| Message ID | 839133dd-8ed4-5fec-c311-ce9f8abf3d5f@samsung.com (mailing list archive) |
|---|---|
| State | Superseded, archived |
| Headers | show |
| Series | [1/2] video: fbdev: amifb: remove dead APUS support | expand |
Hi Geert! On 6/2/20 12:37 PM, Bartlomiej Zolnierkiewicz wrote: > > On 5/14/20 10:21 PM, Geert Uytterhoeven wrote: > >> These #ifdefs are relics from APUS (Amiga Power-Up System), which >> added a PPC board. APUS support was killed off a long time ago, >> when arch/ppc/ was still king, but these #ifdefs were missed, because >> they didn't test for CONFIG_APUS. > > Reported-by: Al Viro <viro@zeniv.linux.org.uk> > Reported-by: Geert Uytterhoeven <geert@linux-m68k.org> > Signed-off-by: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com> > --- > drivers/video/fbdev/amifb.c | 63 -------------------------------------------- > 1 file changed, 63 deletions(-) What do you mean with the sentence "when arch/ppc/ was still king"? Does that mean - in the case we would re-add APUS support in the future, that these particular changes would not be necessary? I assume there could be new affordable PowerPC upgrade cards for the Amiga in the future as the PowerPC cards are still sought after by the Amiga community, so there is still demand for those on the market. Adrian
Hi Adrian, On Tue, Jun 2, 2020 at 12:41 PM John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de> wrote: > On 6/2/20 12:37 PM, Bartlomiej Zolnierkiewicz wrote: > >> These #ifdefs are relics from APUS (Amiga Power-Up System), which > >> added a PPC board. APUS support was killed off a long time ago, > >> when arch/ppc/ was still king, but these #ifdefs were missed, because > >> they didn't test for CONFIG_APUS. > > > > Reported-by: Al Viro <viro@zeniv.linux.org.uk> > > Reported-by: Geert Uytterhoeven <geert@linux-m68k.org> > > Signed-off-by: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com> > > --- > > drivers/video/fbdev/amifb.c | 63 -------------------------------------------- > > 1 file changed, 63 deletions(-) > > What do you mean with the sentence "when arch/ppc/ was still king"? Ah, Bartl copied that from my email ;-) There used to be APUS support under arch/ppc/. Later, 32-bit arch/ppc/ and 64-bit arch/ppc64/ were merged in a new\ architecture port under arch/powerpc/, and the old ones were dropped. APUS was never converted, and thus dropped. > Does that mean - in the case we would re-add APUS support in the future, that > these particular changes would not be necessary? They would still be necessary, as PowerPC doesn't grok m68k instructions. Alternatively, we could just drop the m68k inline asm, and retain the C version instead? I have no idea how big of a difference that would make on m68k, using a more modern compiler than when the code was written originally. Note that all of this is used only for cursor handling, which I doubt is actually used by any user space application. The only exception is the DIVUL() macro, which is used once during initialization, thus also not performance critical. Gr{oetje,eeting}s, Geert -- Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that. -- Linus Torvalds
Hi! On 6/2/20 1:04 PM, Geert Uytterhoeven wrote: >> What do you mean with the sentence "when arch/ppc/ was still king"? > > Ah, Bartl copied that from my email ;-) > > There used to be APUS support under arch/ppc/. > Later, 32-bit arch/ppc/ and 64-bit arch/ppc64/ were merged in a new\ > architecture port under arch/powerpc/, and the old ones were dropped. > APUS was never converted, and thus dropped. Ah, yes. Similar to the merge with x86. >> Does that mean - in the case we would re-add APUS support in the future, that >> these particular changes would not be necessary? > > They would still be necessary, as PowerPC doesn't grok m68k instructions. > Alternatively, we could just drop the m68k inline asm, and retain the C > version instead? I have no idea how big of a difference that would make > on m68k, using a more modern compiler than when the code was written > originally. Hmm, no idea. I would keep the assembly for the time being. This was just a question out of curiosity. We could still consider such a change if someone should consider working on APUS support again. > Note that all of this is used only for cursor handling, which I doubt is > actually used by any user space application. The only exception is the > DIVUL() macro, which is used once during initialization, thus also not > performance critical. I see, thanks. Adrian
