RE: [Blackfin removal] [PATCH 01/28] Blackfin arch: Remove Blackfin CPU arch general support

From: Wu, Aaron
Date: Fri Mar 16 2018 - 02:53:04 EST




> -----Original Message-----
> From: arndbergmann@xxxxxxxxx [mailto:arndbergmann@xxxxxxxxx] On
> Behalf Of Arnd Bergmann
> Sent: 2018å3æ15æ 22:46
> To: Arnd Bergmann <arnd@xxxxxxxx>; linux-arch <linux-arch@xxxxxxxxxxxxxxx>;
> Linux Kernel Mailing List <linux-kernel@xxxxxxxxxxxxxxx>; Lennox Wu
> <lennox.wu@xxxxxxxxx>; Guenter Roeck <linux@xxxxxxxxxxxx>; Al Viro
> <viro@xxxxxxxxxxxxxxxxxx>; Jonas Bonn <jonas@xxxxxxxxxxxx>; Stefan
> Kristiansson <stefan.kristiansson@xxxxxxxxxxxxx>; Stafford Horne
> <shorne@xxxxxxxxx>; David Howells <dhowells@xxxxxxxxxx>; Peter Zijlstra
> <peterz@xxxxxxxxxxxxx>; Wu, Aaron <Aaron.Wu@xxxxxxxxxx>
> Subject: Re: [Blackfin removal] [PATCH 01/28] Blackfin arch: Remove Blackfin
> CPU arch general support
>
> > Signed-off-by: Aaron Wu <aaron.wu@xxxxxxxxxx>
> >
> > Remove Blackfin CPU arch general support
>
> Hi Aaron,
>
> thanks a lot for your series. As you know, I've had almost the same patches in
> my tree already, but now I've cross-checked the two versions, and found small
> omissions or bugs in both. I ended up staying with my version, as that was
> already reviewed by a number of people, and it had the conflicts with the other
> arch removal resolved, but I've folded in everything that your version had that I
> was missing, see below for the diff between my original version and the new
> one.
>
> I've also added an 'Acked-by: Aaron Wu <aaron.wu@xxxxxxxxxx>'
> to each patch that we both had.
>
> I hope that all works for you.
>
> Arnd
>

Hi Arnd,

Sure, this works perfect for me. May patches are there as additional reference,
exactly what we discussed earlier. Thanks very much for help handing/merging this.

By the way, the patches email get bounced from vger, I may give it another try on
reset of them except this arch one so more people can help to review/comment
rest of them, mostly drivers.

