* PowerPC: Create new MMA instruction masks and use them
@ 2022-07-21 0:01 Peter Bergner
2022-07-21 4:10 ` Alan Modra
0 siblings, 1 reply; 3+ messages in thread
From: Peter Bergner @ 2022-07-21 0:01 UTC (permalink / raw)
To: Binutils
The MMA instructions use XX3_MASK|3<<21 as an instruction mask, but that
misses the RC bit/bit 31, so if we disassemble a .long that represents an
MMA instruction except that it also has bit 31 set, we will erroneously
disassemble it to that MMA instruction. We create new masks defines that
contain bit 31 so that doesn't happen anymore.
This passed testing on powerpc64le-linux with no regressions.
I plan on pushing this tomorrow unless I hear some objections.
Peter
opcodes/
* ppc-opc.c (XACC_MASK, XX3ACC_MASK): New defines.
(P_GER_MASK, xxmfacc, xxmtacc, xxsetaccz, xvi8ger4pp, xvi8ger4,
xvf16ger2pp, xvf16ger2, xvf32gerpp, xvf32ger, xvi4ger8pp, xvi4ger8,
xvi16ger2spp, xvi16ger2s, xvbf16ger2pp, xvbf16ger2, xvf64gerpp,
xvf64ger, xvi16ger2, xvf16ger2np, xvf32gernp, xvi8ger4spp, xvi16ger2pp,
xvbf16ger2np, xvf64gernp, xvf16ger2pn, xvf32gerpn, xvbf16ger2pn,
xvf64gerpn, xvf16ger2nn, xvf32gernn, xvbf16ger2nn, xvf64gernn: Use them.
diff --git a/opcodes/ppc-opc.c b/opcodes/ppc-opc.c
index f5bd6dac50f..7637d3e349e 100644
--- a/opcodes/ppc-opc.c
+++ b/opcodes/ppc-opc.c
@@ -3876,7 +3876,7 @@ const unsigned int num_powerpc_operands = (sizeof (powerpc_operands)
#define P_U8XX4_MASK (P_XX4_MASK & ~(0xffULL << 32))
/* MMIRR:XX3-form 8-byte outer product instructions. */
-#define P_GER_MASK ((-1ULL << 40) | XX3_MASK | (3 << 21) | 1)
+#define P_GER_MASK ((-1ULL << 40) | XX3ACC_MASK)
#define P_GER2_MASK (P_GER_MASK & ~(3ULL << 46))
#define P_GER4_MASK (P_GER_MASK & ~(15ULL << 44))
#define P_GER8_MASK (P_GER_MASK & ~(255ULL << 40))
@@ -4404,6 +4404,12 @@ const unsigned int num_powerpc_operands = (sizeof (powerpc_operands)
/* The mask for an XX3 form instruction with the BF bits specified. */
#define XX3BF_MASK (XX3 (0x3f, 0xff) | (3 << 21) | (1))
+/* An X_MASK with an accumulator register and the RA and RB fields fixed. */
+#define XACC_MASK (X_MASK | RA_MASK | RB_MASK | (3 << 21))
+
+/* The mask for an XX3 form instruction with an accumulator register. */
+#define XX3ACC_MASK (XX3 (0x3f, 0xff) | (3 << 21) | (1))
+
/* The mask for an XX3 form instruction with the DM or SHW bits
specified. */
#define XX3DM_MASK (XX3 (0x3f, 0x1f) | (1 << 10))
@@ -7154,9 +7160,9 @@ const struct powerpc_opcode powerpc_opcodes[] = {
{"msgclrp", XRTRA(31,174,0,0), XRTRA_MASK, POWER8, 0, {RB}},
{"dcbtlse", X(31,174), X_MASK, PPCCHLK, E500MC, {CT, RA0, RB}},
-{"xxmfacc", XVA(31,177,0), XRARB_MASK|3<<21, POWER10, 0, {ACC}},
-{"xxmtacc", XVA(31,177,1), XRARB_MASK|3<<21, POWER10, 0, {ACC}},
-{"xxsetaccz", XVA(31,177,3), XRARB_MASK|3<<21, POWER10, 0, {ACC}},
+{"xxmfacc", XVA(31,177,0), XACC_MASK, POWER10, 0, {ACC}},
+{"xxmtacc", XVA(31,177,1), XACC_MASK, POWER10, 0, {ACC}},
+{"xxsetaccz", XVA(31,177,3), XACC_MASK, POWER10, 0, {ACC}},
{"mtmsrd", X(31,178), XRLARB_MASK, PPC64, 0, {RS, A_L}},
@@ -8797,8 +8803,8 @@ const struct powerpc_opcode powerpc_opcodes[] = {
{"dqua", ZRC(59,3,0), Z2_MASK, POWER6, PPCVLE, {FRT,FRA,FRB,RMC}},
{"dqua.", ZRC(59,3,1), Z2_MASK, POWER6, PPCVLE, {FRT,FRA,FRB,RMC}},
-{"xvi8ger4pp", XX3(59,2), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
-{"xvi8ger4", XX3(59,3), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
