abawo/pico-sdk
3
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Adapt assembler syntax to eliminate Clang errors. (LLVM 14.0.0) (#798)

Browse Source 
Co-authored-by: jeremyd <>
This commit is contained in:
eskimo-software 2022-05-09 22:11:59 +02:00 committed by GitHub
parent e1c5fd34e4
commit f260477802
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
5 changed files with 69 additions and 68 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

30
src/rp2_common/pico_divider/divider.S
View File

@ -103,7 +103,7 @@ regular_func div_s32s32
regular_func divmod_s32s32 regular_func divmod_s32s32
#if !PICO_DIVIDER_DISABLE_INTERRUPTS #if !PICO_DIVIDER_DISABLE_INTERRUPTS
// to support IRQ usage (or context switch) we must save/restore divider state around call if state is dirty // to support IRQ usage (or context switch) we must save/restore divider state around call if state is dirty
ldr r2, =(SIO_BASE) ldr r2, =SIO_BASE
ldr r3, [r2, #SIO_DIV_CSR_OFFSET] ldr r3, [r2, #SIO_DIV_CSR_OFFSET]
lsrs r3, #SIO_DIV_CSR_DIRTY_SHIFT_FOR_CARRY lsrs r3, #SIO_DIV_CSR_DIRTY_SHIFT_FOR_CARRY
bcs divmod_s32s32_savestate bcs divmod_s32s32_savestate
@ -114,7 +114,7 @@ regular_func divmod_s32s32_unsafe
// are the hardware_divider functions that can be used instead anyway // are the hardware_divider functions that can be used instead anyway
regular_func divmod_s32s32_unsafe regular_func divmod_s32s32_unsafe
// to avoid worrying about IRQs (or context switches), simply disable interrupts around call // to avoid worrying about IRQs (or context switches), simply disable interrupts around call
ldr r2, =(SIO_BASE) ldr r2, =SIO_BASE
mrs r3, PRIMASK mrs r3, PRIMASK
cpsid i cpsid i
#endif /* !PICO_DIVIDER_DISABLE_INTERRUPTS */ #endif /* !PICO_DIVIDER_DISABLE_INTERRUPTS */
@ -167,7 +167,7 @@ wrapper_func __aeabi_uidiv
wrapper_func __aeabi_uidivmod wrapper_func __aeabi_uidivmod
#if !PICO_DIVIDER_DISABLE_INTERRUPTS #if !PICO_DIVIDER_DISABLE_INTERRUPTS
// to support IRQ usage (or context switch) we must save/restore divider state around call if state is dirty // to support IRQ usage (or context switch) we must save/restore divider state around call if state is dirty
ldr r2, =(SIO_BASE) ldr r2, =SIO_BASE
ldr r3, [r2, #SIO_DIV_CSR_OFFSET] ldr r3, [r2, #SIO_DIV_CSR_OFFSET]
lsrs r3, #SIO_DIV_CSR_DIRTY_SHIFT_FOR_CARRY lsrs r3, #SIO_DIV_CSR_DIRTY_SHIFT_FOR_CARRY
bcs divmod_u32u32_savestate bcs divmod_u32u32_savestate
