diff options
Diffstat (limited to 'arch/c6x/lib')
-rw-r--r-- | arch/c6x/lib/Makefile | 8 | ||||
-rw-r--r-- | arch/c6x/lib/checksum.c | 11 | ||||
-rw-r--r-- | arch/c6x/lib/csum_64plus.S | 414 | ||||
-rw-r--r-- | arch/c6x/lib/divi.S | 41 | ||||
-rw-r--r-- | arch/c6x/lib/divremi.S | 34 | ||||
-rw-r--r-- | arch/c6x/lib/divremu.S | 75 | ||||
-rw-r--r-- | arch/c6x/lib/divu.S | 86 | ||||
-rw-r--r-- | arch/c6x/lib/llshl.S | 25 | ||||
-rw-r--r-- | arch/c6x/lib/llshr.S | 26 | ||||
-rw-r--r-- | arch/c6x/lib/llshru.S | 26 | ||||
-rw-r--r-- | arch/c6x/lib/memcpy_64plus.S | 43 | ||||
-rw-r--r-- | arch/c6x/lib/mpyll.S | 37 | ||||
-rw-r--r-- | arch/c6x/lib/negll.S | 19 | ||||
-rw-r--r-- | arch/c6x/lib/pop_rts.S | 20 | ||||
-rw-r--r-- | arch/c6x/lib/push_rts.S | 19 | ||||
-rw-r--r-- | arch/c6x/lib/remi.S | 52 | ||||
-rw-r--r-- | arch/c6x/lib/remu.S | 70 | ||||
-rw-r--r-- | arch/c6x/lib/strasgi.S | 77 | ||||
-rw-r--r-- | arch/c6x/lib/strasgi_64plus.S | 27 |
19 files changed, 0 insertions, 1110 deletions
diff --git a/arch/c6x/lib/Makefile b/arch/c6x/lib/Makefile deleted file mode 100644 index e182004f82fe..000000000000 --- a/arch/c6x/lib/Makefile +++ /dev/null @@ -1,8 +0,0 @@ -# SPDX-License-Identifier: GPL-2.0-only -# -# Makefile for arch/c6x/lib/ -# - -lib-y := divu.o divi.o pop_rts.o push_rts.o remi.o remu.o strasgi.o llshru.o -lib-y += llshr.o llshl.o negll.o mpyll.o divremi.o divremu.o -lib-y += checksum.o csum_64plus.o memcpy_64plus.o strasgi_64plus.o diff --git a/arch/c6x/lib/checksum.c b/arch/c6x/lib/checksum.c deleted file mode 100644 index dff2e2ec6e64..000000000000 --- a/arch/c6x/lib/checksum.c +++ /dev/null @@ -1,11 +0,0 @@ -// SPDX-License-Identifier: GPL-2.0-or-later -/* - */ -#include <linux/module.h> -#include <net/checksum.h> - -/* These are from csum_64plus.S */ -EXPORT_SYMBOL(csum_partial); -EXPORT_SYMBOL(csum_partial_copy_nocheck); -EXPORT_SYMBOL(ip_compute_csum); -EXPORT_SYMBOL(ip_fast_csum); diff --git a/arch/c6x/lib/csum_64plus.S b/arch/c6x/lib/csum_64plus.S deleted file mode 100644 index 57148866d8d3..000000000000 --- a/arch/c6x/lib/csum_64plus.S +++ /dev/null @@ -1,414 +0,0 @@ -; SPDX-License-Identifier: GPL-2.0-only -; -; linux/arch/c6x/lib/csum_64plus.s -; -; Port on Texas Instruments TMS320C6x architecture -; -; Copyright (C) 2006, 2009, 2010, 2011 Texas Instruments Incorporated -; Author: Aurelien Jacquiot (aurelien.jacquiot@jaluna.com) -; -#include <linux/linkage.h> - -; -;unsigned int csum_partial_copy_nocheck(const char *src, char * dst, -; int len, int sum) -; -; A4: src -; B4: dst -; A6: len -; B6: sum -; return csum in A4 -; - - .text -ENTRY(csum_partial_copy_nocheck) - MVC .S2 ILC,B30 - - ZERO .D1 A9 ; csum (a side) -|| ZERO .D2 B9 ; csum (b side) -|| SHRU .S2X A6,2,B5 ; len / 4 - - ;; Check alignment and size - AND .S1 3,A4,A1 -|| AND .S2 3,B4,B0 - OR .L2X B0,A1,B0 ; non aligned condition -|| MVC .S2 B5,ILC -|| MVK .D2 1,B2 -|| MV .D1X B5,A1 ; words condition - [!A1] B .S1 L8 - [B0] BNOP .S1 L6,5 - - SPLOOP 1 - - ;; Main loop for aligned words - LDW .D1T1 *A4++,A7 - NOP 4 - MV .S2X A7,B7 -|| EXTU .S1 A7,0,16,A16 - STW .D2T2 B7,*B4++ -|| MPYU .M2 B7,B2,B8 -|| ADD .L1 A16,A9,A9 - NOP - SPKERNEL 8,0 -|| ADD .L2 B8,B9,B9 - - ZERO .D1 A1 -|| ADD .L1X A9,B9,A9 ; add csum from a and b sides - -L6: - [!A1] BNOP .S1 L8,5 - - ;; Main loop for non-aligned words - SPLOOP 2 - || MVK .L1 1,A2 - - LDNW .D1T1 *A4++,A7 - NOP 3 - - NOP - MV .S2X A7,B7 - || EXTU .S1 A7,0,16,A16 - || MPYU .M1 A7,A2,A8 - - ADD .L1 A16,A9,A9 - SPKERNEL 6,0 - || STNW .D2T2 B7,*B4++ - || ADD .L1 A8,A9,A9 - -L8: AND .S2X 2,A6,B5 - CMPGT .L2 B5,0,B0 - [!B0] BNOP .S1 L82,4 - - ;; Manage half-word - ZERO .L1 A7 -|| ZERO .D1 A8 - -#ifdef CONFIG_CPU_BIG_ENDIAN - - LDBU .D1T1 *A4++,A7 - LDBU .D1T1 *A4++,A8 - NOP 3 - SHL .S1 A7,8,A0 - ADD .S1 A8,A9,A9 - STB .D2T1 A7,*B4++ -|| ADD .S1 A0,A9,A9 - STB .D2T1 A8,*B4++ - -#else - - LDBU .D1T1 *A4++,A7 - LDBU .D1T1 *A4++,A8 - NOP 3 - ADD .S1 A7,A9,A9 - SHL .S1 A8,8,A0 - - STB .D2T1 A7,*B4++ -|| ADD .S1 A0,A9,A9 - STB .D2T1 A8,*B4++ - -#endif - - ;; Manage eventually the last byte -L82: AND .S2X 1,A6,B0 - [!B0] BNOP .S1 L9,5 - -|| ZERO .L1 A7 - -L83: LDBU .D1T1 *A4++,A7 - NOP 4 - - MV .L2X A7,B7 - -#ifdef CONFIG_CPU_BIG_ENDIAN - - STB .D2T2 B7,*B4++ -|| SHL .S1 A7,8,A7 - ADD .S1 A7,A9,A9 - -#else - - STB .D2T2 B7,*B4++ -|| ADD .S1 A7,A9,A9 - -#endif - - ;; Fold the csum -L9: SHRU .S2X A9,16,B0 - [!B0] BNOP .S1 L10,5 - -L91: SHRU .S2X A9,16,B4 -|| EXTU .S1 A9,16,16,A3 - ADD .D1X A3,B4,A9 - - SHRU .S1 A9,16,A0 - [A0] BNOP .S1 L91,5 - -L10: MV .D1 A9,A4 - - BNOP .S2 B3,4 - MVC .S2 B30,ILC -ENDPROC(csum_partial_copy_nocheck) - -; -;unsigned short -;ip_fast_csum(unsigned char *iph, unsigned int ihl) -;{ -; unsigned int checksum = 0; -; unsigned short *tosum = (unsigned short *) iph; -; int len; -; -; len = ihl*4; -; -; if (len <= 0) -; return 0; -; -; while(len) { -; len -= 2; -; checksum += *tosum++; -; } -; if (len & 1) -; checksum += *(unsigned char*) tosum; -; -; while(checksum >> 16) -; checksum = (checksum & 0xffff) + (checksum >> 16); -; -; return ~checksum; -;} -; -; A4: iph -; B4: ihl -; return checksum in A4 -; - .text - -ENTRY(ip_fast_csum) - ZERO .D1 A5 - || MVC .S2 ILC,B30 - SHL .S2 B4,2,B0 - CMPGT .L2 B0,0,B1 - [!B1] BNOP .S1 L15,4 - [!B1] ZERO .D1 A3 - - [!B0] B .S1 L12 - SHRU .S2 B0,1,B0 - MVC .S2 B0,ILC - NOP 3 - - SPLOOP 1 - LDHU .D1T1 *A4++,A3 - NOP 3 - NOP - SPKERNEL 5,0 - || ADD .L1 A3,A5,A5 - -L12: SHRU .S1 A5,16,A0 - [!A0] BNOP .S1 L14,5 - -L13: SHRU .S2X A5,16,B4 - EXTU .S1 A5,16,16,A3 - ADD .D1X A3,B4,A5 - SHRU .S1 A5,16,A0 - [A0] BNOP .S1 L13,5 - -L14: NOT .D1 A5,A3 - EXTU .S1 A3,16,16,A3 - -L15: BNOP .S2 B3,3 - MVC .S2 B30,ILC - MV .D1 A3,A4 -ENDPROC(ip_fast_csum) - -; -;unsigned short -;do_csum(unsigned char *buff, unsigned int len) -;{ -; int odd, count; -; unsigned int result = 0; -; -; if (len <= 0) -; goto out; -; odd = 1 & (unsigned long) buff; -; if (odd) { -;#ifdef __LITTLE_ENDIAN -; result += (*buff << 8); -;#else -; result = *buff; -;#endif -; len--; -; buff++; -; } -; count = len >> 1; /* nr of 16-bit words.. */ -; if (count) { -; if (2 & (unsigned long) buff) { -; result += *(unsigned short *) buff; -; count--; -; len -= 2; -; buff += 2; -; } -; count >>= 1; /* nr of 32-bit words.. */ -; if (count) { -; unsigned int carry = 0; -; do { -; unsigned int w = *(unsigned int *) buff; -; count--; -; buff += 4; -; result += carry; -; result += w; -; carry = (w > result); -; } while (count); -; result += carry; -; result = (result & 0xffff) + (result >> 16); -; } -; if (len & 2) { -; result += *(unsigned short *) buff; -; buff += 2; -; } -; } -; if (len & 1) -;#ifdef __LITTLE_ENDIAN -; result += *buff; -;#else -; result += (*buff << 8); -;#endif -; result = (result & 0xffff) + (result >> 16); -; /* add up carry.. */ -; result = (result & 0xffff) + (result >> 16); -; if (odd) -; result = ((result >> 8) & 0xff) | ((result & 0xff) << 8); -;out: -; return result; -;} -; -; A4: buff -; B4: len -; return checksum in A4 -; - -ENTRY(do_csum) - CMPGT .L2 B4,0,B0 - [!B0] BNOP .S1 L26,3 - EXTU .S1 A4,31,31,A0 - - MV .L1 A0,A3 -|| MV .S1X B3,A5 -|| MV .L2 B4,B3 -|| ZERO .D1 A1 - -#ifdef CONFIG_CPU_BIG_ENDIAN - [A0] SUB .L2 B3,1,B3 -|| [A0] LDBU .D1T1 *A4++,A1 -#else - [!A0] BNOP .S1 L21,5 -|| [A0] LDBU .D1T1 *A4++,A0 - SUB .L2 