+{"xvi8ger4pp", XX3(59,2), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
+{"xvi8ger4", XX3(59,3), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
{"fdivs", A(59,18,0), AFRC_MASK, PPC, PPCEFS|PPCVLE, {FRT, FRA, FRB}},
{"fdivs.", A(59,18,1), AFRC_MASK, PPC, PPCEFS|PPCVLE, {FRT, FRA, FRB}},
@@ -8849,8 +8855,8 @@ const struct powerpc_opcode powerpc_opcodes[] = {
{"dquai", ZRC(59,67,0), Z2_MASK, POWER6, PPCVLE, {TE, FRT,FRB,RMC}},
{"dquai.", ZRC(59,67,1), Z2_MASK, POWER6, PPCVLE, {TE, FRT,FRB,RMC}},
-{"xvf16ger2pp", XX3(59,18), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
-{"xvf16ger2", XX3(59,19), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
+{"xvf16ger2pp", XX3(59,18), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
+{"xvf16ger2", XX3(59,19), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
{"dscri", ZRC(59,98,0), Z_MASK, POWER6, PPCVLE, {FRT, FRA, SH16}},
{"dscri.", ZRC(59,98,1), Z_MASK, POWER6, PPCVLE, {FRT, FRA, SH16}},
@@ -8858,31 +8864,31 @@ const struct powerpc_opcode powerpc_opcodes[] = {
{"drintx", ZRC(59,99,0), Z2_MASK, POWER6, PPCVLE, {R, FRT, FRB, RMC}},
{"drintx.", ZRC(59,99,1), Z2_MASK, POWER6, PPCVLE, {R, FRT, FRB, RMC}},
-{"xvf32gerpp", XX3(59,26), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
-{"xvf32ger", XX3(59,27), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
+{"xvf32gerpp", XX3(59,26), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
+{"xvf32ger", XX3(59,27), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
{"dcmpo", X(59,130), X_MASK, POWER6, PPCVLE, {BF, FRA, FRB}},
-{"xvi4ger8pp", XX3(59,34), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
-{"xvi4ger8", XX3(59,35), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
+{"xvi4ger8pp", XX3(59,34), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
+{"xvi4ger8", XX3(59,35), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
{"dtstex", X(59,162), X_MASK, POWER6, PPCVLE, {BF, FRA, FRB}},
-{"xvi16ger2spp", XX3(59,42), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
-{"xvi16ger2s", XX3(59,43), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
+{"xvi16ger2spp",XX3(59,42), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
+{"xvi16ger2s", XX3(59,43), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
{"dtstdc", Z(59,194), Z_MASK, POWER6, PPCVLE, {BF, FRA, DCM}},
-{"xvbf16ger2pp",XX3(59,50), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
-{"xvbf16ger2", XX3(59,51), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
+{"xvbf16ger2pp",XX3(59,50), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
+{"xvbf16ger2", XX3(59,51), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
{"dtstdg", Z(59,226), Z_MASK, POWER6, PPCVLE, {BF, FRA, DGM}},
{"drintn", ZRC(59,227,0), Z2_MASK, POWER6, PPCVLE, {R, FRT, FRB, RMC}},
{"drintn.", ZRC(59,227,1), Z2_MASK, POWER6, PPCVLE, {R, FRT, FRB, RMC}},
-{"xvf64gerpp", XX3(59,58), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6ap, XB6a}},
-{"xvf64ger", XX3(59,59), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6ap, XB6a}},
+{"xvf64gerpp", XX3(59,58), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6ap, XB6a}},
+{"xvf64ger", XX3(59,59), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6ap, XB6a}},
{"dctdp", XRC(59,258,0), X_MASK, POWER6, PPCVLE, {FRT, FRB}},