@ -178,7 +178,7 @@ regular_func divmod_u32u32_unsafe
// are the hardware_divider functions that can be used instead anyway // are the hardware_divider functions that can be used instead anyway
regular_func divmod_u32u32_unsafe regular_func divmod_u32u32_unsafe
// to avoid worrying about IRQs (or context switches), simply disable interrupts around call // to avoid worrying about IRQs (or context switches), simply disable interrupts around call
ldr r2, =(SIO_BASE) ldr r2, =SIO_BASE
mrs r3, PRIMASK mrs r3, PRIMASK
cpsid i cpsid i
#endif /* !PICO_DIVIDER_DISABLE_INTERRUPTS */ #endif /* !PICO_DIVIDER_DISABLE_INTERRUPTS */
@ -227,7 +227,7 @@ regular_func divmod_s64s64
#if !PICO_DIVIDER_DISABLE_INTERRUPTS #if !PICO_DIVIDER_DISABLE_INTERRUPTS
// to support IRQ usage (or context switch) we must save/restore divider state around call if state is dirty // to support IRQ usage (or context switch) we must save/restore divider state around call if state is dirty
mov ip, r2 mov ip, r2
ldr r2, =(SIO_BASE) ldr r2, =SIO_BASE
ldr r2, [r2, #SIO_DIV_CSR_OFFSET] ldr r2, [r2, #SIO_DIV_CSR_OFFSET]
lsrs r2, #SIO_DIV_CSR_DIRTY_SHIFT_FOR_CARRY lsrs r2, #SIO_DIV_CSR_DIRTY_SHIFT_FOR_CARRY
mov r2, ip mov r2, ip
@ -255,7 +255,7 @@ regular_func divmod_u64u64
#if !PICO_DIVIDER_DISABLE_INTERRUPTS #if !PICO_DIVIDER_DISABLE_INTERRUPTS
// to support IRQ usage (or context switch) we must save/restore divider state around call if state is dirty // to support IRQ usage (or context switch) we must save/restore divider state around call if state is dirty
mov ip, r2 mov ip, r2
ldr r2, =(SIO_BASE) ldr r2, =SIO_BASE
ldr r2, [r2, #SIO_DIV_CSR_OFFSET] ldr r2, [r2, #SIO_DIV_CSR_OFFSET]
lsrs r2, #SIO_DIV_CSR_DIRTY_SHIFT_FOR_CARRY lsrs r2, #SIO_DIV_CSR_DIRTY_SHIFT_FOR_CARRY
mov r2, ip mov r2, ip
@ -278,7 +278,7 @@ regular_func divmod_u64u64_savestate
.macro dneg lo,hi .macro dneg lo,hi
mvns \hi,\hi mvns \hi,\hi
rsbs \lo,#0 negs \lo,\lo
bne l\@_1 bne l\@_1
adds \hi,#1 adds \hi,#1
l\@_1: l\@_1:
@ -352,7 +352,7 @@ regular_func divmod_u64u64_unsafe
cmp r2,#0 cmp r2,#0
beq 2f @ x==0? beq 2f @ x==0?
mov r12,r7 mov r12,r7
ldr r7,=#SIO_BASE ldr r7,=SIO_BASE
str r0,[r7,#SIO_DIV_UDIVIDEND_OFFSET] str r0,[r7,#SIO_DIV_UDIVIDEND_OFFSET]
str r2,[r7,#SIO_DIV_UDIVISOR_OFFSET] str r2,[r7,#SIO_DIV_UDIVISOR_OFFSET]
movs r1,#0 movs r1,#0
@ -367,7 +367,7 @@ regular_func divmod_u64u64_unsafe
cmp r0,#0 @ y==0? cmp r0,#0 @ y==0?
