Skip to navigation


Drawing ships: LL9 (Part 8 of 12)

[BBC Micro disc version, Docked]

Name: LL9 (Part 8 of 12) [Show more] Type: Subroutine Category: Drawing ships Summary: Draw ship: Calculate the screen coordinates of visible vertices Deep dive: Drawing ships
Context: See this subroutine in context in the source code Variations: See code variations for this subroutine in the different versions References: This subroutine is called as follows: * LL62 calls via LL66

This section projects the coordinate of the vertex into screen coordinates and stores them on the XX3 heap. By the end of this part, the XX3 heap contains four bytes containing the 16-bit screen coordinates of the current vertex, in the order: x_lo, x_hi, y_lo, y_hi. When we reach here, we are looping through the vertices, and we've just worked out the coordinates of the vertex in our normal coordinate system, as follows XX15(2 0) (x_sign x_lo) = x-coordinate of the current vertex XX15(5 3) (y_sign y_lo) = y-coordinate of the current vertex (U T) (z_sign z_lo) = z-coordinate of the current vertex Note that U is always zero when we get to this point, as the vertex is always in front of us (so it has a positive z-coordinate, into the screen).
Other entry points: LL70+1 Contains an RTS (as the first byte of an LDA instruction) LL66 A re-entry point into the ship-drawing routine, used by the LL62 routine to store 128 - (U R) on the XX3 heap
.LL60 LDA T \ Set Q = z_lo STA Q LDA XX15 \ Set A = x_lo CMP Q \ If x_lo < z_lo jump to LL69 BCC LL69 JSR LL61 \ Call LL61 to calculate: \ \ (U R) = 256 * A / Q \ = 256 * x / z \ \ which we can do as x >= z JMP LL65 \ Jump to LL65 to skip the division for x_lo < z_lo .LL69 JSR LL28 \ Call LL28 to calculate: \ \ R = 256 * A / Q \ = 256 * x / z \ \ Because x < z, the result fits into one byte, and we \ also know that U = 0, so (U R) also contains the \ result .LL65 \ At this point we have: \ \ (U R) = x / z \ \ so (U R) contains the vertex's x-coordinate projected \ on screen \ \ The next task is to convert (U R) to a pixel screen \ coordinate and stick it on the XX3 heap. \ \ We start with the x-coordinate. To convert the \ x-coordinate to a screen pixel we add 128, the \ x-coordinate of the centre of the screen, because the \ projected value is relative to an origin at the centre \ of the screen, but the origin of the screen pixels is \ at the top-left of the screen LDX CNT \ Fetch the pointer to the end of the XX3 heap from CNT \ into X LDA XX15+2 \ If x_sign is negative, jump up to LL62, which will BMI LL62 \ store 128 - (U R) on the XX3 heap and return by \ jumping down to LL66 below LDA R \ Calculate 128 + (U R), starting with the low bytes CLC ADC #128 STA XX3,X \ Store the low byte of the result in the X-th byte of \ the heap at XX3 INX \ Increment the heap pointer in X to point to the next \ byte LDA U \ And then add the high bytes ADC #0 STA XX3,X \ Store the high byte of the result in the X-th byte of \ the heap at XX3 .LL66 \ We've just stored the screen x-coordinate of the \ vertex on the XX3 heap, so now for the y-coordinate TXA \ Store the heap pointer in X on the stack (at this PHA \ it points to the last entry on the heap, not the first \ free byte) LDA #0 \ Set U = 0 STA U LDA T \ Set Q = z_lo STA Q LDA XX15+3 \ Set A = y_lo CMP Q \ If y_lo < z_lo jump to LL67 BCC LL67 JSR LL61 \ Call LL61 to calculate: \ \ (U R) = 256 * A / Q \ = 256 * y / z \ \ which we can do as y >= z JMP LL68 \ Jump to LL68 to skip the division for y_lo < z_lo .LL70 \ This gets called from below when y_sign is negative LDA #Y \ Calculate #Y + (U R), starting with the low bytes CLC ADC R STA XX3,X \ Store the low byte of the result in the X-th byte of \ the heap at XX3 INX \ Increment the heap pointer in X to point to the next \ byte LDA #0 \ And then add the high bytes ADC U STA XX3,X \ Store the high byte of the result in the X-th byte of \ the heap at XX3 JMP LL50 \ Jump to LL50 to move on to the next vertex .LL67 JSR LL28 \ Call LL28 to calculate: \ \ R = 256 * A / Q \ = 256 * y / z \ \ Because y < z, the result fits into one byte, and we \ also know that U = 0, so (U R) also contains the \ result .LL68 \ At this point we have: \ \ (U R) = y / z \ \ so (U R) contains the vertex's y-coordinate projected \ on screen \ \ We now want to convert this to a screen y-coordinate \ and stick it on the XX3 heap, much like we did with \ the x-coordinate above. Again, we convert the \ coordinate by adding or subtracting the y-coordinate \ of the centre of the screen, which is in the constant \ #Y, but this time we do the opposite, as a positive \ projected y-coordinate, i.e. up the space y-axis and \ up the screen, converts to a low y-coordinate, which \ is the opposite way round to the x-coordinates PLA \ Restore the heap pointer from the stack into X TAX INX \ When we stored the heap pointer, it pointed to the \ last entry on the heap, not the first free byte, so we \ increment it so it does point to the next free byte LDA XX15+5 \ If y_sign is negative, jump up to LL70, which will BMI LL70 \ store #Y + (U R) on the XX3 heap and return by jumping \ down to LL50 below LDA #Y \ Calculate #Y - (U R), starting with the low bytes SEC SBC R STA XX3,X \ Store the low byte of the result in the X-th byte of \ the heap at XX3 INX \ Increment the heap pointer in X to point to the next \ byte LDA #0 \ And then subtract the high bytes SBC U STA XX3,X \ Store the high byte of the result in the X-th byte of \ the heap at XX3 .LL50 \ By the time we get here, the XX3 heap contains four \ bytes containing the screen coordinates of the current \ vertex, in the order: x_lo, x_hi, y_lo, y_hi CLC \ Set CNT = CNT + 4, so the heap pointer points to the LDA CNT \ next free byte on the heap ADC #4 STA CNT LDA XX17 \ Set A to the offset of the current vertex's data, \ which we set in part 6 ADC #6 \ Set Y = A + 6, so Y now points to the data for the TAY \ next vertex BCS LL72 \ If the addition just overflowed, meaning we just tried \ to access vertex #43, jump to LL72, as the maximum \ number of vertices allowed is 42 CMP XX20 \ If Y >= number of vertices * 6 (which we stored in BCS LL72 \ XX20 in part 6), jump to LL72, as we have processed \ all the vertices for this ship JMP LL48 \ Loop back to LL48 in part 6 to calculate visibility \ and screen coordinates for the next vertex