Best Regards,
Aaron

> ---
> Documentation/driver-api/usb/writing_musb_glue_layer.rst | 3
> drivers/misc/echo/echo.c | 73 ------------
> drivers/misc/echo/fir.h | 50 --------
> drivers/usb/gadget/function/f_uac1_legacy.c | 2
> drivers/usb/gadget/function/u_uac1_legacy.c | 2
> drivers/usb/gadget/function/u_uac1_legacy.h | 2
> drivers/usb/gadget/legacy/audio.c | 2
> drivers/usb/host/isp1362.h | 4
> drivers/usb/musb/musb_core.c | 2
> drivers/usb/musb/musb_core.h | 13 --
> include/linux/cpuhotplug.h | 1
> include/linux/fb.h | 3
> 14 files changed, 3 insertions(+), 248 deletions(-)
>
> diff --git a/drivers/misc/echo/echo.c b/drivers/misc/echo/echo.c index
> 9597e9523cac..8a5adc0d2e88 100644
> --- a/drivers/misc/echo/echo.c
> +++ b/drivers/misc/echo/echo.c
> @@ -115,78 +115,6 @@
>
> /* adapting coeffs using the traditional stochastic descent (N)LMS algorithm */
>
> -#ifdef __bfin__
> -static inline void lms_adapt_bg(struct oslec_state *ec, int clean, int shift) -{
> - int i;
> - int offset1;
> - int offset2;
> - int factor;
> - int exp;
> - int16_t *phist;
> - int n;
> -
> - if (shift > 0)
> - factor = clean << shift;
> - else
> - factor = clean >> -shift;
> -
> - /* Update the FIR taps */
> -
> - offset2 = ec->curr_pos;
> - offset1 = ec->taps - offset2;
> - phist = &ec->fir_state_bg.history[offset2];
> -
> - /* st: and en: help us locate the assembler in echo.s */
> -
> - /* asm("st:"); */
> - n = ec->taps;
> - for (i = 0; i < n; i++) {
> - exp = *phist++ * factor;
> - ec->fir_taps16[1][i] += (int16_t) ((exp + (1 << 14)) >> 15);
> - }
> - /* asm("en:"); */
> -
> - /* Note the asm for the inner loop above generated by Blackfin gcc
> - 4.1.1 is pretty good (note even parallel instructions used):
> -
> - R0 = W [P0++] (X);
> - R0 *= R2;
> - R0 = R0 + R3 (NS) ||
> - R1 = W [P1] (X) ||
> - nop;
> - R0 >>>= 15;
> - R0 = R0 + R1;
> - W [P1++] = R0;
> -
> - A block based update algorithm would be much faster but the
> - above can't be improved on much. Every instruction saved in
> - the loop above is 2 MIPs/ch! The for loop above is where the
> - Blackfin spends most of it's time - about 17 MIPs/ch measured
> - with speedtest.c with 256 taps (32ms). Write-back and
> - Write-through cache gave about the same performance.
> - */
> -}
> -
> -/*
> - IDEAS for further optimisation of lms_adapt_bg():
> -
> - 1/ The rounding is quite costly. Could we keep as 32 bit coeffs
> - then make filter pluck the MS 16-bits of the coeffs when filtering?
> - However this would lower potential optimisation of filter, as I
> - think the dual-MAC architecture requires packed 16 bit coeffs.
> -
> - 2/ Block based update would be more efficient, as per comments above,
> - could use dual MAC architecture.
> -
> - 3/ Look for same sample Blackfin LMS code, see if we can get dual-MAC
> - packing.
> -
> - 4/ Execute the whole e/c in a block of say 20ms rather than sample
> - by sample. Processing a few samples every ms is inefficient.
> -*/
> -
> -#else
> static inline void lms_adapt_bg(struct oslec_state *ec, int clean, int shift) {
> int i;
> @@ -215,7 +143,6 @@ static inline void lms_adapt_bg(struct oslec_state *ec,
> int clean, int shift)
> ec->fir_taps16[1][i] += (int16_t) ((exp + (1 << 14)) >> 15);
> }
> }
> -#endif
>
> static inline int top_bit(unsigned int bits) { diff --git a/drivers/misc/echo/fir.h
> b/drivers/misc/echo/fir.h index 7b9fabf1fea5..4e0f365f0577 100644
> --- a/drivers/misc/echo/fir.h
> +++ b/drivers/misc/echo/fir.h
> @@ -27,14 +27,6 @@
> #define _FIR_H_
>
> /*
> - Blackfin NOTES & IDEAS:
> -
> - A simple dot product function is used to implement the filter.
> This performs
> - just one MAC/cycle which is inefficient but was easy to implement as a first
> - pass. The current Blackfin code also uses an unrolled form of the filter
> - history to avoid 0 length hardware loop issues. This is wasteful of
> - memory.
> -
> Ideas for improvement:
>
> 1/ Rewrite filter for dual MAC inner loop. The issue here is handling @@ -
> 94,21 +86,13 @@ static inline const int16_t *fir16_create(struct fir16_state_t
> *fir,
> fir->taps = taps;
> fir->curr_pos = taps - 1;
> fir->coeffs = coeffs;
> -#if defined(__bfin__)
> - fir->history = kcalloc(2 * taps, sizeof(int16_t), GFP_KERNEL); -#else
> fir->history = kcalloc(taps, sizeof(int16_t), GFP_KERNEL); -#endif
> return fir->history;
> }
>
> static inline void fir16_flush(struct fir16_state_t *fir) { -#if defined(__bfin__)
> - memset(fir->history, 0, 2 * fir->taps * sizeof(int16_t)); -#else
> memset(fir->history, 0, fir->taps * sizeof(int16_t)); -#endif }
>
> static inline void fir16_free(struct fir16_state_t *fir) @@ -116,42 +100,9 @@
> static inline void fir16_free(struct fir16_state_t *fir)
> kfree(fir->history);
> }
>
> -#ifdef __bfin__
> -static inline int32_t dot_asm(short *x, short *y, int len) -{
> - int dot;
> -
> - len--;
> -
> - __asm__("I0 = %1;\n\t"
> - "I1 = %2;\n\t"
> - "A0 = 0;\n\t"
> - "R0.L = W[I0++] || R1.L = W[I1++];\n\t"
> - "LOOP dot%= LC0 = %3;\n\t"
> - "LOOP_BEGIN dot%=;\n\t"
> - "A0 += R0.L * R1.L (IS) || R0.L = W[I0++] || R1.L = W[I1++];\n\t"
> - "LOOP_END dot%=;\n\t"
> - "A0 += R0.L*R1.L (IS);\n\t"
> - "R0 = A0;\n\t"
> - "%0 = R0;\n\t"
> - : "=&d"(dot)
> - : "a"(x), "a"(y), "a"(len)
> - : "I0", "I1", "A1", "A0", "R0", "R1"
> - );
> -
> - return dot;
> -}
> -#endif
> -
> static inline int16_t fir16(struct fir16_state_t *fir, int16_t sample) {
> int32_t y;
> -#if defined(__bfin__)
> - fir->history[fir->curr_pos] = sample;
> - fir->history[fir->curr_pos + fir->taps] = sample;
> - y = dot_asm((int16_t *) fir->coeffs, &fir->history[fir->curr_pos],
> - fir->taps);
> -#else
> int i;
> int offset1;
> int offset2;
> @@ -165,7 +116,6 @@ static inline int16_t fir16(struct fir16_state_t *fir,
> int16_t sample)
> y += fir->coeffs[i] * fir->history[i - offset1];
> for (; i >= 0; i--)
> y += fir->coeffs[i] * fir->history[i + offset2]; -#endif
> if (fir->curr_pos <= 0)
> fir->curr_pos = fir->taps;
> fir->curr_pos--;
> diff --git a/drivers/usb/gadget/function/f_uac1_legacy.c
> b/drivers/usb/gadget/function/f_uac1_legacy.c
> index 04f4b2862256..24c086bcdeaa 100644
> --- a/drivers/usb/gadget/function/f_uac1_legacy.c
> +++ b/drivers/usb/gadget/function/f_uac1_legacy.c
> @@ -4,8 +4,6 @@
> *
> * Copyright (C) 2008 Bryan Wu <cooloney@xxxxxxxxxx>
> * Copyright (C) 2008 Analog Devices, Inc
> - *
> - * Enter bugs at http://blackfin.uclinux.org/
> */
>
> #include <linux/slab.h>
> diff --git a/drivers/usb/gadget/function/u_uac1_legacy.c
> b/drivers/usb/gadget/function/u_uac1_legacy.c
> index cbc868d117af..5393e5c37a4b 100644
> --- a/drivers/usb/gadget/function/u_uac1_legacy.c
> +++ b/drivers/usb/gadget/function/u_uac1_legacy.c
> @@ -4,8 +4,6 @@
> *
> * Copyright (C) 2008 Bryan Wu <cooloney@xxxxxxxxxx>
> * Copyright (C) 2008 Analog Devices, Inc
> - *
> - * Enter bugs at http://blackfin.uclinux.org/
> */
>
> #include <linux/kernel.h>
> diff --git a/drivers/usb/gadget/function/u_uac1_legacy.h
> b/drivers/usb/gadget/function/u_uac1_legacy.h
> index dd69e408a3d9..5c1bdf46fe32 100644
> --- a/drivers/usb/gadget/function/u_uac1_legacy.h