{"dctdp.", XRC(59,258,1), X_MASK, POWER6, PPCVLE, {FRT, FRB}},
@@ -8893,22 +8899,22 @@ const struct powerpc_opcode powerpc_opcodes[] = {
{"ddedpd", XRC(59,322,0), X_MASK, POWER6, PPCVLE, {SP, FRT, FRB}},
{"ddedpd.", XRC(59,322,1), X_MASK, POWER6, PPCVLE, {SP, FRT, FRB}},
-{"xvi16ger2", XX3(59,75), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
+{"xvi16ger2", XX3(59,75), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
-{"xvf16ger2np", XX3(59,82), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
+{"xvf16ger2np", XX3(59,82), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
{"dxex", XRC(59,354,0), X_MASK, POWER6, PPCVLE, {FRT, FRB}},
{"dxex.", XRC(59,354,1), X_MASK, POWER6, PPCVLE, {FRT, FRB}},
-{"xvf32gernp", XX3(59,90), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
+{"xvf32gernp", XX3(59,90), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
-{"xvi8ger4spp", XX3(59,99), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
+{"xvi8ger4spp", XX3(59,99), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
-{"xvi16ger2pp", XX3(59,107), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
+{"xvi16ger2pp", XX3(59,107), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
-{"xvbf16ger2np",XX3(59,114), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
+{"xvbf16ger2np",XX3(59,114), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
-{"xvf64gernp", XX3(59,122), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6ap, XB6a}},
+{"xvf64gernp", XX3(59,122), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6ap, XB6a}},
{"dsub", XRC(59,514,0), X_MASK, POWER6, PPCVLE, {FRT, FRA, FRB}},
{"dsub.", XRC(59,514,1), X_MASK, POWER6, PPCVLE, {FRT, FRA, FRB}},
@@ -8916,18 +8922,18 @@ const struct powerpc_opcode powerpc_opcodes[] = {
{"ddiv", XRC(59,546,0), X_MASK, POWER6, PPCVLE, {FRT, FRA, FRB}},
{"ddiv.", XRC(59,546,1), X_MASK, POWER6, PPCVLE, {FRT, FRA, FRB}},
-{"xvf16ger2pn", XX3(59,146), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
+{"xvf16ger2pn", XX3(59,146), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
-{"xvf32gerpn", XX3(59,154), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
+{"xvf32gerpn", XX3(59,154), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
{"dcmpu", X(59,642), X_MASK, POWER6, PPCVLE, {BF, FRA, FRB}},
{"dtstsf", X(59,674), X_MASK, POWER6, PPCVLE, {BF, FRA, FRB}},
{"dtstsfi", X(59,675), X_MASK|1<<22,POWER9, PPCVLE, {BF, UIM6, FRB}},
-{"xvbf16ger2pn",XX3(59,178), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
+{"xvbf16ger2pn",XX3(59,178), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
-{"xvf64gerpn", XX3(59,186), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6ap, XB6a}},
+{"xvf64gerpn", XX3(59,186), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6ap, XB6a}},
{"drsp", XRC(59,770,0), X_MASK, POWER6, PPCVLE, {FRT, FRB}},
{"drsp.", XRC(59,770,1), X_MASK, POWER6, PPCVLE, {FRT, FRB}},
@@ -8938,7 +8944,7 @@ const struct powerpc_opcode powerpc_opcodes[] = {
{"denbcd", XRC(59,834,0), X_MASK, POWER6, PPCVLE, {S, FRT, FRB}},
{"denbcd.", XRC(59,834,1), X_MASK, POWER6, PPCVLE, {S, FRT, FRB}},
-{"xvf16ger2nn", XX3(59,210), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
+{"xvf16ger2nn", XX3(59,210), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
{"fcfids", XRC(59,846,0), XRA_MASK, POWER7|PPCA2, PPCVLE, {FRT, FRB}},
