Considering a move to mastodon.gamedev.place from mastodon.social. I'm sure the people over there love my DOS adventures :p
How bad of an idea is it?
| web | https://marioslab.io |
| https://twitter.com/badlogicgames | |
| github | https://github.com/badlogic |
Mario Zechner
@badlogicgames
- 6 Followers
- 238 Following
- 431 Posts
libGDX, Beginning Android Games, RoboVM, Spine. Member of the council of Warios.
Well, hello there. I suppose now I can really call it "Rendering like it's 1996"
I actually remember the freaking interrupts...
This was my bible: http://www2.ift.ulaval.ca/~marchand/ift17583/dosints.pdf
RayJust published rTexPacker v2.5, a simple and easy-to-use sprites packer and font atlas generator. Engine-agnostic, full-featured, multiplatform, portable, powerful CLI... contained in a single 1MB executable!
It can also be used online for free: https://raylibtech.itch.io/rtexpacker
Feedback is welcome! 😀
Want to view the control flow graph of some disassembly from Godbolt's Compiler Explorer? Here you go:
https://marioslab.io/projects/cfg
Made it a little prettier. Now also pretty decent on mobile. ARM assembly support in the works. There be dragons.
I dearly miss a CFG/basic block view in Compiler Explorer. So I built myself one with duct tape.
https://marioslab.io/projects/cfg/
Paste the Compiler Explorer disassembly on the left, select the function and view the CFG on the right.
Should work with output from MSVC/Clang/GCC
It's part 3. MSVC made me so mad, I can't even count gud anymore.
Seriously though, MSVC is really not good at auto-vectorizing trivial loops. It fails to vectorize both `r96_rect()` and `blit()`, but it generates a pretty good vectorized version of `blit_keyed()`. Which beats `r96_rect()`. WTF
Compiler Explorer - C
typedef struct r96_image { int32_t width, height; uint32_t *pixels; } r96_image; void r96_rect(r96_image *image, int32_t x1, int32_t y1, int32_t width, int32_t height, uint32_t color) { if (width <= 0) return; if (height <= 0) return; int32_t x2 = x1 + width - 1; int32_t y2 = y1 + height - 1; if (x1 >= image->width) return; if (x2 < 0) return; if (y1 >= image->height) return; if (y2 < 0) return; if (x1 < 0) x1 = 0; if (y1 < 0) y1 = 0; if (x2 >= image->width) x2 = image->width - 1; if (y2 >= image->height) y2 = image->height - 1; int32_t clipped_width = x2 - x1 + 1; int32_t next_row = image->width - clipped_width; uint32_t *pixel = image->pixels + y1 * image->width + x1; for (int y = y1; y <= y2; y++) { for (int i = 0; i < clipped_width; i++) { *pixel++ = color; } pixel += next_row; } } void blit(r96_image *dst, r96_image *src, int x, int y) { int32_t dst_x1 = x; int32_t dst_y1 = y; int32_t dst_x2 = x + src->width - 1; int32_t dst_y2 = y + src->height - 1; int32_t src_x1 = 0; int32_t src_y1 = 0; if (dst_x1 >= dst->width) return; if (dst_x2 < 0) return; if (dst_y1 >= dst->height) return; if (dst_y2 < 0) return; if (dst_x1 < 0) { src_x1 -= dst_x1; dst_x1 = 0; } if (dst_y1 < 0) { src_y1 -= dst_y1; dst_y1 = 0; } if (dst_x2 >= dst->width) dst_x2 = dst->width - 1; if (dst_y2 >= dst->height) dst_y2 = dst->height - 1; int32_t clipped_width = dst_x2 - dst_x1 + 1; int32_t dst_next_row = dst->width - clipped_width; int32_t src_next_row = src->width - clipped_width; uint32_t *dst_pixel = dst->pixels + dst_y1 * dst->width + dst_x1; uint32_t *src_pixel = src->pixels + src_y1 * src->width + src_x1; for (int y = dst_y1; y <= dst_y2; y++) { for (int i = 0; i < clipped_width; i++) { *dst_pixel++ = *src_pixel++; } dst_pixel += dst_next_row; src_pixel += src_next_row; } } void blit_keyed(r96_image *dst, r96_image *src, int x, int y, uint32_t color_key) { int32_t dst_x1 = x; int32_t dst_y1 = y; int32_t dst_x2 = x + src->width - 1; int32_t dst_y2 = y + src->height - 1; int32_t src_x1 = 0; int32_t src_y1 = 0; if (dst_x1 >= dst->width) return; if (dst_x2 < 0) return; if (dst_y1 >= dst->height) return; if (dst_y2 < 0) return; if (dst_x1 < 0) { src_x1 -= dst_x1; dst_x1 = 0; } if (dst_y1 < 0) { src_y1 -= dst_y1; dst_y1 = 0; } if (dst_x2 >= dst->width) dst_x2 = dst->width - 1; if (dst_y2 >= dst->height) dst_y2 = dst->height - 1; int32_t clipped_width = dst_x2 - dst_x1 + 1; int32_t dst_next_row = dst->width - clipped_width; int32_t src_next_row = src->width - clipped_width; uint32_t *dst_pixel = dst->pixels + dst_y1 * dst->width + dst_x1; uint32_t *src_pixel = src->pixels + src_y1 * src->width + src_x1; for (y = dst_y1; y <= dst_y2; y++) { for (int i = 0; i < clipped_width; i++) { uint32_t src_color = *src_pixel; uint32_t dst_color = *dst_pixel; *dst_pixel = src_color != color_key ? src_color : dst_color; src_pixel++; dst_pixel++; } dst_pixel += dst_next_row; src_pixel += src_next_row; } }
Seriously though, MSVC is really not good at auto-vectorizing trivial loops. It fails to vectorize both `r96_rect()` and `blit()`, but it generates a pretty good vectorized version of `blit_keyed()`. Which beats `r96_rect()`. WTF
Compiler Explorer - C
typedef struct r96_image { int32_t width, height; uint32_t *pixels; } r96_image; void r96_rect(r96_image *image, int32_t x1, int32_t y1, int32_t width, int32_t height, uint32_t color) { if (width <= 0) return; if (height <= 0) return; int32_t x2 = x1 + width - 1; int32_t y2 = y1 + height - 1; if (x1 >= image->width) return; if (x2 < 0) return; if (y1 >= image->height) return; if (y2 < 0) return; if (x1 < 0) x1 = 0; if (y1 < 0) y1 = 0; if (x2 >= image->width) x2 = image->width - 1; if (y2 >= image->height) y2 = image->height - 1; int32_t clipped_width = x2 - x1 + 1; int32_t next_row = image->width - clipped_width; uint32_t *pixel = image->pixels + y1 * image->width + x1; for (int y = y1; y <= y2; y++) { for (int i = 0; i < clipped_width; i++) { *pixel++ = color; } pixel += next_row; } } void blit(r96_image *dst, r96_image *src, int x, int y) { int32_t dst_x1 = x; int32_t dst_y1 = y; int32_t dst_x2 = x + src->width - 1; int32_t dst_y2 = y + src->height - 1; int32_t src_x1 = 0; int32_t src_y1 = 0; if (dst_x1 >= dst->width) return; if (dst_x2 < 0) return; if (dst_y1 >= dst->height) return; if (dst_y2 < 0) return; if (dst_x1 < 0) { src_x1 -= dst_x1; dst_x1 = 0; } if (dst_y1 < 0) { src_y1 -= dst_y1; dst_y1 = 0; } if (dst_x2 >= dst->width) dst_x2 = dst->width - 1; if (dst_y2 >= dst->height) dst_y2 = dst->height - 1; int32_t clipped_width = dst_x2 - dst_x1 + 1; int32_t dst_next_row = dst->width - clipped_width; int32_t src_next_row = src->width - clipped_width; uint32_t *dst_pixel = dst->pixels + dst_y1 * dst->width + dst_x1; uint32_t *src_pixel = src->pixels + src_y1 * src->width + src_x1; for (int y = dst_y1; y <= dst_y2; y++) { for (int i = 0; i < clipped_width; i++) { *dst_pixel++ = *src_pixel++; } dst_pixel += dst_next_row; src_pixel += src_next_row; } } void blit_keyed(r96_image *dst, r96_image *src, int x, int y, uint32_t color_key) { int32_t dst_x1 = x; int32_t dst_y1 = y; int32_t dst_x2 = x + src->width - 1; int32_t dst_y2 = y + src->height - 1; int32_t src_x1 = 0; int32_t src_y1 = 0; if (dst_x1 >= dst->width) return; if (dst_x2 < 0) return; if (dst_y1 >= dst->height) return; if (dst_y2 < 0) return; if (dst_x1 < 0) { src_x1 -= dst_x1; dst_x1 = 0; } if (dst_y1 < 0) { src_y1 -= dst_y1; dst_y1 = 0; } if (dst_x2 >= dst->width) dst_x2 = dst->width - 1; if (dst_y2 >= dst->height) dst_y2 = dst->height - 1; int32_t clipped_width = dst_x2 - dst_x1 + 1; int32_t dst_next_row = dst->width - clipped_width; int32_t src_next_row = src->width - clipped_width; uint32_t *dst_pixel = dst->pixels + dst_y1 * dst->width + dst_x1; uint32_t *src_pixel = src->pixels + src_y1 * src->width + src_x1; for (y = dst_y1; y <= dst_y2; y++) { for (int i = 0; i < clipped_width; i++) { uint32_t src_color = *src_pixel; uint32_t dst_color = *dst_pixel; *dst_pixel = src_color != color_key ? src_color : dst_color; src_pixel++; dst_pixel++; } dst_pixel += dst_next_row; src_pixel += src_next_row; } }