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

FPU: Do renormalization from DO_ZERO_DEN state

Browse Source 
Instead of having the various DO_* states (DO_FMUL, DO_FDIV, etc.)
handle checking for denormalized inputs, we now have DO_ZERO_DEN state
check for denormalized inputs and branch to RENORM_{A,B,C} to handle
them.

This also meant some changes were needed in how fsqrt and frsqrte
handled inputs with odd exponent.  The DO_FSQRT and DO_FRSQRTE states
were very similar and have been combined into one.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
pull/442/head
Paul Mackerras 1 year ago
parent 8648ddb64f
commit 70819c4c39
1 changed files with 79 additions and 140 deletions
Show Stats Download Patch File Download Diff File

219
fpu.vhdl
Unescape Escape View File

@ -53,7 +53,7 @@ architecture behaviour of fpu is
DO_FCFID, DO_FCTI, DO_FCFID, DO_FCTI,
DO_FRSP, DO_FRSP_2, DO_FRI, DO_FRSP, DO_FRSP_2, DO_FRI,
DO_FADD, DO_FMUL, DO_FDIV, DO_FSQRT, DO_FMADD, DO_FADD, DO_FMUL, DO_FDIV, DO_FSQRT, DO_FMADD,
DO_FRE, DO_FRSQRTE, DO_FRE,
DO_FSEL, DO_FSEL,
DO_IDIVMOD, DO_IDIVMOD,
FRI_1, FRI_1,
@ -62,10 +62,9 @@ architecture behaviour of fpu is
MULT_1, MULT_1,
FMADD_0, FMADD_1, FMADD_2, FMADD_3, FMADD_0, FMADD_1, FMADD_2, FMADD_3,
FMADD_4, FMADD_5, FMADD_6, FMADD_4, FMADD_5, FMADD_6,
LOOKUP,
DIV_2, DIV_3, DIV_4, DIV_5, DIV_6, DIV_2, DIV_3, DIV_4, DIV_5, DIV_6,
FRE_1, FRE_1,
RSQRT_1, SQRT_ODD, RSQRT_1,
FTDIV_1, FTDIV_1,
SQRT_1, SQRT_2, SQRT_3, SQRT_4, SQRT_1, SQRT_2, SQRT_3, SQRT_4,
SQRT_5, SQRT_6, SQRT_7, SQRT_8, SQRT_5, SQRT_6, SQRT_7, SQRT_8,
@ -76,9 +75,8 @@ architecture behaviour of fpu is
ROUND_UFLOW, ROUND_OFLOW, ROUND_UFLOW, ROUND_OFLOW,
ROUNDING, ROUNDING_2, ROUNDING_3, ROUNDING, ROUNDING_2, ROUNDING_3,
DENORM, DENORM,
RENORM_A, RENORM_A2, RENORM_A, RENORM_B, RENORM_C,
RENORM_B, RENORM_B2, RENORM_1, RENORM_2,
RENORM_C, RENORM_C2,
IDIV_NORMB, IDIV_NORMB2, IDIV_NORMB3, IDIV_NORMB, IDIV_NORMB2, IDIV_NORMB3,
IDIV_CLZA, IDIV_CLZA2, IDIV_CLZA3, IDIV_CLZA, IDIV_CLZA2, IDIV_CLZA3,
IDIV_NR0, IDIV_NR1, IDIV_NR2, IDIV_USE0_5, IDIV_NR0, IDIV_NR1, IDIV_NR2, IDIV_USE0_5,
@ -174,6 +172,7 @@ architecture behaviour of fpu is
res_int : std_ulogic; res_int : std_ulogic;
exec_state : state_t; exec_state : state_t;
cycle_1 : std_ulogic; cycle_1 : std_ulogic;
regsel : std_ulogic_vector(1 downto 0);
end record; end record;


type lookup_table is array(0 to 1023) of std_ulogic_vector(17 downto 0); type lookup_table is array(0 to 1023) of std_ulogic_vector(17 downto 0);
@ -309,7 +308,7 @@ architecture behaviour of fpu is
2#10110# => DO_FSQRT, 2#10110# => DO_FSQRT,
2#11000# => DO_FRE, 2#11000# => DO_FRE,
2#11001# => DO_FMUL, 2#11001# => DO_FMUL,
2#11010# => DO_FRSQRTE, 2#11010# => DO_FSQRT,
2#11100# => DO_FMADD, 2#11100# => DO_FMADD,
2#11101# => DO_FMADD, 2#11101# => DO_FMADD,
2#11110# => DO_FMADD, 2#11110# => DO_FMADD,
@ -870,6 +869,7 @@ begin
variable rsgn_op : std_ulogic_vector(1 downto 0); variable rsgn_op : std_ulogic_vector(1 downto 0);
variable is_nan_inf : std_ulogic; variable is_nan_inf : std_ulogic;
variable is_zero_den : std_ulogic; variable is_zero_den : std_ulogic;
variable set_reg_ind : std_ulogic;
begin begin
v := r; v := r;
v.complete := '0'; v.complete := '0';
@ -1170,6 +1170,7 @@ begin
mult_mask := '0'; mult_mask := '0';
rnd_b32 := '0'; rnd_b32 := '0';
illegal := '0'; illegal := '0';
set_reg_ind := '0';


re_sel1 <= REXP1_ZERO; re_sel1 <= REXP1_ZERO;
re_sel2 <= REXP2_CON; re_sel2 <= REXP2_CON;
@ -1208,6 +1209,7 @@ begin
v.x := '0'; v.x := '0';
v.old_exc := r.fpscr(FPSCR_VX downto FPSCR_XX); v.old_exc := r.fpscr(FPSCR_VX downto FPSCR_XX);
set_s := '1'; set_s := '1';
v.regsel := AIN_R;


when DO_NAN_INF => when DO_NAN_INF =>
-- At least one floating-point operand is infinity or NaN -- At least one floating-point operand is infinity or NaN
@ -1331,6 +1333,14 @@ begin
-- This will trigger for fmul as well as fmadd/sub, but -- This will trigger for fmul as well as fmadd/sub, but
-- it doesn't matter since r.is_subtract = 0 for fmul. -- it doesn't matter since r.is_subtract = 0 for fmul.
rsgn_op := RSGN_SUB; rsgn_op := RSGN_SUB;
end if;
if r.a.denorm = '1' and (r.is_multiply = '1' or r.is_inverse = '1') then
v.state := RENORM_A;
elsif r.c.denorm = '1' then
v.state := RENORM_C;
elsif r.b.denorm = '1' and (r.is_inverse = '1' or r.is_sqrt = '1') then
v.state := RENORM_B;
elsif r.is_multiply = '1' and r.b.class = ZERO then
v.state := DO_FMUL; v.state := DO_FMUL;
else else
v.state := r.exec_state; v.state := r.exec_state;
@ -1639,16 +1649,8 @@ begin
re_sel1 <= REXP1_A; re_sel1 <= REXP1_A;
re_sel2 <= REXP2_C; re_sel2 <= REXP2_C;
re_set_result <= '1'; re_set_result <= '1';
-- Renormalize denorm operands f_to_multiply.valid <= '1';
if r.a.denorm = '1' then v.state := MULT_1;
v.state := RENORM_A;
elsif r.c.denorm = '1' then
opsel_a <= AIN_C;
v.state := RENORM_C;
else
f_to_multiply.valid <= '1';
v.state := MULT_1;
end if;


when DO_FDIV => when DO_FDIV =>
opsel_a <= AIN_A; opsel_a <= AIN_A;
@ -1658,16 +1660,8 @@ begin
re_neg2 <= '1'; re_neg2 <= '1';
re_set_result <= '1'; re_set_result <= '1';
v.count := "00"; v.count := "00";
-- Renormalize denorm operands v.first := '1';
if r.a.denorm = '1' then v.state := DIV_2;
v.state := RENORM_A;
elsif r.b.denorm = '1' then
opsel_a <= AIN_B;
v.state := RENORM_B;
else
v.first := '1';
v.state := DIV_2;
end if;