beq 3f @ then pass 0 to __aeabi_ldiv0 beq 3f @ then pass 0 to __aeabi_ldiv0
udiv0: udiv0:
ldr r0,=#0xffffffff ldr r0,=0xffffffff
movs r1,r0 @ pass 2^64-1 to __aeabi_ldiv0 movs r1,r0 @ pass 2^64-1 to __aeabi_ldiv0
3: 3:
push {r14} push {r14}
@ -402,7 +402,7 @@ y64:
cmp r2,#0 cmp r2,#0
beq udiv0 @ x==0? exit as with y!=0 case above beq udiv0 @ x==0? exit as with y!=0 case above
push {r7} push {r7}
ldr r7,=#SIO_BASE ldr r7,=SIO_BASE
str r1,[r7,#SIO_DIV_UDIVIDEND_OFFSET] str r1,[r7,#SIO_DIV_UDIVIDEND_OFFSET]
str r2,[r7,#SIO_DIV_UDIVISOR_OFFSET] str r2,[r7,#SIO_DIV_UDIVISOR_OFFSET]
wait_div 4 wait_div 4
@ -464,9 +464,9 @@ y64_x32:
lsrs r4,r2,#15 lsrs r4,r2,#15
adds r4,#1 @ x1=(x0>>15)+1; 2^16<x1<=2^17 adds r4,#1 @ x1=(x0>>15)+1; 2^16<x1<=2^17
ldr r7,=#SIO_BASE ldr r7,=SIO_BASE
str r4,[r7,#SIO_DIV_UDIVISOR_OFFSET] str r4,[r7,#SIO_DIV_UDIVISOR_OFFSET]
ldr r4,=#0xffffffff ldr r4,=0xffffffff
str r4,[r7,#SIO_DIV_UDIVIDEND_OFFSET] str r4,[r7,#SIO_DIV_UDIVIDEND_OFFSET]
lsrs r6,r1,#16 lsrs r6,r1,#16
uxth r3,r2 @ x0l uxth r3,r2 @ x0l
@ -687,9 +687,9 @@ y64_x48:
adcs r4,r4 adcs r4,r4
adds r4,#1 @ x1=(ui32)(x0>>31)+1; // 2^16<x1<=2^17 adds r4,#1 @ x1=(ui32)(x0>>31)+1; // 2^16<x1<=2^17
ldr r7,=#SIO_BASE ldr r7,=SIO_BASE
str r4,[r7,#SIO_DIV_UDIVISOR_OFFSET] str r4,[r7,#SIO_DIV_UDIVISOR_OFFSET]
ldr r4,=#0xffffffff ldr r4,=0xffffffff
str r4,[r7,#SIO_DIV_UDIVIDEND_OFFSET] str r4,[r7,#SIO_DIV_UDIVIDEND_OFFSET]
lsrs r6,r1,#16 lsrs r6,r1,#16
wait_div 1 wait_div 1
@ -821,7 +821,7 @@ y64_x64:
adds r5,r3,#1 adds r5,r3,#1
beq 1f beq 1f
ldr r7,=#SIO_BASE ldr r7,=SIO_BASE
str r5,[r7,#SIO_DIV_UDIVISOR_OFFSET] str r5,[r7,#SIO_DIV_UDIVISOR_OFFSET]
str r1,[r7,#SIO_DIV_UDIVIDEND_OFFSET] str r1,[r7,#SIO_DIV_UDIVIDEND_OFFSET]
wait_div 0 wait_div 0

6
src/rp2_common/pico_double/double_aeabi.S
View File

@ -290,7 +290,7 @@ wrapper_func __aeabi_cdcmple
wrapper_func __aeabi_cdcmpeq wrapper_func __aeabi_cdcmpeq
push {r0-r7,r14} push {r0-r7,r14}
__aeabi_dfcmple_guts: __aeabi_dfcmple_guts:
ldr r7,=#0x7ff @ flush NaNs and denormals ldr r7,=0x7ff @ flush NaNs and denormals
lsls r4,r1,#1 lsls r4,r1,#1
lsrs r4,#21 lsrs r4,#21
beq 1f beq 1f
@ -450,7 +450,7 @@ wrapper_func __aeabi_i2d
lsls r2, r4 lsls r2, r4
lsls r0, r2, #20 lsls r0, r2, #20
lsrs r2, #12 lsrs r2, #12
ldr r1,=#1055 ldr r1,=1055
