diff options
Diffstat (limited to 'arch/mips/kernel/r4k_fpu.S')
| -rw-r--r-- | arch/mips/kernel/r4k_fpu.S | 436 |
1 files changed, 259 insertions, 177 deletions
diff --git a/arch/mips/kernel/r4k_fpu.S b/arch/mips/kernel/r4k_fpu.S index 1d88af26ba82..f09546ee2cdc 100644 --- a/arch/mips/kernel/r4k_fpu.S +++ b/arch/mips/kernel/r4k_fpu.S @@ -13,6 +13,7 @@ * Copyright (C) 1999, 2001 Silicon Graphics, Inc. */ #include <asm/asm.h> +#include <asm/asmmacro.h> #include <asm/errno.h> #include <asm/fpregdef.h> #include <asm/mipsregs.h> @@ -35,6 +36,14 @@ .set noreorder +/** + * _save_fp_context() - save FP context from the FPU + * @a0 - pointer to fpregs field of sigcontext + * @a1 - pointer to fpc_csr field of sigcontext + * + * Save FP context, including the 32 FP data registers and the FP + * control & status register, from the FPU to signal context. + */ LEAF(_save_fp_context) .set push SET_HARDFLOAT @@ -54,117 +63,60 @@ LEAF(_save_fp_context) nop #endif /* Store the 16 odd double precision registers */ - EX sdc1 $f1, SC_FPREGS+8(a0) - EX sdc1 $f3, SC_FPREGS+24(a0) - EX sdc1 $f5, SC_FPREGS+40(a0) - EX sdc1 $f7, SC_FPREGS+56(a0) - EX sdc1 $f9, SC_FPREGS+72(a0) - EX sdc1 $f11, SC_FPREGS+88(a0) - EX sdc1 $f13, SC_FPREGS+104(a0) - EX sdc1 $f15, SC_FPREGS+120(a0) - EX sdc1 $f17, SC_FPREGS+136(a0) - EX sdc1 $f19, SC_FPREGS+152(a0) - EX sdc1 $f21, SC_FPREGS+168(a0) - EX sdc1 $f23, SC_FPREGS+184(a0) - EX sdc1 $f25, SC_FPREGS+200(a0) - EX sdc1 $f27, SC_FPREGS+216(a0) - EX sdc1 $f29, SC_FPREGS+232(a0) - EX sdc1 $f31, SC_FPREGS+248(a0) + EX sdc1 $f1, 8(a0) + EX sdc1 $f3, 24(a0) + EX sdc1 $f5, 40(a0) + EX sdc1 $f7, 56(a0) + EX sdc1 $f9, 72(a0) + EX sdc1 $f11, 88(a0) + EX sdc1 $f13, 104(a0) + EX sdc1 $f15, 120(a0) + EX sdc1 $f17, 136(a0) + EX sdc1 $f19, 152(a0) + EX sdc1 $f21, 168(a0) + EX sdc1 $f23, 184(a0) + EX sdc1 $f25, 200(a0) + EX sdc1 $f27, 216(a0) + EX sdc1 $f29, 232(a0) + EX sdc1 $f31, 248(a0) 1: .set pop #endif .set push SET_HARDFLOAT /* Store the 16 even double precision registers */ - EX sdc1 $f0, SC_FPREGS+0(a0) - EX sdc1 $f2, SC_FPREGS+16(a0) - EX sdc1 $f4, SC_FPREGS+32(a0) - EX sdc1 $f6, SC_FPREGS+48(a0) - EX sdc1 $f8, SC_FPREGS+64(a0) - EX sdc1 $f10, SC_FPREGS+80(a0) - EX sdc1 $f12, SC_FPREGS+96(a0) - EX sdc1 $f14, SC_FPREGS+112(a0) - EX sdc1 $f16, SC_FPREGS+128(a0) - EX sdc1 $f18, SC_FPREGS+144(a0) - EX sdc1 $f20, SC_FPREGS+160(a0) - EX sdc1 $f22, SC_FPREGS+176(a0) - EX sdc1 $f24, SC_FPREGS+192(a0) - EX sdc1 $f26, SC_FPREGS+208(a0) - EX sdc1 $f28, SC_FPREGS+224(a0) - EX sdc1 $f30, SC_FPREGS+240(a0) - EX sw t1, SC_FPC_CSR(a0) + EX sdc1 $f0, 0(a0) + EX sdc1 $f2, 16(a0) + EX sdc1 $f4, 32(a0) + EX