On 6/2/20 1:07 PM, John Paul Adrian Glaubitz wrote: > Hi! > > On 6/2/20 1:04 PM, Geert Uytterhoeven wrote: >>> What do you mean with the sentence "when arch/ppc/ was still king"? >> >> Ah, Bartl copied that from my email ;-) >> >> There used to be APUS support under arch/ppc/. >> Later, 32-bit arch/ppc/ and 64-bit arch/ppc64/ were merged in a new\ >> architecture port under arch/powerpc/, and the old ones were dropped. >> APUS was never converted, and thus dropped. > > Ah, yes. Similar to the merge with x86. > >>> Does that mean - in the case we would re-add APUS support in the future, that >>> these particular changes would not be necessary? >> >> They would still be necessary, as PowerPC doesn't grok m68k instructions. >> Alternatively, we could just drop the m68k inline asm, and retain the C >> version instead? I have no idea how big of a difference that would make >> on m68k, using a more modern compiler than when the code was written >> originally. > > Hmm, no idea. I would keep the assembly for the time being. This was just > a question out of curiosity. We could still consider such a change if > someone should consider working on APUS support again. > >> Note that all of this is used only for cursor handling, which I doubt is >> actually used by any user space application. The only exception is the >> DIVUL() macro, which is used once during initialization, thus also not >> performance critical. > I see, thanks. Since the code in question is not performance critical it indeed seems to be good idea to use C version. However it still would need be tested on the hardware (or emulator at least) so for the time being I think that we should just add another FIXME comment instead of doing real code changes.. Best regards, -- Bartlomiej Zolnierkiewicz Samsung R&D Institute Poland Samsung Electronics
Hi Bartlomiej, On Tue, 2 Jun 2020 at 11:37, Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com> wrote: > > > On 5/14/20 10:21 PM, Geert Uytterhoeven wrote: > > > These #ifdefs are relics from APUS (Amiga Power-Up System), which > > added a PPC board. APUS support was killed off a long time ago, > > when arch/ppc/ was still king, but these #ifdefs were missed, because > > they didn't test for CONFIG_APUS. > > Reported-by: Al Viro <viro@zeniv.linux.org.uk> > Reported-by: Geert Uytterhoeven <geert@linux-m68k.org> > Signed-off-by: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com> > --- > drivers/video/fbdev/amifb.c | 63 -------------------------------------------- > 1 file changed, 63 deletions(-) > A quick look through my checkout (drm-misc/next aka 5.8 ish), shows multiple other places which check for the define. And a single place where it's being set - the Makefile below. Should those be addressed as well? Or perhaps they are and I've got an old tree. HTH Emil $ git grep -c __mc68000__ arch/m68k/Makefile:1 drivers/block/floppy.c:2 drivers/ide/ide-probe.c:2 drivers/input/misc/hp_sdc_rtc.c:1 drivers/input/serio/hp_sdc.c:3 drivers/input/serio/hp_sdc_mlc.c:1 drivers/net/ethernet/i825xx/82596.c:8 drivers/tty/vt/keyboard.c:1 drivers/video/fbdev/amifb.c:11 include/linux/a.out.h:1 include/linux/hp_sdc.h:1 include/uapi/linux/a.out.h:1 lib/fonts/fonts.c:2 lib/mpi/longlong.h:1
Hi Emil, On Mon, Jun 15, 2020 at 10:38 PM Emil Velikov <emil.l.velikov@gmail.com> wrote: > On Tue, 2 Jun 2020 at 11:37, Bartlomiej Zolnierkiewicz > <b.zolnierkie@samsung.com> wrote: > > On 5/14/20 10:21 PM, Geert Uytterhoeven wrote: > > > These #ifdefs are relics from APUS (Amiga Power-Up System), which > > > added a PPC board. APUS support was killed off a long time ago, > > > when arch/ppc/ was still king, but these #ifdefs were missed, because > > > they didn't test for CONFIG_APUS. > > > > Reported-by: Al Viro <viro@zeniv.linux.org.uk> > > Reported-by: Geert Uytterhoeven <geert@linux-m68k.org> > > Signed-off-by: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com> > > --- > > drivers/video/fbdev/amifb.c | 63 -------------------------------------------- > > 1 file changed, 63 deletions(-) > > > A quick look through my checkout (drm-misc/next aka 5.8 ish), shows > multiple other places which check for the define. > And a single place where it's being set - the Makefile below. > > Should those be addressed as well? Or perhaps they are and I've got an old tree. Only the above apply to APUS support. All other below are probably legitimate. > $ git grep -c __mc68000__ > arch/m68k/Makefile:1 > drivers/block/floppy.c:2 > drivers/ide/ide-probe.c:2 > drivers/input/misc/hp_sdc_rtc.c:1 > drivers/input/serio/hp_sdc.c:3 > drivers/input/serio/hp_sdc_mlc.c:1 > drivers/net/ethernet/i825xx/82596.c:8 > drivers/tty/vt/keyboard.c:1 > drivers/video/fbdev/amifb.c:11 > include/linux/a.out.h:1 > include/linux/hp_sdc.h:1 > include/uapi/linux/a.out.h:1 > lib/fonts/fonts.c:2 > lib/mpi/longlong.h:1 Gr{oetje,eeting}s, Geert