subs r1, r4 subs r1, r4
lsls r1, #20 lsls r1, #20
orrs r1, r3 orrs r1, r3
@ -552,7 +552,7 @@ regular_func double2int64_z
cmp r1, #0 cmp r1, #0
bmi 1f bmi 1f
movs r2, #0 movs r2, #0
rsbs r0, #0 negs r0, r0
sbcs r2, r1 sbcs r2, r1
mov r1, r2 mov r1, r2
pop {pc} pop {pc}

75
src/rp2_common/pico_double/double_v1_rom_shim.S
View File

@ -144,7 +144,7 @@ l\@_2:
lsls \re,#21 lsls \re,#21
bcc l\@_1 @ skip on positive bcc l\@_1 @ skip on positive
mvns \rb,\rb @ negate mantissa mvns \rb,\rb @ negate mantissa
rsbs \ra,#0 negs \ra,\ra
bcc l\@_1 bcc l\@_1
adds \rb,#1 adds \rb,#1
l\@_1: l\@_1:
@ -218,7 +218,7 @@ da_pack:
mvns r1,r1 @ negate mantissa mvns r1,r1 @ negate mantissa
mvns r0,r0 mvns r0,r0
movs r2,#0 movs r2,#0
rsbs r4,#0 negs r4,r4
adcs r0,r2 adcs r0,r2
adcs r1,r2 adcs r1,r2
1: 1:
@ -263,7 +263,7 @@ da_3:
da_7: da_7:
@ here exponent overflow: return signed infinity @ here exponent overflow: return signed infinity
lsls r1,r3,#31 lsls r1,r3,#31
ldr r3,=#0x7ff00000 ldr r3,=0x7ff00000
orrs r1,r3 orrs r1,r3
b 1f b 1f
da_6: da_6:
@ -524,7 +524,7 @@ double_section dmul_shim
adcs r1,r1 adcs r1,r1
subs r3,#1 @ correct exponent subs r3,#1 @ correct exponent
1: 1:
ldr r6,=#0x3ff ldr r6,=0x3ff
subs r3,r6 @ correct for exponent bias subs r3,r6 @ correct for exponent bias
lsls r6,#1 @ 0x7fe lsls r6,#1 @ 0x7fe
cmp r3,r6 cmp r3,r6
@ -811,7 +811,7 @@ dd_1:
muls r3,r0 @ zL*xH Q73 muls r3,r0 @ zL*xH Q73
subs r4,r3 subs r4,r3
mul32_32_64 r2,r0, r2,r3, r5,r6,r7,r2,r3 @ xL*zL mul32_32_64 r2,r0, r2,r3, r5,r6,r7,r2,r3 @ xL*zL
rsbs r2,#0 @ borrow from low half? negs r2,r2 @ borrow from low half?
sbcs r4,r3 @ y-xz Q73 (remainder bits 52..73) sbcs r4,r3 @ y-xz Q73 (remainder bits 52..73)
cmp r4,#0 cmp r4,#0
@ -830,9 +830,9 @@ dd_2:
mov r2,r12 mov r2,r12
lsls r7,r2,#31 @ result sign lsls r7,r2,#31 @ result sign
asrs r2,#2 @ result exponent asrs r2,#2 @ result exponent
ldr r3,=#0x3fd ldr r3,=0x3fd
adds r2,r3 adds r2,r3
ldr r3,=#0x7fe ldr r3,=0x7fe
cmp r2,r3 cmp r2,r3
bhs dd_3 @ over- or underflow? bhs dd_3 @ over- or underflow?
lsls r2,#20 lsls r2,#20
@ -936,7 +936,7 @@ dq_0:
b dq_4 b dq_4
dq_3: dq_3:
ldr r1,=#0x7ff ldr r1,=0x7ff
lsls r1,#20 @ return +Inf lsls r1,#20 @ return +Inf
dq_4: dq_4:
movs r0,#0 movs r0,#0
@ -949,7 +949,7 @@ regular_func dsqrt_shim
bcs dq_0 @ negative? bcs dq_0 @ negative?
