AuxXxilium/linux_dsm_epyc7002
1
0
Fork 0
mirror of https://github.com/AuxXxilium/linux_dsm_epyc7002.git synced 2024-12-26 19:45:22 +07:00
Code Issues Actions Packages Projects Releases Wiki Activity
3f07c01441
linux_dsm_epyc7002/arch/sh/kernel/cpu/sh4/fpu.c

430 lines
12 KiB
C
Raw Normal View History

sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
/*
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
* Save/restore floating point context for signal handlers.
*
* This file is subject to the terms and conditions of the GNU General Public
* License. See the file "COPYING" in the main directory of this archive
* for more details.
*
* Copyright (C) 1999, 2000 Kaz Kojima & Niibe Yutaka
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
* Copyright (C) 2006 ST Microelectronics Ltd. (denorm support)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
*
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
* FIXME! These routines have not been tested for big endian case.
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
*/
#include <linux/sched.h>
#include <linux/signal.h>
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
#include <linux/io.h>
sh: migrate to arch/sh/include/ This follows the sparc changes a439fe51a1f8eb087c22dd24d69cebae4a3addac. Most of the moving about was done with Sam's directions at: http://marc.info/?l=linux-sh&m=121724823706062&w=2 with subsequent hacking and fixups entirely my fault. Signed-off-by: Sam Ravnborg <sam@ravnborg.org> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2008-07-29 06:09:44 +07:00
#include <cpu/fpu.h>
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
#include <asm/processor.h>
sh: Fix occasional FPU register corruption under preempt. Presently with preempt enabled there's the possibility to be preempted after the TIF_USEDFPU test and the register save, leading to bogus state post-__switch_to(). Use an explicit preempt_disable()/enable() pair around unlazy_fpu()/clear_fpu() to avoid this. Follows the x86 change. Reported-by: Takuo Koguchi <takuo.koguchi.sw@hitachi.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2008-03-26 17:02:47 +07:00
#include <asm/fpu.h>
sh: fix up fallout from system.h disintegration. Quite a bit of fallout all over the place, nothing terribly exciting. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2012-03-30 17:29:57 +07:00
#include <asm/traps.h>
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
/* The PR (precision) bit in the FP Status Register must be clear when
* an frchg instruction is executed, otherwise the instruction is undefined.
* Executing frchg with PR set causes a trap on some SH4 implementations.
*/
#define FPSCR_RCHG 0x00000000
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
extern unsigned long long float64_div(unsigned long long a,
unsigned long long b);
extern unsigned long int float32_div(unsigned long int a, unsigned long int b);
extern unsigned long long float64_mul(unsigned long long a,
unsigned long long b);
extern unsigned long int float32_mul(unsigned long int a, unsigned long int b);
extern unsigned long long float64_add(unsigned long long a,
unsigned long long b);
extern unsigned long int float32_add(unsigned long int a, unsigned long int b);
extern unsigned long long float64_sub(unsigned long long a,
unsigned long long b);
extern unsigned long int float32_sub(unsigned long int a, unsigned long int b);
sh: Subnormal double to float conversion This patch adds support for the SH4 to convert a subnormal double into a float by catching the FPE and implementing the FCNVDS instruction in software. Signed-off-by: Carl Shaw <carl.shaw@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2008-09-05 14:36:19 +07:00
extern unsigned long int float64_to_float32(unsigned long long a);
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
static unsigned int fpu_exception_flags;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
/*
* Save FPU registers onto task structure.