B3,1,B3 -|| SHL .S1 A0,8,A1 -L21: -#endif - SHR .S2 B3,1,B0 - [!B0] BNOP .S1 L24,3 - MVK .L1 2,A0 - AND .L1 A4,A0,A0 - - [!A0] BNOP .S1 L22,5 -|| [A0] LDHU .D1T1 *A4++,A0 - SUB .L2 B0,1,B0 -|| SUB .S2 B3,2,B3 -|| ADD .L1 A0,A1,A1 -L22: - SHR .S2 B0,1,B0 -|| ZERO .L1 A0 - - [!B0] BNOP .S1 L23,5 -|| [B0] MVC .S2 B0,ILC - - SPLOOP 3 - SPMASK L1 -|| MV .L1 A1,A2 -|| LDW .D1T1 *A4++,A1 - - NOP 4 - ADD .L1 A0,A1,A0 - ADD .L1 A2,A0,A2 - - SPKERNEL 1,2 -|| CMPGTU .L1 A1,A2,A0 - - ADD .L1 A0,A2,A6 - EXTU .S1 A6,16,16,A7 - SHRU .S2X A6,16,B0 - NOP 1 - ADD .L1X A7,B0,A1 -L23: - MVK .L2 2,B0 - AND .L2 B3,B0,B0 - [B0] LDHU .D1T1 *A4++,A0 - NOP 4 - [B0] ADD .L1 A0,A1,A1 -L24: - EXTU .S2 B3,31,31,B0 -#ifdef CONFIG_CPU_BIG_ENDIAN - [!B0] BNOP .S1 L25,4 -|| [B0] LDBU .D1T1 *A4,A0 - SHL .S1 A0,8,A0 - ADD .L1 A0,A1,A1 -L25: -#else - [B0] LDBU .D1T1 *A4,A0 - NOP 4 - [B0] ADD .L1 A0,A1,A1 -#endif - EXTU .S1 A1,16,16,A0 - SHRU .S2X A1,16,B0 - NOP 1 - ADD .L1X A0,B0,A0 - SHRU .S1 A0,16,A1 - ADD .L1 A0,A1,A0 - EXTU .S1 A0,16,16,A1 - EXTU .S1 A1,16,24,A2 - - EXTU .S1 A1,24,16,A0 -|| MV .L2X A3,B0 - - [B0] OR .L1 A0,A2,A1 -L26: - NOP 1 - BNOP .S2X A5,4 - MV .L1 A1,A4 -ENDPROC(do_csum) - -;__wsum csum_partial(const void *buff, int len, __wsum wsum) -;{ -; unsigned int sum = (__force unsigned int)wsum; -; unsigned int result = do_csum(buff, len); -; -; /* add in old sum, and carry.. */ -; result += sum; -; if (sum > result) -; result += 1; -; return (__force __wsum)result; -;} -; -ENTRY(csum_partial) - MV .L1X B3,A9 -|| CALLP .S2 do_csum,B3 -|| MV .S1 A6,A8 - BNOP .S2X A9,2 - ADD .L1 A8,A4,A1 - CMPGTU .L1 A8,A1,A0 - ADD .L1 A1,A0,A4 -ENDPROC(csum_partial) - -;unsigned short -;ip_compute_csum(unsigned char *buff, unsigned int len) -; -; A4: buff -; B4: len -; return checksum in A4 - -ENTRY(ip_compute_csum) - MV .L1X B3,A9 -|| CALLP .S2 do_csum,B3 - BNOP .S2X A9,3 - NOT .S1 A4,A4 - CLR .S1 A4,16,31,A4 -ENDPROC(ip_compute_csum) diff --git a/arch/c6x/lib/divi.S b/arch/c6x/lib/divi.S deleted file mode 100644 index d1764ae0b519..000000000000 --- a/arch/c6x/lib/divi.S +++ /dev/null @@ -1,41 +0,0 @@ -;; SPDX-License-Identifier: GPL-2.0-or-later -;; Copyright 2010 Free Software Foundation, Inc. -;; Contributed by Bernd Schmidt <bernds@codesourcery.com>. -;; - -#include <linux/linkage.h> - - ;; ABI considerations for the divide functions - ;; The following registers are call-used: - ;; __c6xabi_divi A0,A1,A2,A4,A6,B0,B1,B2,B4,B5 - ;; __c6xabi_divu A0,A1,A2,A4,A6,B0,B1,B2,B4 - ;; __c6xabi_remi A1,A2,A4,A5,A6,B0,B1,B2,B4 - ;; __c6xabi_remu A1,A4,A5,A7,B0,B1,B2,B4 - ;; - ;; In our implementation, divu and remu are leaf functions, - ;; while both divi and remi call into divu. - ;; A0 is not clobbered by any of the functions. - ;; divu does not clobber B2 either, which is taken advantage of - ;; in remi. - ;; divi uses B5 to hold the original return address during - ;; the call to divu. - ;; remi uses B2 and A5 to hold the input values during the - ;; call to divu. It stores B3 in on the stack. - - .text -ENTRY(__c6xabi_divi) - call .s2 __c6xabi_divu -|| mv .d2 B3, B5 -|| cmpgt .l1 0, A4, A1 -|| cmpgt .l2 0, B4, B1 - - [A1] neg .l1 A4, A4 -|| [B1] neg .l2 B4, B4 -|| xor .s1x A1, B1, A1 - [A1] addkpc .s2 _divu_ret, B3, 4 -_divu_ret: - neg .l1 A4, A4 -|| mv .l2 B3,B5 -|| ret .s2 B5 - nop 5 -ENDPROC(__c6xabi_divi) diff --git a/arch/c6x/lib/divremi.S b/arch/c6x/lib/divremi.S deleted file mode 100644 index 575fc57a8a76..000000000000 --- a/arch/c6x/lib/divremi.S +++ /dev/null @@ -1,34 +0,0 @@ -;; SPDX-License-Identifier: GPL-2.0-or-later -;; Copyright 2010 Free Software Foundation, Inc. -;; Contributed by Bernd Schmidt <bernds@codesourcery.com>. -;; - -#include <linux/linkage.h> - - .text -ENTRY(__c6xabi_divremi) - stw .d2t2 B3, *B15--[2] -|| cmpgt .l1 0, A4, A1 -|| cmpgt .l2 0, B4, B2 -|| mv .s1 A4, A5 -|| call .s2 __c6xabi_divu - - [A1] neg .l1 A4, A4 -|| [B2] neg .l2 B4, B4 -|| xor .s2x B2, A1, B0 -|| mv .d2 B4, B2 - - [B0] addkpc .s2 _divu_ret_1, B3, 1 - [!B0] addkpc .s2 _divu_ret_2, B3, 1 - nop 2 -_divu_ret_1: - neg .l1 A4, A4 -_divu_ret_2: - ldw .d2t2 *++B15[2], B3 - - mpy32 .m1x A4, B2, A6 - nop 3 - ret .s2 B3 - sub .l1 A5, A6, A5 - nop 4 -ENDPROC(__c6xabi_divremi) diff --git a/arch/c6x/lib/divremu.S b/arch/c6x/lib/divremu.S deleted file mode 100644 index 5f6a6a2997ae..000000000000 --- a/arch/c6x/lib/divremu.S +++ /dev/null @@ -1,75 +0,0 @@ -;; SPDX-License-Identifier: GPL-2.0-or-later -;; Copyright 2011 Free Software Foundation, Inc. -;; Contributed by Bernd Schmidt <bernds@codesourcery.com>. -;; - -#include <linux/linkage.h> - - .text -ENTRY(__c6xabi_divremu) - ;; We use a series of up to 31 subc instructions. First, we find - ;; out how many leading zero bits there are in the divisor. This - ;; gives us both a shift count for aligning (shifting) the divisor - ;; to the, and the number of times we have to execute subc. - - ;; At the end, we have both the remainder and most of the quotient - ;; in A4. The top bit of the quotient is computed first and is - ;; placed in A2. - - ;; Return immediately if the dividend is zero. Setting B4 to 1 - ;; is a trick to allow us to leave the following insns in the jump - ;; delay slot without affecting the result. - mv .s2x A4, B1 - - [b1] lmbd .l2 1, B4, B1 -||[!b1] b .s2 B3 ; RETURN A -||[!b1] mvk .d2 1, B4 - -||[!b1] zero .s1 A5 - mv .l1x B1, A6 -|| shl .s2 B4, B1, B4 - - ;; The loop performs a maximum of 28 steps, so we do the - ;; first 3 here. - cmpltu .l1x A4, B4, A2 - [!A2] sub .l1x A4, B4, A4 -|| shru .s2 B4, 1, B4 -|| xor .s1 1, A2, A2 - - shl .s1 A2, 31, A2 -|| [b1] subc .l1x A4,B4,A4 -|| [b1] add .s2 -1, B1, B1 - [b1] subc .l1x A4,B4,A4 -|| [b1] add .s2 -1, B1, B1 - - ;; RETURN A may happen here (note: must happen before the next branch) -__divremu0: - cmpgt .l2 B1, 7, B0 -|| [b1] subc .l1x A4,B4,A4 -|| [b1] add .s2 -1, B1, B1 - [b1] subc .l1x A4,B4,A4 -|| [b1] add .s2 -1, B1, B1 -|| [b0] b .s1 __divremu0 - [b1] subc .l1x A4,B4,A4 -|| [b1] add .s2 -1, B1, B1 - [b1] subc .l1x A4,B4,A4 -|| [b1] add .s2 -1, B1, B1 - [b1] subc .l1x A4,B4,A4 -|| [b1] add .s2 -1, B1, B1 - [b1] subc .l1x A4,B4,A4 -|| [b1] add .s2 -1, B1, B1 - [b1] subc .l1x A4,B4,A4 -|| [b1] add .s2 -1, B1, B1 - ;; loop backwards branch happens here - - ret .s2 B3 -|| mvk .s1 32, A1 - sub .l1 A1, A6, A6 -|| extu .s1 A4, A6, A5 - shl .s1 A4, A6, A4 - shru .s1 A4, 1, A4 -|| sub .l1 A6, 1, A6 - or .l1 A2, A4, A4 - shru .s1 A4, A6, A4 - nop -ENDPROC(__c6xabi_divremu) diff --git a/arch/c6x/lib/divu.S b/arch/c6x/lib/divu.S deleted file mode 100644 index f0f6082944c2..000000000000 --- a/arch/c6x/lib/divu.S +++ /dev/null @@ -1,86 +0,0 @@ -;; SPDX-License-Identifier: GPL-2.0-or-later -;; Copyright 2010 Free Software Foundation, Inc. -;; Contributed by Bernd Schmidt <bernds@codesourcery.com>. -;; - -#include <linux/linkage.h> - - ;; ABI considerations for the divide functions - ;; The following registers are call-used: - ;; __c6xabi_divi A0,A1,A2,A4,A6,B0,B1,B2,B4,B5 - ;; __c6xabi_divu A0,A1,A2,A4,A6,B0,B1,B2,B4 - ;; __c6xabi_remi A1,A2,A4,A5,A6,B0,B1,B2,B4 - ;; __c6xabi_remu A1,A4,A5,A7,B0,B1,B2,B4 - ;; - ;; In our implementation, divu and remu are leaf functions, - ;; while both divi and remi call into divu. - ;; A0 is not clobbered by any of the functions. - ;; divu does not clobber B2 either, which