lsrs r2,#21 @ extract exponent lsrs r2,#21 @ extract exponent
subs r2,#1 subs r2,#1
ldr r3,=#0x7fe ldr r3,=0x7fe
cmp r2,r3 cmp r2,r3
bhs dq_2 @ catches 0 and +Inf bhs dq_2 @ catches 0 and +Inf
push {r4-r7,r14} push {r4-r7,r14}
@ -968,8 +968,8 @@ regular_func dsqrt_shim
@ here @ here
@ r0:r1 y mantissa Q52 [1,4) @ r0:r1 y mantissa Q52 [1,4)
@ r12 result exponent @ r12 result exponent
.equ drsqrtapp_minus_8, (drsqrtapp-8)
adr r4,drsqrtapp-8 @ first eight table entries are never accessed because of the mantissa's leading 1 adr r4,drsqrtapp_minus_8 @ first eight table entries are never accessed because of the mantissa's leading 1
lsrs r2,r1,#17 @ y Q3 lsrs r2,r1,#17 @ y Q3
ldrb r2,[r4,r2] @ initial approximation to reciprocal square root a0 Q8 ldrb r2,[r4,r2] @ initial approximation to reciprocal square root a0 Q8
lsrs r3,r1,#4 @ first Newton-Raphson iteration lsrs r3,r1,#4 @ first Newton-Raphson iteration
@ -1100,7 +1100,7 @@ dq_5:
adds r2,r3 @ r1:r2 a5^2 Q106 adds r2,r3 @ r1:r2 a5^2 Q106
lsls r0,#22 @ y Q84 lsls r0,#22 @ y Q84
rsbs r1,#0 negs r1,r1
sbcs r0,r2 @ remainder y-a5^2 sbcs r0,r2 @ remainder y-a5^2
bmi 1f @ y<a5^2: no need to increment a5 bmi 1f @ y<a5^2: no need to increment a5
movs r3,#0 movs r3,#0
@ -1290,7 +1290,8 @@ ret_dzero:
bx r14 bx r14
.weak d2fix_a // weak because it exists in float code too .weak d2fix_a // weak because it exists in float code too
regular_func d2fix_a .thumb_func
d2fix_a:
@ here @ here
@ r0:r1 two's complement mantissa @ r0:r1 two's complement mantissa
@ r2 unbaised exponent @ r2 unbaised exponent
@ -1304,7 +1305,7 @@ regular_func d2fix_a
movs r4,r0 movs r4,r0
lsls r1,r2 lsls r1,r2
lsls r0,r2 lsls r0,r2
rsbs r2,#0 negs r2,r2
adds r2,#32 @ complementary shift adds r2,#32 @ complementary shift
lsrs r4,r2 lsrs r4,r2
orrs r1,r4 orrs r1,r4
@ -1319,7 +1320,7 @@ regular_func d2fix_a
bmi 1f @ long shift? bmi 1f @ long shift?
mov r4,r1 mov r4,r1
lsls r4,r2 lsls r4,r2
rsbs r2,#0 negs r2,r2
adds r2,#32 @ complementary shift adds r2,#32 @ complementary shift
asrs r1,r2 asrs r1,r2
lsrs r0,r2 lsrs r0,r2
@ -1331,7 +1332,7 @@ regular_func d2fix_a
asrs r1,#31 @ shift down 32 places asrs r1,#31 @ shift down 32 places
adds r2,#32 adds r2,#32
bmi 1f @ very long shift? bmi 1f @ very long shift?
rsbs r2,#0 negs r2,r2
adds r2,#32 adds r2,#32
asrs r0,r2 asrs r0,r2
pop {r4,r15} pop {r4,r15}
@ -1344,7 +1345,7 @@ double_section double2float_shim
regular_func double2float_shim regular_func double2float_shim
lsls r2,r1,#1 lsls r2,r1,#1
lsrs r2,#21 @ exponent lsrs r2,#21 @ exponent
ldr r3,=#0x3ff-0x7f ldr r3,=0x3ff-0x7f
subs r2,r3 @ fix exponent bias subs r2,r3 @ fix exponent bias
ble 1f @ underflow or zero ble 1f @ underflow or zero
cmp r2,#0xff cmp r2,#0xff
@ -1432,7 +1433,7 @@ regular_func fix642double_shim
sbcs r1,r3 sbcs r1,r3
uf2d: uf2d:
push {r4,r5,r14} push {r4,r5,r14}
