 49.1s -> 45.5s
The full import process has a lot of other overhead besides actual
OBJ file parsing (mostly creating actual blender objects out of
parsed data). In pure parsing, in the monkey test scene above, the
parsing part goes 1.0s -> 0.6s.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14936
2022-05-12 13:48:55 +03:00
|
|
|
p = drop_whitespace(p, end);
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
}
|
OBJ: improve new importer file parsing performance on windows
The OBJ parser was primarily using StringRef for convenience, with
functions like "skip whitespace" or "parse a number" taking an input
stringref, representing an input line, and returning a new stringref,
representing the remainder of the line. This is convenient, but does
more work than strictly needed -- while parsing, only the "beginning"
of the line ever changes by moving forward; the end of the line
always stays the same. We can change the code to take a pair of
pointers (begin of line, end of line) as input, and make the
functions return the new begin of line pointer. This makes the return
value neatly fit into a processor register, which StringRef did not.
On Windows, this does result in non-trivial speedups in the actual
OBJ file parsing part, due to Windows calling convention where return
values larger than 64 bits are returned via memory. Does not
measurably affect performance on Mac/Linux, because the calling
convention there uses a pair of 64-bit registers to return a
StringRef.
End-to-end times of importing several test files, on Windows
(VS2022 build, Ryzen 5950X):
- Monkey subdivided to level 6, no normals (220MB file): 1.25s -> 0.85s
- Rungholt minecraft level (270MB file): 7.0s -> 5.8s
- Blender 3 splash scene (2.4GB file): 49.1s -> 45.5s
The full import process has a lot of other overhead besides actual
OBJ file parsing (mostly creating actual blender objects out of
parsed data). In pure parsing, in the monkey test scene above, the
parsing part goes 1.0s -> 0.6s.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14936
2022-05-12 13:48:55 +03:00
|
|
|
p = drop_plus(p, end);
|
|
|
|
|
fast_float::from_chars_result res = fast_float::from_chars(p, end, dst);
|
2022-09-23 15:24:20 +10:00
|
|
|
if (ELEM(res.ec, std::errc::invalid_argument, std::errc::result_out_of_range)) {
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
dst = fallback;
|
|
|
|
|
}
|
2022-07-07 11:34:04 +03:00
|
|
|
else if (require_trailing_space && res.ptr < end && !is_whitespace(*res.ptr)) {
|
|
|
|
|
/* If there are trailing non-space characters, do not eat up the number. */
|
|
|
|
|
dst = fallback;
|
|
|
|
|
return p;
|
|
|
|
|
}
|
OBJ: improve new importer file parsing performance on windows
The OBJ parser was primarily using StringRef for convenience, with
functions like "skip whitespace" or "parse a number" taking an input
stringref, representing an input line, and returning a new stringref,
representing the remainder of the line. This is convenient, but does
more work than strictly needed -- while parsing, only the "beginning"
of the line ever changes by moving forward; the end of the line
always stays the same. We can change the code to take a pair of
pointers (begin of line, end of line) as input, and make the
functions return the new begin of line pointer. This makes the return
value neatly fit into a processor register, which StringRef did not.
On Windows, this does result in non-trivial speedups in the actual
OBJ file parsing part, due to Windows calling convention where return
values larger than 64 bits are returned via memory. Does not
measurably affect performance on Mac/Linux, because the calling
convention there uses a pair of 64-bit registers to return a
StringRef.
End-to-end times of importing several test files, on Windows
(VS2022 build, Ryzen 5950X):
- Monkey subdivided to level 6, no normals (220MB file): 1.25s -> 0.85s
- Rungholt minecraft level (270MB file): 7.0s -> 5.8s
- Blender 3 splash scene (2.4GB file): 49.1s -> 45.5s
The full import process has a lot of other overhead besides actual
OBJ file parsing (mostly creating actual blender objects out of
parsed data). In pure parsing, in the monkey test scene above, the
parsing part goes 1.0s -> 0.6s.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14936
2022-05-12 13:48:55 +03:00
|
|
|
return res.ptr;
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
}
|
|
|
|
|
|
2022-07-07 11:34:04 +03:00
|
|
|
const char *parse_floats(const char *p,
|
|
|
|
|
const char *end,
|
|
|
|
|
float fallback,
|
|
|
|
|
float *dst,
|
|
|
|
|
int count,
|
|
|
|
|
bool require_trailing_space)
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
{
|
|
|
|
|
for (int i = 0; i < count; ++i) {
|
2022-07-07 11:34:04 +03:00
|
|
|
p = parse_float(p, end, fallback, dst[i], true, require_trailing_space);
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
}
|
OBJ: improve new importer file parsing performance on windows
The OBJ parser was primarily using StringRef for convenience, with
functions like "skip whitespace" or "parse a number" taking an input
stringref, representing an input line, and returning a new stringref,
representing the remainder of the line. This is convenient, but does
more work than strictly needed -- while parsing, only the "beginning"
of the line ever changes by moving forward; the end of the line
always stays the same. We can change the code to take a pair of
pointers (begin of line, end of line) as input, and make the
functions return the new begin of line pointer. This makes the return
value neatly fit into a processor register, which StringRef did not.