is taken advantage of - ;; in remi. - ;; divi uses B5 to hold the original return address during - ;; the call to divu. - ;; remi uses B2 and A5 to hold the input values during the - ;; call to divu. It stores B3 in on the stack. - - .text -ENTRY(__c6xabi_divu) - ;; We use a series of up to 31 subc instructions. First, we find - ;; out how many leading zero bits there are in the divisor. This - ;; gives us both a shift count for aligning (shifting) the divisor - ;; to the, and the number of times we have to execute subc. - - ;; At the end, we have both the remainder and most of the quotient - ;; in A4. The top bit of the quotient is computed first and is - ;; placed in A2. - - ;; Return immediately if the dividend is zero. - mv .s2x A4, B1 - [B1] lmbd .l2 1, B4, B1 -|| [!B1] b .s2 B3 ; RETURN A -|| [!B1] mvk .d2 1, B4 - mv .l1x B1, A6 -|| shl .s2 B4, B1, B4 - - ;; The loop performs a maximum of 28 steps, so we do the - ;; first 3 here. - cmpltu .l1x A4, B4, A2 - [!A2] sub .l1x A4, B4, A4 -|| shru .s2 B4, 1, B4 -|| xor .s1 1, A2, A2 - - shl .s1 A2, 31, A2 -|| [B1] subc .l1x A4,B4,A4 -|| [B1] add .s2 -1, B1, B1 - [B1] subc .l1x A4,B4,A4 -|| [B1] add .s2 -1, B1, B1 - - ;; RETURN A may happen here (note: must happen before the next branch) -_divu_loop: - cmpgt .l2 B1, 7, B0 -|| [B1] subc .l1x A4,B4,A4 -|| [B1] add .s2 -1, B1, B1 - [B1] subc .l1x A4,B4,A4 -|| [B1] add .s2 -1, B1, B1 -|| [B0] b .s1 _divu_loop - [B1] subc .l1x A4,B4,A4 -|| [B1] add .s2 -1, B1, B1 - [B1] subc .l1x A4,B4,A4 -|| [B1] add .s2 -1, B1, B1 - [B1] subc .l1x A4,B4,A4 -|| [B1] add .s2 -1, B1, B1 - [B1] subc .l1x A4,B4,A4 -|| [B1] add .s2 -1, B1, B1 - [B1] subc .l1x A4,B4,A4 -|| [B1] add .s2 -1, B1, B1 - ;; loop backwards branch happens here - - ret .s2 B3 -|| mvk .s1 32, A1 - sub .l1 A1, A6, A6 - shl .s1 A4, A6, A4 - shru .s1 A4, 1, A4 -|| sub .l1 A6, 1, A6 - or .l1 A2, A4, A4 - shru .s1 A4, A6, A4 - nop -ENDPROC(__c6xabi_divu) diff --git a/arch/c6x/lib/llshl.S b/arch/c6x/lib/llshl.S deleted file mode 100644 index 3272499618e0..000000000000 --- a/arch/c6x/lib/llshl.S +++ /dev/null @@ -1,25 +0,0 @@ -;; SPDX-License-Identifier: GPL-2.0-or-later -;; Copyright (C) 2010 Texas Instruments Incorporated -;; Contributed by Mark Salter <msalter@redhat.com>. -;; - -;; uint64_t __c6xabi_llshl(uint64_t val, uint shift) - -#include <linux/linkage.h> - - .text -ENTRY(__c6xabi_llshl) - mv .l1x B4,A1 - [!A1] b .s2 B3 ; just return if zero shift - mvk .s1 32,A0 - sub .d1 A0,A1,A0 - cmplt .l1 0,A0,A2 - [A2] shru .s1 A4,A0,A0 - [!A2] neg .l1 A0,A5 -|| [A2] shl .s1 A5,A1,A5 - [!A2] shl .s1 A4,A5,A5 -|| [A2] or .d1 A5,A0,A5 -|| [!A2] mvk .l1 0,A4 - [A2] shl .s1 A4,A1,A4 - bnop .s2 B3,5 -ENDPROC(__c6xabi_llshl) diff --git a/arch/c6x/lib/llshr.S b/arch/c6x/lib/llshr.S deleted file mode 100644 index 6bfaacd15e73..000000000000 --- a/arch/c6x/lib/llshr.S +++ /dev/null @@ -1,26 +0,0 @@ -;; SPDX-License-Identifier: GPL-2.0-or-later -;; Copyright (C) 2010 Texas Instruments Incorporated -;; Contributed by Mark Salter <msalter@redhat.com>. -;; - -;; uint64_t __c6xabi_llshr(uint64_t val, uint shift) - -#include <linux/linkage.h> - - .text -ENTRY(__c6xabi_llshr) - mv .l1x B4,A1 - [!A1] b .s2 B3 ; return if zero shift count - mvk .s1 32,A0 - sub .d1 A0,A1,A0 - cmplt .l1 0,A0,A2 - [A2] shl .s1 A5,A0,A0 - nop - [!A2] neg .l1 A0,A4 -|| [A2] shru .s1 A4,A1,A4 - [!A2] shr .s1 A5,A4,A4 -|| [A2] or .d1 A4,A0,A4 - [!A2] shr .s1 A5,0x1f,A5 - [A2] shr .s1 A5,A1,A5 - bnop .s2 B3,5 -ENDPROC(__c6xabi_llshr) diff --git a/arch/c6x/lib/llshru.S b/arch/c6x/lib/llshru.S deleted file mode 100644 index 103128f50770..000000000000 --- a/arch/c6x/lib/llshru.S +++ /dev/null @@ -1,26 +0,0 @@ -;; SPDX-License-Identifier: GPL-2.0-or-later -;; Copyright (C) 2010 Texas Instruments Incorporated -;; Contributed by Mark Salter <msalter@redhat.com>. -;; - -;; uint64_t __c6xabi_llshru(uint64_t val, uint shift) - -#include <linux/linkage.h> - - .text -ENTRY(__c6xabi_llshru) - mv .l1x B4,A1 - [!A1] b .s2 B3 ; return if zero shift count - mvk .s1 32,A0 - sub .d1 A0,A1,A0 - cmplt .l1 0,A0,A2 - [A2] shl .s1 A5,A0,A0 - nop - [!A2] neg .l1 A0,A4 -|| [A2] shru .s1 A4,A1,A4 - [!A2] shru .s1 A5,A4,A4 -|| [A2] or .d1 A4,A0,A4 -|| [!A2] mvk .l1 0,A5 - [A2] shru .s1 A5,A1,A5 - bnop .s2 B3,5 -ENDPROC(__c6xabi_llshru) diff --git a/arch/c6x/lib/memcpy_64plus.S b/arch/c6x/lib/memcpy_64plus.S deleted file mode 100644 index 157a30486bfd..000000000000 --- a/arch/c6x/lib/memcpy_64plus.S +++ /dev/null @@ -1,43 +0,0 @@ -; SPDX-License-Identifier: GPL-2.0-only -; Port on Texas Instruments TMS320C6x architecture -; -; Copyright (C) 2006, 2009, 2010 Texas Instruments Incorporated -; Author: Aurelien Jacquiot (aurelien.jacquiot@jaluna.com) -; - -#include <linux/linkage.h> - - .text - -ENTRY(memcpy) - AND .L1 0x1,A6,A0 - || AND .S1 0x2,A6,A1 - || AND .L2X 0x4,A6,B0 - || MV .D1 A4,A3 - || MVC .S2 ILC,B2 - - [A0] LDB .D2T1 *B4++,A5 - [A1] LDB .D2T1 *B4++,A7 - [A1] LDB .D2T1 *B4++,A8 - [B0] LDNW .D2T1 *B4++,A9 - || SHRU .S2X A6,0x3,B1 - [!B1] BNOP .S2 B3,1 - - [A0] STB .D1T1 A5,*A3++ - ||[B1] MVC .S2 B1,ILC - [A1] STB .D1T1 A7,*A3++ - [A1] STB .D1T1 A8,*A3++ - [B0] STNW .D1T1 A9,*A3++ ; return when len < 8 - - SPLOOP 2 - - LDNDW .D2T1 *B4++,A9:A8 - NOP 3 - - NOP - SPKERNEL 0,0 - || STNDW .D1T1 A9:A8,*A3++ - - BNOP .S2 B3,4 - MVC .S2 B2,ILC -ENDPROC(memcpy) diff --git a/arch/c6x/lib/mpyll.S b/arch/c6x/lib/mpyll.S deleted file mode 100644 index d07c13ec4fd4..000000000000 --- a/arch/c6x/lib/mpyll.S +++ /dev/null @@ -1,37 +0,0 @@ -;; SPDX-License-Identifier: GPL-2.0-or-later -;; Copyright (C) 2010 Texas Instruments Incorporated -;; Contributed by Mark Salter <msalter@redhat.com>. -;; - -#include <linux/linkage.h> - - ;; uint64_t __c6xabi_mpyll(uint64_t x, uint64_t y) - ;; - ;; 64x64 multiply - ;; First compute partial results using 32-bit parts of x and y: - ;; - ;; b63 b32 b31 b0 - ;; ----------------------------- - ;; | 1 | 0 | - ;; ----------------------------- - ;; - ;; P0 = X0*Y0 - ;; P1 = X0*Y1 + X1*Y0 - ;; P2 = X1*Y1 - ;; - ;; result = (P2 << 64) + (P1 << 32) + P0 - ;; - ;; Since the result is also 64-bit, we can skip the P2 term. - - .text -ENTRY(__c6xabi_mpyll) - mpy32u .m1x A4,B4,A1:A0 ; X0*Y0 - b .s2 B3 - || mpy32u .m2x B5,A4,B1:B0 ; X0*Y1 (don't need upper 32-bits) - || mpy32u .m1x A5,B4,A3:A2 ; X1*Y0 (don't need upper 32-bits) - nop - nop - mv .s1 A0,A4 - add .l1x A2,B0,A5 - add .s1 A1,A5,A5 -ENDPROC(__c6xabi_mpyll) diff --git a/arch/c6x/lib/negll.S b/arch/c6x/lib/negll.S deleted file mode 100644 index 9ba434db5366..000000000000 --- a/arch/c6x/lib/negll.S +++ /dev/null @@ -1,19 +0,0 @@ -;; SPDX-License-Identifier: GPL-2.0-or-later -;; Copyright (C) 2010 Texas Instruments Incorporated -;; Contributed by Mark Salter <msalter@redhat.com>. -;; - -;; int64_t __c6xabi_negll(int64_t val) - -#include <linux/linkage.h> - - .text -ENTRY(__c6xabi_negll) - b .s2 B3 - mvk .l1 0,A0 - subu .l1 A0,A4,A3:A2 - sub .l1 A0,A5,A0 -|| ext .s1 A3,24,24,A5 - add .l1 A5,A0,A5 - mv .s1 A2,A4 -ENDPROC(__c6xabi_negll) diff --git a/arch/c6x/lib/pop_rts.S b/arch/c6x/lib/pop_rts.S deleted file mode 100644 index f129e32943c5..000000000000 --- a/arch/c6x/lib/pop_rts.S +++ /dev/null @@ -1,20 +0,0 @@ -;; SPDX-License-Identifier: GPL-2.0-or-later -;; Copyright 2010 Free Software Foundation, Inc. -;; Contributed