sdc1 $f6, 48(a0) + EX sdc1 $f8, 64(a0) + EX sdc1 $f10, 80(a0) + EX sdc1 $f12, 96(a0) + EX sdc1 $f14, 112(a0) + EX sdc1 $f16, 128(a0) + EX sdc1 $f18, 144(a0) + EX sdc1 $f20, 160(a0) + EX sdc1 $f22, 176(a0) + EX sdc1 $f24, 192(a0) + EX sdc1 $f26, 208(a0) + EX sdc1 $f28, 224(a0) + EX sdc1 $f30, 240(a0) + EX sw t1, 0(a1) jr ra li v0, 0 # success .set pop END(_save_fp_context) -#ifdef CONFIG_MIPS32_COMPAT - /* Save 32-bit process floating point context */ -LEAF(_save_fp_context32) - .set push - .set MIPS_ISA_ARCH_LEVEL_RAW - SET_HARDFLOAT - cfc1 t1, fcr31 - -#ifndef CONFIG_CPU_MIPS64_R6 - mfc0 t0, CP0_STATUS - sll t0, t0, 5 - bgez t0, 1f # skip storing odd if FR=0 - nop -#endif - - /* Store the 16 odd double precision registers */ - EX sdc1 $f1, SC32_FPREGS+8(a0) - EX sdc1 $f3, SC32_FPREGS+24(a0) - EX sdc1 $f5, SC32_FPREGS+40(a0) - EX sdc1 $f7, SC32_FPREGS+56(a0) - EX sdc1 $f9, SC32_FPREGS+72(a0) - EX sdc1 $f11, SC32_FPREGS+88(a0) - EX sdc1 $f13, SC32_FPREGS+104(a0) - EX sdc1 $f15, SC32_FPREGS+120(a0) - EX sdc1 $f17, SC32_FPREGS+136(a0) - EX sdc1 $f19, SC32_FPREGS+152(a0) - EX sdc1 $f21, SC32_FPREGS+168(a0) - EX sdc1 $f23, SC32_FPREGS+184(a0) - EX sdc1 $f25, SC32_FPREGS+200(a0) - EX sdc1 $f27, SC32_FPREGS+216(a0) - EX sdc1 $f29, SC32_FPREGS+232(a0) - EX sdc1 $f31, SC32_FPREGS+248(a0) - - /* Store the 16 even double precision registers */ -1: EX sdc1 $f0, SC32_FPREGS+0(a0) - EX sdc1 $f2, SC32_FPREGS+16(a0) - EX sdc1 $f4, SC32_FPREGS+32(a0) - EX sdc1 $f6, SC32_FPREGS+48(a0) - EX sdc1 $f8, SC32_FPREGS+64(a0) - EX sdc1 $f10, SC32_FPREGS+80(a0) - EX sdc1 $f12, SC32_FPREGS+96(a0) - EX sdc1 $f14, SC32_FPREGS+112(a0) - EX sdc1 $f16, SC32_FPREGS+128(a0) - EX sdc1 $f18, SC32_FPREGS+144(a0) - EX sdc1 $f20, SC32_FPREGS+160(a0) - EX sdc1 $f22, SC32_FPREGS+176(a0) - EX sdc1 $f24, SC32_FPREGS+192(a0) - EX sdc1 $f26, SC32_FPREGS+208(a0) - EX sdc1 $f28, SC32_FPREGS+224(a0) - EX sdc1 $f30, SC32_FPREGS+240(a0) - EX sw t1, SC32_FPC_CSR(a0) - cfc1 t0, $0 # implementation/version - EX sw t0, SC32_FPC_EIR(a0) - .set pop - - jr ra - li v0, 0 # success - END(_save_fp_context32) -#endif - -/* - * Restore FPU state: - * - fp gp registers - * - cp1 status/control register +/** + * _restore_fp_context() - restore FP context to the FPU + * @a0 - pointer to fpregs field of sigcontext + * @a1 - pointer to fpc_csr field of sigcontext + * + * Restore FP context, including the 32 FP data registers and the FP + * control & status register, from signal context to the FPU. */ LEAF(_restore_fp_context) - EX lw t1, SC_FPC_CSR(a0) + EX lw t1, 0(a1) #if defined(CONFIG_64BIT) || defined(CONFIG_CPU_MIPS32_R2) || \ defined(CONFIG_CPU_MIPS32_R6) @@ -178,101 +130,231 @@ LEAF(_restore_fp_context) bgez t0, 1f # skip loading odd if FR=0 nop #endif - EX ldc1 $f1, SC_FPREGS+8(a0) - EX ldc1 $f3, SC_FPREGS+24(a0) - EX ldc1 $f5, SC_FPREGS+40(a0) - EX ldc1 $f7, SC_FPREGS+56(a0) - EX ldc1 $f9, SC_FPREGS+72(a0) - EX ldc1 $f11, SC_FPREGS+88(a0) - EX ldc1 $f13, SC_FPREGS+104(a0) - EX ldc1 $f15, SC_FPREGS+120(a0) - EX ldc1 $f17, SC_FPREGS+136(a0) - EX ldc1 $f19, SC_FPREGS+152(a0) - EX ldc1 $f21, SC_FPREGS+168(a0) - EX ldc1 $f23, SC_FPREGS+184(a0) - EX ldc1 $f25, SC_FPREGS+200(a0) - EX ldc1 $f27, SC_FPREGS+216(a0) - EX ldc1 $f29, SC_FPREGS+232(a0) - EX ldc1 $f31, SC_FPREGS+248(a0) + EX ldc1 $f1, 8(a0) + EX ldc1 $f3, 24(a0) + EX ldc1 $f5, 40(a0) + EX ldc1 $f7, 56(a0) + EX ldc1 $f9, 72(a0) + EX ldc1 $f11, 88(a0) + EX ldc1 $f13, 104(a0) + EX ldc1 $f15, 120(a0) + EX ldc1 $f17, 136(a0) + EX ldc1 $f19, 152(a0) + EX ldc1 $f21, 168(a0) + EX ldc1 $f23, 184(a0) + EX ldc1 $f25, 200(a0) + EX ldc1 $f27, 216(a0) + EX ldc1 $f29, 232(a0) + EX ldc1 $f31, 248(a0) 1: .set pop #endif .set push SET_HARDFLOAT - EX ldc1 $f0, SC_FPREGS+0(a0) - EX ldc1 $f2, SC_FPREGS+16(a0) - EX ldc1 $f4, SC_FPREGS+32(a0) - EX ldc1 $f6, SC_FPREGS+48(a0) - EX ldc1 $f8, SC_FPREGS+64(a0) - EX ldc1 $f10, SC_FPREGS+80(a0) - EX ldc1 $f12, SC_FPREGS+96(a0) - EX ldc1 $f14, SC_FPREGS+112(a0) - EX ldc1 $f16, SC_FPREGS+128(a0) - EX ldc1 $f18, SC_FPREGS+144(a0) - EX ldc1 $f20, SC_FPREGS+160(a0) - EX ldc1 $f22, SC_FPREGS+176(a0) - EX ldc1 $f24, SC_FPREGS+192(a0) - EX ldc1 $f26, SC_FPREGS+208(a0) - EX ldc1 $f28, SC_FPREGS+224(a0) - EX ldc1 $f30, SC_FPREGS+240(a0) + EX ldc1 $f0, 0(a0) + EX ldc1 $f2, 16(a0) + EX ldc1 $f4, 32(a0) + EX ldc1 $f6, 48(a0) + EX ldc1 $f8, 64(a0) + EX ldc1 $f10, 80(a0) + EX ldc1 $f12, 96(a0) + EX ldc1 $f14, 112(a0) + EX ldc1 $f16, 128(a0) + EX ldc1 $f18, 144(a0) + EX ldc1 $f20, 160(a0) + EX ldc1 $f22, 176(a0) + EX ldc1 $f24, 192(a0) + EX ldc1 $f26, 208(a0) + EX ldc1 $f28, 224(a0) + EX ldc1 $f30, 240(a0) ctc1 t1, fcr31 .set pop jr ra li v0, 0 # success END(_restore_fp_context) -#ifdef CONFIG_MIPS32_COMPAT -LEAF(_restore_fp_context32) - /* Restore an o32 sigcontext. */ - .set push - SET_HARDFLOAT - EX lw t1, SC32_FPC_CSR(a0) +#ifdef CONFIG_CPU_HAS_MSA -#ifndef CONFIG_CPU_MIPS64_R6 - mfc0 t0, CP0_STATUS - sll t0, t0, 5 - bgez t0, 1f # skip loading odd if FR=0 + .macro op_one_wr op, idx, base + .align 4 +\idx: \op \idx, 0, \base + jr ra nop -#endif + .endm - EX ldc1 $f1, SC32_FPREGS+8(a0) - EX ldc1 $f3, SC32_FPREGS+24(a0) - EX ldc1 $f5, SC32_FPREGS+40(a0) - EX ldc1 $f7, SC32_FPREGS+56(a0) - EX ldc1 $f9, SC32_FPREGS+72(a0) - EX ldc1 $f11, SC32_FPREGS+88(a0) - EX