# quick-n-dirty way to print out floats ######## In hex, following C's %a format # https://www.exploringbinary.com/hexadecimal-floating-point-constants # examples: # 0.5 = 0x3f000000 = 0011| 1111 | 0000 | 0000 | 0000 | 0000 | 0000 | 0000 # = 0 | 01111110 | 00000000000000000000000 # + exponent mantissa # = 0 | 00000000000000000000000 | 01111110 # mantissa exponent # = 0 | 000000000000000000000000 | 01111110 # zero-pad mantissa exponent # = +1.000000 P -01 fn test-print-float-normal { var screen-on-stack: screen var screen/esi: (addr screen) <- address screen-on-stack initialize-screen screen, 5, 0x20 # 32 columns should be more than enough # 0.5 var half/xmm0: float <- rational 1, 2 print-float screen, half check-screen-row screen, 1, "1.000000P-01 ", "F - test-print-float-normal 0.5" # 0.25 clear-screen screen var quarter/xmm0: float <- rational 1, 4 print-float screen, quarter check-screen-row screen, 1, "1.000000P-02 ", "F - test-print-float-normal 0.25" # 0.75 clear-screen screen var three-quarters/xmm0: float <- rational 3, 4 print-float screen, three-quarters check-screen-row screen, 1, "1.800000P-01 ", "F - test-print-float-normal 0.75" # 0.1 clear-screen screen var tenth/xmm0: float <- rational 1, 0xa print-float screen, tenth check-screen-row screen, 1, "1.99999aP-04 ", "F - test-print-float-normal 0.1" } fn test-print-float-integer { var screen-on-stack: screen var screen/esi: (addr screen) <- address screen-on-stack initialize-screen screen, 5, 0x20 # 32 columns should be more than enough # 1 var one-f/xmm0: float <- rational 1, 1 print-float screen, one-f check-screen-row screen, 1, "1.000000P00 ", "F - test-print-float-integer 1" # 2 clear-screen screen var two-f/xmm0: float <- rational 2, 1 print-float screen, two-f check-screen-row screen, 1, "1.000000P01 ", "F - test-print-float-integer 2" # 10 clear-screen screen var ten-f/xmm0: float <- rational 0xa, 1 print-float screen, ten-f check-screen-row screen, 1, "1.400000P03 ", "F - test-print-float-integer 10" # -10 clear-screen screen var minus-ten-f/xmm0: float <- rational -0xa, 1 print-float screen, minus-ten-f check-screen-row screen, 1, "-1.400000P03 ", "F - test-print-float-integer -10" } fn test-print-float-zero { var screen-on-stack: screen var screen/esi: (addr screen) <- address screen-on-stack initialize-screen screen, 5, 0x20 # 32 columns should be more than enough var zero: float print-float screen, zero check-screen-row screen, 1, "0 ", "F - test-print-float-zero" } fn test-print-float-negative-zero { var screen-on-stack: screen var screen/esi: (addr screen) <- address screen-on-stack initialize-screen screen, 5, 0x20 # 32 columns should be more than enough var n: int copy-to n, 0x80000000 var negative-zero/xmm0: float <- reinterpret n print-float screen, negative-zero check-screen-row screen, 1, "-0 ", "F - test-print-float-negative-zero" } fn test-print-float-infinity { var screen-on-stack: screen var screen/esi: (addr screen) <- address screen-on-stack initialize-screen screen, 5, 0x20 # 32 columns should be more than enough var n: int # 0|11111111|00000000000000000000000 # 0111|1111|1000|0000|0000|0000|0000|0000 copy-to n, 0x7f800000 var infinity/xmm0: float <- reinterpret n print-float screen, infinity check-screen-row screen, 1, "Inf ", "F - test-print-float-infinity" } fn test-print-float-negative-infinity { var screen-on-stack: screen var screen/esi: (addr screen) <- address screen-on-stack initialize-screen screen, 5, 0x20 # 32 columns should be more than enough var n: int copy-to n, 0xff800000 var negative-infinity/xmm0: float <- reinterpret n print-float screen, negative-infinity check-screen-row screen, 1, "-Inf ", "F - test-print-float-negative-infinity" } fn test-print-float-not-a-number { var screen-on-stack: screen var screen/esi: (addr screen) <- address screen-on-stack initialize-screen screen, 5, 0x20 # 32 columns should be more than enough var n: int copy-to n, 0xffffffff # exponent must be all 1's, and mantissa must be non-zero var negative-infinity/xmm0: float <- reinterpret n print-float screen, negative-infinity check-screen-row screen, 1, "NaN ", "F - test-print-float-not-a-number" } fn print-float screen: (addr screen), n: float { # - special names var bits/eax: int <- reinterpret n compare bits, 0 { break-if-!