by Bernd Schmidt <bernds@codesourcery.com>. -;; - -#include <linux/linkage.h> - - .text - -ENTRY(__c6xabi_pop_rts) - lddw .d2t2 *++B15, B3:B2 - lddw .d2t1 *++B15, A11:A10 - lddw .d2t2 *++B15, B11:B10 - lddw .d2t1 *++B15, A13:A12 - lddw .d2t2 *++B15, B13:B12 - lddw .d2t1 *++B15, A15:A14 -|| b .s2 B3 - ldw .d2t2 *++B15[2], B14 - nop 4 -ENDPROC(__c6xabi_pop_rts) diff --git a/arch/c6x/lib/push_rts.S b/arch/c6x/lib/push_rts.S deleted file mode 100644 index 40b0a4fe937c..000000000000 --- a/arch/c6x/lib/push_rts.S +++ /dev/null @@ -1,19 +0,0 @@ -;; SPDX-License-Identifier: GPL-2.0-or-later -;; Copyright 2010 Free Software Foundation, Inc. -;; Contributed by Bernd Schmidt <bernds@codesourcery.com>. -;; - -#include <linux/linkage.h> - - .text - -ENTRY(__c6xabi_push_rts) - stw .d2t2 B14, *B15--[2] - stdw .d2t1 A15:A14, *B15-- -|| b .s2x A3 - stdw .d2t2 B13:B12, *B15-- - stdw .d2t1 A13:A12, *B15-- - stdw .d2t2 B11:B10, *B15-- - stdw .d2t1 A11:A10, *B15-- - stdw .d2t2 B3:B2, *B15-- -ENDPROC(__c6xabi_push_rts) diff --git a/arch/c6x/lib/remi.S b/arch/c6x/lib/remi.S deleted file mode 100644 index 96a1335eac20..000000000000 --- a/arch/c6x/lib/remi.S +++ /dev/null @@ -1,52 +0,0 @@ -;; SPDX-License-Identifier: GPL-2.0-or-later -;; Copyright 2010 Free Software Foundation, Inc. -;; Contributed by Bernd Schmidt <bernds@codesourcery.com>. -;; - -#include <linux/linkage.h> - - ;; ABI considerations for the divide functions - ;; The following registers are call-used: - ;; __c6xabi_divi A0,A1,A2,A4,A6,B0,B1,B2,B4,B5 - ;; __c6xabi_divu A0,A1,A2,A4,A6,B0,B1,B2,B4 - ;; __c6xabi_remi A1,A2,A4,A5,A6,B0,B1,B2,B4 - ;; __c6xabi_remu A1,A4,A5,A7,B0,B1,B2,B4 - ;; - ;; In our implementation, divu and remu are leaf functions, - ;; while both divi and remi call into divu. - ;; A0 is not clobbered by any of the functions. - ;; divu does not clobber B2 either, which is taken advantage of - ;; in remi. - ;; divi uses B5 to hold the original return address during - ;; the call to divu. - ;; remi uses B2 and A5 to hold the input values during the - ;; call to divu. It stores B3 in on the stack. - - .text - -ENTRY(__c6xabi_remi) - stw .d2t2 B3, *B15--[2] -|| cmpgt .l1 0, A4, A1 -|| cmpgt .l2 0, B4, B2 -|| mv .s1 A4, A5 -|| call .s2 __c6xabi_divu - - [A1] neg .l1 A4, A4 -|| [B2] neg .l2 B4, B4 -|| xor .s2x B2, A1, B0 -|| mv .d2 B4, B2 - - [B0] addkpc .s2 _divu_ret_1, B3, 1 - [!B0] addkpc .s2 _divu_ret_2, B3, 1 - nop 2 -_divu_ret_1: - neg .l1 A4, A4 -_divu_ret_2: - ldw .d2t2 *++B15[2], B3 - - mpy32 .m1x A4, B2, A6 - nop 3 - ret .s2 B3 - sub .l1 A5, A6, A4 - nop 4 -ENDPROC(__c6xabi_remi) diff --git a/arch/c6x/lib/remu.S b/arch/c6x/lib/remu.S deleted file mode 100644 index 428feb9c06c0..000000000000 --- a/arch/c6x/lib/remu.S +++ /dev/null @@ -1,70 +0,0 @@ -;; SPDX-License-Identifier: GPL-2.0-or-later -;; Copyright 2010 Free Software Foundation, Inc. -;; Contributed by Bernd Schmidt <bernds@codesourcery.com>. -;; - -#include <linux/linkage.h> - - ;; ABI considerations for the divide functions - ;; The following registers are call-used: - ;; __c6xabi_divi A0,A1,A2,A4,A6,B0,B1,B2,B4,B5 - ;; __c6xabi_divu A0,A1,A2,A4,A6,B0,B1,B2,B4 - ;; __c6xabi_remi A1,A2,A4,A5,A6,B0,B1,B2,B4 - ;; __c6xabi_remu A1,A4,A5,A7,B0,B1,B2,B4 - ;; - ;; In our implementation, divu and remu are leaf functions, - ;; while both divi and remi call into divu. - ;; A0 is not clobbered by any of the functions. - ;; divu does not clobber B2 either, which is taken advantage of - ;; in remi. - ;; divi uses B5 to hold the original return address during - ;; the call to divu. - ;; remi uses B2 and A5 to hold the input values during the - ;; call to divu. It stores B3 in on the stack. - - - .text - -ENTRY(__c6xabi_remu) - ;; The ABI seems designed to prevent these functions calling each other, - ;; so we duplicate most of the divsi3 code here. - mv .s2x A4, B1 - lmbd .l2 1, B4, B1 -|| [!B1] b .s2 B3 ; RETURN A -|| [!B1] mvk .d2 1, B4 - - mv .l1x B1, A7 -|| shl .s2 B4, B1, B4 - - cmpltu .l1x A4, B4, A1 - [!A1] sub .l1x A4, B4, A4 - shru .s2 B4, 1, B4 - -_remu_loop: - cmpgt .l2 B1, 7, B0 -|| [B1] subc .l1x A4,B4,A4 -|| [B1] add .s2 -1, B1, B1 - ;; RETURN A may happen here (note: must happen before the next branch) - [B1] subc .l1x A4,B4,A4 -|| [B1] add .s2 -1, B1, B1 -|| [B0] b .s1 _remu_loop - [B1] subc .l1x A4,B4,A4 -|| [B1] add .s2 -1, B1, B1 - [B1] subc .l1x A4,B4,A4 -|| [B1] add .s2 -1, B1, B1 - [B1] subc .l1x A4,B4,A4 -|| [B1] add .s2 -1, B1, B1 - [B1] subc .l1x A4,B4,A4 -|| [B1] add .s2 -1, B1, B1 - [B1] subc .l1x A4,B4,A4 -|| [B1] add .s2 -1, B1, B1 - ;; loop backwards branch happens here - - ret .s2 B3 - [B1] subc .l1x A4,B4,A4 -|| [B1] add .s2 -1, B1, B1 - [B1] subc .l1x A4,B4,A4 - - extu .s1 A4, A7, A4 - nop 2 -ENDPROC(__c6xabi_remu) diff --git a/arch/c6x/lib/strasgi.S b/arch/c6x/lib/strasgi.S deleted file mode 100644 index 715aeb200792..000000000000 --- a/arch/c6x/lib/strasgi.S +++ /dev/null @@ -1,77 +0,0 @@ -;; SPDX-License-Identifier: GPL-2.0-or-later -;; Copyright 2010 Free Software Foundation, Inc. -;; Contributed by Bernd Schmidt <bernds@codesourcery.com>. -;; - -#include <linux/linkage.h> - - .text - -ENTRY(__c6xabi_strasgi) - ;; This is essentially memcpy, with alignment known to be at least - ;; 4, and the size a multiple of 4 greater than or equal to 28. - ldw .d2t1 *B4++, A0 -|| mvk .s2 16, B1 - ldw .d2t1 *B4++, A1 -|| mvk .s2 20, B2 -|| sub .d1 A6, 24, A6 - ldw .d2t1 *B4++, A5 - ldw .d2t1 *B4++, A7 -|| mv .l2x A6, B7 - ldw .d2t1 *B4++, A8 - ldw .d2t1 *B4++, A9 -|| mv .s2x A0, B5 -|| cmpltu .l2 B2, B7, B0 - -_strasgi_loop: - stw .d1t2 B5, *A4++ -|| [B0] ldw .d2t1 *B4++, A0 -|| mv .s2x A1, B5 -|| mv .l2 B7, B6 - - [B0] sub .d2 B6, 24, B7 -|| [B0] b .s2 _strasgi_loop -|| cmpltu .l2 B1, B6, B0 - - [B0] ldw .d2t1 *B4++, A1 -|| stw .d1t2 B5, *A4++ -|| mv .s2x A5, B5 -|| cmpltu .l2 12, B6, B0 - - [B0] ldw .d2t1 *B4++, A5 -|| stw .d1t2 B5, *A4++ -|| mv .s2x A7, B5 -|| cmpltu .l2 8, B6, B0 - - [B0] ldw .d2t1 *B4++, A7 -|| stw .d1t2 B5, *A4++ -|| mv .s2x A8, B5 -|| cmpltu .l2 4, B6, B0 - - [B0] ldw .d2t1 *B4++, A8 -|| stw .d1t2 B5, *A4++ -|| mv .s2x A9, B5 -|| cmpltu .l2 0, B6, B0 - - [B0] ldw .d2t1 *B4++, A9 -|| stw .d1t2 B5, *A4++ -|| mv .s2x A0, B5 -|| cmpltu .l2 B2, B7, B0 - - ;; loop back branch happens here - - cmpltu .l2 B1, B6, B0 -|| ret .s2 b3 - - [B0] stw .d1t1 A1, *A4++ -|| cmpltu .l2 12, B6, B0 - [B0] stw .d1t1 A5, *A4++ -|| cmpltu .l2 8, B6, B0 - [B0] stw .d1t1 A7, *A4++ -|| cmpltu .l2 4, B6, B0 - [B0] stw .d1t1 A8, *A4++ -|| cmpltu .l2 0, B6, B0 - [B0] stw .d1t1 A9, *A4++ - - ;; return happens here -ENDPROC(__c6xabi_strasgi) diff --git a/arch/c6x/lib/strasgi_64plus.S b/arch/c6x/lib/strasgi_64plus.S deleted file mode 100644 index d10aa2dc3249..000000000000 --- a/arch/c6x/lib/strasgi_64plus.S +++ /dev/null @@ -1,27 +0,0 @@ -;; SPDX-License-Identifier: GPL-2.0-or-later -;; Copyright 2010 Free Software Foundation, Inc. -;; Contributed by Bernd Schmidt <bernds@codesourcery.com>. -;; - -#include <linux/linkage.h> - - .text - -ENTRY(__c6xabi_strasgi_64plus) - shru .s2x a6, 2, b31 -|| mv .s1 a4, a30 -|| mv .d2 b4, b30 - - add .s2 -4, b31, b31 - - sploopd 1 -|| mvc .s2 b31, ilc - ldw .d2t2 *b30++, b31 - nop 4 - mv .s1x b31,a31 - spkernel 6, 0 -|| stw .d1t1 a31, *a30++ - - ret .s2 b3 - nop 5 -ENDPROC(__c6xabi_strasgi_64plus) |