From: Hale Wang Date: Mon, 11 Jul 2016 17:11:31 +0000 (+0000) Subject: lib1funcs.S: Add new wrapper. X-Git-Url: https://git.libre-soc.org/?a=commitdiff_plain;h=827424041e867b19577eb46f8d8651239c41839f;p=gcc.git lib1funcs.S: Add new wrapper. 2016-07-11 Hale Wang Andre Vieira * config/arm/lib1funcs.S: Add new wrapper. Co-Authored-By: Andre Vieira From-SVN: r238215 --- diff --git a/libgcc/ChangeLog b/libgcc/ChangeLog index b6f830fee82..974ac5e5a4c 100644 --- a/libgcc/ChangeLog +++ b/libgcc/ChangeLog @@ -1,3 +1,8 @@ +2016-07-11 Hale Wang + Andre Vieira + + * config/arm/lib1funcs.S: Add new wrapper. + 2016-07-07 Thomas Preud'homme * config/arm/lib1funcs.S (__ARM_ARCH__): Define to 8 for ARMv8-M. diff --git a/libgcc/config/arm/lib1funcs.S b/libgcc/config/arm/lib1funcs.S index 96e206ee542..ba52e7b762f 100644 --- a/libgcc/config/arm/lib1funcs.S +++ b/libgcc/config/arm/lib1funcs.S @@ -311,34 +311,13 @@ LSYM(Lend_fde): #ifdef __ARM_EABI__ .macro THUMB_LDIV0 name signed #ifdef NOT_ISA_TARGET_32BIT - .ifc \signed, unsigned - cmp r0, #0 - beq 1f - mov r0, #0 - mvn r0, r0 @ 0xffffffff -1: - .else - cmp r0, #0 - beq 2f - blt 3f + + push {r0, lr} mov r0, #0 - mvn r0, r0 - lsr r0, r0, #1 @ 0x7fffffff - b 2f -3: mov r0, #0x80 - lsl r0, r0, #24 @ 0x80000000 -2: - .endif - push {r0, r1, r2} - ldr r0, 4f - adr r1, 4f - add r0, r1 - str r0, [sp, #8] + bl SYM(__aeabi_idiv0) @ We know we are not on armv4t, so pop pc is safe. - pop {r0, r1, pc} - .align 2 -4: - .word __aeabi_idiv0 - 4b + pop {r1, pc} + #elif defined(__thumb2__) .syntax unified .ifc \signed, unsigned @@ -950,7 +929,170 @@ LSYM(Lover7): add dividend, work .endif LSYM(Lgot_result): -.endm +.endm + +/* If performance is preferred, the following functions are provided. */ +#if defined(__prefer_thumb__) && !defined(__OPTIMIZE_SIZE__) + +/* Branch to div(n), and jump to label if curbit is lo than divisior. */ +.macro BranchToDiv n, label + lsr curbit, dividend, \n + cmp curbit, divisor + blo \label +.endm + +/* Body of div(n). Shift the divisor in n bits and compare the divisor + and dividend. Update the dividend as the substruction result. */ +.macro DoDiv n + lsr curbit, dividend, \n + cmp curbit, divisor + bcc 1f + lsl curbit, divisor, \n + sub dividend, dividend, curbit + +1: adc result, result +.endm + +/* The body of division with positive divisor. Unless the divisor is very + big, shift it up in multiples of four bits, since this is the amount of + unwinding in the main division loop. Continue shifting until the divisor + is larger than the dividend. */ +.macro THUMB1_Div_Positive + mov result, #0 + BranchToDiv #1, LSYM(Lthumb1_div1) + BranchToDiv #4, LSYM(Lthumb1_div4) + BranchToDiv #8, LSYM(Lthumb1_div8) + BranchToDiv #12, LSYM(Lthumb1_div12) + BranchToDiv #16, LSYM(Lthumb1_div16) +LSYM(Lthumb1_div_large_positive): + mov result, #0xff + lsl divisor, divisor, #8 + rev result, result + lsr curbit, dividend, #16 + cmp curbit, divisor + blo 1f + asr result, #8 + lsl divisor, divisor, #8 + beq LSYM(Ldivbyzero_waypoint) + +1: lsr curbit, dividend, #12 + cmp curbit, divisor + blo LSYM(Lthumb1_div12) + b LSYM(Lthumb1_div16) +LSYM(Lthumb1_div_loop): + lsr divisor, divisor, #8 +LSYM(Lthumb1_div16): + Dodiv #15 + Dodiv #14 + Dodiv #13 + Dodiv #12 +LSYM(Lthumb1_div12): + Dodiv #11 + Dodiv #10 + Dodiv #9 + Dodiv #8 + bcs LSYM(Lthumb1_div_loop) +LSYM(Lthumb1_div8): + Dodiv #7 + Dodiv #6 + Dodiv #5 +LSYM(Lthumb1_div5): + Dodiv #4 +LSYM(Lthumb1_div4): + Dodiv #3 +LSYM(Lthumb1_div3): + Dodiv #2 +LSYM(Lthumb1_div2): + Dodiv #1 +LSYM(Lthumb1_div1): + sub divisor, dividend, divisor + bcs 1f + cpy divisor, dividend + +1: adc result, result + cpy dividend, result + RET + +LSYM(Ldivbyzero_waypoint): + b LSYM(Ldiv0) +.endm + +/* The body of division with negative divisor. Similar with + THUMB1_Div_Positive except that the shift steps are in multiples + of six bits. */ +.macro THUMB1_Div_Negative + lsr result, divisor, #31 + beq 1f + neg divisor, divisor + +1: asr curbit, dividend, #32 + bcc 2f + neg dividend, dividend + +2: eor curbit, result + mov result, #0 + cpy ip, curbit + BranchToDiv #4, LSYM(Lthumb1_div_negative4) + BranchToDiv #8, LSYM(Lthumb1_div_negative8) +LSYM(Lthumb1_div_large): + mov result, #0xfc + lsl divisor, divisor, #6 + rev result, result + lsr curbit, dividend, #8 + cmp curbit, divisor + blo LSYM(Lthumb1_div_negative8) + + lsl divisor, divisor, #6 + asr result, result, #6 + cmp curbit, divisor + blo LSYM(Lthumb1_div_negative8) + + lsl divisor, divisor, #6 + asr result, result, #6 + cmp curbit, divisor + blo LSYM(Lthumb1_div_negative8) + + lsl divisor, divisor, #6 + beq LSYM(Ldivbyzero_negative) + asr result, result, #6 + b LSYM(Lthumb1_div_negative8) +LSYM(Lthumb1_div_negative_loop): + lsr divisor, divisor, #6 +LSYM(Lthumb1_div_negative8): + DoDiv #7 + DoDiv #6 + DoDiv #5 + DoDiv #4 +LSYM(Lthumb1_div_negative4): + DoDiv #3 + DoDiv #2 + bcs LSYM(Lthumb1_div_negative_loop) + DoDiv #1 + sub divisor, dividend, divisor + bcs 1f + cpy divisor, dividend + +1: cpy curbit, ip + adc result, result + asr curbit, curbit, #1 + cpy dividend, result + bcc 2f + neg dividend, dividend + cmp curbit, #0 + +2: bpl 3f + neg divisor, divisor + +3: RET + +LSYM(Ldivbyzero_negative): + cpy curbit, ip + asr curbit, curbit, #1 + bcc LSYM(Ldiv0) + neg dividend, dividend +.endm +#endif /* ARM Thumb version. */ + /* ------------------------------------------------------------------------ */ /* Start of the Real Functions */ /* ------------------------------------------------------------------------ */ @@ -960,6 +1102,7 @@ LSYM(Lgot_result): FUNC_START udivsi3 FUNC_ALIAS aeabi_uidiv udivsi3 +#if defined(__OPTIMIZE_SIZE__) cmp divisor, #0 beq LSYM(Ldiv0) @@ -977,6 +1120,14 @@ LSYM(udivsi3_skip_div0_test): pop { work } RET +/* Implementation of aeabi_uidiv for ARMv6m. This version is only + used in ARMv6-M when we need an efficient implementation. */ +#else +LSYM(udivsi3_skip_div0_test): + THUMB1_Div_Positive + +#endif /* __OPTIMIZE_SIZE__ */ + #elif defined(__ARM_ARCH_EXT_IDIV__) ARM_FUNC_START udivsi3 @@ -1028,12 +1179,21 @@ LSYM(udivsi3_skip_div0_test): FUNC_START aeabi_uidivmod cmp r1, #0 beq LSYM(Ldiv0) +# if defined(__OPTIMIZE_SIZE__) push {r0, r1, lr} bl LSYM(udivsi3_skip_div0_test) POP {r1, r2, r3} mul r2, r0 sub r1, r1, r2 bx r3 +# else + /* Both the quotient and remainder are calculated simultaneously + in THUMB1_Div_Positive. There is no need to calculate the + remainder again here. */ + b LSYM(udivsi3_skip_div0_test) + RET +# endif /* __OPTIMIZE_SIZE__ */ + #elif defined(__ARM_ARCH_EXT_IDIV__) ARM_FUNC_START aeabi_uidivmod cmp r1, #0 @@ -1089,7 +1249,7 @@ LSYM(Lover10): RET #else /* ARM version. */ - + FUNC_START umodsi3 subs r2, r1, #1 @ compare divisor with 1 @@ -1114,8 +1274,9 @@ LSYM(Lover10): #if defined(__prefer_thumb__) - FUNC_START divsi3 + FUNC_START divsi3 FUNC_ALIAS aeabi_idiv divsi3 +#if defined(__OPTIMIZE_SIZE__) cmp divisor, #0 beq LSYM(Ldiv0) @@ -1138,7 +1299,7 @@ LSYM(Lover11): blo LSYM(Lgot_result) THUMB_DIV_MOD_BODY 0 - + mov r0, result mov work, ip cmp work, #0 @@ -1148,6 +1309,22 @@ LSYM(Lover12): pop { work } RET +/* Implementation of aeabi_idiv for ARMv6m. This version is only + used in ARMv6-M when we need an efficient implementation. */ +#else +LSYM(divsi3_skip_div0_test): + cpy curbit, dividend + orr curbit, divisor + bmi LSYM(Lthumb1_div_negative) + +LSYM(Lthumb1_div_positive): + THUMB1_Div_Positive + +LSYM(Lthumb1_div_negative): + THUMB1_Div_Negative + +#endif /* __OPTIMIZE_SIZE__ */ + #elif defined(__ARM_ARCH_EXT_IDIV__) ARM_FUNC_START divsi3 @@ -1159,8 +1336,8 @@ LSYM(Lover12): RET #else /* ARM/Thumb-2 version. */ - - ARM_FUNC_START divsi3 + + ARM_FUNC_START divsi3 ARM_FUNC_ALIAS aeabi_idiv divsi3 cmp r1, #0 @@ -1214,12 +1391,21 @@ LSYM(divsi3_skip_div0_test): FUNC_START aeabi_idivmod cmp r1, #0 beq LSYM(Ldiv0) +# if defined(__OPTIMIZE_SIZE__) push {r0, r1, lr} bl LSYM(divsi3_skip_div0_test) POP {r1, r2, r3} mul r2, r0 sub r1, r1, r2 bx r3 +# else + /* Both the quotient and remainder are calculated simultaneously + in THUMB1_Div_Positive and THUMB1_Div_Negative. There is no + need to calculate the remainder again here. */ + b LSYM(divsi3_skip_div0_test) + RET +# endif /* __OPTIMIZE_SIZE__ */ + #elif defined(__ARM_ARCH_EXT_IDIV__) ARM_FUNC_START aeabi_idivmod cmp r1, #0