ldr r4,=#0x432 ldr r4,=0x432
subs r2,r4,r2 @ form biased exponent subs r2,r4,r2 @ form biased exponent
@ here @ here
@ r0:r1 unnormalised mantissa @ r0:r1 unnormalised mantissa
@ -1455,7 +1456,7 @@ uf2d:
lsrs r4,r1,#21 lsrs r4,r1,#21
bcc 5b bcc 5b
4: 4:
ldr r4,=#0x7fe ldr r4,=0x7fe
cmp r2,r4 cmp r2,r4
bhs 6f @ over/underflow? return signed zero/infinity bhs 6f @ over/underflow? return signed zero/infinity
7: 7:
@ -1506,7 +1507,7 @@ uf2d:
double_section dunpacks double_section dunpacks
regular_func dunpacks regular_func dunpacks
mdunpacks r0,r1,r2,r3,r4 mdunpacks r0,r1,r2,r3,r4
ldr r3,=#0x3ff ldr r3,=0x3ff
subs r2,r3 @ exponent without offset subs r2,r3 @ exponent without offset
bx r14 bx r14
@ -1572,7 +1573,7 @@ double_section dreduce
lsrs r3,r0,#24 lsrs r3,r0,#24
orrs r1,r3 orrs r1,r3
lsls r0,#8 @ r0:r1 Q60, to be shifted down -r2 places lsls r0,#8 @ r0:r1 Q60, to be shifted down -r2 places
rsbs r3,r2,#0 negs r3,r2
adds r2,#32 @ shift down in r3, complementary shift in r2 adds r2,#32 @ shift down in r3, complementary shift in r2
bmi 1f @ long shift? bmi 1f @ long shift?
2: 2:
@ -1663,13 +1664,13 @@ dsincos_internal:
bl dreduce bl dreduce
movs r4,#0 movs r4,#0
ldr r5,=#0x9df04dbb @ this value compensates for the non-unity scaling of the CORDIC rotations ldr r5,=0x9df04dbb @ this value compensates for the non-unity scaling of the CORDIC rotations
ldr r6,=#0x36f656c5 ldr r6,=0x36f656c5
lsls r2,#31 lsls r2,#31
bcc 1f bcc 1f
@ quadrant 2 or 3 @ quadrant 2 or 3
mvns r6,r6 mvns r6,r6
rsbs r5,r5,#0 negs r5,r5
adcs r6,r4 adcs r6,r4
1: 1:
lsls r2,#1 lsls r2,#1
@ -1757,7 +1758,7 @@ regular_func datan2_shim
@ r2:r3 x @ r2:r3 x
push {r4-r7,r14} push {r4-r7,r14}
bl push_r8_r11 bl push_r8_r11
ldr r5,=#0x7ff00000 ldr r5,=0x7ff00000
movs r4,r1 movs r4,r1
ands r4,r5 @ y==0? ands r4,r5 @ y==0?
beq 1f beq 1f
@ -1786,7 +1787,7 @@ regular_func datan2_shim
eors r3,r5 eors r3,r5
eors r1,r5 eors r1,r5
bmi 1f @ quadrant offset=+2 if y was positive bmi 1f @ quadrant offset=+2 if y was positive
rsbs r6,#0 @ quadrant offset=-2 if y was negative negs r6,r6 @ quadrant offset=-2 if y was negative
1: 1:
@ now in quadrant 0 or 3 @ now in quadrant 0 or 3
adds r7,r1,r5 @ r7=-r1 adds r7,r1,r5 @ r7=-r1
@ -1895,8 +1896,8 @@ regular_func datan2_shim
cmp r6,#0 cmp r6,#0
beq 1f beq 1f
ldr r4,=#0x885A308D @ π/2 Q61 ldr r4,=0x885A308D @ π/2 Q61
ldr r5,=#0x3243F6A8 ldr r5,=0x3243F6A8
bpl 2f bpl 2f
mvns r4,r4 @ negative quadrant offset mvns r4,r4 @ negative quadrant offset
mvns r5,r5 mvns r5,r5
@ -1959,8 +1960,8 @@ regular_func dexp_shim
bl dreduce bl dreduce
cmp r1,#0 cmp r1,#0
bge 1f bge 1f
ldr r4,=#0xF473DE6B ldr r4,=0xF473DE6B
ldr r5,=#0x2C5C85FD @ ln2 Q62 ldr r5,=0x2C5C85FD @ ln2 Q62
adds r0,r4 adds r0,r4
adcs r1,r5 adcs r1,r5
subs r2,#1 subs r2,#1
@ -1979,7 +1980,7 @@ regular_func dexp_shim
sbcs r1,r5 sbcs r1,r5
bmi 1f bmi 1f