*/
sh: Minor optimisations to FPU handling A number of small optimisations to FPU handling, in particular: - move the task USEDFPU flag from the thread_info flags field (which is accessed asynchronously to the thread) to a new status field, which is only accessed by the thread itself. This allows locking to be removed in most cases, or can be reduced to a preempt_lock(). This mimics the i386 behaviour. - move the modification of regs->sr and thread_info->status flags out of save_fpu() to __unlazy_fpu(). This gives the compiler a better chance to optimise things, as well as making save_fpu() symmetrical with restore_fpu() and init_fpu(). - implement prepare_to_copy(), so that when creating a thread, we can unlazy the FPU prior to copying the thread data structures. Also make sure that the FPU is disabled while in the kernel, in particular while booting, and for newly created kernel threads, In a very artificial benchmark, the execution time for 2500000 context switches was reduced from 50 to 45 seconds. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2009-09-26 00:25:10 +07:00
void save_fpu(struct task_struct *tsk)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
{
unsigned long dummy;
enable_fpu();
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
asm volatile ("sts.l fpul, @-%0\n\t"
"sts.l fpscr, @-%0\n\t"
"lds %2, fpscr\n\t"
"frchg\n\t"
"fmov.s fr15, @-%0\n\t"
"fmov.s fr14, @-%0\n\t"
"fmov.s fr13, @-%0\n\t"
"fmov.s fr12, @-%0\n\t"
"fmov.s fr11, @-%0\n\t"
"fmov.s fr10, @-%0\n\t"
"fmov.s fr9, @-%0\n\t"
"fmov.s fr8, @-%0\n\t"
"fmov.s fr7, @-%0\n\t"
"fmov.s fr6, @-%0\n\t"
"fmov.s fr5, @-%0\n\t"
"fmov.s fr4, @-%0\n\t"
"fmov.s fr3, @-%0\n\t"
"fmov.s fr2, @-%0\n\t"
"fmov.s fr1, @-%0\n\t"
"fmov.s fr0, @-%0\n\t"
"frchg\n\t"
"fmov.s fr15, @-%0\n\t"
"fmov.s fr14, @-%0\n\t"
"fmov.s fr13, @-%0\n\t"
"fmov.s fr12, @-%0\n\t"
"fmov.s fr11, @-%0\n\t"
"fmov.s fr10, @-%0\n\t"
"fmov.s fr9, @-%0\n\t"
"fmov.s fr8, @-%0\n\t"
"fmov.s fr7, @-%0\n\t"
"fmov.s fr6, @-%0\n\t"
"fmov.s fr5, @-%0\n\t"
"fmov.s fr4, @-%0\n\t"
"fmov.s fr3, @-%0\n\t"
"fmov.s fr2, @-%0\n\t"
"fmov.s fr1, @-%0\n\t"
"fmov.s fr0, @-%0\n\t"
"lds %3, fpscr\n\t":"=r" (dummy)
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
:"0"((char *)(&tsk->thread.xstate->hardfpu.status)),
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
"r"(FPSCR_RCHG), "r"(FPSCR_INIT)
:"memory");
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
sh: SH-2A FPU support. Signed-off-by: Kieran Bingham <kbingham@mpc-data.co.uk> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-26 18:38:36 +07:00
disable_fpu();
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
}
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
void restore_fpu(struct task_struct *tsk)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
{
unsigned long dummy;
sh: SH-2A FPU support. Signed-off-by: Kieran Bingham <kbingham@mpc-data.co.uk> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-26 18:38:36 +07:00
enable_fpu();
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
asm volatile ("lds %2, fpscr\n\t"
"fmov.s @%0+, fr0\n\t"
"fmov.s @%0+, fr1\n\t"
"fmov.s @%0+, fr2\n\t"
"fmov.s @%0+, fr3\n\t"
"fmov.s @%0+, fr4\n\t"
"fmov.s @%0+, fr5\n\t"
"fmov.s @%0+, fr6\n\t"
"fmov.s @%0+, fr7\n\t"
"fmov.s @%0+, fr8\n\t"
"fmov.s @%0+, fr9\n\t"
"fmov.s @%0+, fr10\n\t"
"fmov.s @%0+, fr11\n\t"
"fmov.s @%0+, fr12\n\t"
"fmov.s @%0+, fr13\n\t"
"fmov.s @%0+, fr14\n\t"
"fmov.s @%0+, fr15\n\t"
"frchg\n\t"
"fmov.s @%0+, fr0\n\t"
"fmov.s @%0+, fr1\n\t"
"fmov.s @%0+, fr2\n\t"
"fmov.s @%0+, fr3\n\t"
"fmov.s @%0+, fr4\n\t"
"fmov.s @%0+, fr5\n\t"
"fmov.s @%0+, fr6\n\t"
"fmov.s @%0+, fr7\n\t"
"fmov.s @%0+, fr8\n\t"
"fmov.s @%0+, fr9\n\t"
"fmov.s @%0+, fr10\n\t"
"fmov.s @%0+, fr11\n\t"
"fmov.s @%0+, fr12\n\t"
"fmov.s @%0+, fr13\n\t"
"fmov.s @%0+, fr14\n\t"
"fmov.s @%0+, fr15\n\t"
"frchg\n\t"
"lds.l @%0+, fpscr\n\t"
"lds.l @%0+, fpul\n\t"
:"=r" (dummy)
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
:"0" (tsk->thread.xstate), "r" (FPSCR_RCHG)
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
:"memory");
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
disable_fpu();
}
/**
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
* denormal_to_double - Given denormalized float number,
* store double float
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
*
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
* @fpu: Pointer to sh_fpu_hard structure
* @n: Index to FP register
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
*/
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
static void denormal_to_double(struct sh_fpu_hard_struct *fpu, int n)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
{
unsigned long du, dl;
unsigned long x = fpu->fpul;
int exp = 1023 - 126;
if (x != 0 && (x & 0x7f800000) == 0) {
du = (x & 0x80000000);
while ((x & 0x00800000) == 0) {
x <<= 1;
exp--;
}
x &= 0x007fffff;
du |= (exp << 20) | (x >> 3);
dl = x << 29;
fpu->fp_regs[n] = du;
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
fpu->fp_regs[n + 1] = dl;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
}
}
/**
* ieee_fpe_handler - Handle denormalized number exception
*
* @regs: Pointer to register structure
*
* Returns 1 when it's handled (should not cause exception).