= print-string screen, "0" return } compare bits, 0x80000000 { break-if-!= print-string screen, "-0" return } compare bits, 0x7f800000 { break-if-!= print-string screen, "Inf" return } compare bits, 0xff800000 { break-if-!= print-string screen, "-Inf" return } var exponent/ecx: int <- copy bits exponent <- shift-right 0x17 # 23 bits of mantissa exponent <- and 0xff exponent <- subtract 0x7f compare exponent, 0x80 { break-if-!= print-string screen, "NaN" return } # - regular numbers var sign/edx: int <- copy bits sign <- shift-right 0x1f { compare sign, 1 break-if-!= print-string screen, "-" } $print-float:leading-digit: { # check for subnormal numbers compare exponent, -0x7f { break-if-!= print-string screen, "0." exponent <- increment break $print-float:leading-digit } # normal numbers print-string screen, "1." } var mantissa/ebx: int <- copy bits mantissa <- and 0x7fffff mantissa <- shift-left 1 # pad to whole nibbles print-int32-hex-bits screen, mantissa, 0x18 # print exponent print-string screen, "P" compare exponent, 0 { break-if->= print-string screen, "-" } var exp-magnitude/eax: int <- abs exponent print-int32-hex-bits screen, exp-magnitude, 8 } #? fn main -> _/ebx: int { #? run-tests #? #? test-print-float-negative-zero #? #? print-int32-hex 0, 0 #? #? test-print-float-normal #? return 0 #? } ######## In decimal # Try to keep it short. fn test-print-float-decimal-approximate-normal { var screen-on-stack: screen var screen/esi: (addr screen) <- address screen-on-stack initialize-screen screen, 5, 0x20 # 32 columns should be more than enough # 0.5 var half/xmm0: float <- rational 1, 2 print-float-decimal-approximate screen, half check-screen-row screen, 1, "0.5 ", "F - test-print-float-decimal-approximate-normal 0.5" # 0.25 clear-screen screen var quarter/xmm0: float <- rational 1, 4 print-float-decimal-approximate screen, quarter check-screen-row screen, 1, "0.25 ", "F - test-print-float-decimal-approximate-normal 0.25" # 0.75 clear-screen screen var three-quarters/xmm0: float <- rational 3, 4 print-float-decimal-approximate screen, three-quarters check-screen-row screen, 1, "0.75 ", "F - test-print-float-decimal-approximate-normal 0.75" # 0.125 clear-screen screen var eighth/xmm0: float <- rational 1, 8 print-float-decimal-approximate screen, eighth check-screen-row screen, 1, "0.125 ", "F - test-print-float-decimal-approximate-normal 0.1" # 0.0625; start truncating past 3 decimal places clear-screen screen var sixteenth/xmm0: float <- rational 1, 0x10 print-float-decimal-approximate screen, sixteenth check-screen-row screen, 1, "0.062 ", "F - test-print-float-decimal-a
== code
#   instruction                     effective address                                                   register    displacement    immediate
# . op          subop               mod             rm32          base        index         scale       r32
# . 1-3 bytes   3 bits              2 bits          3 bits        3 bits      3 bits        2 bits      2 bits      0/1/2/4 bytes   0/1/2/4 bytes

# print 'arr' in hex with a space after every byte
emit-hex-array:  # out: (addr buffered-file), arr: (addr array byte)
    # . prologue
    55/push-ebp
    89/copy                         3/mod/direct    5/rm32/ebp    .           .             .           4/r32/esp   .               .                 # copy esp to ebp
    # . save registers
    50/push-eax
    51/push-ecx
    52/push-edx
    57/push-edi
    # edi = out
    8b/copy                         1/mod/*+disp8   5/rm32/ebp    .           .             .           7/r32/edi   8/disp8         .                 # copy *(ebp+8) to edi
    # edx = arr
    8b/copy                         1/mod/*+disp8   5/rm32/ebp    .           .             .           2/r32/edx   0xc/disp8       .                 # copy *(ebp+12) to edx
    # var curr/ecx: (addr byte) = arr->data
    8d/copy-address                 1/mod/*+disp8   2/rm32/edx    .           .             .