rsbs r6,r7,#0 negs r6,r7
adds r6,#32 @ complementary shift adds r6,#32 @ complementary shift
movs r5,r3 movs r5,r3
asrs r5,r7 asrs r5,r7
@ -2015,7 +2016,7 @@ regular_func dexp_shim
adcs r1,r3 adcs r1,r3
pop {r2} pop {r2}
rsbs r2,#0 negs r2,r2
adds r2,#62 adds r2,#62
bl fix642double_shim @ in principle we can pack faster than this because we know the exponent bl fix642double_shim @ in principle we can pack faster than this because we know the exponent
pop {r4-r7,r15} pop {r4-r7,r15}
@ -2046,7 +2047,7 @@ regular_func dln_shim
movs r3,#0 @ y=0 Q62 movs r3,#0 @ y=0 Q62
3: 3:
rsbs r6,r7,#0 negs r6,r7
adds r6,#32 @ complementary shift adds r6,#32 @ complementary shift
movs r5,r1 movs r5,r1
asrs r5,r7 asrs r5,r7
@ -2086,8 +2087,8 @@ regular_func dln_shim
@ here: @ here:
@ r2:r3 ln m/2 = ln m - ln2 Q62 @ r2:r3 ln m/2 = ln m - ln2 Q62
@ r7 unbiased exponent @ r7 unbiased exponent
.equ dreddata1_plus_4, (dreddata1+4)
adr r4,dreddata1+4 adr r4,dreddata1_plus_4
ldmia r4,{r0,r1,r4} ldmia r4,{r0,r1,r4}
adds r7,#1 adds r7,#1
muls r0,r7 @ Q62 muls r0,r7 @ Q62
@ -2128,12 +2129,12 @@ regular_func dln_shim
pop {r4-r7,r15} pop {r4-r7,r15}
5: 5:
ldr r1,=#0xfff00000 ldr r1,=0xfff00000
movs r0,#0 movs r0,#0
pop {r4-r7,r15} pop {r4-r7,r15}
6: 6:
ldr r1,=#0x7ff00000 ldr r1,=0x7ff00000
movs r0,#0 movs r0,#0
pop {r4-r7,r15} pop {r4-r7,r15}

8
src/rp2_common/pico_float/float_aeabi.S
View File

@ -378,7 +378,7 @@ wrapper_func __aeabi_i2f
lsrs r1, r0, #31 lsrs r1, r0, #31
lsls r1, #31 lsls r1, #31
bpl 1f bpl 1f
rsbs r0, #0 negs r0, r0
1: 1:
cmp r0, #0 cmp r0, #0
beq 7f beq 7f
@ -392,7 +392,7 @@ __aeabi_i2f_main:
pop {r1, r2} pop {r1, r2}
lsls r1, r0 lsls r1, r0
subs r0, #158 subs r0, #158
rsbs r0, #0 negs r0, r0
adds r1,#0x80 @ rounding adds r1,#0x80 @ rounding
bcs 5f @ tripped carry? then have leading 1 in C as required (and result is even so can ignore sticky bits) bcs 5f @ tripped carry? then have leading 1 in C as required (and result is even so can ignore sticky bits)
@ -459,7 +459,7 @@ regular_func float2int_z
bl __aeabi_f2uiz bl __aeabi_f2uiz
cmp r0, #0 cmp r0, #0
bmi 1f bmi 1f
rsbs r0, #0 negs r0, r0
pop {pc} pop {pc}
1: 1:
movs r0, #128 movs r0, #128
@ -534,7 +534,7 @@ regular_func float2int64_z
cmp r1, #0 cmp r1, #0
bmi 1f bmi 1f
movs r2, #0 movs r2, #0
rsbs r0, #0 negs r0, r0
sbcs r2, r1 sbcs r2, r1
mov r1, r2 mov r1, r2
pop {pc} pop {pc}

18
src/rp2_common/pico_float/float_v1_rom_shim.S
View File

@ -100,7 +100,7 @@ regular_func fix642float_shim
movs r5,r0 movs r5,r0
movs r0,r1 movs r0,r1
4: 4:
rsbs r2,#0 negs r2,r2
adds r2,#32+29 adds r2,#32+29
// bl packx // bl packx
@ -132,7 +132,7 @@ regular_func fatan2_shim
bmi 2f @ force y to 0 proper, so result will be zero bmi 2f @ force y to 0 proper, so result will be zero
subs r4,r2,r3 @ calculate shift subs r4,r2,r3 @ calculate shift
bge 1f @ ex>=ey? bge 1f @ ex>=ey?