*/
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
static int ieee_fpe_handler(struct pt_regs *regs)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
{
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
unsigned short insn = *(unsigned short *)regs->pc;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
unsigned short finsn;
unsigned long nextpc;
int nib[4] = {
(insn >> 12) & 0xf,
(insn >> 8) & 0xf,
(insn >> 4) & 0xf,
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
insn & 0xf
};
if (nib[0] == 0xb || (nib[0] == 0x4 && nib[2] == 0x0 && nib[3] == 0xb))
regs->pr = regs->pc + 4; /* bsr & jsr */
if (nib[0] == 0xa || nib[0] == 0xb) {
/* bra & bsr */
nextpc = regs->pc + 4 + ((short)((insn & 0xfff) << 4) >> 3);
finsn = *(unsigned short *)(regs->pc + 2);
} else if (nib[0] == 0x8 && nib[1] == 0xd) {
/* bt/s */
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
if (regs->sr & 1)
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
nextpc = regs->pc + 4 + ((char)(insn & 0xff) << 1);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
else
nextpc = regs->pc + 4;
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
finsn = *(unsigned short *)(regs->pc + 2);
} else if (nib[0] == 0x8 && nib[1] == 0xf) {
/* bf/s */
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
if (regs->sr & 1)
nextpc = regs->pc + 4;
else
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
nextpc = regs->pc + 4 + ((char)(insn & 0xff) << 1);
finsn = *(unsigned short *)(regs->pc + 2);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
} else if (nib[0] == 0x4 && nib[3] == 0xb &&
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
(nib[2] == 0x0 || nib[2] == 0x2)) {
/* jmp & jsr */
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
nextpc = regs->regs[nib[1]];
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
finsn = *(unsigned short *)(regs->pc + 2);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
} else if (nib[0] == 0x0 && nib[3] == 0x3 &&
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
(nib[2] == 0x0 || nib[2] == 0x2)) {
/* braf & bsrf */
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
nextpc = regs->pc + 4 + regs->regs[nib[1]];
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
finsn = *(unsigned short *)(regs->pc + 2);
} else if (insn == 0x000b) {
/* rts */
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
nextpc = regs->pr;
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
finsn = *(unsigned short *)(regs->pc + 2);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
} else {
sh: Fix PC adjustments for varying opcode length. There are a few different cases for figuring out how to size the instruction. We read in the instruction located at regs->pc - 4 when rewinding the opcode to figure out if there's a 32-bit opcode before the faulting instruction, with a default of a - 2 adjustment on a mismatch. In practice this works for the cases where pc - 4 is just another 16-bit opcode, or we happen to have a 32-bit and a 16-bit immediately preceeding the pc value. In the cases where we aren't rewinding, this is much less ugly.. We also don't bother fixing up the places where we're explicitly dealing with 16-bit instructions, since this might lead to confusion regarding the encoding size possibilities on other CPU variants. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-05-08 13:31:48 +07:00
nextpc = regs->pc + instruction_size(insn);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
finsn = insn;
}
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
if ((finsn & 0xf1ff) == 0xf0ad) {
/* fcnvsd */
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
struct task_struct *tsk = current;
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
if ((tsk->thread.xstate->hardfpu.fpscr & FPSCR_CAUSE_ERROR))
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
/* FPU error */
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
denormal_to_double(&tsk->thread.xstate->hardfpu,
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
(finsn >> 8) & 0xf);
else
return 0;
regs->pc = nextpc;
return 1;
} else if ((finsn & 0xf00f) == 0xf002) {
/* fmul */
struct task_struct *tsk = current;
int fpscr;
int n, m, prec;
unsigned int hx, hy;
n = (finsn >> 8) & 0xf;
m = (finsn >> 4) & 0xf;
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
hx = tsk->thread.xstate->hardfpu.fp_regs[n];
hy = tsk->thread.xstate->hardfpu.fp_regs[m];