> +++ b/drivers/usb/gadget/function/u_uac1_legacy.h
> @@ -4,8 +4,6 @@
> *
> * Copyright (C) 2008 Bryan Wu <cooloney@xxxxxxxxxx>
> * Copyright (C) 2008 Analog Devices, Inc
> - *
> - * Enter bugs at http://blackfin.uclinux.org/
> */
>
> #ifndef __U_UAC1_LEGACY_H
> diff --git a/drivers/usb/gadget/legacy/audio.c
> b/drivers/usb/gadget/legacy/audio.c
> index 7b11dce98b94..dd81fd538cb8 100644
> --- a/drivers/usb/gadget/legacy/audio.c
> +++ b/drivers/usb/gadget/legacy/audio.c
> @@ -4,8 +4,6 @@
> *
> * Copyright (C) 2008 Bryan Wu <cooloney@xxxxxxxxxx>
> * Copyright (C) 2008 Analog Devices, Inc
> - *
> - * Enter bugs at http://blackfin.uclinux.org/
> */
>
> /* #define VERBOSE_DEBUG */
> diff --git a/drivers/usb/host/isp1362.h b/drivers/usb/host/isp1362.h index
> 87c1ac97e538..650240846ee2 100644
> --- a/drivers/usb/host/isp1362.h
> +++ b/drivers/usb/host/isp1362.h
> @@ -6,9 +6,7 @@
> */
>
> /* ------------------------------------------------------------------------- */
> -/*
> - * Platform specific compile time options
> - */
> +
> #define MAX_ROOT_PORTS 2
>
> #define USE_32BIT 0
> diff --git a/drivers/usb/musb/musb_core.c b/drivers/usb/musb/musb_core.c
> index 011c8ca7de42..13486588e561 100644
> --- a/drivers/usb/musb/musb_core.c
> +++ b/drivers/usb/musb/musb_core.c
> @@ -2166,7 +2166,7 @@ musb_init_controller(struct device *dev, int nIrq, void
> __iomem *ctrl)
> * - initializes musb->xceiv, usually by otg_get_phy()
> * - stops powering VBUS
> *
> - * There are various transceiver configurations. Blackfin,
> + * There are various transceiver configurations.
> * DaVinci, TUSB60x0, and others integrate them. OMAP3 uses
> * external/discrete ones in various flavors (twl4030 family,
> * isp1504, non-OTG, etc) mostly hooking up through ULPI.
> diff --git a/drivers/usb/musb/musb_core.h b/drivers/usb/musb/musb_core.h
> index ac675b1a34c4..8a74cb2907f8 100644
> --- a/drivers/usb/musb/musb_core.h
> +++ b/drivers/usb/musb/musb_core.h
> @@ -414,19 +414,6 @@ struct musb {
> struct usb_gadget_driver *gadget_driver; /* its driver */
> struct usb_hcd *hcd; /* the usb hcd */
>
> - /*
> - * FIXME: Remove this flag.
> - *
> - * This is only added to allow Blackfin to work
> - * with current driver. For some unknown reason
> - * Blackfin doesn't work with double buffering
> - * and that's enabled by default.
> - *
> - * We added this flag to forcefully disable double
> - * buffering until we get it working.
> - */
> - unsigned double_buffer_not_ok:1;
> -
> const struct musb_hdrc_config *config;
>
> int xceiv_old_state;
> diff --git a/include/linux/cpuhotplug.h b/include/linux/cpuhotplug.h index
> 5b211fe295f0..8796ba387152 100644
> --- a/include/linux/cpuhotplug.h
> +++ b/include/linux/cpuhotplug.h
> @@ -29,7 +29,6 @@ enum cpuhp_state {
> CPUHP_PERF_PREPARE,
> CPUHP_PERF_X86_PREPARE,
> CPUHP_PERF_X86_AMD_UNCORE_PREP,
> - CPUHP_PERF_BFIN,
> CPUHP_PERF_POWER,
> CPUHP_PERF_SUPERH,
> CPUHP_X86_HPET_DEAD,
> diff --git a/include/linux/fb.h b/include/linux/fb.h index
> f577d3c89618..0c20da7ab35f 100644
> --- a/include/linux/fb.h
> +++ b/include/linux/fb.h
> @@ -571,8 +571,7 @@ static inline struct apertures_struct
> *alloc_apertures(unsigned int max_num) {
>
> #elif defined(__i386__) || defined(__alpha__) || defined(__x86_64__) || \
> defined(__hppa__) || defined(__sh__) || defined(__powerpc__) || \
> - defined(__avr32__) || defined(__bfin__) || defined(__arm__) || \
> - defined(__aarch64__)
> + defined(__avr32__) || defined(__arm__) || defined(__aarch64__)
>
> #define fb_readb __raw_readb
> #define fb_readw __raw_readw