cores
/
microwatt
mirror of https://github.com/antonblanchard/microwatt
3
1
Fork 1
Code Issues Projects Releases Wiki Activity

FPU: Implement floating round-to-integer instructions

Browse Source 
This implements frin, friz, frip and frim, and adds tests for them.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
pull/245/head
Paul Mackerras 4 years ago
parent 03d1aa968a
commit 0ad2aa3014
5 changed files with 114 additions and 2 deletions
Show Stats Download Patch File Download Diff File

4
decode1.vhdl
Unescape Escape View File

@ -441,6 +441,10 @@ architecture behaviour of decode1 is
2#100000010# => (FPU, OP_FPOP, NONE, FRB, NONE, FRT, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', RC, '0', '0'), -- 2/8=fmr
2#100000100# => (FPU, OP_FPOP, NONE, FRB, NONE, FRT, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', RC, '0', '0'), -- 4/8=fnabs
2#100001000# => (FPU, OP_FPOP, NONE, FRB, NONE, FRT, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', RC, '0', '0'), -- 8/8=fabs
2#100001100# => (FPU, OP_FPOP, NONE, FRB, NONE, FRT, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', RC, '0', '0'), -- 12/8=frin
2#100001101# => (FPU, OP_FPOP, NONE, FRB, NONE, FRT, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', RC, '0', '0'), -- 13/8=friz
2#100001110# => (FPU, OP_FPOP, NONE, FRB, NONE, FRT, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', RC, '0', '0'), -- 14/8=frip
2#100001111# => (FPU, OP_FPOP, NONE, FRB, NONE, FRT, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', RC, '0', '0'), -- 15/8=frim
2#110000000# => (FPU, OP_FPOP, NONE, FRB, NONE, FRT, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '1', '0', RC, '0', '0'), -- 0/12=frsp
2#111000000# => (FPU, OP_FPOP, NONE, FRB, NONE, FRT, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', RC, '0', '0'), -- 0/14=fctiw
2#111000100# => (FPU, OP_FPOP, NONE, FRB, NONE, FRT, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', RC, '0', '0'), -- 4/14=fctiwu

40
fpu.vhdl
Unescape Escape View File

@ -39,7 +39,8 @@ architecture behaviour of fpu is
DO_MCRFS, DO_MTFSB, DO_MTFSFI, DO_MFFS, DO_MTFSF,
DO_FMR,
DO_FCFID, DO_FCTI,
DO_FRSP,
DO_FRSP, DO_FRI,
FRI_1,
INT_SHIFT, INT_ROUND, INT_ISHIFT,
INT_FINAL, INT_CHECK, INT_OFLOW,
FINISH, NORMALIZE,
@ -461,7 +462,11 @@ begin
v.state := DO_MTFSF;
end if;
when "01000" =>
v.state := DO_FMR;
if e_in.insn(9 downto 8) /= "11" then
v.state := DO_FMR;
else
v.state := DO_FRI;
end if;
when "01100" =>
v.state := DO_FRSP;
when "01110" =>
@ -587,6 +592,31 @@ begin
v.instr_done := '1';
v.state := IDLE;

when DO_FRI => -- fri[nzpm]
opsel_a <= AIN_B;
v.result_class := r.b.class;
v.result_sign := r.b.negative;
v.result_exp := r.b.exponent;
v.fpscr(FPSCR_FR) := '0';
v.fpscr(FPSCR_FI) := '0';
if r.b.class = NAN and r.b.mantissa(53) = '0' then
-- Signalling NAN
v.fpscr(FPSCR_VXSNAN) := '1';
invalid := '1';
end if;
if r.b.class = FINITE then
if r.b.exponent >= to_signed(52, EXP_BITS) then
-- integer already, no rounding required
arith_done := '1';
else
v.shift := r.b.exponent - to_signed(52, EXP_BITS);
v.state := FRI_1;
v.round_mode := '1' & r.insn(7 downto 6);
end if;
else
arith_done := '1';
end if;

when DO_FRSP =>
opsel_a <= AIN_B;
v.result_class := r.b.class;
@ -749,6 +779,12 @@ begin
invalid := '1';
arith_done := '1';

when FRI_1 =>
opsel_r <= RES_SHIFT;
set_x := '1';
v.shift := to_signed(-2, EXP_BITS);
v.state := ROUNDING;

when FINISH =>
if r.r(63 downto 54) /= "0000000001" then
renormalize := '1';