fpscr = tsk->thread.xstate->hardfpu.fpscr;
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
prec = fpscr & FPSCR_DBL_PRECISION;
if ((fpscr & FPSCR_CAUSE_ERROR)
&& (prec && ((hx & 0x7fffffff) < 0x00100000
|| (hy & 0x7fffffff) < 0x00100000))) {
long long llx, lly;
/* FPU error because of denormal (doubles) */
llx = ((long long)hx << 32)
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
| tsk->thread.xstate->hardfpu.fp_regs[n + 1];
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
lly = ((long long)hy << 32)
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
| tsk->thread.xstate->hardfpu.fp_regs[m + 1];
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
llx = float64_mul(llx, lly);
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
tsk->thread.xstate->hardfpu.fp_regs[n] = llx >> 32;
tsk->thread.xstate->hardfpu.fp_regs[n + 1] = llx & 0xffffffff;
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
} else if ((fpscr & FPSCR_CAUSE_ERROR)
&& (!prec && ((hx & 0x7fffffff) < 0x00800000
|| (hy & 0x7fffffff) < 0x00800000))) {
/* FPU error because of denormal (floats) */
hx = float32_mul(hx, hy);
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
tsk->thread.xstate->hardfpu.fp_regs[n] = hx;
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
} else
return 0;
regs->pc = nextpc;
return 1;
} else if ((finsn & 0xf00e) == 0xf000) {
/* fadd, fsub */
struct task_struct *tsk = current;
int fpscr;
int n, m, prec;
unsigned int hx, hy;
n = (finsn >> 8) & 0xf;
m = (finsn >> 4) & 0xf;
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
hx = tsk->thread.xstate->hardfpu.fp_regs[n];
hy = tsk->thread.xstate->hardfpu.fp_regs[m];
fpscr = tsk->thread.xstate->hardfpu.fpscr;
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
prec = fpscr & FPSCR_DBL_PRECISION;
if ((fpscr & FPSCR_CAUSE_ERROR)
&& (prec && ((hx & 0x7fffffff) < 0x00100000
|| (hy & 0x7fffffff) < 0x00100000))) {
long long llx, lly;
/* FPU error because of denormal (doubles) */
llx = ((long long)hx << 32)
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
| tsk->thread.xstate->hardfpu.fp_regs[n + 1];
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
lly = ((long long)hy << 32)
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
| tsk->thread.xstate->hardfpu.fp_regs[m + 1];
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
if ((finsn & 0xf00f) == 0xf000)
llx = float64_add(llx, lly);
else
llx = float64_sub(llx, lly);
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
tsk->thread.xstate->hardfpu.fp_regs[n] = llx >> 32;
tsk->thread.xstate->hardfpu.fp_regs[n + 1] = llx & 0xffffffff;
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
} else if ((fpscr & FPSCR_CAUSE_ERROR)
&& (!prec && ((hx & 0x7fffffff) < 0x00800000
|| (hy & 0x7fffffff) < 0x00800000))) {
/* FPU error because of denormal (floats) */
if ((finsn & 0xf00f) == 0xf000)
hx = float32_add(hx, hy);
else
hx = float32_sub(hx, hy);
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
tsk->thread.xstate->hardfpu.fp_regs[n] = hx;
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
} else
return 0;
regs->pc = nextpc;
return 1;
} else if ((finsn & 0xf003) == 0xf003) {
/* fdiv */
struct task_struct *tsk = current;
int fpscr;
int n, m, prec;
unsigned int hx, hy;
n = (finsn >> 8) & 0xf;
m = (finsn >> 4) & 0xf;
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
hx = tsk->thread.xstate->hardfpu.fp_regs[n];
hy = tsk->thread.xstate->hardfpu.fp_regs[m];
fpscr = tsk->thread.xstate->hardfpu.fpscr;
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
prec = fpscr & FPSCR_DBL_PRECISION;
if ((fpscr & FPSCR_CAUSE_ERROR)
&& (prec && ((hx & 0x7fffffff) < 0x00100000
|| (hy & 0x7fffffff) < 0x00100000))) {
long long llx, lly;
/* FPU error because of denormal (doubles) */
llx = ((long long)hx << 32)
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
| tsk->thread.xstate->hardfpu.fp_regs[n + 1];
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
lly = ((long long)hy << 32)
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
| tsk->thread.xstate->hardfpu.fp_regs[m + 1];
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
llx = float64_div(llx, lly);
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
tsk->thread.xstate->hardfpu.fp_regs[n] = llx >> 32;
tsk->thread.xstate->hardfpu.fp_regs[n + 1] = llx & 0xffffffff;
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
} else if ((fpscr & FPSCR_CAUSE_ERROR)
&& (!prec && ((hx & 0x7fffffff) < 0x00800000
|| (hy & 0x7fffffff) < 0x00800000))) {
/* FPU error because of denormal (floats) */
hx = float32_div(hx, hy);
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
tsk->thread.xstate->hardfpu.fp_regs[n] = hx;
sh: Set up correct siginfo structures for page faults. Remove the previous saving of fault codes into the thread_struct as they are never used, and appeared to be inherited from x86. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2006-11-21 11:34:04 +07:00
} else
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
return 0;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
sh: Subnormal double to float conversion This patch adds support for the SH4 to convert a subnormal double into a float by catching the FPE and implementing the FCNVDS instruction in software. Signed-off-by: Carl Shaw <carl.shaw@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2008-09-05 14:36:19 +07:00
regs->pc = nextpc;
return 1;
} else if ((finsn & 0xf0bd) == 0xf0bd) {
/* fcnvds - double to single precision convert */
struct task_struct *tsk = current;
int m;
unsigned int hx;
sh: fcnvds fix with denormalized numbers on SH-4 FPU. This fixes a bug in the FPU exception handler for the FCNVDS instruction. To get the register number the instruction is shifted right by 9, though it should be shifted right by 8. More information at ST Linux bugzilla: https://bugzilla.stlinux.com/show_bug.cgi?id=4892 Signed-off-by: Giuseppe Di Giore <giuseppe.di-giore@st.com> Signed-off-by: Carmelo Amoroso <carmelo.amoroso@st.com> Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2008-10-15 22:48:16 +07:00
m = (finsn >> 8) & 0x7;
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
hx = tsk->thread.xstate->hardfpu.fp_regs[m];
sh: Subnormal double to float conversion This patch adds support for the SH4 to convert a subnormal double into a float by catching the FPE and implementing the FCNVDS instruction in software. Signed-off-by: Carl Shaw <carl.shaw@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2008-09-05 14:36:19 +07:00
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
if ((tsk->thread.xstate->hardfpu.fpscr & FPSCR_CAUSE_ERROR)
sh: Subnormal double to float conversion This patch adds support for the SH4 to convert a subnormal double into a float by catching the FPE and implementing the FCNVDS instruction in software. Signed-off-by: Carl Shaw <carl.shaw@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2008-09-05 14:36:19 +07:00
&& ((hx & 0x7fffffff) < 0x00100000)) {
/* subnormal double to float conversion */
long long llx;
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
llx = ((long long)tsk->thread.xstate->hardfpu.fp_regs[m] << 32)
| tsk->thread.xstate->hardfpu.fp_regs[m + 1];
sh: Subnormal double to float conversion This patch adds support for the SH4 to convert a subnormal double into a float by catching the FPE and implementing the FCNVDS instruction in software. Signed-off-by: Carl Shaw <carl.shaw@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2008-09-05 14:36:19 +07:00
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
tsk->thread.xstate->hardfpu.fpul = float64_to_float32(llx);
sh: Subnormal double to float conversion This patch adds support for the SH4 to convert a subnormal double into a float by catching the FPE and implementing the FCNVDS instruction in software. Signed-off-by: Carl Shaw <carl.shaw@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2008-09-05 14:36:19 +07:00
} else
return 0;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
regs->pc = nextpc;
return 1;
}
return 0;
}
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
void float_raise(unsigned int flags)
{
fpu_exception_flags |= flags;
}
int float_rounding_mode(void)
{
struct task_struct *tsk = current;
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
int roundingMode = FPSCR_ROUNDING_MODE(tsk->thread.xstate->hardfpu.fpscr);
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
return roundingMode;
}
sh: SH-2A FPU support. Signed-off-by: Kieran Bingham <kbingham@mpc-data.co.uk> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-26 18:38:36 +07:00
BUILD_TRAP_HANDLER(fpu_error)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
{
struct task_struct *tsk = current;
sh: SH-2A FPU support. Signed-off-by: Kieran Bingham <kbingham@mpc-data.co.uk> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-26 18:38:36 +07:00