On Windows, this does result in non-trivial speedups in the actual
OBJ file parsing part, due to Windows calling convention where return
values larger than 64 bits are returned via memory. Does not
measurably affect performance on Mac/Linux, because the calling
convention there uses a pair of 64-bit registers to return a
StringRef.
End-to-end times of importing several test files, on Windows
(VS2022 build, Ryzen 5950X):
- Monkey subdivided to level 6, no normals (220MB file): 1.25s -> 0.85s
- Rungholt minecraft level (270MB file): 7.0s -> 5.8s
- Blender 3 splash scene (2.4GB file): 49.1s -> 45.5s
The full import process has a lot of other overhead besides actual
OBJ file parsing (mostly creating actual blender objects out of
parsed data). In pure parsing, in the monkey test scene above, the
parsing part goes 1.0s -> 0.6s.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14936
2022-05-12 13:48:55 +03:00
|
|
|
return p;
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
}
|
|
|
|
|
|
OBJ: improve new importer file parsing performance on windows
The OBJ parser was primarily using StringRef for convenience, with
functions like "skip whitespace" or "parse a number" taking an input
stringref, representing an input line, and returning a new stringref,
representing the remainder of the line. This is convenient, but does
more work than strictly needed -- while parsing, only the "beginning"
of the line ever changes by moving forward; the end of the line
always stays the same. We can change the code to take a pair of
pointers (begin of line, end of line) as input, and make the
functions return the new begin of line pointer. This makes the return
value neatly fit into a processor register, which StringRef did not.
On Windows, this does result in non-trivial speedups in the actual
OBJ file parsing part, due to Windows calling convention where return
values larger than 64 bits are returned via memory. Does not
measurably affect performance on Mac/Linux, because the calling
convention there uses a pair of 64-bit registers to return a
StringRef.
End-to-end times of importing several test files, on Windows
(VS2022 build, Ryzen 5950X):
- Monkey subdivided to level 6, no normals (220MB file): 1.25s -> 0.85s
- Rungholt minecraft level (270MB file): 7.0s -> 5.8s
- Blender 3 splash scene (2.4GB file): 49.1s -> 45.5s
The full import process has a lot of other overhead besides actual
OBJ file parsing (mostly creating actual blender objects out of
parsed data). In pure parsing, in the monkey test scene above, the
parsing part goes 1.0s -> 0.6s.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14936
2022-05-12 13:48:55 +03:00
|
|
|
const char *parse_int(const char *p, const char *end, int fallback, int &dst, bool skip_space)
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
{
|
|
|
|
|
if (skip_space) {
|
OBJ: improve new importer file parsing performance on windows
The OBJ parser was primarily using StringRef for convenience, with
functions like "skip whitespace" or "parse a number" taking an input
stringref, representing an input line, and returning a new stringref,
representing the remainder of the line. This is convenient, but does
more work than strictly needed -- while parsing, only the "beginning"
of the line ever changes by moving forward; the end of the line
always stays the same. We can change the code to take a pair of
pointers (begin of line, end of line) as input, and make the
functions return the new begin of line pointer. This makes the return
value neatly fit into a processor register, which StringRef did not.
On Windows, this does result in non-trivial speedups in the actual
OBJ file parsing part, due to Windows calling convention where return
values larger than 64 bits are returned via memory. Does not
measurably affect performance on Mac/Linux, because the calling
convention there uses a pair of 64-bit registers to return a
StringRef.
End-to-end times of importing several test files, on Windows
(VS2022 build, Ryzen 5950X):
- Monkey subdivided to level 6, no normals (220MB file): 1.25s -> 0.85s
- Rungholt minecraft level (270MB file): 7.0s -> 5.8s
- Blender 3 splash scene (2.4GB file): 49.1s -> 45.5s
The full import process has a lot of other overhead besides actual
OBJ file parsing (mostly creating actual blender objects out of
parsed data). In pure parsing, in the monkey test scene above, the
parsing part goes 1.0s -> 0.6s.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14936
2022-05-12 13:48:55 +03:00
|
|
|
p = drop_whitespace(p, end);
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
}
|
OBJ: improve new importer file parsing performance on windows
The OBJ parser was primarily using StringRef for convenience, with
functions like "skip whitespace" or "parse a number" taking an input
stringref, representing an input line, and returning a new stringref,
representing the remainder of the line. This is convenient, but does
more work than strictly needed -- while parsing, only the "beginning"
of the line ever changes by moving forward; the end of the line
always stays the same. We can change the code to take a pair of
pointers (begin of line, end of line) as input, and make the
functions return the new begin of line pointer. This makes the return
value neatly fit into a processor register, which StringRef did not.
On Windows, this does result in non-trivial speedups in the actual
OBJ file parsing part, due to Windows calling convention where return
values larger than 64 bits are returned via memory. Does not
measurably affect performance on Mac/Linux, because the calling
convention there uses a pair of 64-bit registers to return a
StringRef.
End-to-end times of importing several test files, on Windows
(VS2022 build, Ryzen 5950X):
- Monkey subdivided to level 6, no normals (220MB file): 1.25s -> 0.85s
- Rungholt minecraft level (270MB file): 7.0s -> 5.8s
- Blender 3 splash scene (2.4GB file): 49.1s -> 45.5s
The full import process has a lot of other overhead besides actual
OBJ file parsing (mostly creating actual blender objects out of
parsed data). In pure parsing, in the monkey test scene above, the
parsing part goes 1.0s -> 0.6s.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14936
2022-05-12 13:48:55 +03:00
|
|
|
p = drop_plus(p, end);
|
|
|
|
|
std::from_chars_result res = std::from_chars(p, end, dst);
|
2022-09-23 15:24:20 +10:00
|
|
|
if (ELEM(res.ec, std::errc::invalid_argument, std::errc::result_out_of_range)) {
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
dst = fallback;
|
|
|
|
|
}
|
OBJ: improve new importer file parsing performance on windows
The OBJ parser was primarily using StringRef for convenience, with
functions like "skip whitespace" or "parse a number" taking an input
stringref, representing an input line, and returning a new stringref,
representing the remainder of the line. This is convenient, but does
more work than strictly needed -- while parsing, only the "beginning"
of the line ever changes by moving forward; the end of the line
always stays the same. We can change the code to take a pair of
pointers (begin of line, end of line) as input, and make the
functions return the new begin of line pointer. This makes the return
value neatly fit into a processor register, which StringRef did not.
On Windows, this does result in non-trivial speedups in the actual
OBJ file parsing part, due to Windows calling convention where return
values larger than 64 bits are returned via memory. Does not
measurably affect performance on Mac/Linux, because the calling
convention there uses a pair of 64-bit registers to return a
StringRef.
End-to-end times of importing several test files, on Windows
(VS2022 build, Ryzen 5950X):
- Monkey subdivided to level 6, no normals (220MB file): 1.25s -> 0.85s
- Rungholt minecraft level (270MB file): 7.0s -> 5.8s
- Blender 3 splash scene (2.4GB file): 49.1s -> 45.5s
The full import process has a lot of other overhead besides actual
OBJ file parsing (mostly creating actual blender objects out of
parsed data). In pure parsing, in the monkey test scene above, the
parsing part goes 1.0s -> 0.6s.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14936
2022-05-12 13:48:55 +03:00
|
|
|
return res.ptr;
|
OBJ: further optimize, cleanup and harden the new C++ importer
Continued improvements to the new C++ based OBJ importer.
Performance: about 2x faster.
- Rungholt.obj (several meshes, 263MB file): Windows 12.7s -> 5.9s, Mac 7.7s -> 3.1s.
- Blender 3.0 splash (24k meshes, 2.4GB file): Windows 97.3s -> 53.6s, Mac 137.3s -> 80.0s.
- "Windows" is VS2022, AMD Ryzen 5950X (32 threads), "Mac" is Xcode/clang 13, M1Max (10 threads).
- Slightly reduced memory usage during import as well.
The performance gains are a combination of several things:
- Replacing `std::stof` / `std::stoi` with C++17 `from_chars`.
- Stop reading input file char-by-char using `std::getline`, and instead read in 64kb chunks, and parse from there (taking care of possibly handling lines split mid-way due to chunk boundaries).
- Removing abstractions for splitting a line by some char,
- Avoid tiny memory allocations: instead of storing a vector of polygon corners in each face, store all the corners in one big array, and per-face only store indices "where do corners start, and how many". Likewise, don't store full string names of material/group names for each face; only store indices into overall material/group names arrays.
- Stop always doing mesh validation, which is slow. Do it just like the Alembic importer does: only do validation if found some invalid faces during import, or if requested by the user via an import setting checkbox (which defaults to off).
- Stop doing "collection sync" for each object being added; instead do the collection sync right after creating all the objects.
Cleanup / Robustness:
This reworking of parser (see "removing abstractions" point above) means that all the functions that were in `parser_string_utils` file are gone, and replaced with different set of functions. However they are not OBJ specific, so as pointed out during review of the previous differential, they are now in `source/blender/io/common` library.
Added gtest coverage for said functions as well; something that was only indirectly covered by obj tests previously.
Rework of some bits of parsing made the parser actually better able to deal with invalid syntax. E.g. previously, if a face corner were a `/123` string, it would have incorrectly treated that as a vertex index (since it would get "hey that's one number" after splitting a string by a slash), instead of properly marking it as invalid syntax.
Added gtest coverage for .mtl parsing; something that was not covered by any tests at all previously.
Reviewed By: Howard Trickey
Differential Revision: https://developer.blender.org/D14586
2022-04-17 22:07:43 +03:00
|
|
|
}
|
|
|
|
|
|
2025-02-10 16:56:52 +01:00
|
|
|
} // namespace blender::io
|