when DO_FSEL => when DO_FSEL =>
rsgn_op := RSGN_SEL; rsgn_op := RSGN_SEL;
@ -1691,14 +1685,13 @@ begin
if r.b.negative = '1' then if r.b.negative = '1' then
v.fpscr(FPSCR_VXSQRT) := '1'; v.fpscr(FPSCR_VXSQRT) := '1';
qnan_result := '1'; qnan_result := '1';
elsif r.b.denorm = '1' then end if;
v.state := RENORM_B; if r.b.exponent(0) = '1' then
elsif r.b.exponent(0) = '0' then v.state := SQRT_ODD;
elsif r.is_inverse = '0' then
v.state := SQRT_1; v.state := SQRT_1;
else else
-- set shift to 1 v.state := RSQRT_1;
rs_con2 <= RSCON2_1;
v.state := RENORM_B2;
end if; end if;


when DO_FRE => when DO_FRE =>
@ -1706,29 +1699,7 @@ begin
v.result_class := r.b.class; v.result_class := r.b.class;
re_sel2 <= REXP2_B; re_sel2 <= REXP2_B;
re_set_result <= '1'; re_set_result <= '1';
if r.b.denorm = '1' then v.state := FRE_1;
v.state := RENORM_B;
else
v.state := FRE_1;
end if;

when DO_FRSQRTE =>
opsel_a <= AIN_B;
v.result_class := r.b.class;
re_sel2 <= REXP2_B;
re_set_result <= '1';
-- set shift to 1
rs_con2 <= RSCON2_1;
if r.b.negative = '1' then
v.fpscr(FPSCR_VXSQRT) := '1';
qnan_result := '1';
elsif r.b.denorm = '1' then
v.state := RENORM_B;
elsif r.b.exponent(0) = '0' then
v.state := RSQRT_1;
else
v.state := RENORM_B2;
end if;


when DO_FMADD => when DO_FMADD =>
-- fmadd, fmsub, fnmadd, fnmsub -- fmadd, fmsub, fnmadd, fnmsub
@ -1740,14 +1711,7 @@ begin
re_set_result <= '1'; re_set_result <= '1';
-- put b.exp into shift -- put b.exp into shift
rs_sel1 <= RSH1_B; rs_sel1 <= RSH1_B;
-- Make sure A and C are normalized if (r.a.exponent + r.c.exponent + 1) < r.b.exponent then
if r.a.denorm = '1' then
opsel_a <= AIN_A;
v.state := RENORM_A;
elsif r.c.denorm = '1' then
opsel_a <= AIN_C;
v.state := RENORM_C;
elsif (r.a.exponent + r.c.exponent + 1) < r.b.exponent then
-- addend is bigger, do multiply first -- addend is bigger, do multiply first
-- if subtracting, sign is opposite to initial estimate -- if subtracting, sign is opposite to initial estimate
f_to_multiply.valid <= '1'; f_to_multiply.valid <= '1';
@ -1759,68 +1723,48 @@ begin
end if; end if;


when RENORM_A => when RENORM_A =>
rs_norm <= '1'; -- Get A into R
v.state := RENORM_A2; opsel_a <= AIN_A;

v.regsel := AIN_A;
when RENORM_A2 => re_sel1 <= REXP1_A;
set_a := '1';
re_sel2 <= REXP2_NE;
re_set_result <= '1'; re_set_result <= '1';
if r.is_multiply = '1' then v.a.denorm := '0';
opsel_a <= AIN_C; v.state := RENORM_1;
if r.c.mantissa(UNIT_BIT) = '1' then
if r.is_addition = '0' or r.b.class = ZERO then
v.first := '1';
v.state := MULT_1;
else
v.state := DO_FMADD;
end if;
else
v.state := RENORM_C;
end if;
else
opsel_a <= AIN_B;
if r.b.mantissa(UNIT_BIT) = '1' then
v.first := '1';
v.state := DIV_2;
else
v.state := RENORM_B;
end if;
end if;