TRAP_HANDLER_DECL;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
sh: Minor optimisations to FPU handling A number of small optimisations to FPU handling, in particular: - move the task USEDFPU flag from the thread_info flags field (which is accessed asynchronously to the thread) to a new status field, which is only accessed by the thread itself. This allows locking to be removed in most cases, or can be reduced to a preempt_lock(). This mimics the i386 behaviour. - move the modification of regs->sr and thread_info->status flags out of save_fpu() to __unlazy_fpu(). This gives the compiler a better chance to optimise things, as well as making save_fpu() symmetrical with restore_fpu() and init_fpu(). - implement prepare_to_copy(), so that when creating a thread, we can unlazy the FPU prior to copying the thread data structures. Also make sure that the FPU is disabled while in the kernel, in particular while booting, and for newly created kernel threads, In a very artificial benchmark, the execution time for 2500000 context switches was reduced from 50 to 45 seconds. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2009-09-26 00:25:10 +07:00
__unlazy_fpu(tsk, regs);
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
fpu_exception_flags = 0;
if (ieee_fpe_handler(regs)) {
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
tsk->thread.xstate->hardfpu.fpscr &=
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
~(FPSCR_CAUSE_MASK | FPSCR_FLAG_MASK);
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
tsk->thread.xstate->hardfpu.fpscr |= fpu_exception_flags;
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
/* Set the FPSCR flag as well as cause bits - simply
* replicate the cause */
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
tsk->thread.xstate->hardfpu.fpscr |= (fpu_exception_flags >> 10);
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
grab_fpu(regs);
restore_fpu(tsk);
sh: Minor optimisations to FPU handling A number of small optimisations to FPU handling, in particular: - move the task USEDFPU flag from the thread_info flags field (which is accessed asynchronously to the thread) to a new status field, which is only accessed by the thread itself. This allows locking to be removed in most cases, or can be reduced to a preempt_lock(). This mimics the i386 behaviour. - move the modification of regs->sr and thread_info->status flags out of save_fpu() to __unlazy_fpu(). This gives the compiler a better chance to optimise things, as well as making save_fpu() symmetrical with restore_fpu() and init_fpu(). - implement prepare_to_copy(), so that when creating a thread, we can unlazy the FPU prior to copying the thread data structures. Also make sure that the FPU is disabled while in the kernel, in particular while booting, and for newly created kernel threads, In a very artificial benchmark, the execution time for 2500000 context switches was reduced from 50 to 45 seconds. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2009-09-26 00:25:10 +07:00
task_thread_info(tsk)->status |= TS_USEDFPU;
sh: Move over to dynamically allocated FPU context. This follows the x86 xstate changes and implements a task_xstate slab cache that is dynamically sized to match one of hard FP/soft FP/FPU-less. This also tidies up and consolidates some of the SH-2A/SH-4 FPU fragmentation. Now fpu state restorers are commonly defined, with the init_fpu()/fpu_init() mess reworked to follow the x86 convention. The fpu_init() register initialization has been replaced by xstate setup followed by writing out to hardware via the standard restore path. As init_fpu() now performs a slab allocation a secondary lighterweight restorer is also introduced for the context switch. In the future the DSP state will be rolled in here, too. More work remains for math emulation and the SH-5 FPU, which presently uses its own special (UP-only) interfaces. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2010-01-13 10:51:40 +07:00
if ((((tsk->thread.xstate->hardfpu.fpscr & FPSCR_ENABLE_MASK) >> 7) &
sh: Support denormalization on SH-4 FPU. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:42:27 +07:00
(fpu_exception_flags >> 2)) == 0) {
return;
}
}
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 05:20:36 +07:00
force_sig(SIGFPE, tsk);
}
Reference in New Issue Copy Permalink