Index: b/drivers/video/fbdev/amifb.c =================================================================== --- a/drivers/video/fbdev/amifb.c +++ b/drivers/video/fbdev/amifb.c @@ -576,14 +576,8 @@ static u_short maxfmode, chipset; #define modx(x, v) ((v) & ((x) - 1)) /* if x1 is not a constant, this macro won't make real sense :-) */ -#ifdef __mc68000__ #define DIVUL(x1, x2) ({int res; asm("divul %1,%2,%3": "=d" (res): \ "d" (x2), "d" ((long)((x1) / 0x100000000ULL)), "0" ((long)(x1))); res;}) -#else -/* We know a bit about the numbers, so we can do it this way */ -#define DIVUL(x1, x2) ((((long)((unsigned long long)x1 >> 8) / x2) << 8) + \ - ((((long)((unsigned long long)x1 >> 8) % x2) << 8) / x2)) -#endif #define highw(x) ((u_long)(x)>>16 & 0xffff) #define loww(x) ((u_long)(x) & 0xffff) @@ -1837,11 +1831,7 @@ static int ami_get_var_cursorinfo(struct const struct amifb_par *par) { register u_short *lspr, *sspr; -#ifdef __mc68000__ register u_long datawords asm ("d2"); -#else - register u_long datawords; -#endif register short delta; register u_char color; short height, width, bits, words; @@ -1868,24 +1858,14 @@ static int ami_get_var_cursorinfo(struct for (width = (short)var->width - 1; width >= 0; width--) { if (bits == 0) { bits = 16; --words; -#ifdef __mc68000__ asm volatile ("movew %1@(%3:w:2),%0 ; swap %0 ; movew %1@+,%0" : "=d" (datawords), "=a" (lspr) : "1" (lspr), "d" (delta)); -#else - datawords = (*(lspr + delta) << 16) | (*lspr++); -#endif } --bits; -#ifdef __mc68000__ asm volatile ( "clrb %0 ; swap %1 ; lslw #1,%1 ; roxlb #1,%0 ; " "swap %1 ; lslw #1,%1 ; roxlb #1,%0" : "=d" (color), "=d" (datawords) : "1" (datawords)); -#else - color = (((datawords >> 30) & 2) - | ((datawords >> 15) & 1)); - datawords <<= 1; -#endif put_user(color, data++); } if (bits > 0) { @@ -1893,17 +1873,8 @@ static int ami_get_var_cursorinfo(struct } while (--words >= 0) ++lspr; -#ifdef __mc68000__ asm volatile ("lea %0@(%4:w:2),%0 ; tstl %1 ; jeq 1f ; exg %0,%1\n1:" : "=a" (lspr), "=a" (sspr) : "0" (lspr), "1" (sspr), "d" (delta)); -#else - lspr += delta; - if (sspr) { - u_short *tmp = lspr; - lspr = sspr; - sspr = tmp; - } -#endif } return 0; } @@ -1912,11 +1883,7 @@ static int ami_set_var_cursorinfo(struct u_char __user *data, struct amifb_par *par) { register u_short *lspr, *sspr; -#ifdef __mc68000__ register u_long datawords asm ("d2"); -#else - register u_long datawords; -#endif register short delta; u_short fmode; short height, width, bits, words; @@ -1958,60 +1925,30 @@ static int ami_set_var_cursorinfo(struct unsigned long tdata = 0; get_user(tdata, data); data++; -#ifdef __mc68000__ asm volatile ( "lsrb #1,%2 ; roxlw #1,%0 ; swap %0 ; " "lsrb #1,%2 ; roxlw #1,%0 ; swap %0" : "=d" (datawords) : "0" (datawords), "d" (tdata)); -#else - datawords = ((datawords << 1) & 0xfffefffe); - datawords |= tdata & 1; - datawords |= (tdata & 2) << (16 - 1); -#endif if (--bits == 0) { bits = 16; --words; -#ifdef __mc68000__ asm volatile ("swap %2 ; movew %2,%0@(%3:w:2) ; swap %2 ; movew %2,%0@+" : "=a" (lspr) : "0" (lspr), "d" (datawords), "d" (delta)); -#else - *(lspr + delta) = (u_short) (datawords >> 16); - *lspr++ = (u_short) (datawords & 0xffff); -#endif } } if (bits < 16) { --words; -#ifdef __mc68000__ asm volatile ( "swap %2 ; lslw %4,%2 ; movew %2,%0@(%3:w:2) ; " "swap %2 ; lslw %4,%2 ; movew %2,%0@+" : "=a" (lspr) : "0" (lspr), "d" (datawords), "d" (delta), "d" (bits)); -#else - *(lspr + delta) = (u_short) (datawords >> (16 + bits)); - *lspr++ = (u_short) ((datawords & 0x0000ffff) >> bits); -#endif } while (--words >= 0) { -#ifdef __mc68000__ asm volatile ("moveql #0,%%d0 ; movew %%d0,%0@(%2:w:2) ; movew %%d0,%0@+" : "=a" (lspr) : "0" (lspr), "d" (delta) : "d0"); -#else - *(lspr + delta) = 0; - *lspr++ = 0; -#endif } -#ifdef __mc68000__ asm volatile ("lea %0@(%4:w:2),%0 ; tstl %1 ; jeq 1f ; exg %0,%1\n1:" : "=a" (lspr), "=a" (sspr) : "0" (lspr), "1" (sspr), "d" (delta)); -#else - lspr += delta; - if (sspr) { - u_short *tmp = lspr; - lspr = sspr; - sspr = tmp; - } -#endif } par->crsr.height = var->height; par->crsr.width = var->width;