rsbs r4,#0 @ make shift positive negs r4,r4 @ make shift positive
asrs r0,r4 asrs r0,r4
cmp r4,#28 cmp r4,#28
blo 3f blo 3f
@ -159,8 +159,8 @@ regular_func fatan2_shim
ldr r3, =0x2cfc @ &pi_q29, circular coefficients ldr r3, =0x2cfc @ &pi_q29, circular coefficients
cmp r0,#0 @ x negative cmp r0,#0 @ x negative
bge 5f bge 5f
rsbs r0,#0 @ rotate to 1st/4th quadrants negs r0,r0 @ rotate to 1st/4th quadrants
rsbs r1,#0 negs r1,r1
ldr r2,[r3] @ pi Q29 ldr r2,[r3] @ pi Q29
5: 5:
movs r4,#1 @ m=1 movs r4,#1 @ m=1
@ -272,7 +272,7 @@ d2fix_a:
movs r4,r0 movs r4,r0
lsls r1,r2 lsls r1,r2
lsls r0,r2 lsls r0,r2
rsbs r2,#0 negs r2,r2
adds r2,#32 @ complementary shift adds r2,#32 @ complementary shift
lsrs r4,r2 lsrs r4,r2
orrs r1,r4 orrs r1,r4
@ -287,7 +287,7 @@ d2fix_a:
bmi 1f @ long shift? bmi 1f @ long shift?
mov r4,r1 mov r4,r1
lsls r4,r2 lsls r4,r2
rsbs r2,#0 negs r2,r2
adds r2,#32 @ complementary shift adds r2,#32 @ complementary shift
asrs r1,r2 asrs r1,r2
lsrs r0,r2 lsrs r0,r2
@ -299,7 +299,7 @@ d2fix_a:
asrs r1,#31 @ shift down 32 places asrs r1,#31 @ shift down 32 places
adds r2,#32 adds r2,#32
bmi 1f @ very long shift? bmi 1f @ very long shift?
rsbs r2,#0 negs r2,r2
adds r2,#32 adds r2,#32
asrs r0,r2 asrs r0,r2
pop {r4,r15} pop {r4,r15}
@ -329,7 +329,7 @@ regular_func float2double_shim
cmp r2,#0xff @ Inf? cmp r2,#0xff @ Inf?
beq 2f beq 2f
lsrs r1,#4 @ exponent and top 20 bits of mantissa lsrs r1,#4 @ exponent and top 20 bits of mantissa
ldr r2,=#(0x3ff-0x7f)<<20 @ difference in exponent offsets ldr r2,=(0x3ff-0x7f)<<20 @ difference in exponent offsets
adds r1,r2 adds r1,r2
orrs r1,r3 orrs r1,r3
lsls r0,#29 @ bottom 3 bits of mantissa lsls r0,#29 @ bottom 3 bits of mantissa
@ -340,7 +340,7 @@ regular_func float2double_shim
movs r0,#0 movs r0,#0
bx r14 bx r14
2: 2:
ldr r1,=#0x7ff00000 @ return signed infinity ldr r1,=0x7ff00000 @ return signed infinity
adds r1,r3 adds r1,r3
b 3b b 3b