71
tests/fpu/fpu.c
Unescape Escape View File

@ -753,6 +753,76 @@ int fpu_test_10(void)
return trapit(0, test10);
}

struct frivals {
unsigned long val;
unsigned long nval;
unsigned long zval;
unsigned long pval;
unsigned long mval;
} frivals[] = {
{ 0x0000000000000000, 0, 0, 0, 0 },
{ 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000 },
{ 0x3fdfffffffffffff, 0, 0, 0x3ff0000000000000, 0 },
{ 0x3ff0000000000000, 0x3ff0000000000000, 0x3ff0000000000000, 0x3ff0000000000000, 0x3ff0000000000000 },
{ 0xbff0000000000000, 0xbff0000000000000, 0xbff0000000000000, 0xbff0000000000000, 0xbff0000000000000 },
{ 0x402123456789abcd, 0x4022000000000000, 0x4020000000000000, 0x4022000000000000, 0x4020000000000000 },
{ 0x406123456789abcd, 0x4061200000000000, 0x4061200000000000, 0x4061400000000000, 0x4061200000000000 },
{ 0x409123456789abcd, 0x4091240000000000, 0x4091200000000000, 0x4091240000000000, 0x4091200000000000 },
{ 0x41c123456789abcd, 0x41c1234567800000, 0x41c1234567800000, 0x41c1234568000000, 0x41c1234567800000 },
{ 0x41d123456789abcd, 0x41d1234567800000, 0x41d1234567800000, 0x41d1234567c00000, 0x41d1234567800000 },
{ 0x41e123456789abcd, 0x41e1234567800000, 0x41e1234567800000, 0x41e1234567a00000, 0x41e1234567800000 },
{ 0x41f123456789abcd, 0x41f1234567900000, 0x41f1234567800000, 0x41f1234567900000, 0x41f1234567800000 },
{ 0xc1f123456789abcd, 0xc1f1234567900000, 0xc1f1234567800000, 0xc1f1234567800000, 0xc1f1234567900000 },
{ 0xc1f1234567880000, 0xc1f1234567900000, 0xc1f1234567800000, 0xc1f1234567800000, 0xc1f1234567900000 },
{ 0x432123456789abcd, 0x432123456789abce, 0x432123456789abcc, 0x432123456789abce, 0x432123456789abcc },
{ 0x433123456789abcd, 0x433123456789abcd, 0x433123456789abcd, 0x433123456789abcd, 0x433123456789abcd },
{ 0x434123456789abcd, 0x434123456789abcd, 0x434123456789abcd, 0x434123456789abcd, 0x434123456789abcd },
{ 0x43c123456789abcd, 0x43c123456789abcd, 0x43c123456789abcd, 0x43c123456789abcd, 0x43c123456789abcd },
{ 0x43d123456789abcd, 0x43d123456789abcd, 0x43d123456789abcd, 0x43d123456789abcd, 0x43d123456789abcd },
{ 0x43e123456789abcd, 0x43e123456789abcd, 0x43e123456789abcd, 0x43e123456789abcd, 0x43e123456789abcd },
{ 0x43f123456789abcd, 0x43f123456789abcd, 0x43f123456789abcd, 0x43f123456789abcd, 0x43f123456789abcd },
{ 0xc3f123456789abcd, 0xc3f123456789abcd, 0xc3f123456789abcd, 0xc3f123456789abcd, 0xc3f123456789abcd },
{ 0x7ff0000000000000, 0x7ff0000000000000, 0x7ff0000000000000, 0x7ff0000000000000, 0x7ff0000000000000 },
{ 0xfff0000000000000, 0xfff0000000000000, 0xfff0000000000000, 0xfff0000000000000, 0xfff0000000000000 },
{ 0x7ff123456789abcd, 0x7ff923456789abcd, 0x7ff923456789abcd, 0x7ff923456789abcd, 0x7ff923456789abcd },
{ 0xfff923456789abcd, 0xfff923456789abcd, 0xfff923456789abcd, 0xfff923456789abcd, 0xfff923456789abcd },
};

int test11(long arg)
{
long i;
unsigned long results[4];
struct frivals *vp = frivals;

for (i = 0; i < sizeof(frivals) / sizeof(frivals[0]); ++i, ++vp) {
set_fpscr(FPS_RN_FLOOR);
asm("lfd 3,0(%0); frin 4,3; stfd 4,0(%1)"
: : "b" (&vp->val), "b" (results) : "memory");
set_fpscr(FPS_RN_NEAR);
asm("friz 5,3; stfd 5,8(%0)" : : "b" (results) : "memory");
set_fpscr(FPS_RN_ZERO);
asm("frip 5,3; stfd 5,16(%0)" : : "b" (results) : "memory");
set_fpscr(FPS_RN_CEIL);
asm("frim 5,3; stfd 5,24(%0)" : : "b" (results) : "memory");
if (results[0] != vp->nval || results[1] != vp->zval ||
results[2] != vp->pval || results[3] != vp->mval) {
print_hex(i, 2, "\r\n");
print_hex(results[0], 16, " ");
print_hex(results[1], 16, " ");
print_hex(results[2], 16, " ");
print_hex(results[3], 16, " ");
return i + 1;
}
}
return 0;
}

int fpu_test_11(void)
{
enable_fp();
return trapit(0, test11);
}

int fail = 0;

void do_test(int num, int (*test)(void))
@ -788,6 +858,7 @@ int main(void)
do_test(8, fpu_test_8);
do_test(9, fpu_test_9);
do_test(10, fpu_test_10);
do_test(11, fpu_test_11);

return fail;
}

BIN
tests/test_fpu.bin
View File

Binary file not shown.

1
tests/test_fpu.console_out
Unescape Escape View File

@ -8,3 +8,4 @@ test 07:PASS
test 08:PASS
test 09:PASS
test 10:PASS
test 11:PASS

Write Preview
Loading…
Cancel
Save