Mirror of the gdb-patches mailing list
 help / color / mirror / Atom feed
From: Andrew Burgess <andrew.burgess@embecosm.com>
To: Jozef Lawrynowicz <jozef.l@mittosystems.com>
Cc: gdb-patches@sourceware.org
Subject: Re: [PATCH] MSP430: sim: Fix incorrect simulation of unsigned widening multiply
Date: Tue, 4 Aug 2020 15:55:36 +0100	[thread overview]
Message-ID: <20200804145536.GY853475@embecosm.com> (raw)
In-Reply-To: <20200731141100.adrjos6aixi65r2y@jozef-acer-manjaro>

* Jozef Lawrynowicz <jozef.l@mittosystems.com> [2020-07-31 15:11:00 +0100]:

> Operand sizes used for simulation of MSP430 hardware multiply
> operations are not aligned with the sizes used on the target, resulting
> in the simulator storing signed operands with too much precision.
> 
> Additionally, simulation of unsigned multiplication is missing explicit
> casts to prevent any implicit sign extension.
> 
> gcc.c-torture/execute/pr91450-1.c uses unsigned widening multiplication
> of 32-bit operands -4 and 2, to a 64-bit result:
> 0xffff fffc * 0x2 = 0x1 ffff fff8
> 
> If -4 is stored in 64-bit precision, then the multiplication is
> essentially signed and the result is -8 in 64-bit precision
> (0xffff ffff ffff fffc), which is not correct.
> 
> Successfully regtested the GCC and G++ DejaGNU testsuites for msp430-elf in the
> default configuration, and with -mhwmult=f5series.
> This patch fixes 158 execution failures for -mhwmult=f5series.
> 
> Ok to apply?

Approved.

