holey-bytes/lang/tests/son_tests_string_flip.txt

main:
    ADDI64 r254, r254, -40d
    LI64 r16, 1d
    LI64 r15, 4d
    ADDI64 r17, r254, 0d
    CP r14, r0
    CP r13, r14
 6: JNE r13, r15, :0
    ADDI64 r18, r254, 32d
    LI64 r19, 2d
    CP r13, r14
 4: LD r15, r254, 16a, 8h
    JNE r13, r16, :1
    CP r1, r15
    JMP :2
 1: ADD64 r15, r13, r16
    SUB64 r20, r19, r15
    MUL64 r20, r20, r19
    MUL64 r21, r13, r19
    CP r13, r14
 5: JNE r13, r19, :3
    CP r13, r15
    JMP :4
 3: ADD64 r22, r21, r13
    ADD64 r23, r20, r13
    MULI64 r22, r22, 8d
    MULI64 r23, r23, 8d
    ADD64 r22, r17, r22
    ADD64 r23, r17, r23
    BMC r22, r18, 8h
    BMC r23, r22, 8h
    BMC r18, r23, 8h
    ADD64 r13, r13, r16
    JMP :5
 0: MULI64 r18, r13, 8d
    ADD64 r18, r17, r18
    ST r13, r18, 0a, 8h
    ADD64 r13, r13, r16
    JMP :6
 2: ADDI64 r254, r254, 40d
    JALA r0, r31, 0a
code size: 264
ret: 0
status: Ok(())
fixing stack alloc overoptimization 2024-10-24 12:57:36 -05:00			`main:`
fixed a stack prelude postlude being needlesly generated + struct can now be compared 2024-11-17 03:06:10 -06:00			`ADDI64 r254, r254, -40d`
making the instruction scheduling smarter the instructions that are only depended by phis are pushed to the end of the block, which usually saves copy instructions Signed-off-by: Jakub Doka <jakub.doka2@gmail.com> 2024-12-19 04:00:19 -06:00			`LI64 r16, 1d`
changing the gcm to not mutate nodes in recursive functions 2024-11-23 12:47:17 -06:00			`LI64 r15, 4d`
making the instruction scheduling smarter the instructions that are only depended by phis are pushed to the end of the block, which usually saves copy instructions Signed-off-by: Jakub Doka <jakub.doka2@gmail.com> 2024-12-19 04:00:19 -06:00			`ADDI64 r17, r254, 0d`
maybe fixed mandelbrot 2024-11-15 15:35:03 -06:00			`CP r14, r0`
			`CP r13, r14`
changing the gcm to not mutate nodes in recursive functions 2024-11-23 12:47:17 -06:00			`6: JNE r13, r15, :0`
making the instruction scheduling smarter the instructions that are only depended by phis are pushed to the end of the block, which usually saves copy instructions Signed-off-by: Jakub Doka <jakub.doka2@gmail.com> 2024-12-19 04:00:19 -06:00			`ADDI64 r18, r254, 32d`
			`LI64 r19, 2d`
maybe fixed mandelbrot 2024-11-15 15:35:03 -06:00			`CP r13, r14`
			`4: LD r15, r254, 16a, 8h`
making the instruction scheduling smarter the instructions that are only depended by phis are pushed to the end of the block, which usually saves copy instructions Signed-off-by: Jakub Doka <jakub.doka2@gmail.com> 2024-12-19 04:00:19 -06:00			`JNE r13, r16, :1`
maybe fixed mandelbrot 2024-11-15 15:35:03 -06:00			`CP r1, r15`
fixing stack alloc overoptimization 2024-10-24 12:57:36 -05:00			`JMP :2`
making the instruction scheduling smarter the instructions that are only depended by phis are pushed to the end of the block, which usually saves copy instructions Signed-off-by: Jakub Doka <jakub.doka2@gmail.com> 2024-12-19 04:00:19 -06:00			`1: ADD64 r15, r13, r16`
			`SUB64 r20, r19, r15`
			`MUL64 r20, r20, r19`
			`MUL64 r21, r13, r19`
maybe fixed mandelbrot 2024-11-15 15:35:03 -06:00			`CP r13, r14`
making the instruction scheduling smarter the instructions that are only depended by phis are pushed to the end of the block, which usually saves copy instructions Signed-off-by: Jakub Doka <jakub.doka2@gmail.com> 2024-12-19 04:00:19 -06:00			`5: JNE r13, r19, :3`
			`CP r13, r15`
fixing stack alloc overoptimization 2024-10-24 12:57:36 -05:00			`JMP :4`
making the instruction scheduling smarter the instructions that are only depended by phis are pushed to the end of the block, which usually saves copy instructions Signed-off-by: Jakub Doka <jakub.doka2@gmail.com> 2024-12-19 04:00:19 -06:00			`3: ADD64 r22, r21, r13`
			`ADD64 r23, r20, r13`
			`MULI64 r22, r22, 8d`
maybe fixed mandelbrot 2024-11-15 15:35:03 -06:00			`MULI64 r23, r23, 8d`
making the instruction scheduling smarter the instructions that are only depended by phis are pushed to the end of the block, which usually saves copy instructions Signed-off-by: Jakub Doka <jakub.doka2@gmail.com> 2024-12-19 04:00:19 -06:00			`ADD64 r22, r17, r22`
			`ADD64 r23, r17, r23`
			`BMC r22, r18, 8h`
			`BMC r23, r22, 8h`
			`BMC r18, r23, 8h`
			`ADD64 r13, r13, r16`
fixing stack alloc overoptimization 2024-10-24 12:57:36 -05:00			`JMP :5`
making the instruction scheduling smarter the instructions that are only depended by phis are pushed to the end of the block, which usually saves copy instructions Signed-off-by: Jakub Doka <jakub.doka2@gmail.com> 2024-12-19 04:00:19 -06:00			`0: MULI64 r18, r13, 8d`
			`ADD64 r18, r17, r18`
			`ST r13, r18, 0a, 8h`
			`ADD64 r13, r13, r16`
fixing stack alloc overoptimization 2024-10-24 12:57:36 -05:00			`JMP :6`
fixed a stack prelude postlude being needlesly generated + struct can now be compared 2024-11-17 03:06:10 -06:00			`2: ADDI64 r254, r254, 40d`
fixing stack alloc overoptimization 2024-10-24 12:57:36 -05:00			`JALA r0, r31, 0a`
making the instruction scheduling smarter the instructions that are only depended by phis are pushed to the end of the block, which usually saves copy instructions Signed-off-by: Jakub Doka <jakub.doka2@gmail.com> 2024-12-19 04:00:19 -06:00			`code size: 264`
making better peepholes and fixing overoptimization on memory swaps 2024-11-12 05:20:08 -06:00			`ret: 0`
fixing stack alloc overoptimization 2024-10-24 12:57:36 -05:00			`status: Ok(())`