{"fcfids.", XRC(59,846,1), XRA_MASK, POWER7|PPCA2, PPCVLE, {FRT, FRB}},
@@ -8946,14 +8952,14 @@ const struct powerpc_opcode powerpc_opcodes[] = {
{"diex", XRC(59,866,0), X_MASK, POWER6, PPCVLE, {FRT, FRA, FRB}},
{"diex.", XRC(59,866,1), X_MASK, POWER6, PPCVLE, {FRT, FRA, FRB}},
-{"xvf32gernn", XX3(59,218), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
+{"xvf32gernn", XX3(59,218), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
-{"xvbf16ger2nn",XX3(59,242), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
+{"xvbf16ger2nn",XX3(59,242), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6a, XB6a}},
{"fcfidus", XRC(59,974,0), XRA_MASK, POWER7|PPCA2, PPCVLE, {FRT, FRB}},
{"fcfidus.", XRC(59,974,1), XRA_MASK, POWER7|PPCA2, PPCVLE, {FRT, FRB}},
-{"xvf64gernn", XX3(59,250), XX3_MASK|3<<21, POWER10, PPCVLE, {ACC, XA6ap, XB6a}},
+{"xvf64gernn", XX3(59,250), XX3ACC_MASK, POWER10, PPCVLE, {ACC, XA6ap, XB6a}},
{"xsaddsp", XX3(60,0), XX3_MASK, PPCVSX2, PPCVLE, {XT6, XA6, XB6}},
{"xsmaddasp", XX3(60,1), XX3_MASK, PPCVSX2, PPCVLE, {XT6, XA6, XB6}},
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: PowerPC: Create new MMA instruction masks and use them
2022-07-21 0:01 PowerPC: Create new MMA instruction masks and use them Peter Bergner
@ 2022-07-21 4:10 ` Alan Modra
2022-07-21 20:17 ` Peter Bergner
0 siblings, 1 reply; 3+ messages in thread
From: Alan Modra @ 2022-07-21 4:10 UTC (permalink / raw)
To: Peter Bergner; +Cc: Binutils
On Wed, Jul 20, 2022 at 07:01:35PM -0500, Peter Bergner wrote:
> The MMA instructions use XX3_MASK|3<<21 as an instruction mask, but that
> misses the RC bit/bit 31, so if we disassemble a .long that represents an
> MMA instruction except that it also has bit 31 set, we will erroneously
> disassemble it to that MMA instruction. We create new masks defines that
> contain bit 31 so that doesn't happen anymore.
>
> This passed testing on powerpc64le-linux with no regressions.
> I plan on pushing this tomorrow unless I hear some objections.
2.39 branch too, please.
--
Alan Modra
Australia Development Lab, IBM
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: PowerPC: Create new MMA instruction masks and use them
2022-07-21 4:10 ` Alan Modra
@ 2022-07-21 20:17 ` Peter Bergner
0 siblings, 0 replies; 3+ messages in thread
From: Peter Bergner @ 2022-07-21 20:17 UTC (permalink / raw)
To: Alan Modra; +Cc: Binutils
On 7/20/22 11:10 PM, Alan Modra wrote:
> On Wed, Jul 20, 2022 at 07:01:35PM -0500, Peter Bergner wrote:
>> The MMA instructions use XX3_MASK|3<<21 as an instruction mask, but that
>> misses the RC bit/bit 31, so if we disassemble a .long that represents an
>> MMA instruction except that it also has bit 31 set, we will erroneously
>> disassemble it to that MMA instruction. We create new masks defines that
>> contain bit 31 so that doesn't happen anymore.
>>
>> This passed testing on powerpc64le-linux with no regressions.
>> I plan on pushing this tomorrow unless I hear some objections.
>
> 2.39 branch too, please.
Ok, pushed to trunk and the 2.39 branch (after testing).
Peter
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2022-07-21 20:22 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-07-21 0:01 PowerPC: Create new MMA instruction masks and use them Peter Bergner
2022-07-21 4:10 ` Alan Modra
2022-07-21 20:17 ` Peter Bergner
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).