Russell, -----Original Message----- From: Russell King - ARM Linux [mailto:linux@xxxxxxxxxxxxxxxx] Sent: Wednesday, May 26, 2010 12:28 AM To: DebBarma, Tarun Kanti Cc: linux-omap@xxxxxxxxxxxxxxx; linux-arm-kernel@xxxxxxxxxxxxxxxxxxx Subject: Re: [PATCH] ARM:VFPv3:enable {d16-d31} access On Tue, May 25, 2010 at 02:39:17PM +0530, DebBarma, Tarun Kanti wrote: > #ifdef CONFIG_VFPv3 > @ d16 - d31 registers > - .irp dr,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15 > -1: mrrc p11, 3, r0, r1, c\dr @ fmrrd r0, r1, d\dr > + .irp dr,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31 > +1: fmrrd r0, r1, d\dr The existing code is correct. For every fmrrd instruction, there is a corresponding mrrc version which assembles to exactly the same opcode. mrrc instructions take: 1. Co-processor number, range 0-15. 2. Opcode number N, range 0-15. 3. Destination register 1, range 0-15. 4. Destination register 2, range 0-15. 5. Co-processor register number R, range 0-15. For fmrrd encodings, the first 16 registers are encoded using N=1 with R=0 to 15. The second 16 registers are encoded using N=3 with R=0 to 15. Specifying a co-processor register number greater than 15 is illegal, hence why the 'irp' specifies the numbers 0 to 15. If we look at the instruction encodings, for MRRC: 3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 | cond |1 1 0 0 0 1 0 1| Rn | Rm |CP Num | N | R | For FMRRD: 3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 | cond |1 1 0 0 0 1 0 1| Rn | Rm |1 0 1 1 0 0 M 1| R | where "M" and "R" together define the register. As I said above, the existing code is correct. What problem are you actually trying to solve here? ###################### I have a test case which exercise all VFP general purpose registers by writing a known value and reading it back using vfp_put_double() and vfp_get_double() APIs. long long d1=777, d2=0; int i=31; for (; i>=0; i--){ vfp_put_double(d1, i); d2 = vfp_get_double(i); printk("D%d read=%lld\n",i, (long long)d2); d2 = 0.0; } 1) With the existing implementation I am not able to correctly write/read {d0-d15} but not the {d16-d31} set 2) With my changes I am able to write/read correctly. Tarun -- To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