ldc1 $f13, SC32_FPREGS+104(a0) - EX ldc1 $f15, SC32_FPREGS+120(a0) - EX ldc1 $f17, SC32_FPREGS+136(a0) - EX ldc1 $f19, SC32_FPREGS+152(a0) - EX ldc1 $f21, SC32_FPREGS+168(a0) - EX ldc1 $f23, SC32_FPREGS+184(a0) - EX ldc1 $f25, SC32_FPREGS+200(a0) - EX ldc1 $f27, SC32_FPREGS+216(a0) - EX ldc1 $f29, SC32_FPREGS+232(a0) - EX ldc1 $f31, SC32_FPREGS+248(a0) + .macro op_msa_wr name, op +LEAF(\name) + .set push + .set noreorder + sll t0, a0, 4 + PTR_LA t1, 0f + PTR_ADDU t0, t0, t1 + jr t0 + nop + op_one_wr \op, 0, a1 + op_one_wr \op, 1, a1 + op_one_wr \op, 2, a1 + op_one_wr \op, 3, a1 + op_one_wr \op, 4, a1 + op_one_wr \op, 5, a1 + op_one_wr \op, 6, a1 + op_one_wr \op, 7, a1 + op_one_wr \op, 8, a1 + op_one_wr \op, 9, a1 + op_one_wr \op, 10, a1 + op_one_wr \op, 11, a1 + op_one_wr \op, 12, a1 + op_one_wr \op, 13, a1 + op_one_wr \op, 14, a1 + op_one_wr \op, 15, a1 + op_one_wr \op, 16, a1 + op_one_wr \op, 17, a1 + op_one_wr \op, 18, a1 + op_one_wr \op, 19, a1 + op_one_wr \op, 20, a1 + op_one_wr \op, 21, a1 + op_one_wr \op, 22, a1 + op_one_wr \op, 23, a1 + op_one_wr \op, 24, a1 + op_one_wr \op, 25, a1 + op_one_wr \op, 26, a1 + op_one_wr \op, 27, a1 + op_one_wr \op, 28, a1 + op_one_wr \op, 29, a1 + op_one_wr \op, 30, a1 + op_one_wr \op, 31, a1 + .set pop + END(\name) + .endm -1: EX ldc1 $f0, SC32_FPREGS+0(a0) - EX ldc1 $f2, SC32_FPREGS+16(a0) - EX ldc1 $f4, SC32_FPREGS+32(a0) - EX ldc1 $f6, SC32_FPREGS+48(a0) - EX ldc1 $f8, SC32_FPREGS+64(a0) - EX ldc1 $f10, SC32_FPREGS+80(a0) - EX ldc1 $f12, SC32_FPREGS+96(a0) - EX ldc1 $f14, SC32_FPREGS+112(a0) - EX ldc1 $f16, SC32_FPREGS+128(a0) - EX ldc1 $f18, SC32_FPREGS+144(a0) - EX ldc1 $f20, SC32_FPREGS+160(a0) - EX ldc1 $f22, SC32_FPREGS+176(a0) - EX ldc1 $f24, SC32_FPREGS+192(a0) - EX ldc1 $f26, SC32_FPREGS+208(a0) - EX ldc1 $f28, SC32_FPREGS+224(a0) - EX ldc1 $f30, SC32_FPREGS+240(a0) - ctc1 t1, fcr31 + op_msa_wr read_msa_wr_b, st_b + op_msa_wr read_msa_wr_h, st_h + op_msa_wr read_msa_wr_w, st_w + op_msa_wr read_msa_wr_d, st_d + + op_msa_wr write_msa_wr_b, ld_b + op_msa_wr write_msa_wr_h, ld_h + op_msa_wr write_msa_wr_w, ld_w + op_msa_wr write_msa_wr_d, ld_d + +#endif /* CONFIG_CPU_HAS_MSA */ + +#ifdef CONFIG_CPU_HAS_MSA + + .macro save_msa_upper wr, off, base + .set push + .set noat +#ifdef CONFIG_64BIT + copy_u_d \wr, 1 + EX sd $1, \off(\base) +#elif defined(CONFIG_CPU_LITTLE_ENDIAN) + copy_u_w \wr, 2 + EX sw $1, \off(\base) + copy_u_w \wr, 3 + EX sw $1, (\off+4)(\base) +#else /* CONFIG_CPU_BIG_ENDIAN */ + copy_u_w \wr, 2 + EX sw $1, (\off+4)(\base) + copy_u_w \wr, 3 + EX sw $1, \off(\base) +#endif + .set pop + .endm + +LEAF(_save_msa_all_upper) + save_msa_upper 0, 0x00, a0 + save_msa_upper 1, 0x08, a0 + save_msa_upper 2, 0x10, a0 + save_msa_upper 3, 0x18, a0 + save_msa_upper 4, 0x20, a0 + save_msa_upper 5, 0x28, a0 + save_msa_upper 6, 0x30, a0 + save_msa_upper 7, 0x38, a0 + save_msa_upper 8, 0x40, a0 + save_msa_upper 9, 0x48, a0 + save_msa_upper 10, 0x50, a0 + save_msa_upper 11, 0x58, a0 + save_msa_upper 12, 0x60, a0 + save_msa_upper 13, 0x68, a0 + save_msa_upper 14, 0x70, a0 + save_msa_upper 15, 0x78, a0 + save_msa_upper 16, 0x80, a0 + save_msa_upper 17, 0x88, a0 + save_msa_upper 18, 0x90, a0 + save_msa_upper 19, 0x98, a0 + save_msa_upper 20, 0xa0, a0 + save_msa_upper 21, 0xa8, a0 + save_msa_upper 22, 0xb0, a0 + save_msa_upper 23, 0xb8, a0 + save_msa_upper 24, 0xc0, a0 + save_msa_upper 25, 0xc8, a0 + save_msa_upper 26, 0xd0, a0 + save_msa_upper 27, 0xd8, a0 + save_msa_upper 28, 0xe0, a0 + save_msa_upper 29, 0xe8, a0 + save_msa_upper 30, 0xf0, a0 + save_msa_upper 31, 0xf8, a0 jr ra - li v0, 0 # success - .set pop - END(_restore_fp_context32) + li v0, 0 + END(_save_msa_all_upper) + + .macro restore_msa_upper wr, off, base + .set push + .set noat +#ifdef CONFIG_64BIT + EX ld $1, \off(\base) + insert_d \wr, 1 +#elif defined(CONFIG_CPU_LITTLE_ENDIAN) + EX lw $1, \off(\base) + insert_w \wr, 2 + EX lw $1, (\off+4)(\base) + insert_w \wr, 3 +#else /* CONFIG_CPU_BIG_ENDIAN */ + EX lw $1, (\off+4)(\base) + insert_w \wr, 2 + EX lw $1, \off(\base) + insert_w \wr, 3 #endif + .set pop + .endm + +LEAF(_restore_msa_all_upper) + restore_msa_upper 0, 0x00, a0 + restore_msa_upper 1, 0x08, a0 + restore_msa_upper 2, 0x10, a0 + restore_msa_upper 3, 0x18, a0 + restore_msa_upper 4, 0x20, a0 + restore_msa_upper 5, 0x28, a0 + restore_msa_upper 6, 0x30, a0 + restore_msa_upper 7, 0x38, a0 + restore_msa_upper 8, 0x40, a0 + restore_msa_upper 9, 0x48, a0 + restore_msa_upper 10, 0x50, a0 + restore_msa_upper 11, 0x58, a0 + restore_msa_upper 12, 0x60, a0 + restore_msa_upper 13, 0x68, a0 + restore_msa_upper 14, 0x70, a0 + restore_msa_upper 15, 0x78, a0 + restore_msa_upper 16, 0x80, a0 + restore_msa_upper 17, 0x88, a0 + restore_msa_upper 18, 0x90, a0 + restore_msa_upper 19, 0x98, a0 + restore_msa_upper 20, 0xa0, a0 + restore_msa_upper 21, 0xa8, a0 + restore_msa_upper 22, 0xb0, a0 + restore_msa_upper 23, 0xb8, a0 + restore_msa_upper 24, 0xc0, a0 + restore_msa_upper 25, 0xc8, a0 + restore_msa_upper 26, 0xd0, a0 + restore_msa_upper 27, 0xd8, a0 + restore_msa_upper 28, 0xe0, a0 + restore_msa_upper 29, 0xe8, a0 + restore_msa_upper 30, 0xf0, a0 + restore_msa_upper 31, 0xf8, a0 + jr ra + li v0, 0 + END(_restore_msa_all_upper) + +#endif /* CONFIG_CPU_HAS_MSA */ .set reorder |