when RENORM_B => when RENORM_B =>
rs_norm <= '1'; -- Get B into R
renorm_sqrt := r.is_sqrt; opsel_a <= AIN_B;
v.state := RENORM_B2; v.regsel := AIN_B;

re_sel2 <= REXP2_B;
when RENORM_B2 => re_set_result <= '1';
set_b := '1'; v.b.denorm := '0';
-- For fdiv, we need to increase result_exp by shift rather v.state := RENORM_1;
-- than decreasing it as for fre/frsqrte and fsqrt.
-- We do that by negating r.shift in this cycle and then
-- setting result_exp to new_exp in the next cycle
if r.use_a = '1' then
rs_sel1 <= RSH1_S;
rs_neg1 <= '1';
else
re_sel2 <= REXP2_NE;
re_set_result <= '1';
end if;
v.state := LOOKUP;


when RENORM_C => when RENORM_C =>
-- Get C into R
opsel_a <= AIN_C;
v.regsel := AIN_C;
re_sel2 <= REXP2_C;
re_set_result <= '1';
v.c.denorm := '0';
v.state := RENORM_1;

when RENORM_1 =>
rs_norm <= '1'; rs_norm <= '1';
v.state := RENORM_C2; renorm_sqrt := r.is_sqrt;
v.state := RENORM_2;


when RENORM_C2 => when RENORM_2 =>
set_c := '1'; set_reg_ind := '1';
re_sel2 <= REXP2_NE; if r.c.denorm = '1' then
re_set_result <= '1'; -- must be either fmul or fmadd/sub
if r.is_addition = '0' or r.b.class = ZERO then v.state := RENORM_C;
v.first := '1'; elsif r.b.denorm = '1' and r.is_addition = '0' then
v.state := MULT_1; v.state := RENORM_B;
elsif r.is_multiply = '1' and r.b.class = ZERO then
v.state := DO_FMUL;
else else
v.state := DO_FMADD; v.state := r.exec_state;
end if; end if;


when ADD_1 => when ADD_1 =>
@ -2017,28 +1961,6 @@ begin
v.state := NORMALIZE; v.state := NORMALIZE;
end if; end if;


when LOOKUP =>
-- wait one cycle for inverse_table[B] lookup
-- if this is a division, compute exponent
-- (see comment on RENORM_B2 above)
opsel_a <= AIN_B;
if r.use_a = '1' then
re_sel2 <= REXP2_NE;
re_set_result <= '1';
end if;
v.first := '1';
if r.is_sqrt = '1' then
if r.is_inverse = '1' then
v.state := RSQRT_1;
else
v.state := SQRT_1;
end if;
elsif r.use_a = '1' then
v.state := DIV_2;
else
v.state := FRE_1;
end if;

when DIV_2 => when DIV_2 =>
-- compute Y = inverse_table[B] (when count=0); P = 2 - B * Y -- compute Y = inverse_table[B] (when count=0); P = 2 - B * Y
msel_1 <= MUL1_B; msel_1 <= MUL1_B;
@ -2135,6 +2057,12 @@ begin
v.doing_ftdiv := "10"; v.doing_ftdiv := "10";
end if; end if;


when SQRT_ODD =>
-- set shift to 1
rs_con2 <= RSCON2_1;
v.regsel := AIN_B;
v.state := RENORM_2;

when RSQRT_1 => when RSQRT_1 =>
opsel_r <= RES_MISC; opsel_r <= RES_MISC;
misc_sel <= "101"; misc_sel <= "101";
@ -3344,6 +3272,17 @@ begin
end case; end case;
end if; end if;


if set_reg_ind = '1' then
case r.regsel is
when AIN_A =>
set_a := '1';
when AIN_B =>
set_b := '1';
when AIN_C =>
set_c := '1';
when others =>
end case;
end if;
if set_a = '1' or set_a_exp = '1' then if set_a = '1' or set_a_exp = '1' then
v.a.exponent := new_exp; v.a.exponent := new_exp;
end if; end if;

Write Preview
Loading…
Cancel
Save