Thanks,
Andrew

> 
> Thanks,
> Jozef

> From 68a3cab29e116dc91a07add7ee16d19175db36a1 Mon Sep 17 00:00:00 2001
> From: Jozef Lawrynowicz <jozef.l@mittosystems.com>
> Date: Tue, 28 Jul 2020 10:36:10 +0100
> Subject: [PATCH] MSP430: sim: Fix incorrect simulation of unsigned widening
>  multiply
> 
> Operand sizes used for simulation of MSP430 hardware multiply
> operations are not aligned with the sizes used on the target, resulting
> in the simulator storing signed operands with too much precision.
> 
> Additionally, simulation of unsigned multiplication is missing explicit
> casts to prevent any implicit sign extension.
> 
> gcc.c-torture/execute/pr91450-1.c uses unsigned widening multiplication
> of 32-bit operands -4 and 2, to a 64-bit result:
> 0xffff fffc * 0x2 = 0x1 ffff fff8
> 
> If -4 is stored in 64-bit precision, then the multiplication is
> essentially signed and the result is -8 in 64-bit precision
> (0xffff ffff ffff fffc), which is not correct.
> 
> sim/msp430/ChangeLog:
> 
> 2020-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
> 
> 	* msp430-sim.c (put_op): For unsigned multiplication, explicitly cast
> 	operands to the unsigned type before multiplying.
> 	* msp430-sim.h (struct msp430_cpu_state): Fix types used to store hwmult
> 	operands.
> ---
>  sim/msp430/msp430-sim.c | 28 ++++++++++++++++++++--------
>  sim/msp430/msp430-sim.h |  8 ++++----
>  2 files changed, 24 insertions(+), 12 deletions(-)
> 
> diff --git a/sim/msp430/msp430-sim.c b/sim/msp430/msp430-sim.c
> index e21c8cf6a64..a330c6caf5d 100644
> --- a/sim/msp430/msp430-sim.c
> +++ b/sim/msp430/msp430-sim.c
> @@ -566,8 +566,13 @@ put_op (SIM_DESC sd, MSP430_Opcode_Decoded *opc, int n, int val)
>  	      switch (HWMULT (sd, hwmult_type))
>  		{
>  		case UNSIGN_32:
> -		  HWMULT (sd, hwmult_result) = HWMULT (sd, hwmult_op1) * HWMULT (sd, hwmult_op2);
> -		  HWMULT (sd, hwmult_signed_result) = (signed) HWMULT (sd, hwmult_result);
> +		  a = HWMULT (sd, hwmult_op1);
> +		  b = HWMULT (sd, hwmult_op2);
> +		  /* For unsigned 32-bit multiplication of 16-bit operands, an
> +		     explicit cast is required to prevent any implicit
> +		     sign-extension.  */
> +		  HWMULT (sd, hwmult_result) = (unsigned32) a * (unsigned32) b;
> +		  HWMULT (sd, hwmult_signed_result) = a * b;
>  		  HWMULT (sd, hwmult_accumulator) = HWMULT (sd, hwmult_signed_accumulator) = 0;
>  		  break;
>  
> @@ -575,13 +580,16 @@ put_op (SIM_DESC sd, MSP430_Opcode_Decoded *opc, int n, int val)
>  		  a = sign_ext (HWMULT (sd, hwmult_op1), 16);
>  		  b = sign_ext (HWMULT (sd, hwmult_op2), 16);
>  		  HWMULT (sd, hwmult_signed_result) = a * b;
> -		  HWMULT (sd, hwmult_result) = (unsigned) HWMULT (sd, hwmult_signed_result);
> +		  HWMULT (sd, hwmult_result) = (unsigned32) a * (unsigned32) b;
>  		  HWMULT (sd, hwmult_accumulator) = HWMULT (sd, hwmult_signed_accumulator) = 0;
>  		  break;
>  
>  		case UNSIGN_MAC_32:
> -		  HWMULT (sd, hwmult_accumulator) += HWMULT (sd, hwmult_op1) * HWMULT (sd, hwmult_op2);
> -		  HWMULT (sd, hwmult_signed_accumulator) += HWMULT (sd, hwmult_op1) * HWMULT (sd, hwmult_op2);
> +		  a = HWMULT (sd, hwmult_op1);
> +		  b = HWMULT (sd, hwmult_op2);
> +		  HWMULT (sd, hwmult_accumulator)
> +		    += (unsigned32) a * (unsigned32) b;
> +		  HWMULT (sd, hwmult_signed_accumulator) += a * b;
>  		  HWMULT (sd, hwmult_result) = HWMULT (sd, hwmult_accumulator);
>  		  HWMULT (sd, hwmult_signed_result) = HWMULT (sd, hwmult_signed_accumulator);
>  		  break;
> @@ -589,7 +597,8 @@ put_op (SIM_DESC sd, MSP430_Opcode_Decoded *opc, int n, int val)
>  		case SIGN_MAC_32:
>  		  a = sign_ext (HWMULT (sd, hwmult_op1), 16);
>  		  b = sign_ext (HWMULT (sd, hwmult_op2), 16);
> -		  HWMULT (sd, hwmult_accumulator) += a * b;
> +		  HWMULT (sd, hwmult_accumulator)
> +		    += (unsigned32) a * (unsigned32) b;
>  		  HWMULT (sd, hwmult_signed_accumulator) += a * b;
>  		  HWMULT (sd, hwmult_result) = HWMULT (sd, hwmult_accumulator);
>  		  HWMULT (sd, hwmult_signed_result) = HWMULT (sd, hwmult_signed_accumulator);
> @@ -648,10 +657,13 @@ put_op (SIM_DESC sd, MSP430_Opcode_Decoded *opc, int n, int val)
>  	      switch (HWMULT (sd, hw32mult_type))
>  		{
>  		case UNSIGN_64:
> -		  HWMULT (sd, hw32mult_result) = HWMULT (sd, hw32mult_op1) * HWMULT (sd, hw32mult_op2);
> +		  HWMULT (sd, hw32mult_result)
> +		    = (unsigned64) HWMULT (sd, hw32mult_op1)
> +		    * (unsigned64) HWMULT (sd, hw32mult_op2);
>  		  break;
>  		case SIGN_64:
> -		  HWMULT (sd, hw32mult_result) = sign_ext (HWMULT (sd, hw32mult_op1), 32)
> +		  HWMULT (sd, hw32mult_result)
> +		    = sign_ext (HWMULT (sd, hw32mult_op1), 32)
>  		    * sign_ext (HWMULT (sd, hw32mult_op2), 32);
>  		  break;
>  		}
> diff --git a/sim/msp430/msp430-sim.h b/sim/msp430/msp430-sim.h
> index ad83e5b6ae6..7c486c2f350 100644
> --- a/sim/msp430/msp430-sim.h
> +++ b/sim/msp430/msp430-sim.h
> @@ -31,16 +31,16 @@ struct msp430_cpu_state
>    int cio_buffer;
>  
>    hwmult_type  hwmult_type;
> -  unsigned32   hwmult_op1;
> -  unsigned32   hwmult_op2;
> +  unsigned16   hwmult_op1;
> +  unsigned16   hwmult_op2;
>    unsigned32   hwmult_result;
>    signed32     hwmult_signed_result;
>    unsigned32   hwmult_accumulator;
>    signed32     hwmult_signed_accumulator;
>  
>    hw32mult_type  hw32mult_type;
> -  unsigned64     hw32mult_op1;
> -  unsigned64     hw32mult_op2;
> +  unsigned32     hw32mult_op1;
> +  unsigned32     hw32mult_op2;
>    unsigned64     hw32mult_result;
>  };
>  
> -- 
> 2.27.0
> 



      reply	other threads:[~2020-08-04 14:55 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-07-31 14:11 Jozef Lawrynowicz
2020-08-04 14:55 ` Andrew Burgess [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20200804145536.GY853475@embecosm.com \
    --to=andrew.burgess@embecosm.com \
    --cc=gdb-patches@sourceware.org \
    --cc=jozef.l@mittosystems.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox